R for Windows Changelog

What's new in R for Windows 4.3.3

Mar 1, 2024

NEW FEATURES:
iconv() now fixes up variant encoding names such as "utf8" case-insensitively.
DEPRECATED AND DEFUNCT:
The legacy encoding = "MacRoman" is deprecated in pdf() and postscript(): support was incomplete in earlier versions of R.
BUG FIXES:
Arguments are now properly forwarded to methods on S4 generics with ... in the middle of their formal arguments. This was broken for the case when a method introduced an argument but did not include ... in its own formals. Thanks to Herv'e Pag`es for the report PR#18538.
Some invalid file arguments to pictex(), postscript() and xfig() opened a file called NA rather than throw an error. These included postscript(NULL) (which some people expected to work like pdf(NULL)).
Passing filename = NA to svg(), cairo_pdf(), cairo_ps() or the Cairo-based bitmap devices opened a file called NA: it now throws an error.
quartz(file = NA) opened a file called NA, including when used as a Quartz-based bitmap device. It now gives an error.
rank(<long vector>) now works, fixing PR#18617, thanks to Ilia Kats.
seq.int() did not adequately check its length.out argument.
match(<POSIXct>, .) is correct again for differing time zones, ditto for "POSIXlt", fixing PR#18618 reported by Bastian Klein.
drop.terms(*, dropx = <0-length>) now works, fixing PR#18563 as proposed by Mikael Jagan.
drop.terms(*) keeps + offset(.) terms when it should, PR#18565, and drop.terms() no longer makes up a response, PR#18566, fixing both bugs thanks to Mikael Jagan.
getS3method("t", "test") no longer finds the t.test() function, fixing PR#18627.
pdf() and postscript() support for the documented Adobe encodings "Greek" and "Cyrilllic" was missing (although the corresponding Windows' codepages could be used).
Computations of glyph metric information for pdf() and postscript() did not take into account that transliteration could replace one character by two or more (only seen on macOS 14) and typically warned that the information was not known.
rank(x) no longer overflows during integer addition, when computing rank average for largish but not-yet long vector x, fixing PR#18630, thanks to Ilia Kats.
list.files() on Windows now returns also files with names longer that 260 bytes (the Windows limit is 260 characters). Previously, some file names particularly with 'East Asian' characters were omitted.
cov2cor(<0 x 0>) now works, fixing PR#18423 thanks to Mikael Jagan and Elin Waring.
cov2cor(<negative diagonal>) and similar now give one warning instead of two, with better wording, fixing PR#18424 thanks to Mikael Jagan.
tools:: startDynamicHelp() now ensures port is in proper range, fixing PR#18645.
pbeta(x, a,b) is correct now for x=0 or 1 in the boundary cases where a or b or both are 0, fixing PR#18672 thanks to Michael Fay.
pmatch(x, table) for large table, also called for data frame rowselection, dfrm[nm, ], is now interruptible, fixing PR#18656.
predict(<rank-deficient lm>, newdata=*) fix computing of nbasis, see Russ Lenth's comment 29 in PR#16158.
Added a work-around for a bug in macOS 14.3.1 and higher which prevents R plots in the Quartz Cocoa device from updating on screen.

New in R for Windows 4.3.2 (Nov 1, 2023)

NEW FEATURES:
The default initialization of the "repos" option from the repositories file at startup can be skipped by setting environment variable R_REPOSITORIES to NULL such that getOption("repos") is empty if not set elsewhere.
qr.X() is now an implicit S4 generic in methods.
iconv(to = "ASCII//TRANSLIT") is emulated using substitution on platforms which do not support it (notably Alpine Linux). This should give a human-readable conversion in ASCII on all platforms (rather than NA_character_).
trans3d() gains options continuous and verbose addressing the problem of possible "wrap around" when projecting too long curves, as reported by Achim Zeileis in PR#18537.
tools::showNonASCII() has been rewritten to work better on macOS 14 (which has a changed implementation of iconv()).
tiff(type = "quartz") (the default on macOS) now warns if compression is specified: it continues to be ignored.
INSTALLATION on a UNIX-ALIKE:
There is some support for building with Intel's LLVM-based compilers on x86_64 Linux, such as (C) icx, (C++) ipcx and (Fortran) ifx from oneAPI 2023.x.y.
There is support for using LLVM's flang-new as the Fortran compiler from LLVM 16.0.x (preferably 17.0.0 or later).
UTILITIES:
R CMD check reports the use of the Fortran 90 random number generator RANDOM_NUMBER() and the subroutines to initialize it. 'Writing R Extensions' has example code to use R's RNGs from Fortran.
BUG FIXES:
substr(x, n, L) <- cc now works (more) correctly for multibyte UTF-8 strings x when L > nchar(x), thanks to a report and patch by 'Architect 95'.
contrib.url(character()) now returns 0-length character() as documented, which also avoids spurious warnings from available.packages() et al. in the edge case of an empty vector of repository URLs.
readChar(., 4e8) no longer fails, thanks to Kodi Arfer's report (PR#18557).
lapply(<list>, as.data.frame) no longer warns falsely for some base vector components.
Communication between parent and child processes in the multicore part of parallel could fail on platforms that do not support an arbitrarily large payload in system functions read()/write() on pipes (seen on macOS where a restriction to INT_MAX bytes is documented, without doing a partial read unlike Linux). The payload is now split into 1Gb chunks to avoid that problem.
(PR#18571)
qqplot(x,y, conf.level=.) gives better confidence bounds when length(x) != length(y), thanks to Alexander Ploner's report and patch proposal (PR#18557).
norm(<0-length>, "2") now gives zero instead of an error, as all the other norm types, thanks to Mikael Jagan's PR#18542.
Build-stage Rd macros packageAuthor and packageMaintainer now process Authors@R, fixing NA results when the package DESCRIPTION omits Author and Maintainer fields.
Formatting and printing complex numbers could give things like 0.1683-0i because of rounding error: -0i is now replaced by +0i.
postscript() refused to accept a title comment containing the letter "W" (PR#18599).
isoreg(c(1,Inf)) signals an error instead of segfaulting, fixing PR#18603.
tiff(type = "Xlib") was only outputting the last page of multi-page plots.
tools::latexToUtf8() again knows about ~{n} and other letters with tilde, fixing a regression in R 4.3.0, and about ^{i} as an alternative to ^{i} (similarly with other accents). Furthermore, LaTeX codes for accented I letters are now correctly converted, also fixing related mistakes in tools::encoded_text_to_latex().
tar(*, tar = "internal") no longer creates out-of-spec tar files in the very rare case of user or group names longer than 32
bytes, fixing PR#17871 with thanks to Ivan Krylov.
When using the "internal" timezone datetime code, adding a fraction of a second no longer adds one second, fixing PR#16856
from a patch by Ivan Krylov.
tools::checkRd() no longer produces spurious notes about "unnecessary braces" from multi-line Rd results of Sexpr macros.

New in R for Windows 4.3.1 (Jun 16, 2023)

New in R for Windows 4.3.0 (Apr 22, 2023)

SIGNIFICANT USER-VISIBLE CHANGES:
Calling && or || with LHS or (if evaluated) RHS of length greater than one is now always an error, with a report of the form
'length = 4' in coercion to 'logical(1)'
Environment variable _R_CHECK_LENGTH_1_LOGIC2_ no longer has any effect.
NEW FEATURES:
The included BLAS sources have been updated to those shipped with LAPACK version 3.10.1. (This caused some platform-dependent changes to package check output.) And then to the sources from LAPACK version 3.11.0 (with changes only to double complex subroutines).
The included LAPACK sources have been updated to include the four Fortran 90 routines rather than their Fortran 77 predecessors. This may give some different signs in SVDs or eigendecompositions.. (This completes the transition to LAPACK 3.10.x begun in R 4.2.0.)
The LAPACK sources have been updated to version 3.11.0. (No new subroutines have been added, so this almost entirely bug fixes: Those fixes do affect some computations with NaNs, including R's NA.)
The parser now signals _classed_ errors, notably in case of the pipe operator |>. The error object and message now give line and column numbers, mostly as proposed and provided by Duncan Murdoch in PR#18328.
toeplitz() is now generalized for asymmetric cases, with a toeplitz2() variant.
xy.coords() and xyz.coords() and consequently, e.g., plot(x,y, log = "y") now signal a _classed_ warning about negative values of y (where log(.) is NA). Such a warning can be specificallysuppressed or caught otherwise.
Regular expression functions now check more thoroughly whether their inputs are valid strings (in their encoding, e.g. in UTF-8).
The performance of grep(), sub(), gsub() and strsplit() has been improved, particularly with perl = TRUE and fixed = TRUE. Use of useBytes = TRUE for performance reasons should no longer be needed and is discouraged: it may lead to incorrect results.
apropos() gains an argument dot_internals which is used by the completion (help(rcompgen)) engine to also see base internals such as .POSIXct().
Support in tools::Rdiff() for comparing uncompressed PDF files is further reduced - see its help page.
qqplot(x, y, ...) gains conf.level and conf.args arguments for computing and plotting a confidence band for the treatment function transforming the distribution of x into the distribution of y (Switzer, 1976, _Biometrika_). Contributed by Torsten Hothorn.
Performance of package_dependencies() has been improved for cases when the number of dependencies is large.
Strings newly created by gsub(), sub() and strsplit(), when any of the inputs is marked as "bytes", are also marked as "bytes". This reduces the risk of creating invalid strings and accidental substitution of bytes deemed invalid.
Support for readLines(encoding = "bytes") has been added to allow processing special text files byte-by-byte, without creating invalid strings.
iconv(from = "") now takes into account any declared encoding of the input elements and uses it in preference to the native encoding. This reduces the risk of accidental creation of invalid strings, particularly when different elements of the input have different encoding (including "bytes").
Package repositories in getOption("repos") are now initialized from the repositories file when utils is loaded (if not already set, e.g., in .Rprofile). (From a report and patch proposal by Gabriel Becker in PR#18405.)
compactPDF() gets a verbose option.
type.convert() and hence read.table() get new option tryLogical = TRUE with back compatible default. When set to false, converts "F" or "T" columns to character.
Added new unit prefixes "R" and "Q" for abbreviating (unrealistically large) sizes beyond 10^{27} in standard = "SI", thanks to Henrik Bengtsson's PR#18435.
as.data.frame()'s default method now also works fine with atomic objects inheriting from classes such as "roman", "octmode" and "hexmode", such fulfilling the wish of PR#18421, by Benjamin Feakins.
The as.data.frame.vector() utility now errors for wrong-length row.names. It warned for almost six years, with "Will be an error!".
sessionInfo() now also contains La_version() and reports codepage and timezone when relevant, in both print() and toLatex() methods which also get new option tzone for displaying timezone information when locale = FALSE.
New function R_compiled_by() reports the C and Fortran compilers used to build R, if known.
predict(<lm>, newdata = *) no longer unnecessarily creates an offset of all 0s.
solve() for complex inputs now uses argument tol and by default checks for 'computational singularity' (as it long has done for numeric inputs).
predict(<rank-deficient lm>, newdata=*) now obeys a new argument rankdeficient, with new default "warnif", warning only if there are non-estimable cases in newdata. Other options include rankdeficient = "NA", predicting NA for non-estimable newdata cases. This addresses PR#15072 by Russ Lenth and is based on his original proposal and discussions in PR#16158 also by David Firth and Elin Waring. Still somewhat experimental.
Rgui console implementation now works better with the NVDA screen reader when the full blinking cursor is selected. The underlying improvements in cursor handling may help also other screenreaders on Windows.
The drop-field control in GraphApp can now be left with the TAB key and all controls can be navigated in the reverse order using the Shift+TAB key, improving accessibility of the Rgui configuration editor.
qnorm(<very large negative>, log.p=TRUE) is now fully accurate (instead of to "only" minimally five digits).
demo(error.catching) now also shows off withWarnings() and tryCatchWEMs().
As an experimental feature the placeholder _ can now also be used in the rhs of a forward pipe |> expression as the first argument in an extraction call, such as _$coef. More generally, it can be used as the head of a chain of extractions, such as _$coef[[2]].
Spaces in the environment variable used to choose the R session's temporary directory (TMPDIR, TMP and TEMP are tried in turn) are now fatal. (On Windows the 'short path' version of the path is tried and used if that does not contain a space.)
all.equal.numeric() gets a new optional switch giveErr to return the numeric error as attribute. Relatedly, stopifnot(all.equal<some>(a, b, ..)) is as "smart" now, as stopifnot(all.equal(....)) has been already, thus allowing
customized all.equal<Some>() wrappers.
R on Windows is now able to work with path names longer than 260 characters when these are enabled in the system (requires at least Windows 10 version 1607). Packages should be updated to work with long paths as well, instead of assuming PATH_MAX to be the maximum length. Custom front-ends and applications embedding
R need to update their manifests if they wish to allow thisfeature. See <https://blog.r-project.org/2023/03/07/path-length-limit-on-windows> for more information.
'Object not found' and 'Missing argument' errors now give a more accurate error context. Patch provided by Lionel Henry in PR#18241.
The @ operator is now an S3 generic. Based on contributions by Tomasz Kalinowski in PR#18482.
New generic chooseOpsMethod() provides a mechanism for objects to resolve cases where two suitable methods are found for an Ops Group Generic. This supports experimenting with alternative object systems. Based on contributions by Tomasz Kalinowski in PR#18484.
inherits(x, what) now accepts values other than a simple character vector for argument what. A new generic, nameOfClass(), is called to resolve the class name from what. This supports experimenting with alternative object systems. Based on contributions by Tomasz Kalinowski in PR#18485.
Detection of BLAS/LAPACK in use (sessionInfo()) with FlexiBLAS now reports the current backend.
The "data.frame" method for subset() now warns about extraneous arguments, typically catching the use of = instead of == in the subset expression.
Calling a:b when numeric a or b is longer than one may now be made into an error by setting environment variable _R_CHECK_LENGTH_COLON_ to a true value, along the proposal in PR#18419 by Henrik Bengtsson.
density(x, weights = *) now warns if automatic bandwidth selection happens without using weights; new optional warnWbw may suppress the warning. Prompted by Christoph Dalitz' PR#18490 and its discussants.
rm(list = *) is faster and more readable thanks to Kevin Ushey's PR#18492.
The plot.lm() function no longer produces a normal Q-Q plot for GLMs. Instead it plots a half-normal Q-Q plot of the absolute value of the standardized deviance residuals.
The print() method for class "summary.glm" no longer shows summary statistics for the deviance residuals by default. Its optional argument show.residuals can be used to show them if required.
The tapply() function now accepts a data frame as its X argument, and allows INDEX to be a formula in that case. by.data.frame() similarly allows INDICES to be a formula.
The performance of df[j] <- value (including for missing j) and write.table(df) has been improved for data frames df with a large number of columns. (Thanks to Gabriel Becker's PR#18500, PR#18503 and discussants, prompted by a report from Toby Dylan Hocking on the R-devel mailing list.)
The matrix multiply operator %*% is now an S3 generic, belonging to new group generic matrixOps. From Tomasz Kalinowski's contribution in PR#18483.
New function array2DF() to convert arrays to data frames, particularly useful for the list arrays created by tapply().
DATES and TIMES:
On platforms where (non-UTC) datetimes before 1902 (or before 1900 as with system functions on recent macOS) are guessed by extrapolating time zones from 1902-2037, there is a warning at the first use of extrapolation in a session. (As all time zones post 2037 are extrapolation, we do not warn on those.)
(Platforms using --with-internal-tzone, including Windows and by default macOS). How years are printed in dates or date-times can be controlled by environment variable R_PAD_YEARS_BY_ZERO. The default remains to pad to 4 digits by zeroes, but setting value no gives no padding (as used by default by glibc).
strftime() tries harder to determine the offset for the "%z" format, and succeeds on the mainstream R platforms.
strftime() has a limit of 2048 bytes on the string produced - attempting to exceed this is an error. (Previously it silently truncated at 255 bytes.)
sessionInfo() records (and by default prints) the system time zone as part of the locale information. Also, the source (system/internal) of the date-time conversion and printing functions.
Objects of class "POSIXlt" created in this version of R always have 11 components: component zone is always set, and component gmtoff is set for times in UTC and usually set on the (almost all) platforms which have C-level support, otherwise is NA.
There are comprehensive validity checks on the structure of objects of class "POSIXlt" when converting (including formatting and printing). (This avoids mis-conversions of hand-crafted objects.)
There is some support for using the native date-time routines on macOS: this is only viable on recent versions (e.g. 12.6 and 13) and does get wrong some historical changes (before 1900, during WWII). Use of --with-internal-tzone remains the default.
as.POSIXct(<numeric>) and as.POSIXlt(.) (without specifying origin) now work. So does as.Date(<numeric>).
as.Date.POSIXct(., tz) now treats several tz values, notably "GMT" as equivalent to "UTC", proposed and improved by Michael Chirico and Joshua Ulrich in PR#17674.
Experimental balancePOSIXlt() utility allows using "ragged" and or out-of-range "POSIXlt" objects more correctly, e.g., in subsetting and subassignments. Such objects are now documented.
More experimentally, a "POSIXlt" object may have an attribute "balanced" indicating if it is known to be filled or fully balanced.
Functions axis.Date() and axis.POSIXct() are rewritten to gain better default tick locations and better default formats by using prettyDate(). Thanks to Swetlana Herbrandt.
The mapping of Windows' names for time zones to IANA's 'Olson' names has been updated. When ICU is available (it is by
default), it is used to get a mapping for the current region set in Windows. This can be overridden by setting environment variable TZ to the desired Olson name - see OlsonNames() for those currently available.
GRAPHICS:
The graphics engine version, R_GE_version, has been bumped to 16 and so packages that provide graphics devices should be reinstalled.
The grDevices and grid packages have new functions for rendering typeset glyphs, primarily: grDevices::glyphInfo() and
grid::grid.glyph().
Rendering of typeset glyphs is only supported so far on the Cairo-based graphics devices and on the pdf() and quartz() devices.
The defined behaviour for "clear" and "source" compositing operators (via grid::grid.group()) has been changed (to align better with simple interpretation of original Porter-Duff definitions).
Support for gradients, patterns, clipping paths, masks, groups, compositing operators, and affine transformations has been added to the quartz() device.
INSTALLATION on a UNIX-ALIKE:
A system installation of generic LAPACK 3.10.0 or later will be preferred to the version in the R sources.
configure option --with-lapack=no (equivalently --without-lapack) forces compilation of the internal LAPACK sources.
If --with-lapack is not specified, a system liblapack is looked for and used if it reports version 3.10.0 or later and does not
contain BLAS routines.
Packages using LAPACK will need to be reinstalled if this changes to using an external library.
On aarch64 Linux platforms using GCC, configure now defaults to -fPIC (instead of -fpic), as desired in PR#18326.
configure now checks conversion of datetimes between POSIXlt and POSIXct around year 2020. Failure (which has been seen on platforms missing tzdata) is fatal.
If configure option --with-valgrind-instrumentation is given value 1 or 2, option --with-system-valgrind-headers is now the
default and ignored (with a warning). It is highly recommended that the system headers are installed alongside valgrind: they are part of its packaging on some Linux distributions and packaged separately (e.g. in the valgrind-devel RPM) on others. configure will give a warning if they are not found.
The system headers will be required in a future release of R to build with valgrind instrumentation.
libcurl 8.x is now accepted by configure: despite a change in major version number it changes neither API nor ABI.
INSTALLATION on WINDOWS:
The makefiles and installer scripts for Windows have been tailored to Rtools43, an update of the Rtools42 toolchain. It is based on gcc 12 and newer versions of MinGW-W64, binutils and libraries. At this time R-devel can still be built using Rtools42 without changes, but when R-devel is installed via the installer, it will by default look for Rtools43.
Old make targets rsync-extsoft and 32-bit ones that are no longer needed have been removed.
Default builds (including for packages) no longer select C99. Thus the C standard used is the default for the compiler, which
for the toolchain in Rtools43 is C17. (This is consistent with Unix builds.)
PACKAGE INSTALLATION:
The default C++ standard has been changed to C++17 where available (which it is on all currently checked platforms): if not C++14 or C++11 is used if available otherwise C++ is not supported.
USE_FC_LEN_T is the default: this uses the correct (compiler-dependent) prototypes for Fortran BLAS/LAPACK routines
called from C/C++, and requires adjustment of many such calls - see 'Writing R Extensions' SS6.6.1.
There is initial support for C++23 as several compilers are now supporting -std=c++23 or -std=c++2b or similar. As for C++20, there no additional configure checks for C++23 features beyond a check that the compiler reports a __cplusplus value greater than that in the C++20 standard. C++ feature tests should be used.
There is support for a package to indicate the version of the C standard which should be used to compile it, and for the installing user to specify this. In most cases R defaults to the C compiler's default standard which is C17 (a `bug-fix' of C11) -
earlier versions of R or compilers may have defaulted to C99.
Current options are:
USE_C17 Use a standard that is at most C17. The intention is to allow legacy packages to still be installed when later C standards become the default, including packages using new keywords as identifiers or with K&R-style function declarations. This will use C17 if available, falling back to C11.
USE_C90 Use the C90 (aka C89) standard. (As that standard did not require compilers to identify that version, all we can verify is that the compiler does not claim to be using a later standard. It may accept C99 features - for example clang accepts // to make comments.)
USE_C99 Use the C99 standard. This should be rarely needed - it avoids the few new features of C11/C17 which can be useful if a package assumes them if C17 is specified and they are not implemented.
USE_C23 Use C23 (or in future, later). Compiler/library support for C23 is still being implemented, but LLVM clang from 15.0.0 and the upcoming GCC 13 have quite extensive support.
These can be specified as part of the SystemRequirements field in the package's DESCRIPTION file or _via_ options --use-C17 and so on of R CMD INSTALL and R CMD SHLIB.
For further details see "Writing R Extensions" SS1.2.5.
(Windows) A src/Makefile.ucrt or src/Makefile.win file is no included after R_HOME/etcR_ARCH/Makeconf and so no longer needs to include that file itself. Installation of a package with such a file now uses a site Makevars file in the same way as a package with a src/Makevars.win file would.
configure is now passed crucial variables such as CC and CFLAGS in its environment, as many packages were not setting them (as documented in 'Writing R Extensions' SS1.2).
This has most effect where configure is used to compile parts of the package - most often by cmake or libtool which obfuscate the actual compile commands used.
Also used for configure.win and configure.ucrt on Windows.
FORTRAN FLAGS:
The flag -fno-optimize-sibling-calls is no longer forced for gfortran 7 and later. It should no longer be needed now using 'hidden' character-length arguments when calling BLAS/LAPACK routines from C/C++ is the default even for packages. (Unless perhaps packages call Fortran code from C/C++ without using R's headers and without allowing for these arguments.)
C-LEVEL FACILITIES:
The deprecated S-compatibility macros DOUBLE_* in R_ext/Constants.h (included by R.h) have been removed.
The deprecated legacy typedefs of Sint and Sfloat in header R.h are no longer defined, and that header no longer includes header limits.h from C nor climits from C++.
New macro CAD5R() is provided in Rinternals.h and used in a few places in the R sources.
ALTREP now supports VECSXP vectors. Contributed by Gabor Csardi in PR#17620.
The Rcomplex definition (in header R_ext/Complex.h) has been extended to prevent possible mis-compilation when interfacin with Fortran (PR#18430). The new definition causes compiler warnings with static initializers such as {1, 2}, which can b changed to {.r=1, .i=2}.
Using the new definition from C++ depends on compiler extensions supporting C features that have not been incorporated into the C++ standards but are available in g++ and clang++: this may result in C++ compiler warnings but these have been worked around for recent versions of common compilers (GCC, Apple/LLVM clang, Intel).
It is intended to change the inclusion of header R_ext/Complex.h by other R headers, so C/C++ code files which make use of Rcomplex should include that header explicitly.
UTILITIES:
R CMD check does more checking of package .Rd files, warning about invalid email addresses and (some) invalid URIs and noting empty item labels in description lists.
R CMD check now also reports problems when reading package news in md (file NEWS.md) and (optionally) plain text (file NEWS) formats.
_R_CHECK_TIMINGS_ defaults to a value from the environment even for R CMD check --as-cran; this allows for exceptionally fast or slow platforms.
It now applies to checking PDF and HTML versions of the manuals, and 'checking CRAN incoming feasibility'.
R CMD check can optionally (but included in --as-cran) check whether HTML math rendering _via_ KaTeX works for the package .Rd files.
Non-interactive debugger invocations can be trapped by setting the environment variable _R_CHECK_BROWSER_NONINTERACTIVE_ to a true value. This is enabled by R CMD check --as-cran to detect the use of leftover browser() statements in the package.
The use of sprintf and vsprintf from C/C++ has been deprecated in macOS 13 and is a known security risk. R CMD check now reports (on all platforms) if their use is found in compiled code replace by snprintf or vsnprintf respectively. [*NB:* whether such calls get compiled into the package is platform-dependent.]
Where recorded at installation, R CMD check reports the C and Fortran compilers used to build R.
It reports the OS in use (if known, as given by osVersion) as well as that R was built for.
It notes if a C++ standard was specified which is older than the current default: many packages have used C++11 to mean 'not C++98' - as C++11 is the minimum supported since R 4.0.0, that specification can probably be removed.
R CMD INSTALL reports the compilers (and on macOS, the SDK) used, and this is copied to the output of R CMD check.
Where a C++ standard is specified, it is reported.
R CMD check's 'checking compilation flags in Makevars' has been relaxed to accept the use of flags such as -std=f2008 in PKG_FFLAGS.
tools::buildVignettes() has a new argument skip, which is used by R CMD check to skip (and note) vignettes with unavailable VignetteDepends (PR#18318).
New generic .AtNames() added to enable class-specific completions after @. The formerly internal function findMatches() is now exported, mainly for use in methods for .DollarNames() and.AtNames().
DEPRECATED AND DEFUNCT:
default.stringsAsFactors() is defunct.
Calling as.data.frame.<class>() directly (for 12 atomic classes) is going to be formally deprecated, currently activated by
setting the environment variable_R_CHECK_AS_DATA_FRAME_EXPLICIT_METHOD_ to non-empty, which also happens in R CMD check --as-cran.
BUG FIXES:
Hashed environments with sizes less than 5 can now grow. (Reported to R-devel by Duncan Garmonsway.)
as.character(<Rd>, deparse = TRUE) failed to re-escape curly braces in LaTeX-like text. (Reported by Hadley Wickham in PR#18324.)
library() now passes its lib.loc argument when requiring Depends packages; reported (with fix) in PR#18331 by Mikael Jagan.
R CMD Stangle: improved message about 'Output' files.
head(x, n) and tail(x, n) now signal an error if n is not numeric, instead of incidentally "working" sometimes returning all of x. Reported and discussed by Colin Fay, in PR#18357.
The "lm" method for summary() now gives the correct F-statisti when the model contains an offset. Reported in PR#18008.
C() and `contrasts<-`() now preserve factor level names when given a function object (as opposed a function name which did preserve names). Reported in PR#17616.
c(a = 1, 2)[[]] no longer matches 2 but rather signals a_classed_ error. Reported and analysed by Davis Vaughan i PR#18367, a duplicate of PR#18004, by Jan Meis et al. For consistency, NULL[[]] is also erroneous now. x[[]] <- v gives an error of the same class "MissingSubscriptError".
The relist() function of utils now supports NULL elements in the skeleton (PR#15854).
ordered(levels = *) (missing x) now works analogously to factor(, ordered=TRUE); reported (with fix) by Achim Zeileis in PR#18389.
User-defined Rd macro definitions can now span multiple lines, thanks to a patch from Duncan Murdoch. Previously, the Rd parser silently ignored everything after the first line.
Plain-text help (tools::Rd2txt()) now preserves an initial blank line for text following description list items.
tools::Rd2HTML() and tools::Rd2latex() no longer split arguments and value lists at Rd comments.
tools::Rd2latex() now correctly handles optional text outside items of argument lists as well as bracketed text at the beginning of sections, e.g., value{[NULL]}.
as.character(<POSIXt>) now behaves more in line with the method for atomic vectors such as numbers, and is no longer influenced by options(). Ditto for as.character(<Date>). The as.character() method gets arguments digits and OutDec with
defaults _not_ depending on options(). Use of as.character(*, format = .) now warns.
Similarly, the as.character.hexmode() and *.octmode() methods also behave as good citizen methods and back compatibility option keepStr = TRUE.
The as.POSIXlt(<POSIXlt>) and as.POSIXct(<POSIXct>) default methods now do obey their tz argument, also in this case.
as.POSIXlt(<Date>) now does apply a tz (time zone) argument, as does as.POSIXct(); partly suggested by Roland Fuss on the R-devel mailing list.
as.Date.POSIXlt(x) now also works when the list components are of unequal length, aka "partially filled" or "ragged".
expand.model.frame() looked up variables in the wrong environment when applied to models fitted without data. Reported in PR#18414.
time() now (also) uses the ts.eps = getOption("ts.eps") argument and thus by default rounds values very close to the start (or end) of a year. Based on a proposal by Andre"i V. Kostyrka on R-help.
Printing of a factanal() result with just one factor and sort = TRUE now works regularly, fixing PR#17863 by Timothy Bates, thanks to the 'R Contributors' working group.
Printing 0-length objects of class "factor", "roman", "hexmode", "octmode", "person", "bibentry", or "citation" now prints something better, one of which fixes PR#18422, reported by Benjamin Feakins.
Sys.timezone() queries timedatectl only if systemd is loaded; addressing a report by Jan Gorecki in PR#17421.
The formula method of cor.test() had scoping problems when environment(formula) was not the calling environment; reported with a patch proposal by Mao Kobayashi in PR#18439.
attach() of an environment with active bindings now preserves the active bindings. Reported by Kevin Ushey in PR#18425.
BLAS detection now works also with system-provided libraries not available as regular files. This fixes detection of the Accelerate framework on macOS since Big Sur. Reported by David Novgorodsky.
download.file() gives a helpful error message in case of an invalid download.file.method option, thanks to Colin Fay's report in PR#18455.
Sporadic crashes of Rterm when using completion have been fixed.
Rprof() is now more reliable. A livelock in thread initialization with too short sampling interval has been fixed on macOS. A deadlock in using the C runtime has been fixed on Windows. A potential deadlock has been prevented on Unix.
Cursor placement in Rgui now works even after a fixed-width font is selected.
Mandatory options (options()) are now set on startup so that saving and restoring them always works (PR#18372).
Package installation, R CMD INSTALL or install.packages(*), now parses each of the <pkg>/R/*.R files individually instead of first concatenating and then parse()ing the large resulting file. This allows parser or syntax errors to be diagnosed with correct file names and line numbers, thanks to Simon Dedman's report and Bill Dunlap's patch in PR#17859.
This _does_ require syntactically self contained R source files now, fixing another inadvertent bug.
predict.lm(<model with offset>) now finds the offset in the correct environment, thanks to Andr'e Gillibert's report and patch in PR#18456.
getInitial(<formula>) now finds the selfStart model in the correct environment. (Reported by Ivan Krylov in PR#18368.)
Fix for possible segfault when using recently-added graphics features, such as gradients, clipping paths, masks, and groups with pdf(file=NULL).
class(m) <- class(m) no longer changes a matrix m by adding a class _attribute_.
packageDate(pkg) now only warns once if there is no pkg.
When ts() creates a multivariate time series, "mts", it also inherits from "array" now, and is.mts() is documented _and stricter.
Rd2txt() now preserves line breaks of verb Rd content and from duplicated cr. The former also fixes the rendering of verbatim output from Rd Sexpr in plain-text help.
uniroot(f, interval) should no longer wrongly converge _outside_ the interval in some cases where abs(f(x)) == Inf for an x at the interval boundary, thanks to posts by Ben Bolker and Serguei Sokol on R-devel.
Vectorized alpha handling in palette functions such as in gray(), rainbow(), or hcl.colors() works correctly now, thanks to Achim Zeileis' report and patch in PR#18476.
Formatting and print()ing of bibentry objects has dropped the deprecated citation.bibtex.max argument, such that the bibtex argument's default for print.bibentry() depends directly on the citation.bibtex.max option, whereas in format.bibentry() the option no longer applies.
Attempting to use a character string naming a foreign function entry point in a foreign function call in a package will no signal an error if the packages has called R_forceSymbols to specify that symbols must be used.
An error in table() could permanently set options(warn=2) promoting all subsequent warnings to errors.
The sigma() function gave misleading results for binary GLMs. A new method for objects of class "glm" returns the square root of the estimate of the dispersion parameter using the same calculation as summary.glm().
bs() and ns() in the (typical) case of automatic knot construction, when some of the supposedly inner knots coincide with boundary knots, now moves them inside (with a warning), building on PR#18442 by Ben Bolker.
R CMD on Windows now skips the site profile with --no-site-file and --vanilla even when R_PROFILE is set (PR#18512, from Kevin Ushey).

New in R for Windows 4.2.3 (Mar 17, 2023)

C-LEVEL FACILITIES:
The definition of DL_FUNC in R_ext/Rdynload.h has been changed to be fully C-compliant. This means that functions loaded _via_ for example R_GetCCallable need to be cast to an appropriate type if they have any arguments.
.Machine has a new element sizeof.time_t to identify old systems with a 32-bit type and hence a limited range of date-times (and limited support for dates millions of years from present).
PACKAGE INSTALLATION:
(Windows) The default C++ standard had accidentally been left at C++11 when it was changed to C++14 on Unix.
BUG FIXES:
As "POSIXlt" objects may be "partially filled" and their list components meant to be recycled, length() now is the length of the longest component.
as.POSIXlt.Date() could underflow for dates in the far past (more than half a million years BCE).
as.Date.POSIXlt(x) would return "1970-01-01" instead of NA in R 4.2.2, e.g., for
x <- as.POSIXlt(c("2019-01-30","2001-1-1")) x$mon <- c(0L, NA); as.Date(x)
R CMD check failed to apply enabled _R_CHECK_SUGGESTS_ONLY_ to examples and vignettes (regression in R 4.2.0).
R CMD check did not re-build vignettes in separate processes by default (regression in R 4.2.0).
Running examples from HTML documentation now restores previous knitr settings and options (PR#18420).
Quartz: fonts are now located using Core Graphics API instead of deprecated ATS which is no longer supported in the macOS 13 SDK (PR#18426). This also addresses an issue where the currently used font in the Quartz device context was not correctly
retained.
(Windows) Math symbols in text drawing functions are again rendered correctly (PR#18440). This fixes a regression in R 4.2.1 caused by a fix in PR#18382 which uncovered an issue in GraphApp due to which the symbol charset was not used with TT
Symbol font face.
(Windows) Installing a package with a src/Makefile.{win,ucrt} file includes ~/.R/Makevars.win64 in the search for user makevars, as documented in "R Installation and Administration" and done for packages with a src/Makevars.{win,ucrt} file.
format(<POSIXlt_w/_unbalanced_sec>, "....%OS<n>") with n > 0 no longer accidentally uses the unbalanced seconds, thanks to Suharto Anggono's report (including patch) in PR#18448.
solve.default(a, b) works around issues with some versions of LAPACK when a contains NA or NaN values.
When UseMethod() cannot dispatch, it no longer segfaults producing the error message in case of a long class(), thanks to Joris Vankerschaver's report (including patch) in PR#18447.
When example(foo, ..) produces graphics on an interactive device it needs to open itself, it now leaves devAskNewPage() unchanged even when it was FALSE, thus fixing a 14 years old '<FIXME>'.
packageDescription() again catches errors from encoding conversions. This also fixes broken packageVersion() in C locale on systems where iconv does not support transliteration.

New in R for Windows 4.2.2 (Nov 1, 2022)

NEW FEATURES:
tools::Rdiff(useDiff = TRUE) checks for the presence of an external diff command and switches to useDiff = FALSE if none is found. This allows R CMD Rdiff to always work.
On Windows, environment variable R_LIBCURL_SSL_REVOKE_BEST_EFFORT can be used to switch to only 'best-effort' SSL certificate revocation checks with the default "libcurl" download method.This reduces security, but may be needed for downloads to work with MITM proxies (PR#18379).
(macOS) The run-time check for libraries from XQuartz for X11 and Tcl/Tk no longer uses otool from the Apple Developer Tools (PR#18400).
The LaTeX style for producing the PDF manuals, Rd.sty, now loads the standard amsmath, amsfonts and amssymb packages for greater coverage of math commands in the Rd eqn and deqn macros. The mathscr LaTeX command is also provided (via the mathrsfs package, if available, or the amsfonts bundle otherwise), fulfilling the wish of PR#18398.
(Windows) The default format of readClipboard() and writeClipboard() has been changed to 13 (CF_UNICODETEXT).
INSTALLATION on a UNIX-ALIKE:
The PDF manuals (if built) can be compacted by the new target make compact-pdf (at the top level or in directory doc/manual).
There is now configure support for LLVM clang 15 on Linux, which defaults to position-independent (PIE) executables whereas gfortran does not.
Many small changes to ease compilation (and suppress warnings) with LLVM clang 15.
BUG FIXES:
Rscript -e would fail if stdin were closed (Reported by Henrik Bengtsson.)
qt(*, log.p=TRUE) in outer tails no longer produces NaN in its final steps, thus fixing PR#18360.
tools::Rd2latex() now escapes hashes and ampersands when writing URLs, fixing LaTeX errors with such URLs in tabular.
When isGeneric(f, fdef=*) is used with mismatching names, the warning is better understandable; reported (with fix) in PR#18370 by Gabe Becker.
poly(x, n) now works again (and is now documented) when x is a "Date" or "POSIXct" object, or of another class while fulfilling mode(x) == "numeric". This also enables poly(x, *, raw=TRUE) for such variables. Reported by Michael Chirico to R-devel.
write.table(), write.csv() and write.csv2() restore their numerical precision (internal equivalent of digits = 15) after an interrupt (PR#18384).
One can now read also byte FF from a clipboard connection (PR#18385).
source("") and source(character()) now give more helpful error messages.
R CMD check --as-cran set _R_CHECK_TIMINGS_ too late to have the intended effect.
as.POSIXlt(x) now also works with very large dates x, fixing PR#18401 reported by Hannes M"uhleisen.
Files can now be extracted even from very large zip archives (PR#18390, thanks to Martin Jakt).
Non-finite objects of class "POSIXlt" are now correctly coerced to classes "Date" and "POSIXct"; following up on the extension to format() them correctly.
Added methods for is.finite(), is.infinite() and is.nan() for "POSIXlt" date-time objects.
BUG FIXES on Windows:
Non-ASCII characters are now properly displayed on Windows in windows created using GraphApp via e.g. winDialogString thanks to a workaround for an at least surprising Windows behavior with UTF-8 as the system encoding (PR#18382).
Find and replace operations work again in the script editor in Rgui on Windows.
Computation of window size based on requested client size in GraphApp when running in a multi-byte locale on Windows has been fixed (regression in R 4.2.0 for users of systems where R 4.1 used a single-byte locale). Rgui again respects the number of console rows and columns given in Rconsole file.
Rterm support for Alt+xxx sequences has been fixed to produce the corresponding character (only) once. This fixes pasting text with tilde on Italian keyboard (PR#18391).

New in R for Windows 4.2.1 (Jun 24, 2022)

New in R for Windows 4.2.0 (May 11, 2022)

SIGNIFICANT USER-VISIBLE CHANGES:
The formula method of aggregate() now matches the generic innaming its first argument x (resolving PR#18299 by Thomas Soeiro).
This means that calling aggregate() with a formula as a named first argument requires name formula in earlier versions of R and name x now, so portable code should not name the argument (code in many packages did).
Calling && or || with either argument of length greater than one now gives a warning (which it is intended will become an error).
Calling if() or while() with a condition of length greater than one gives an error rather than a warning. Consequently, environment variable _R_CHECK_LENGTH_1_CONDITION_ no longer has any effect.
Windows users should consult the WINDOWS section below for some profound changes including:
Support for 32-bit builds has been dropped.
UTF-8 locales are used where available.
The default locations for the R installation and personal
library folder have been changed.
Thanks to Tomas Kalibera for months of work on the Windows port
for this release.
NEW FEATURES:
Matrix(x, n, m) now warns in more cases where length(x) differs from n * m, as suggested by Abby Spurdle and Wolfgang Huber in Feb 2021 on the R-devel mailing list.
This warning can be turned into an error by setting environment variable _R_CHECK_MATRIX_DATA_ to TRUE: R CMD check --as-cran does so unless it is already set.
Function file_test() in package utils gains tests for symlinks, readability and writability.
Capabilities("libxml") is now false.
The description of capabilities("http/ftp") now reflects that it refers to the default method, no longer the internal one.
Simplify2array() gains an except argument for controlling the exceptions used by sapply().
Environment variables R_LIBS_USER and R_LIBS_SITE are both now set to the R system default if unset or empty, and can be set to NULL to indicate an empty list of user or site library directories.
The warning for axis()(-like) calls in cases of relatively small ranges (typically in log-scale situations) is slightly improved _and_ suppressed from explicit calls to .axisPars() as has always been the intention.
The contrasts setter function `contrasts<-` gains an explicit default how.many = NULL rather than just using missing(how.many).
Grid.pretty() gains a new optional argument n = 5.
There is a new function .pretty() with option bounds as a technical-utility version of pretty(). It and pretty() gain a new argument f.min with a better than back-compatible default.
Function grDevices::axisTicks() and related functions such as graphics::axis() work better, notably for the log scale; partly because of the pretty() improvements, but also because care is taken e.g., when ylim is finite but diff(ylim) is infinite.
Nclass.FD() gains a digits option.
The R Mathlib internal C function bd0() (called indirectly from a dozen probability density and distribution functions such as dpois(), dbinom(), dgamma(), pgamma() _etc_) has been complemented by a more sophisticated and (mostly) more accurate C function ebd0(), currently called only by internal dpois_raw() improving accuracy for R level dpois() and potentially others calling it such as dnbinom(), dgamma() or pgamma(). (Thanks to Morten Welinder's PR#15628.)
Write.ftable() gains sep = " " argument as suggested by Thomas Soeiro.
The names of the locale categories supported by R's Sys.getlocale() and Sys.setlocale() are now provided by variable .LC.categories in the base namespace.
The Date and POSIXt methods for hist() and the histogram method for plot() now also use the new default col = "lightgray" in consistency with the corresponding change to hist()'s default for R 4.0.0.
Hist.default() gains new fuzz argument, and the histogram plot method no longer uses fractional axis ticks when displaying counts ("Frequency").
Mapply() and hence Map() now also obey the "max-or-0-if-any" recycling rule, such that, e.g., Map(`+`, 1:3, 1[0]) is valid now.
As.character(<obj>) for "hexmode" or "octmode" objects now fulfils the important basic rule as.character(x)[j] === as.character(x[j]).
The set utility functions, notably intersect() have been tweaked to be more consistent and symmetric in their two set arguments, also preserving a common mode.
Substr(ch, start,end) <- new now e.g., preserves names(ch); ditto for substring(), thanks to a patch from Brodie Gaslam.
Plot(<lm>) gains a extend.ylim.f argument, in partial response to PR#15285; further PR#17784 is fixed thanks to several contributors and a patch by Elin Waring. The Cook's dist contours get customizable via cook.col and cook.lty with a different default color and their legend is nicer by default and customizable via cook.legendChanges.
Attempting to subset an object that is not subsettable now signals an error of class notSubsettableError. The non-subsettable object is contained in the object field of theerror condition.
Subscript-out-of-bounds errors are now signaled as errors of class subscriptOutOfBoundsError.
Stack-overflow errors are now signaled as errors inheriting from class stackOverflowError. See ?stackOverflowError for more details.
New partly experimental Sys.setLanguage() utility, solving the main problem of PR#18055.
Gettext() and gettextf() get a new option trim = TRUE which when set to false allows translations for strings such as "Execution haltedn" typical for C code.
An experimental implementation of hash tables is now available. See ?hashtab for more details.
Identical() gains a extptr.as.ref argument for requesting that external pointer objects be compared as reference objects.
Reorder() gets an argument decreasing which it passes to sort() for level creation; based on the wish and patch by Thomas Soeiro in PR#18243.
As.vector() gains a data.frame method which returns a simple named list, also clearing a long standing 'FIXME' to enable as.vector(<data.frame>, mode="list"). This breaks code relying on as.vector(<data.frame>) to return the unchanged data frame.
Legend() is now vectorized for arguments cex, x.intersp, and text.width. The latter can now also be specified as a vector (one element for each column of the legend) or as NA for computing a proper column wise maximum value of strwidth(legend).
The argument y.intersp can be specified as a vector with one entry for each row of the legend.
Legend() also gains new arguments title.cex and title.font. Thanks to Swetlana Herbrandt.
Deparsing no longer remaps attribute names dim, dimnames, levels, names and tsp to historical S-compatible names (which structure() maps back).
Sample() and sample.int() have additional sanity checks on their size and n arguments.
All.equal.numeric() gains a sanity check on its tolerance argument - calling all.equal(a, b, c) for three numeric vectors is a surprisingly common error.
Mean(na.rm =), rank(na.last =), barplot(legend.text =), boxplot(), contour(drawlabels =), polygon(border =) and methods::is(class2 =) have more robust sanity checks on their arguments.
R CMD Rd2pdf (used by R CMD check) has a more robust sanity check on the format of alias{} commands.
Psigamma(x, deriv) for negative x now also works for deriv = 4 and 5; their underlying C level dpsifn() is documented in 'Writing R Extensions'.
The HTML help system now uses HTML5 (wish of PR#18149).
Ks.test() now provides exact p-values also with ties and MC p-values in the two-sample (Smirnov) case. By Torsten Hothorn.
Ks.test() gains a formula interface, with y ~ 1 for the one-sample (Kolmogorov) test and y ~ group for the two-sample (Smirnov) test. Contributed by Torsten Hothorn.
The return value from ks.test() now has class c("ks.test", "htest") - packages using try() need to take care to use inherits() and not == on the class.
New functions psmirnov(), qsmirnov() and rsmirnov() in package stats implementing the asymptotic and exact distributions of the two-sample Smirnov statistic.
Iconv() now allows sub = "c99" to use C99-style escapes for UTF-8 inputs which cannot be converted to encoding to.
In a forward pipe |> expression it is now possible to use a named argument with the placeholder _ in the rhs call to specify where the lhs is to be inserted. The placeholder can only appear once on the rhs.
The included LAPACK sources have been updated to version 3.10.0, except for the four Fortran 77 routines which 3.10.0 has re-implemented in Fortran 90 (where the older versions have been retained as the R build process does not support Fortran 90).
Path.expand() and most other uses of tilde expansion now warn if a path would be too long if expanded. (An exception is file.exists(), which silently returns false.)
Lrunc(<Date>, *) now supports units = "months" or "years" for consistency with the POSIXt method, thanks to Dirk Eddelbuettel's proposal in PR#18099.
List2DF() now checks that its arguments are of the same length, rather than use recycling.
The HTML help system has several new features: LaTeX-like math can be typeset using either KaTeX or MathJax, usage and example code is highlighted using Prism, and for dynamic help the output of examples and demos can be shown within the browser if the knitr package is installed. These features can be disabled by setting the environment variable _R_HELP_ENABLE_ENHANCED_HTML_ to a false value.
GRAPHICS:
The graphics engine version, R_GE_version, has been bumped to 15 and so packages that provide graphics devices should be reinstalled.
The grid package now allows the user to specify a "vector" of pattern fills. The fill argument to gpar() accepts a list of gradients and/or patterns and the functions linearGradient(), radialGradient(), and pattern() have a new group argument.
Points grobs (data symbols) can now also have a pattern fill.
The grobCoords() function now returns a more informative and complex result.
The grid package has new functions for drawing isolated groups: grid.group(), grid.define(), and grid.use(). These functions add compositing operators and affine transformations to R's graphics capabilities.
The grid package also has new functions for stroking and filling paths: grid.stroke(), grid.fill(), and grid.fillStroke().
A new function as.path() allows the user to specify the fill rule for a path that is to be used for clipping, stroking, or filling; available options are "winding" and "evenodd". A new function as.mask() allows the user to specify the type of a mask; available options are "alpha" and "luminance".
These new features are only supported so far (at most) on the Cairo-based graphics devices and on the pdf() device.
Dev.capabilities() reports on device support for the new features.
Par() now warns about unnamed non-character arguments to prevent misuse such as {usr <- par("usr"); par(usr)}.
WINDOWS:
R uses UTF-8 as the native encoding on recent Windows systems (at least Windows 10 version 1903, Windows Server 2022 or Windows Server 1903). As a part of this change, R uses UCRT as the C runtime. UCRT should be installed manually on systems older than Windows 10 or Windows Server 2016 before installing R.
The default personal library on Windows, folder Rwin-libraryx.y where x.y stands for R release x.y.z, is now a subdirectory of Local Application Data directory (usually a hidden directory C:UsersusernameAppDataLocal). Use shell.exec(.libPaths()[1]) from R to open the personal library in Explorer when it is first in the list (PR#17842).
R uses a new 64-bit Tcl/Tk bundle. The previous 32-bit/64-bit bundle had a different layout and can no longer be used.
Make files and installer scripts for Windows have been tailored to Rtools42, the newly recommended 64-bit gcc 10.3 MinGW-W64 UCRT toolchain.
Rtools42 by default uses the Windows security features ASLR and DEP; hence CRAN builds of R and packages also do.
R now supports files Makevars.ucrt, Makefile.ucrt, configure.ucrt and cleanup.ucrt in packages, which are used in preference to the .win variants. This allows keeping the .win files around to support older versions of R. This feature will be removed in the future once support for older versions of R would no longer be needed.
R.version gains a new field crt (only on Windows) to denote the C runtime. The value is "ucrt".
On Windows, download.file(method = "auto") and url(method = "default") now follow Unix in using "libcurl" for all except file:// URIs.
Rtools42 includes an unpatched Msys2 build of GNU tar. Paths including drive letters can be made to work by adding --force-local to environment variable TAR_OPTIONS. (Rtools40 and earlier included a patched version which defaulted to this option.)
Installer builds of R automatically find the Rtools42 software collection as well as the compiler toolchain. No PATH setting is required from the user.
The default installation directory of R for a user-only installation has been changed to the User Program Files directory (usually a hidden directory C:UsersusernameAppDataLocalPrograms) to follow Windows conventions. Use shell.exec(R.home()) from R to open the R installation directory in Explorer (PR#17842).
R now supports installation-time patching of packages. Patches may be installed from a supplied URL or a local directory or disabled. Patches are included into the installed packages for reference. This experimental feature may be removed in the future.
Libcurl is now required for building from source.
The clipboard connection now works also with text in other than the current native encoding (PR#18267, with Hiroaki Yutani). Text is always pasted to the clipboard in UTF16-LE and the encoding argument is ignored.
The internal case-changing functions are now used by default on Windows - this circumvents problems (for example with E acute) of the UCRT Windows' runtime.
R on Windows now uses the system memory allocator. Doug Lea's allocator was used since R 1.2.0 to mitigate performance limitations seen with system allocators on earlier versions of Windows.
Memory.limit() and memory.size() are now stubs on Windows (as on Unix-alikes).
Applications embedding R on Windows can now use additional callbacks, which have so far only been available only on Unix (PR#18286).
INSTALLATION:
Facilities for accessing ftp:// sites are no longer tested (except _pro tem_ for curlGetHeaders()) as modern browsers have removed support.
R can now be built with DEFS = -DSTRICT_R_HEADERS .
PACKAGE INSTALLATION:
R CMD INSTALL no longer tangles vignettes. This completes an R CMD build change in R 3.0.0 and affects packages built before R 3.0.2. Such packages should be re-made with R CMD build to have the tangled R code of vignettes shipped with the tarball.
USE_FC_LEN_T will become the default: this uses the correct prototypes for Fortran BLAS/LAPACK routines called from C/C++, and requires adjustment of most such calls - see 'Writing R Extensions' SS6.6.2. (This has been supported since R 3.6.2.)
Package installation speed for packages installed with keep.source has been improved. This resolve the issue reported by Ofek Shilon in PR#18236.
UTILITIES:
R CMD check can optionally report files/directories left behind in home, /tmp (even though TMPDIR is set) and other directories. See the "R Internals" manual for details.
R CMD check now reports byte-compilation errors during installation. These are not usually fatal but may result in parts of the package not being byte-compiled.
_R_CHECK_DEPENDS_ONLY_ can be applied selectively to examples, tests and/or vignettes in R CMD check: see the "R Internals" manual.
_R_CHECK_SRC_MINUS_W_IMPLICIT_ now defaults to true: recent versions of Apple clang on macOS have made implicit function declarations in C into a compilation error.
R CMD check --as-cran makes use of the environment variable AUTORECONF. See the "R Internals" manual SS8 for further details.
R CMD check --use-valgrind also uses valgrind when re-building vignettes as some non-Sweave vignettes unhelpfully comment out all their code when R CMD check runs vignettes.
Errors in re-building vignettes (unless there are LaTeX errors) are reported by R CMD check as ERROR rather than WARNING when running vignettes has been skipped (as it frequently is in CRAN checks and by --as-cran).
R CMD Rd2pdf gains a --quiet option that is used by R CMD build when building the PDF package manual.
R CMD Rd2pdf now always runs LaTeX in batch mode, consistent with Texinfo >= 6.7. The --batch option is ignored.
R CMD build and R CMD check now include the Rd file name and line numbers in the error message of an Sexpr evaluation failure.
For packages using the doi Rd macro (now an install-time Sexpr) but no other dynamic Rd content, R CMD build now produces a smaller tarball and is considerably faster - skipping temporary package installation.
C-LEVEL FACILITIES:
The non-API header R_ext/R-ftp-http.h is no longer provided, as the entry points it covered are now all defunct.
A number of non-API declarations and macro definitions have been moved from the installed header Rinternals.h to the internal header Defn.h. Packages that only use entry points and definitions documented to be part of the API as specified in
'Writing R Extensions' SS6 should not be affected.
The macro USE_RINTERNALS no longer has any effect when compiling package code. Packages which also use R_NO_REMAP will need to ensure that the remapped names are used for calls to API functions that were formerly also made available as macros.
The deprecated legacy S-compatibility macros PROBLEM, MESSAGE, ERROR, WARN, WARNING, RECOVER, ... are no longer defined in R_exts/RS.h (included by R.h). Replace these by calls to Rf_error and Rf_warning (defined in header R_ext/Error.h included
by R.h).
Header R_ext/RS.h no longer includes R_ext/Error.h.
Header R_ext/Constants.h (included by R.h) when included from C++ now includes the C++ header cfloat rather than the C header float.h (now possible as C++11 is required).
The legacy S-compatibility macros DOUBLE_* in R_ext/Constants.h (included by R.h) are deprecated.
The deprecated S-compatibility macros SINGLE_* in R_ext/Constants.h (included by R.h) have been removed.
R_Calloc, R_Free and R_Realloc are preferred to their unprefixed forms and error messages now use the prefix. These forms were introduced in R 3.4.0 and are available even when STRICT_R_HEADERS is defined.
Rmultinom has been documented in 'Writing R Extensions' SS6 so is now part of the R API.
Similarly, Rtanpi, called from R level tanpi() is now part of the R API.
The long-deprecated, undocumented and non-API entry point call_R is no longer declared in R_ext/RS.h (included by R.h).
The header S.h which has been unsupported since Jan 2016 has been removed. Use R.h instead.
DEPRECATED AND DEFUNCT:
The (non-default and deprecated) method = "internal" for download.file() and url() no longer supports http:// nor ftp:// URIs. (It is used only for file:// URIs.)
On Windows, download.file(method = "wininet") no longer supports ftp:// URIs. (It is no longer the default method, which is "libcurl" and does.)
On Windows, the deprecated method = "wininet" now gives a warning for http:// and https:// URIs for both download.file() and url(). (It is no longer the default method.)
On Windows, the command-line option --max-mem-size and environment variable R_MAX_MEM_SIZE are defunct. The memory allocation limit was important for 32-bit builds, but these are no longer supported.
Default.stringsAsFactors() is now formally deprecated, where that was only mentioned on its regular help page, previously. So it now gives a warning if called.
Unix.time() is defunct now; it had been deprecated since R 3.4.0.
BUG FIXES:
Setting digits = 0 in format(), print.default() (and hence typically print()) or options() is again invalid. Its behaviour was platform-dependent, and it is unclear what "zero significant digits" should mean (PR#18098).
Messages from C code in the cairo section of package grDevices are now also offered for translation, thanks to Michael Chirico's PR#18123.
Mean(x) with finite x now is finite also without "long.double" capability.
R CMD Rd2pdf no longer leaves an empty build directory behind when it aborts due to an already existing output file. (Thanks to Sebastian Meyer's PR#18141.)
Density(x, weights = w, na.rm = TRUE) when anyNA(x) is true, now removes weights "in parallel" to x, fixing PR#18151, reported by Matthias Gondan. Additionally, it gets a subdensity option.
Conversion of Sexpr[]{<expR>} to LaTeX or HTML no longer produces long blocks of empty lines when <expR> itself contains several lines all producing empty output. Thanks to a report and patch by Ivan Krylov posted to R-devel.
R CMD build no longer fails if a package vignette uses child documents and inst/doc exists. (Thanks to Sebastian Meyer's PR#18156.)
When an R documentation ('help' source) file man/foo.Rd in a package has donttest{..} examples with a syntax error, it is now signalled as ERROR and with correct line numbers relating to the Ex.R file, thanks to Duncan Murdoch and Sebastian Meyer's
reports and patch proposals in PR#17501.
improved determination of the correct translation domain in non-base packages, addressing the combination of PR#18092 and PR#17998 (#c6) with reports and _augmented_ patch #2904 by Suharto Anggono.
Note that "R-base" is no longer the default domain e.g., for top-level calls to gettext(); rather translation needs explicit domain = * specification in such cases.
Identical(attrib.as.set=FALSE) now works correctly with data frames with default row names (Thanks to Charlie Gao's PR#18179).
TxtProgressBar() now enforces a non-zero width for argument char, without which no progress can be visible.
Dimnames(table(d)) is more consistent in the case where d is a list with a single component, thanks to Thomas Soeiro's report to R-devel.
Further, table(d1, d2) now gives an error when d1 and d2 are data frames as suggested by Thomas in PR#18224.
Fix for drawing semi-transparent lines and fills on the native Windows graphics device (PR#18219 and PR#16694). Thanks to Nick Ray for helpful diagnosis on Bugzilla.
The deparser now wraps sub-expressions such as if(A) .. with parentheses when needed; thanks to Duncan Murdoch's PR#18232 and Lionel Henry's patches there.
Remove.packages() no longer tries to uninstall Priority: base packages, thanks to a report and suggestions by Colin Fay in PR#18227.
Win.metafile() now has xpinch and ypinch arguments so that the user can override Windows' (potentially wrong) guess at device dimensions.
X[i] and x[[i]] for non-integer i should now behave in all cases as always documented: the index used is equivalent to as.integer(i) unless that would overflow where trunc(i) is used instead; thanks to Suharto Anggono's report and patch proposals in PR#17977.
AsOneSidedFormula() now associates the resulting formula with the
global environment rather than the evaluation environment created for the call.
<bibentry>$name now matches the field name case-insensitively, consistent with bibentry() creation and the replacement method.
Cbind() failed to detect some length mismatches with a mixture of time-series and non-time-series inputs.
The default LaTeX style file Sweave.sty used by the RweaveLatex driver no longer loads the obsolete ae package; thanks to a report by Thomas Soeiro in PR#18271. Furthermore, it now skips usepackage[T1]{fontenc} for engines other than pdfTeX (if
detected) or if the new [nofontenc] option is used.
Smooth.spline() now stores its logical cv argument more safely, fixing a rare bug when printing, and also stores n.
Smooth.spline(x,y,*) now computes the cv.crit statistic correctly, also when is.unsorted(x), fixing PR#18294.
The data.frame method of rbind() now warns when binding not-wholly-recycling vectors, by analogy to the default method (for matrices).
SetAs() finds the correct class for name to when multiple packages define a class with that name. Thanks to Gabor Csardi for the report.
Fix for detaching a package when two classes of the same name are present in method signatures for the same generic. Thanks to Gabor Csardi for the report.
Match.arg("", c("", "a", "B")) gives a better error message, in part from PR#17959, thanks to Elin Waring.
R CMD Sweave --clean no longer removes pre-existing files or subdirectories (PR#18242).
The quartz() device no longer splits polylines into subpaths. That has caused narrowly-spaced lines with many points to always look solid even when dashed line type was used due to dash phase restarts.
Deparsing constructs such as quote(1 + `!`(2) + 3) works again as before R 3.5.0, thanks to the report and patch in PR#18284 by Suharto Anggono.
As.list(f) for a factor f now keeps names(f), fixing PR#18309.
Qbeta(.001, .9, .009) and analogous qf() calls now return a correct value instead of NaN or wrongly 1, all with a warning; thanks to the report by Ludger Goeminne in PR#18302.
Plot.lm() failed to produce the plot of residuals vs. factor levels (i.e., which=5 when leverages are constant) for models with character predictors (PR#17840).
Interaction.plot(..., xtick = TRUE) misplaced the x-axis line (PR#18305).
Not strictly fixing a bug, format()ing and print()ing of non-finite Date and POSIXt values NaN and +/-Inf no longer show as NA but the respective string, e.g., Inf, for consistency with numeric vector's behaviour, fulfilling the wish of PR#18308.
R CMD check no longer runs test scripts generated from corresponding .Rin files twice and now signals an ERROR if processing an .Rin script fails.
Tools::Rd2txt() used for plain-text help pages now renders hrefs (if tools::Rd2txt_options(showURLs = TRUE)) and urls with percent-encoding and standards-compliant delimiting style (angle brackets and no URL: prefix). email is now rendered with a mailto: prefix.

New in R for Windows 4.1.3 (Mar 11, 2022)

NEW FEATURES:
The default version of Bioconductor has been changed to 3.14. (This is used by setRepositories and the menus in GUIs.)
UTILITIES:
R CMD check --as-cran has a workaround for a bug in versions of file up to at least 5.41 which mis-identify DBF files last changed in 2022 as executables.
C-LEVEL FACILITIES:
The legacy S-compatibility macros SINGLE_* in ‘R_ext/Constants.h’ (included by ‘R.h’) are deprecated and will be removed in R 4.2.0.
BUG FIXES:
Initialization of self-starting nls() models with initialization functions following the pre-R-4.1.0 API (without the ... argument) works again for now, with a deprecation warning.
Fixed quoting of ~autodetect~ in Java setting defaults to avoid inadvertent user lookup due to leading ~, reported in PR#18231 by Harold Gutch.
substr(., start, stop) <- v now treats negative stop values correctly. Reported with a patch in PR#18228 by Brodie Gaslam.
Subscripting an array x without dimnames by a length(dim(x))-column character matrix gave "random" non-sense, now an error; reported in PR#18244 by Mikael Jagan.
...names() now matches names(list(...)) closely, fixing PR#18247.
all.equal(*, scale = s) now works as intended when length(s) > 1, partly thanks to Michael Chirico's PR#18272.
print(x) for long vectors x now also works for named atomic vectors or lists and prints the correct number when reaching the getOption("max.print") limit; partly thanks to a report and proposal by Hugh Parsonage to the R-devel list.
all.equal(<selfStart>, *) no longer signals a deprecation warning.
reformulate(*, response=r) gives a helpful error message now when length(r) > 1, thanks to Bill Dunlap's PR#18281.
Modifying globalCallingHandlers inside withCallingHandlers() now works or fails correctly, thanks to Henrik Bengtsson's PR#18257.
hist(<Date>, breaks = "days") and hist(<POSIXt>, breaks = "secs") no longer fail for inputs of length 1.
qbeta(.001, .9, .009) and similar cases now converge correctly thanks to Ben Bolker's report in PR#17746.
window(x, start, end) no longer wrongly signals “'start' cannot be after 'end'”, fixing PR#17527 and PR#18291.
data() now checks that its (rarely used) list argument is a character vector – a couple of packages passed other types and gave incorrect results.
which() now checks its arr.ind argument is TRUE rather coercing to logical and taking the first element – which gave incorrect results in package code.
model.weights() and model.offset() more carefully extract their model components, thanks to Ben Bolker and Tim Taylor's R-devel post.
list.files(recursive = TRUE) now shows all broken symlinks (previously, some of them may have been omitted, PR#18296).

New in R for Windows 4.1.2 (Dec 27, 2021)

C-LEVEL FACILITIES:
The workaround in headers R.h and Rmath.h (using namespace std;) for the Oracle Developer Studio compiler is no longer needed now C++11 is required so has been removed. A couple more usages of log() (which should have been std::log()) with an int argument are reported on Solaris.
The undocumented limit of 4095 bytes on messages from the S-compatibility macros PROBLEM and MESSAGE is now documented and longer messages will be silently truncated rather than potentially causing segfaults.
If the R_NO_SEGV_HANDLER environment variable is non-empty, the signal handler for SEGV/ILL/BUS signals (which offers recovery user interface) is not set. This allows more reliable debugging of crashes that involve the console.
DEPRECATED AND DEFUNCT:
The legacy S-compatibility macros PROBLEM, MESSAGE, ERROR, WARN, WARNING, RECOVER, ... are deprecated and will be hidden in R 4.2.0. R's native interface of Rf_error and Rf_warning has long been preferred.
BUG FIXES:
.mapply(F, dots, .) no longer segfaults when dots is not a list and uses match.fun(F) as always documented; reported by Andrew Simmons in PR#18164.
Hist(<Date>, ...) and hist(<POSIXt>, ...) no longer pass arguments for rect() (such as col and density) to axis(). (Thanks to Sebastian Meyer's PR#18171.)
Sexpr{ch} now preserves Encoding(ch). (Thanks to report and patch by Jeroen Ooms in PR#18152.)
Setting the RNG to "Marsaglia-Multicarry" e.g., by RNGkind(), now warns in more places, thanks to Andr'e Gillibert's report and patch in PR#18168.
Gray(numeric(), alpha=1/2) no longer segfaults, fixing PR#18183, reported by Till Krenz.
Fixed dnbinom(x, size=<very_small>, .., log=TRUE) regression, reported by Martin Morgan.
As.Date.POSIXlt(x) now keeps names(x), thanks to Davis Vaughan's report and patch in PR#18188.
Model.response() now strips an "AsIs" class typically, thanks to Duncan Murdoch's report and other discussants in PR#18190.
Try() is considerably faster in case of an error and long call, as e.g., from some do.call(). Thanks to Alexander Kaever's suggestion posted to R-devel.
Qqline(y = <object>) such as y=I(.), now works, see also PR#18190.
Non-integer mgp par() settings are now handled correctly in axis() and mtext(), thanks to Mikael Jagan and Duncan Murdoch's report and suggestion in PR#18194.
FormatC(x) returns length zero character() now, rather than "" when x is of length zero, as documented, thanks to Davis Vaughan's post to R-devel.
RemoveSource(fn) now retains (other) attributes(fn).

New in R for Windows 4.1.1 (Aug 11, 2021)

NEW FEATURES:
require(pkg, quietly = TRUE) is quieter and in particular does not warn if the package is not found.
DEPRECATED AND DEFUNCT:
Use of ftp:// URIs should be regarded as deprecated, with on-going support confined to method = "libcurl" and not routinely
tested. (Nowadays no major browser supports them.)
The non-default method = "internal" is deprecated for http:// and ftp:// URIs for both download.file and url.
On Windows, method = "wininet" is deprecated for http://, https:// and ftp:// URIs for both download.file and url. (A warning is only given for ftp://.)
For ftp:// URIs the default method is now "libcurl" if available (which it is on CRAN builds).
method = "wininet" remains the default for http:// and https:// URIs but if libcurl is available, using method = "libcurl" is preferred.
INSTALLATION:
make check now works also without a LaTeX installation. (Thanks to Sebastian Meyer's PR#18103.)
BUG FIXES:
make check-devel works again in an R build configured with --without-recommended-packages.
qnbinom(p, size, mu) for large size/mu is correct now in a range of cases (PR#18095); similarly for the (size, prob) parametrization of the negative binomial. Also qpois() and qbinom() are better and or faster for extreme cases. The underlying C code has been modularized and is common to all four cases of discrete distributions.
gap.axis is now part of the axis() arguments which are passed from bxp(), and hence boxplot(). (Thanks to Martin Smith's report and suggestions in PR#18109.)
.First and .Last can again be set from the site profile.
seq.int(from, to, *) and seq.default(..) now work better in large range cases where from-to is infinite where the two boundaries are finite.
all.equal(x,y) now returns TRUE correctly also when several entries of abs(x) and abs(y) are close to .Machine$double.xmax, the largest finite numeric.
model.frame() now clears the object bit when removing the class attribute of a value via na.action (PR#18100).
charClass() now works with multi-character strings on Windows (PR#18104, fixed by Bill Dunlap).
encodeString() on Solaris now works again in Latin-1 encoding on characters represented differently in UTF-8. Support for surrogate pairs on Solaris has been improved.
file.show() on Windows now works with non-ASCII path names representable in the current native encoding (PR#18132).
Embedded R on Windows can now find R home directory via the registry even when installed only for the current user (PR#18135).
pretty(x) with finite x now returns finite values also in the case where the extreme x values are close in size to the maximal representable number .Machine$double.xmax.
Also, it's been tweaked for very small ranges and when a boundary is close (or equal) to zero; e.g., pretty(c(0,1e-317)) no longer has negative numbers, currently still warning about a very small range, and pretty(2^-(1024 - 2^-1/(c(24,10)))) is more accurate.
The error message for not finding vignette files when weaving has correct file sizes now. (Thanks to Sebastian Meyer's PR#18154.)
dnbinom(20, <large>, 1) now correctly gives 0, and similar cases are more accurate with underflow precaution. (Reported by Francisco Vera Alcivar in PR#18072.)

New in R for Windows 4.1.0 (May 26, 2021)

FUTURE DIRECTIONS:
It is planned that the 4.1.x series will be the last to support 32-bit Windows, with production of binary packages for that series continuing until early 2023.
SIGNIFICANT USER-VISIBLE CHANGES:
Data set esoph in package datasets now provides the correct numbers of controls; previously it had the numbers of cases added to these. (Reported by Alexander Fowler in PR#17964.)
NEW FEATURES:
www.omegahat.net is no longer one of the repositories known by default to setRepositories(). (Nowadays it only provides source packages and is often unavailable.)
Function package_dependencies() (in package tools) can now use different dependency types for direct and recursive dependencies.
The checking of the size of tarball in R CMD check --as-cran <pkg> may be tweaked via the new environment variable _R_CHECK_CRAN_INCOMING_TARBALL_THRESHOLD_, as suggested in PR#17777 by Jan Gorecki.
Using c() to combine a factor with other factors now gives a factor, an ordered factor when combining ordered factors wit identical levels.
apply() gains a simplify argument to allow disabling of simplification of results.
The format() method for class "ftable" gets a new option justify. (Suggested by Thomas Soeiro.)
New ...names() utility. (Proposed by Neal Fultz in PR#17705.)
type.convert() now warns when its as.is argument is not specified, as the help file always said it _should_. In that case, the default is changed to TRUE in line with its change in read.table() (related to stringsAsFactor) in R 4.0.0.
When printing list arrays, classed objects are now shown _via_ their format() value if this is a short enough character string, or by giving the first elements of their class vector and their length.
capabilities() gets new entry "Rprof" which is TRUE when R has been configured with the equivalent of --enable-R-profiling (as it is by default). (Related to Michael Orlitzky's report PR#17836.)
str(xS4) now also shows extraneous attributes of an S4 object xS4.
Rudimentary support for vi-style tags in rtags() and R CMD rtags has been added. (Based on a patch from Neal Fultz in PR#17214.)
checkRdContents() is now exported from tools; it and also checkDocFiles() have a new option chkInternal allowing to check Rd files marked with keyword "internal" as well. The latter can be activated for R CMD check via environment variable _R_CHECK_RD_INTERNAL_TOO_.
New functions numToBits() and numToInts() extend the raw conversion utilities to (double precision) numeric.
Functions URLencode() and URLdecode() in package utils now work on vectors of URIs. (Based on patch from Bob Rudis submitted with PR#17873.)
path.expand() can expand ~user on most Unix-alikes even when readline is not in use. It tries harder to expand ~, for example should environment variable HOME be unset.
For HTML help (both dynamic and static), Rd file links to help pages in external packages are now treated as references to topics rather than file names, and fall back to a file link only if the topic is not found in the target package. The earlier rule which prioritized file names over topics can be restored by setting the environment variable _R_HELP_LINKS_TO_TOPICS_ to a false value.
c() now removes NULL arguments before dispatching to methods, thus simplifying the implementation of c() methods, _but_ for back compatibility keeps NULL when it is the first argument. (From a report and patch proposal by Lionel Henry in PR#17900.)
Vectorize()'s result function's environment no longer keeps unneeded objects.
Function ...elt() now propagates visibility consistently with ..n. (Thanks to Lionel Henry's PR#17905.)
capture.output() no longer uses non-standard evaluation to evaluate its arguments. This makes evaluation of functions like
parent.frame() more consistent. (Thanks to Lionel Henry's PR#17907.)
packBits(bits, type="double") now works as inverse of numToBits(). (Thanks to Bill Dunlap's proposal in PR#17914.)
curlGetHeaders() has two new arguments, timeout to specify the timeout for that call (overriding getOption("timeout")) and TLS to specify the minimum TLS protocol version to be used for https:// URIs (_inter alia_ providing a means to check for sites using deprecated TLS versions 1.0 and 1.1).
For nls(), an optional constant scaleOffset may be added to the denominator of the relative offset convergence test for cases where the fit of a model is expected to be exact, thanks to a proposal by John Nash. nls(*, trace=TRUE) now also shows the convergence criterion.
Numeric differentiation _via_ numericDeriv() gets new optional arguments eps and central, the latter for taking central divided differences. The latter can be activated for nls() via nls.control(nDcentral = TRUE).
nls() now passes the trace and control arguments to getInitial(), notably for all self-starting models, so these can also be fit in zero-noise situations via a scaleOffset. For this reason, the initial function of a selfStart model must now have ... in its argument list.
bquote(splice = TRUE) can now splice expression vectors with attributes: this makes it possible to splice the result of parse(keep.source = TRUE). (Report and patch provided by Lionel Henry in PR#17869.)
textConnection() gets an optional name argument.
get(), exists(), and get0() now signal an error if the first argument has length greater than 1. Previously additional elements were silently ignored. (Suggested by Antoine Fabri on R-devel.)
R now provides a shorthand notation for creating functions, e.g. (x) x + 1 is parsed as function(x) x + 1.
R now provides a simple native forward pipe syntax |>. The simple form of the forward pipe inserts the left-hand side as the first argument in the right-hand side call. The pipe implementation as a syntax transformation was motivated by suggestions from Jim Hester and Lionel Henry.
all.equal(f, g) for functions now by default also compares their environment(.)s, notably via new all.equal method for class function. Comparison of nls() fits, e.g., may now need all.equal(m1, m2, check.environment = FALSE).
.libPaths() gets a new option include.site, allowing to _not_ include the site library. (Thanks to Dario Strbenac's suggestion and Gabe Becker's PR#18016.)
Lithuanian translations are now available. (Thanks to Rimantas Zakauskas.)
names() now works for DOTSXP objects. On the other hand, in R-lang, the R language manual, we now warn against relying on the structure or even existence of such dot-dot-dot objects.
all.equal() no longer gives an error on DOTSXP objects.
capabilities("cairo") now applies only to the file-based devices as it is now possible (if very unusual) to build R with Cairo support for those but not for X11().
There is optional support for tracing the progress of loadNamespace() - see its help.
(Not Windows.) l10n_info() reports an additional element, the name of the encoding as reported by the OS (which may differ from the encoding part (if any) of the result from Sys.getlocale("LC_CTYPE").
New function gregexec() which generalizes regexec() to find _all_ disjoint matches and well as all substrings corresponding to parenthesized subexpressions of the given regular expression. (Contributed by Brodie Gaslam.)
New function charClass() in package utils to query the wide-character classification functions in use (such as iswprint).
The names of quantile()'s result no longer depend on the global getOption("digits"), but quantile() gets a new optional argument digits = 7 instead.
grep(), sub(), regexp and variants work considerably faster for long factors with few levels. (Thanks to Michael Chirico's PR#18063.)
Provide grouping of x11() graphics windows within a window manager such as Gnome or Unity; thanks to a patch by Ivan Krylov posted to R-devel.
The split() method for class data.frame now allows the f argument to be specified as a formula.
sprintf now warns on arguments unused by the format string.
New palettes "Rocket" and "Mako" for hcl.colors() (approximating palettes of the same name from the 'viridisLite' package).
Contributed by Achim Zeileis.
The base environment and its namespace are now locked (so one can no longer add bindings to these or remove from these).
Rterm handling of multi-byte characters has been improved, allowing use of such characters when supported by the current locale.
Rterm now accepts ALT+ +xxxxxxxx sequences to enter Unicode characters as hex digits.
Environment variable LC_ALL on Windows now takes precedence over LC_CTYPE and variables for other supported categories, matching the POSIX behaviour.
duplicated() and anyDuplicated() are now optimized for integer and real vectors that are known to be sorted via the ALTREP
framework. Contributed by Gabriel Becker via PR#17993.
GRAPHICS:
The graphics engine version, R_GE_version, has been bumped to 14 and so packages that provide graphics devices should be reinstalled.
Graphics devices should now specify deviceVersion to indicate what version of the graphics engine they support.
Graphics devices can now specify deviceClip. If TRUE, the graphics engine will never perform any clipping of output itself. The clipping that the graphics engine does perform (for both canClip = TRUE and canClip = FALSE) has been improved to avoid producing unnecessary artifacts in clipped output.
The grid package now allows gpar(fill) to be a linearGradient(), a radialGradient(), or a pattern(). The viewport(clip) can now also be a grob, which defines a clipping path, and there is a new viewport(mask) that can also be a grob, which defines a mask.
These new features are only supported so far on the Cairo-based graphics devices and on the pdf() device.
(Not Windows.) A warning is given when a Cairo-based type is specified for a png(), jpeg(), tiff() or bmp() device but Cairo is unsupported (so type = "Xlib" is tried instead).
grSoftVersion() now reports the versions of FreeType and FontConfig if they are used directly (not _via_ Pango), as is most commonly done on macOS.
C-LEVEL FACILITIES:
The _standalone_ libRmath math library and R's C API now provide log1pexp() again as documented, and gain log1mexp().
INSTALLATION on a UNIX-ALIKE:
configure checks for a program pkgconf if program pkg-config is not found. These are now only looked for on the path (like almost all other programs) so if needed specify a full path to the command in PKG_CONFIG, for example in file config.site.
C99 function iswblank is required - it was last seen missing ca 2003 so the workaround has been removed.
There are new configure options --with-internal-iswxxxxx, --with-internal-towlower and --with-internal-wcwidth which allows the system functions for wide-character classification case-switching and width (wcwidth and wcswidth) to be replaced b internal ones. The first has long been used on macOS, AIX (and Windows) but this enables it to be unselected there and selected for other platforms (it is the new default on Solaris). The second is new in this version of R and is selected by default on macOS and Solaris. The third has long been the default and remains so as it contains customizations for East Asian languages.
System versions of these functions are often minimally implemented (sometimes only for ASCII characters) and may not cover the full range of Unicode points: for example Solaris (and Windows) only cover the Basic Multilingual Plane.
Cairo installations without X11 are more likely to be detected by configure, when the file-based Cairo graphics devices will be available but not X11(type = "cairo").
There is a new configure option --with-static-cairo which is the default on macOS. This should be used when only static cairo (and where relevant, Pango) libraries are available.
Cairo-based graphics devices on platforms without Pango but with FreeType/FontConfig will make use of the latter for font selection.
LINK-TIME OPTIMIZATION on a UNIX-ALIKE:
Configuring with flag --enable-lto=R now also uses LTO when installing the recommended packages.
R CMD INSTALL and R CMD SHLIB have a new flag --use-LTO to use LTO when compiling code, for use with R configured with --enable-lto=R. For R configured with --enable-lto, they have the new flag --no-use-LTO. Packages can opt in or out of LTO compilation _via_ a UseLTO field in the DESCRIPTION file. (As usual this can be overridden by the command-line flags.)
BUILDING R on Windows:
for GCC >= 8, FC_LEN_T is defined in config.h and hence character lengths are passed from C to Fortran in _inter alia_ BLAS and LAPACK calls.
There is a new text file src/gnuwin32/README.compilation, which outlines how C/Fortran code compilation is organized and documents new features:
R can be built with Link-Time Optimization with a suitable compiler - doing so with GCC 9.2 showed several inconsistencies which have been corrected.
There is support for cross-compiling the C and Fortran code in R and standard packages on suitable (Linux) platforms. This is mainly intended to allow developers to test later versions of compilers - for example using GCC 9.2 or 10.x has detected issues that GCC 8.3 in Rtools40 does not.
There is experimental support for cross-building R packages with C, C++ and/or Fortran code.
The R installer can now be optionally built to support a single architecture (only 64-bit or only 32-bit).
PACKAGE INSTALLATION:
The default C++ standard has been changed to C++14 where available (which it is on all currently checked platforms): if not (as before) C++11 is used if available otherwise C++ is not supported. Packages which specify C++11 will still be installed using C++11.
C++14 compilers may give deprecation warnings, most often for std::random_shuffle (deprecated in C++14 and removed in C++17). Either specify C++11 (see 'Writing R Extensions') or modernize the code and if needed specify C++14. The latter has been supported since R 3.4.0 so the package's DESCRIPTION would need to include something like
Depends: R (>= 3.4)
PACKAGE INSTALLATION on Windows:
R CMD INSTALL and R CMD SHLIB make use of their flag --use-LTO when the LTO_OPT make macro is set in file etc/${R_ARCH}/Makeconf or in a personal/site Makevars file. (For details see 'Writing R Extensions' SS4.5.)
This provides a valuable check on code consistency. It does work with GCC 8.3 as in Rtools40, but that does not detect everything the CRAN checks with current GCC do.
PACKAGE INSTALLATION on macOS:
The default personal library directory on builds with --enable-aqua (including CRAN builds) now differs by CPU type,one of
~/Library/R/x86_64/x.y/library
~/Library/R/arm64/x.y/library
This uses the CPU type R (and hence the packages) were built for, so when a x86_64 build of R is run under Rosetta emulation on an arm64 Mac, the first is used.
UTILITIES:
R CMD check can now scan package functions for bogus return statements, which were possibly intended as return() calls (wish of PR#17180, patch by Sebastian Meyer). This check can be activated via the new environment variable _R_CHECK_BOGUS_RETURN_, true for --as-cran.
R CMD build omits tarballs and binaries of previous builds from the top-level package directory. (PR#17828, patch by Sebastian Meyer.)
R CMD check now runs sanity checks on the use of LazyData, for example that a data directory is present and that LazyDataCompression is not specified without LazyData and has a documented value. For packages with large LazyData databases without specifying LazyDataCompression, there is a reference to the code given in 'Writing R Extensions' SS1.1.6 to test the choice of compression (as in all the CRAN packages tested a non-default method was preferred).
R CMD build removes LazyData and LazyDataCompression fields from the DESCRIPTION file of packages without a data directory.
ENCODING-RELATED CHANGES:
The parser now treats Unnnnnnnn escapes larger than the upper limit for Unicode points (U10FFFF) as an error as they cannot be represented by valid UTF-8. Where such escapes are used for outputting non-printable (including unassigned) characters, 6 hex digits are used (rather than 8 with leading zeros). For clarity, braces are used, for example U{0effff}.
The parser now looks for non-ASCII spaces on Solaris (as previously on most other OSes).
There are warnings (including from the parser) on the use of unpaired surrogate Unicode points such as uD834. (These cannot be converted to valid UTF-8.)
Functions nchar(), tolower(), toupper() and chartr() and those using regular expressions have more support for inputs with a marked Latin-1 encoding.
The character-classification functions used (by default) to replace the system iswxxxxx functions on Windows, macOS and AIX have been updated to Unicode 13.0.0.
The character-width tables have been updated to include new assignments in Unicode 13.0.0.
The code for evaluating default (extended) regular expressions now uses the same character-classification functions as the rest of R (previously they differed on Windows, macOS and AIX).
There is a build-time option to replace the system's wide-character wctrans C function by tables shipped with R: use configure option --with-internal-towlower or (on Windows) -DUSE_RI18N_CASE in CFLAGS when building R. This may be needed to allow tolower() and toupper() to work with Unicode characters beyond the Basic Multilingual Plane where not supported by system functions (e.g. on Solaris where it is the new default).
R is more careful when truncating UTF-8 and other multi-byte strings that are too long to be printed, passed to the system or libraries or placed into an internal buffer. Truncation will no longer produce incomplete multibyte characters.
DEPRECATED AND DEFUNCT:
Function plclust() from the package stats and package.dependencies(), pkgDepends(), getDepList(), installFoundDepends(), and vignetteDepends() from package tools are defunct.
Defunct functions checkNEWS() and readNEWS() from package tools and CRAN.packages() from utils have been removed.
R CMD config CXXCPP is defunct (it was deprecated in R 3.6.2).
parallel::detectCores() drops support for Irix (retired in 2013).
The LINPACK argument to chol.default(), chol2inv(), solve.default() and svd() has been defunct since R 3.1.0. It was silently ignored up to R 4.0.3 but now gives an error.
Subsetting/indexing, such as ddd[*] or ddd$x on a DOTSXP(dot-dot-dot) object ddd has been disabled; it worked by accident only and was undocumented.
BUG FIXES:
Many more C-level allocations (mainly by malloc and strdup) are checked for success with suitable alternative actions.
Bug fix for replayPlot(); this was turning off graphics engine display list recording if a recorded plot was replayed in the same session. The impact of the bug became visible if resize the device after replay OR if attempted another savePlot() after replay (empty display list means empty screen on resize or empty saved plot).
R CMD check etc now warn when a package exports non-existing S4 classes or methods, also in case of no "methods" presence. (Reported by Alex Bertram; reproducible example and patch by Sebastian Meyer in PR#16662.)
boxplot() now also accepts calls for labels such as ylab, the same as plot(). (Reported by Marius Hofert.)
The help page for xtabs() now correctly states that addNA is setting na.action = na.pass among others. (Reported as PR#17770 by Thomas Soeiro.)
The R CMD check <pkg> gives a longer and more comprehensible message when DESCRIPTION misses dependencies, e.g., in Imports:. (Thanks to the contributors of PR#17179.)
update.default() now calls the generic update() on the formula to work correctly for models with extended formulas. (As reported and suggested by Neal Fultz in PR#17865.)
The horizontal position of leaves in a dendrogram is now correct also with center = FALSE. (PR#14938, patch from Sebastian Meyer.)
all.equal.POSIXt() no longer warns about and subsequently ignores inconsistent "tzone" attributes, but describes the difference in its return value (PR#17277). This check can be disabled _via_ the new argument check.tzone = FALSE as suggested by Sebastian Meyer.
as.POSIXct() now populates the "tzone" attribute from its tz argument when x is a logical vector consisting entirely of NA values.
x[[2^31]] <- v now works. (Thanks to the report and patch by Suharto Anggono in PR#17330.)
In log-scale graphics, axis() ticks and label positions are now computed more carefully and symmetrically in their range, typically providing _more_ ticks, fulfilling wishes in PR#17936. The change really corresponds to an improved axisTicks() (package grDevices), potentially influencing grid and lattice, for example.
qnorm(<very large negative>, log.p=TRUE) is now correct to at least five digits where it was catastrophically wrong, previously.
sum(df) and similar "Summary"- and "Math"-group member functions now work for data frames df with logical columns, notably also of zero rows. (Reported to R-devel by Martin "b706".)
unsplit() had trouble with tibbles due to unsound use of rep(NA, len)-indexing, which should use NA_integer_ (Reported to R-devel by Mario Annau.)
pnorm(x, log.p = TRUE) underflows to -Inf slightly later.
show(<hidden S4 generic>) prints better and without quotes for non-hidden S4 generics.
read.table() and relatives treated an "NA" column name as missing when check.names = FALSE PR#18007.
Parsing strings containing UTF-16 surrogate pairs such as "uD834uDD1E" works better on some (uncommon) platforms. sprintf("%X", utf8ToInt("uD834uDD1E")) should now give "1D11E" on all platforms.
identical(x,y) is no longer true for differing DOTSXP objects, fixing PR#18032.
str() now works correctly for DOTSXP and related exotics, even when these are doomed. Additionally, it no longer fails for lists with a class and "irregular" method definitions such that e.g. lapply(*) will necessarily fail, as currently for different igraph objects.
Too long lines in environment files (e.g. Renviron) no longer crash R. This limit has been increased to 100,000 bytes. (PR#18001.)
There is a further workaround for FreeType giving incorrect italic font faces with cairo-based graphics devices on macOS.
add_datalist(*, force = TRUE) (from package tools) now actually updates an existing data/datalist file for new content. (Thanks to a report and patch by Sebastian Meyer in PR#18048.)
cut.Date() and cut.POSIXt() could produce an empty last interval for breaks = "months" or breaks = "years". (Reported as PR#18053 by Christopher Carbone.)
Detection of the encoding of 'regular' macOS locales such as en_US (which is UTF-8) had been broken by a macOS change: fortunately these are now rarely used with en_US.UTF-8 being preferred.
sub() and gsub(pattern, repl, x, *) now keep attributes of x such as names() also when pattern is NA (PR#18079).
Time differences ("difftime" objects) get a replacement and a rep() method to keep "units" consistent. (Thanks to a report and patch by Nicolas Bennett in PR#18066.)
The RdOpts macro, setting defaults for Sexpr options in an Rd file, had been ineffective since R 2.12.0: it now works again. (Thanks to a report and patch by Sebastian Meyer in PR#18073.)
mclapply and pvec no longer accidentally terminate parallel processes started before by mcparallel or related calls in package parallel (PR#18078).
grep and other functions for evaluating (extended) regular expressions handle in Unicode also strings not explicitly flagged
UTF-8, but flagged native when running in UTF-8 locale.
Fixed a crash in fifo implementation on Windows (PR#18031).
Binary mode in fifo on Windows is now properly detected from argument open (PR#15600, PR#18031).

New in R for Windows 4.0.4 (Feb 16, 2021)

NEW FEATURES:
File share/texmf/tex/latex/jss.cls has been updated to work with LaTeX versions since Oct 2020
Unicode character width tables (as used by nchar(, type = "w")) have been updated to Unicode 12.1 by Brodie Gaslam (PR#17781) including many emoji
The internal table for iswprint (used on Windows, macOS and AIX) has been updated to include many recent Unicode characters INSTALLATION on a UNIX-ALIKE If an external BLAS is specified by --with-blas=foo or _via_ environment variable BLAS_LIBS is not found, this is now a configuration error. The previous behaviour was not clear from
the documentation: it was to continue the search as if with-blas=yes was specified
BUG FIXES:
all.equal(x,y) now "sees" the two different NAs in factors thanks to Bill Dunlap and others in PR#17897
(NULL)[1] and similar formula subsetting now works, thanks to a report and patch by Henrik Bengtsson in PR#17935. Additionally subsetting leaving an empty formula now works too, thanks to suggestions by Suharto Anggono
traceback(n) keeps source references again, as before R 4.0.0 fixing a regression; introduced by the PR#17580, reported
including two patch proposals by Brodie Gaslam
unlist(plst, recursive=FALSE) no longer drops content for pairlists with list components, thanks to the report and patch by Suharto Anggono in PR#17950
iconvlist() now also works on MUSL based (Linux) systems, from a report and patch suggestion by Wesley Chan in PR#17970
round() and signif() no longer tolerate wrong argument names notably in 1-argument calls; reported by Shane Mueller on R-devel mailing list); later reported as PR#17976
Machine has longdouble.* elements only if capabilities("long.double") is true, as documented. (Previously they were included if the platform had long double identical to double, as ARM does.)
p.adjust(numeric(), n=0) now works, fixing PR#18002
identical(x,y) no longer prints "Unknown Type .." for typeof(x) "..." objects
Fix (auto-)print()ing of named complex vectors, see PR#17868 and PR#18019
all.equal(<language>, <...>) now works, fixing PR#18029
as.data.frame.list(L, row.names=NULL) now behaves in line with data.frame(), disregarding names of components of L, fixing PR#18034, reported by Kevin Tappe
checkRdaFiles(ff)$version is now correct also when ff contains files of different versions, thanks to a report and patch from Sebastian Meyer in PR#18041

New in R for Windows 4.0.3 (Oct 10, 2020)

NEW FEATURES:
On platforms using configure option --with-internal-tzcode, additional values "internal" and (on macOS only) "macOS" are accepted for the environment variable TZDIR. (See ?TZDIR.)
On macOS, "macOS" is used by default if the system timezone database is a newer version than that in the R installation.
When install.packages(type = "source") fails to find a package in a repository it mentions package versions which are excluded by their R version requirement and links to hints on why a package might not be found.
The default value for options("timeout") can be set from enviromnent variable R_DEFAULT_INTERNET_TIMEOUT, still defaulting to 60 (seconds) if that is not set or invalid.
This may be needed when child R processes are doing downloads, for example during the installation of source packages which download jars or other forms of data.
LINK-TIME OPTIMIZATION on a UNIX-ALIKE:
There is now support for parallelized Link-Time Optimization (LTO) with GCC and for ‘thin’ LTO with clang via setting the LTO macro.
There is support for setting a different LTO flag for the Fortran compiler, including to empty when mixing clang and gfortran (as on macOS). See file ‘config.site’.
There is a new LTO_LD macro to set linker options for LTO compilation, for example to select an alternative linker or to parallelize thin LTO.
DEPRECATED AND DEFUNCT:
The LINPACK argument to chol.default(), chol2inv(), solve.default() and svd() has been defunct since R 3.1.0. Using it now gives a warning which will become an error in R 4.1.0.
BUG FIXES:
The code mitigating stack overflow with PCRE regexps on very long strings is enabled for PCRE2 < 10.30 also when JIT is enabled, since stack overflows have been seen in that case.
Fix to correctly show the group labels in dotchart() (which where lost in the ylab improvement for R 4.0.0).
addmargins(*, ..) now also works when fn() is a local function, thanks to bug report and patch PR#17124 from Alex Bertram.
rank(x) and hence sort(x) now work when x is an object (as per is.object(x)) of type "raw" and provides a valid `[` method, e.g., for gmp::as.bigz(.) numbers.
chisq.test(*, simulate.p.value=TRUE) and r2dtable() now work correctly for large table entries (in the millions). Reported by Sebastian Meyer and investigated by more helpers in PR#16184.
Low-level socket read/write operations have been fixed to correctly signal communication errors. Previously, such errors could lead to a segfault due to invalid memory access. Reported and debugged by Dmitriy Selivanov in PR#17850.
quantile(x, pr) works more consistently for pr values slightly outside [0,1], thanks to Suharto Anggono's PR#17891.
Further, quantile(x, prN, names=FALSE) now works even when prN contains NAs, thanks to Anggono's PR#17892. Ditto for ordered factors or Date objects when type = 1 or 3, thanks to PR#17899.
Libcurl-based internet access, including curlGetHeaders(), was not respecting the "timeout" option. If this causes unanticipated timeouts, consider increasing the default by setting R_DEFAULT_INTERNET_TIMEOUT.
as.Date(<char>) now also works with an initial "", thanks to Michael Chirico's PR#17909.
isS3stdGeneric(f) now detects an S3 generic also when it it is trace()d, thanks to Gabe Becker's PR#17917.
R_allocLD() has been fixed to return memory aligned for long double type PR#16534.
fisher.test() no longer segfaults when called again after its internal stack has been exceeded PR#17904.
Accessing a long vector represented by a compact integer sequence no longer segfaults (reported and debugged by Hugh Parsonage).
duplicated() now works also for strings with multiple encodings inside a single vector PR#17809.
phyper(11, 15, 0, 12, log.p=TRUE) no longer gives NaN; reported as PR#17271 by Alexey Stukalov.
Fix incorrect calculation in logLik.nls() PR#16100, patch from Sebastian Meyer.
A very old bug could cause a segfault in model.matrix() when terms involved logical variables. Part of PR#17879.
model.frame.default() allowed data = 1, leading to involuntary variable capture (rest of PR#17879).
tar() no longer skips non-directory files, thanks to a patch by Sebastian Meyer, fixing the remaining part of PR#16716.

New in R for Windows 4.0.2 (Jun 22, 2020)

New in R for Windows 4.0.1 (Jun 7, 2020)

New in R for Windows 4.0.0 (Apr 24, 2020)

SIGNIFICANT USER-VISIBLE CHANGES:
Packages need to be (re-)installed under this version (4.0.0) of R
matrix objects now also inherit from class "array", so e.g., class(diag(1)) is c("matrix", "array"). This invalidates code incorrectly assuming that class(matrix_obj)) has length one
S3 methods for class "array" are now dispatched for matrix objects
There is a new syntax for specifying raw character constants similar to the one used in C++: r"(...)" with ... any character sequence not containing the sequence )". This makes it easier to write strings that contain backslashes or both single and double quotes. For more details see ?Quotes
R now uses a stringsAsFactors = FALSE default, and hence by default no longer converts strings to factors in calls to data.frame() and read.table()
A large number of packages relied on the previous behaviour and so have needed/will need updating
The plot() S3 generic function is now in package base rather than package graphics, as it is reasonable to have methods that do not use the graphics package. The generic is currently re-exported from the graphics namespace to allow packages importing it from there to continue working, but this may change in future
Packages which define S4 generics for plot() should be re-installed and package code using such generics from other packages needs to ensure that they are imported rather than rely on their being looked for on the search path (as in a namespace, the base namespace has precedence over the search path)
REFERENCE COUNTING:
Reference counting is now used instead of the NAMED mechanism for determining when objects can be safely mutated in base C code. This reduces the need for copying in some cases and should allow further optimizations in the future. It should help make the internal code easier to maintain
This change is expected to have almost no impact on packages using supported coding practices in their C/C++ code
MIGRATION TO PCRE2
This version of R is built against the PCRE2 library for Perl-like regular expressions, if available. (On non-Windows platforms PCRE1 can optionally be used if PCRE2 is not available at build time.) The version of PCRE in use can be obtained via extSoftVersion(): PCRE1 (formerly known as ‘PCRE’) has versions <= 8, PCRE2 versions >= 10
Making PCRE2 available when building R from source is strongly recommended (preferably version 10.30 or later) as PCRE1 is no longer developed: version 8.44 is ‘likely to be the final release’
PCRE2 reports errors for some regular expressions that were accepted by PCRE1. A hyphen now has to be escaped in a character class to be interpreted as a literal (unless first or last in the class definition). R, B and X are no longer allowed in character classes (PCRE1 treated these as literals)
Option PCRE_study is no longer used with PCRE2, and is reported as FALSE when that is in use
NEW FEATURES:
assertError() and assertWarning() (in package tools) can now check for specific error or warning classes via the new optional second argument classes (which is not back compatible with previous use of an unnamed second argument)
DF2formula(), the utility for the data frame method of formula(), now works without parsing and explicit evaluation, starting from Suharto Anggono's suggestion in PR#17555
approxfun() and approx() gain a new argument na.rm defaulting to true. If set to false, missing y values now propagate into the interpolated values
Long vectors are now supported as the seq argument of a for() loop
str(x) gets a new deparse.lines option with a default to speed it up when x is a large call object
The internal traceback object produced when an error is signalled (.Traceback), now contains the calls rather than the deparse()d calls, deferring the deparsing to the user-level functions .traceback() and traceback(). This fulfils the wish of PR#17580, reported including two patch proposals by Brodie Gaslam
data.matrix() now converts character columns to factors and from this to integers
package.skeleton() now explicitly lists all exports in the ‘NAMESPACE’ file
New function .S3method() to register S3 methods in R scripts
file.path() has some support for file paths not in the session encoding, e.g. with UTF-8 inputs in a non-UTF-8 locale the output is marked as UTF-8
Most functions with file-path inputs will give an explicit error if a file-path input in a marked encoding cannot be translated (to the native encoding or in some cases on Windows to UTF-8), rather than translate to a different file path using escapes. Some (such as dir.exists(), file.exists(), file.access(), file.info(), list.files(), normalizePath() and path.expand()) treat this like any other non-existent file, often with a warning
There is a new help document accessed by help("file path encoding") detailing how file paths with marked encodings are handled
New function list2DF() for creating data frames from lists of variables
iconv() has a new option sub = "Unicode" to translate UTF-8 input invalid in the to encoding using <U+xxxx> escapes
There is a new function infoRDS() providing information about the serialization format of a serialized object
S3 method lookup now by default skips the elements of the search path between the global and base environments
Added an argument add_datalist(*, small.size = 0) to allow the creation of a ‘data/datalist’ file even when the total size of the data sets is small
The backquote function bquote() has a new argument splice to enable splicing a computed list of values into an expression, like ,@ in LISP's backquote
The formula interface to t.test() and wilcox.test() has been extended to handle one-sample and paired tests
The palette() function has a new default set of colours (which are less saturated and have better accessibility properties). There are also some new built-in palettes, which are listed by the new palette.pals() function. These include the old default palette under the name "R3". Finally, the new palette.colors() function allows a subset of colours to be selected from any of the built-in palettes
n2mfrow() gains an option asp = 1 to specify the aspect ratio, fulfilling the wish and extending the proposal of Michael Chirico in PR#17648
For head(x, n) and tail() the default and other S3 methods notably for vector n, e.g. to get a “corner” of a matrix, has been extended to array's of higher dimension thanks to the patch proposal by Gabe Becker in PR#17652. Consequently, optional argument addrownums is deprecated and replaced by the (more general) argument keepnums. An invalid second argument n now leads to typically more easily readable error messages
New function .class2() provides the full character vector of class names used for S3 method dispatch
Printing methods(..) now uses a new format() method
sort.list(x) now works for non-atomic objects x and method = "auto" (the default) or "radix" in cases order(x) works
Where they are available, writeBin() allows long vectors
New function deparse1() produces one string, wrapping deparse(), to be used typically in deparse1(substitute(*)), e.g., to fix PR#17671
wilcox.test() enhancements: In the (non-paired) two-sample case, Inf values are treated as very large for robustness consistency. If exact computations are used, the result now has "exact" in the method element of its return value. New arguments tol.root and digits.rank where the latter may be used for stability to treat very close numbers as ties
readBin() and writeBin() now report an error for an invalid endian value. The affected code needs to be fixed with care as the old undocumented behavior was to swap endian-ness in such cases
sequence() is now an S3 generic with an internally implemented default method, and gains arguments to generate more complex sequences. Based on code from the S4Vectors Bioconductor package and the advice of Hervé Pagès
print()'s default method and many other methods (by calling the default eventually and passing ...) now make use of a new optional width argument, avoiding the need for the user to set and reset options("width")
memDecompress() supports the RFC 1952 format (e.g. in-memory copies of gzip-compressed files) as well as RFC 1950
memCompress() and memDecompress() support long raw vectors for types "gzip" and "zx"
sweep() and slice.index() can now use names of dimnames for their MARGIN argument (apply has had this for almost a decade)
New function proportions() and marginSums(). These should replace the unfortunately named prop.table() and margin.table(). They are drop-in replacements, but also add named-margin functionality. The old function names are retained as aliases for back-compatibility
Functions rbinom(), rgeom(), rhyper(), rpois(), rnbinom(), rsignrank() and rwilcox() which have returned integer since R 3.0.0 and hence NA when the numbers would have been outside the integer range, now return double vectors (without NAs, typically) in these cases
matplot(x,y) (and hence matlines() and matpoints()) now call the corresponding methods of plot() and lines(), e.g, when x is a "Date" or "POSIXct" object; prompted by Spencer Graves' suggestion
stopifnot() now allows customizing error messages via argument names, thanks to a patch proposal by Neal Fultz in PR#17688
unlink() gains a new argument expand to disable wildcard and tilde expansion. Elements of x of value "~" are now ignored
mle() in the stats4 package has had its interface extended so that arguments to the negative log-likelihood function can be one or more vectors, with similar conventions applying to bounds, start values, and parameter values to be kept fixed. This required a minor extension to class "mle", so saved objects from earlier versions may need to be recomputed
The default for pdf() is now useDingbats = FALSE
The default fill colour for hist() and boxplot() is now col = "lightgray"
The default order of the levels on the y-axis for spineplot() and cdplot() has been reversed
If the R_ALWAYS_INSTALL_TESTS environment variable is set to a true value, R CMD INSTALL behaves as if the --install-tests option is always specified. Thanks to Reinhold Koch for the suggestion
New function R_user_dir() in package tools suggests paths appropriate for storing R-related user-specific data, configuration and cache files
capabilities() gains a new logical option Xchk to avoid warnings about X11-related capabilities
The internal implementation of grid units has changed, but the only visible effects at user-level should be
a slightly different print format for some units (especially unit arithmetic)
faster performance (for unit operations) and
two new functions unitType() and unit.psum()
Based on code contributed by Thomas Lin Pedersen
When internal dispatch for rep.int() and rep_len() fails, there is an attempt to dispatch on the equivalent call to rep()
Object .Machine now contains new longdouble.* entries (when R uses long doubles internally)
news() has been enhanced to cover the news on R 3.x and 2.x
For consistency, N <- NULL; N[[1]] <- val now turns N into a list also when val) has length one. This enables dimnames(r1)[[1]] <- "R1" for a 1-row matrix r1, fixing PR#17719 reported by Serguei Sokol
deparse(..), dump(..), and dput(x, control = "all") now include control option "digits17" which typically ensures 1:1 invertibility. New option control = "exact" ensures numeric exact invertibility via "hexDigits"
When loading data sets via read.table(), data() now uses LC_COLLATE=C to ensure locale-independent results for possible string-to-factor conversions
A server socket connection, a new connection type representing a listening server socket, is created via serverSocket() and can accept multiple socket connections via socketAccept()
New function socketTimeout() changes the connection timeout of a socket connection
The time needed to start a homogeneous PSOCK cluster on localhost with many nodes has been significantly reduced (package parallel)
New globalCallingHandlers() function to establish global condition handlers. This allows registering default handlers for specific condition classes. Developed in collaboration with Lionel Henry
New function tryInvokeRestart() to invoke a specified restart if one is available and return without signaling an error if no such restart is found. Contributed by Lionel Henry in PR#17598
str(x) now shows the length of attributes in some cases for a data frame x
Rprof() gains a new argument filter.callframes to request that intervening call frames due to lazy evaluation or explicit eval() calls be omitted from the recorded profile data. Contributed by Lionel Henry in PR#17595
The handling of ${FOO-bar} and ${FOO:-bar} in ‘Renviron’ files now follows POSIX shells (at least on a Unix-alike), so the first treats empty environment variables as set and the second does not. Previously both ignored empty variables. There are several uses of the first form in ‘etc/Renviron’
New classes argument for suppressWarnings() and suppressMessages() to selectively suppress only warnings or messages that inherit from particular classes. Based on patch from Lionel Henry submitted with PR#17619
New function activeBindingFunction() retrieves the function of an active binding
New "cairoFT" and "pango" components in the output of grSoftVersion()
New argument symbolfamily in cairo-based graphics devices and new function cairoSymbolFont() that can be used to provide the value for that argument
Windows
Rterm now works also when invoked from MSYS2 terminals. Line editing is possible when command winpty is installed
normalizePath() now resolves symbolic links and normalizes case of long names of path elements in case-insensitive folders (PR#17165)
md5sum() supports UTF-8 file names with characters that cannot be translated to the native encoding (PR#17633)
Rterm gains a new option --workspace to specify the workspace to be restored. This allows equals to be part of the name when opening via Windows file associations (reported by Christian Asseburg)
Rterm now accepts ALT+xxx sequences also with NumLock on. Tilde can be pasted with an Italian keyboard (PR#17679)
R falls back to copying when junction creation fails during package checking (patch from Duncan Murdoch)
DEPRECATED AND DEFUNCT
Make macro F77_VISIBILITY has been removed and replaced by F_VISIBILITY
Make macros F77, FCPIFCPLAGS and SHLIB_OPENMP_FCFLAGS have been removed and replaced by FC, FPICFLAGS and SHLIB_OPENMP_FFLAGS respectively. (Most make programs will set F77 to the value of FC, which is set for package compilation. But portable code should not rely on this.)
The deprecated support for specifying C++98 for package installation has been removed
R CMD config no longer knows about the unused settings F77 and FCPIFCPLAGS, nor CXX98 and similar
Either PCRE2 or PCRE1 >= 8.32 (Nov 2012) is required: the deprecated provision for 8.20–8.31 has been removed
Defunct functions mem.limits(), .readRDS(), .saveRDS(),..find.package(), and .path.package() from package base and allGenerics(), getAccess(), getAllMethods(), getClassName(), getClassPackage(), getExtends(), getProperties(), getPrototype(), getSubclasses(), getVirtual(), mlistMetaName(), removeMethodsObject(), seemsS4Object(), traceOff(), and traceOn() from methods have been removed
C-LEVEL FACILITIES:
installChar is now remapped in ‘Rinternals.h’ to installTrChar, of which it has been a wrapper since R 3.6.0. Neither are part of the API, but packages using installChar can replace it if they depend on R >= 3.6.2
Header ‘R_ext/Print.h’ defines R_USE_C99_IN_CXX and hence exposes Rvprintf and REvprintf if used with a C++11 (or later) compiler
There are new Fortran subroutines dblepr1, realpr1 and intpr1 to print a scalar variable (gfortran 10 enforces the distinction between scalars and length-one arrays). Also labelpr to print just a label
R_withCallingErrorHandler is now available for establishing a calling handler in C code for conditions inheriting from class error
INSTALLATION on a UNIX-ALIKE
User-set DEFS (e.g., in ‘config.site’) is now used for compiling packages (including base packages)
There is a new variant option --enable-lto=check for checking consistency of BLAS/LAPACK/LINPACK calls — see ‘Writing R Extensions’
A C++ compiler default is set only if the C++11 standard is supported: it no longer falls back to C++98
PCRE2 is used if available. To make use of PCRE1 if PCRE2 is unavailable, configure with option --with-pcre1
The minimum required version of libcurl is now 7.28.0 (Oct 2012)
New make target distcheck checks
R can be rebuilt from the tarball created by make dist
the build from the tarball passes make check-all
the build installs and uninstalls
the source files are properly cleaned by make distclean
UTILITIES:
R --help now mentions the option --no-echo (renamed from --slave) and its previously undocumented short form -s
R CMD check now optionally checks configure and cleanup scripts for non-Bourne-shell code (‘bashisms’)
R CMD check --as-cran now runs donttest examples (which are run by example()) instead of instructing the tester to do so. This can be temporarily circumvented during development by setting environment variable _R_CHECK_DONTTEST_EXAMPLES_ to a false value
PACKAGE INSTALLATION
There is the beginnings of support for the recently approved C++20 standard, specified analogously to C++14 and C++17. There is currently only limited support for this in compilers, with flags such as -std=c++20 and -std=c++2a. For the time being the configure test is of accepting one of these flags and compiling C++17 code
BUG FIXES:
formula(x) with length(x) > 1 character vectors, is deprecated now. Such use has been rare, and has ‘worked’ as expected in some cases only. In other cases, wrong x have silently been truncated, not detecting previous errors
Long-standing issue where the X11 device could lose events shortly after startup has been addressed (PR#16702)
The data.frame method for rbind() no longer drops <NA> levels from factor columns by default (PR#17562)
available.packages() and hence install.packages() now pass their ... argument to download.file(), fulfilling the wish of PR#17532; subsequently, available.packages() gets new argument quiet, solving PR#17573
stopifnot() gets new argument exprObject to allow an R object of class expression (or other ‘language’) to work more consistently, thanks to suggestions by Suharto Anggono
conformMethod() now works correctly in cases containing a “&& logic” bug, reported by Henrik Bengtsson. It now creates methods with "missing" entries in the signature. Consequently, rematchDefinition() is amended to use appropriate .local() calls with named arguments where needed
format.default(*, scientific = FALSE) now corresponds to a practically most extreme options(scipen = n) setting rather than arbitrary n = 100
format(as.symbol("foo")) now works (returning "foo")
postscript(.., title = *) now signals an error when the title string contains a character which would produce corrupt PostScript, thanks to PR#17607 by Daisuko Ogawa
Certain Ops (notably comparison such as ==) now also work for 0-length data frames, after reports by Hilmar Berger
methods(class = class(glm(..))) now warns more usefully and only once
write.dcf() no longer mangles field names (PR#17589)
Primitive replacement functions no longer mutate a referenced first argument when used outside of a complex assignment context
A better error message for contour(*, levels = Inf)
The return value of contourLines() is no longer invisible()
The Fortran code for calculating the coefficients component in lm.influence() was very inefficient. It has (for now) been replaced with much faster R code (PR#17624)
cm.colors(n) etc no longer append the code for alpha = 1, "FF", to all colors. Hence all eight *.colors() functions and rainbow() behave consistently and have the same non-explicit default (PR#17659)
dnorm had a problematic corner case with sd == -Inf or negative sd which was not flagged as an error in all cases. Thanks to Stephen D. Weigand for reporting and Wang Jiefei for analyzing this; similar change has been made in dlnorm()
The optional iter.smooth argument of plot.lm(), (the plot() method for lm and glm fits) now defaults to 0 for all glm fits. Especially for binary observations with high or low fitted probabilities, this effectively deleted all observations of 1 or 0. Also, the type of residuals used in the glm case has been switched to "pearson" since deviance residuals do not in general have approximately zero mean
In plot.lm, Cook's distance was computed from unweighted residuals, leading to inconsistencies. Replaced with usual weighted version. (PR#16056)
Time-series ts(*, start, end, frequency) with fractional frequency are supported more consistently; thanks to a report from Johann Kleinbub and analysis and patch by Duncan Murdoch in PR#17669
In case of errors mcmapply() now preserves attributes of returned "try-error" objects and avoids simplification, overriding SIMPLIFY to FALSE. (PR#17653)
as.difftime() gets new optional tz = "UTC" argument which should fix behaviour during daylight-savings-changeover days, fixing PR#16764, thanks to proposals and analysis by Johannes Ranke and Kirill Müller
round() does a better job of rounding “to nearest” by measuring and “to even”; thanks to a careful algorithm originally prompted by the report from Adam Wheeler and then others, in PR#17668
round(x, dig) for negative digits is much more rational now, notably for large |dig|
Inheritance information on S4 classes is maintained more consistently, particularly in the case of class unions (in part due to PR#17596 and a report from Ezra Tucker)
is() behaves more robustly when its argument class2 is a classRepresentation object
The warning message when attempting to export an nonexistent class is now more readable; thanks to Thierry Onkelinx for recognizing the problem
choose() misbehaved in corner cases where it switched n - k for k and n was only nearly integer (report from Erik Scott Wright)
mle() in the stats4 package had problems combining use of box constraints and fixed starting values (in particular, confidence intervals were affected)
Operator ? now has lower precedence than = to work as documented, so = behaves like <- in help expressions (PR#16710)
smoothEnds(x) now returns integer type in both cases when x is integer, thanks to a report and proposal by Bill Dunlap PR#17693
The methods package does a better job of tracking inheritance relationships across packages
norm(diag(c(1, NA)), "2") now works
subset() had problems with 0-col dataframes (reported by Bill Dunlap, PR#17721)
Several cases of integer overflow detected by the ‘undefined behaviour sanitizer’ of clang 10 have been circumvented. One in rhyper() may change the generated value for large input values
dotchart() now places the y-axis label (ylab) much better, not overplotting labels, thanks to a report and suggestion by Alexey Shipunov
A rare C-level array overflow in chull() has been worked around
Some invalid specifications of the day-of-the-year (via %j, e.g. day 366 in 2017) or week plus day-of-the-week are now detected by strptime(). They now return NA but give a warning as they may have given random results or corrupted memory in earlier versions of R
socketConnection(server = FALSE) now respects the connection timeout also on Linux
socketConnection(server = FALSE) no longer leaks a connection that is available right away without waiting (e.g. on localhost)
Socket connections are now robust against spurious readability and spurious availability of an incoming connection
blocking = FALSE is now respected also on the server side of a socket connection, allowing non-blocking read operations
anova.glm() and anova.glmlist() computed incorrect score (Rao) tests in no-intercept cases. (André Gillibert, PR#17734)
summaryRprof() now should work correctly for the Rprof(*, memory.profiling=TRUE) case with small chunk size (and "tseries" or similar) thanks to a patch proposal by Benjamin Tyner, in PR#15886
xgettext() ignores strings passed to ngettext(), since the latter is handled by xngettext(). Thanks to Daniele Medri for the report and all the recent work he has done on the Italian translations
data(package = "P") for P in base and stats no longer reports the data sets from package datasets (which it did for back compatibility for 16 years), fixing PR#17730
x[[Inf]] (returning NULL) no longer leads to undefined behavior, thanks to a report by Kirill Müller in PR#17756. Further, x[[-Inf]] and x[[-n]] now give more helpful error messages
Gamma() family sometimes had trouble storing link name PR#15891
BUG FIXES (Windows)
Sys.glob() now supports all characters from the Unicode Basic Multilingual Plane, no longer corrupting some (less commonly used) characters (PR#17638)
Rterm now correctly displays multi-byte-coded characters representable in the current native encoding (at least on Windows 10 they were sometimes omitted, PR#17632)
scan() issues with UTF-8 data when running in a DBCS locale have been resolved (PR#16520, PR#16584)
RTerm now accepts enhanced/arrow keys also with ConPTY
R can can now be started via the launcher icon in a user documents directory whose path is not representable in the system encoding
socketConnection(server = FALSE) now returns instantly also on Windows when connection failure is signalled
Problems with UTF-16 surrogate pairs have been fixed in several functions, including tolower() and toupper() (PR#17645)

New in R for Windows 3.6.3 (Feb 29, 2020)

New in R for Windows 3.6.2 (Dec 12, 2019)

NEW FEATURES:
runmed(x, *) gains a new option na.action determining _how_ to handle NaN or NA in x.
dotchart() gains new options ann, xaxt, frame.plot and log.
INSTALLATION on a UNIX-ALIKE:
Detection of the C stack direction has been moved from run-time to configure: this is safer with LTO builds and allows the detection to be overridden - see file config.site.
Source-code changes enable installation on platforms using gcc-fno-common (the expected default for gcc 10.x).
C-LEVEL FACILITIES:
installTrChar (which is nowadays is wrapped by installChar) is defined in Rinternals.h. (Neither are part of the API.)
PACKAGE INSTALLATION:
Header Rconfig.h contains the value of FC_LEN_T deduced at installation which is used by the prototypes in headers R_ext/BLAS.h and R_ext/Lapack.h but to avoid extensive breakage this is only exposed when USE_FC_LEN_T is defined.
If a package's C/C++ calls to BLAS/LAPACK allow for the 'hidden' arguments used by most Fortran compilers to pass the lengths of Fortran character arguments, define USE_FC_LEN_T and include Rconfig.h (possibly _via_ R.h) before including R_ext/BLAS.h or R_ext/Lapack.h.
A package with Fortran source code and perhaps C (but not C++) sources can request for its shared object/DLL to be linked by the Fortran compiler by including a line USE_FC_TO_LINK= in src/Makevars[.win] and using $(SHLIB_OPENMP_FFLAGS) as part of PKG_LIBS.
The known reason for doing so is a package which uses Fortran (only) OpenMP on a platform where the Fortran OpenMP runtime is incompatible with the C one (e.g. gfortran 9.x with clang).
UTILITIES:
R CMD check has a new option to mitigate checks leaving files/directories in /tmp. See the 'R Internals' manual - this is part of --as-cran.
Windows:
The default standard for C++ in package installation is C++11 (as it has been on other platforms where available since R 3.6.0: the default toolchain on Windows was defaulting to C++98).
DEPRECATED AND DEFUNCT:
Support for specifying C++98 in package installation is deprecated.
Support in R CMD config for F77, FCPIFCPLAGS, CPP, CXXCPP and CXX98 and similar is deprecated. (CPP is found from the system make and may well not be set.)
Use $CC -E and $CXX -E instead of CPP and CXXCPP.
BUG FIXES:
runmed(x, *) when x contains missing values now works consistently for both algorithm="Stuetzle" and "Turlach", and no longer segfaults for "Turlach", as reported by Hilmar Berger.
apply(diag(3), 2:3, mean) now gives a helpful error message.
dgamma(x, shape, log=TRUE) now longer overflows to Inf for shape < 1 and very small x, fixing PR#17577, reported by Jonathan Rougier.
Buffer overflow in building error messages fixed. Reported by Benjamin Tremblay.
options(str = .) is correctly initialized at package utils load time, now. A consequence is that str() in scripts now is more consistent to interactive use, e.g., when displaying function(**) argument lists.
as.numeric(<call>) now gives correct error message.
Printing ls.str() no longer wrongly shows "<missing>" in rare cases.
Auto-printing S4 objects no longer duplicates the object, for faster speed and reduced memory consumption. Reported by Aaron Lun.
pchisq(<LRG>, <LRG>, ncp=100) no longer takes practically forever in some cases. Hence ditto for corresponding qchisq() calls.
x %% L for finite x no longer returns NaN when L is infinite, nor suffers from cancellation for large finite L, thanks to Long Qu's
Analogously, x %/% L and L %/% x suffer less from cancellation and return values corresponding to limits for large L.
grepl(NA, *) now returns logical as documented.
options(warn=1e11) is an error now, instead of later leading to C stack overflow because of infinite recursion.
R_tryCatch no longer transfers control for all conditions. Reported and patch provided by Lionel Henry in PR#17617.
format(object.size(.), digits=NULL) now works, fixing PR#17628 reported by Jonathan Carroll.
get_all_vars(f, d) now also works for cases, e.g. where d contains a matrix. Reported by Simon Wood in 2009 and patch provided by Ben Bolker in PR#13624.
Additionally, it now also works when some variables are data frames, fixing PR#14905, reported by Patrick Breheny.
barplot() could get spacings wrong if there were exactly two bars PR#15522. Patch by Michael Chirico.
power.t.test() works in more cases when returning values of n smaller than 2.
dotchart(*, pch=., groups=.) now works better. Reported by Robert and confirmed by Nic Rochette in PR#16953.
canCoerce(obj, cl) no longer assumes length(class(obj)) == 1.
plot.formula(*, subset = *) now also works in a boundary case reported by Robert Schlicht (TU Dresden).
readBin() and writeBin() of a rawConnection() now also work inlarge cases, thanks to a report and proposal by Taeke Harkema in

New in R for Windows 3.6.0 (May 1, 2019)

SIGNIFICANT USER-VISIBLE CHANGES:
Serialization format version 3 becomes the default for serialization and saving of the workspace (save(), serialize(), saveRDS(), compiler::cmpfile()). Serialized data in format 3 cannot be read by versions of R prior to version 3.5.0. Serialization format version 2 is still supported and can be selected by version = 2 in the save/serialization functions. The default can be changed back for the whole R session by setting environment variables R_DEFAULT_SAVE_VERSION and R_DEFAULT_SERIALIZE_VERSION to 2. For maximal back-compatibility, files ‘vignette.rds’ and ‘partial.rdb’ generated by R CMD build are in serialization format version 2, and resave by default produces files in serialization format version 2 (unless the original is already in format version 3).
The default method for generating from a discrete uniform distribution (used in sample(), for instance) has been changed. This addresses the fact, pointed out by Ottoboni and Stark, that the previous method made sample() noticeably non-uniform on large populations. See PR#17494 for a discussion. The previous method can be requested using RNGkind() or RNGversion() if necessary for reproduction of old results. Thanks to Duncan Murdoch for contributing the patch and Gabe Becker for further assistance.
The output of RNGkind() has been changed to also return the ‘kind’ used by sample().
NEW FEATURES:
Sys.setFileTime() has been vectorized so arguments path and time of length greater than one are now supported.
axis() gets new option gap.axis = NA for specifying a multiplication factor for the minimal “gap” (distance) between axis labels drawn. Its default is 1 for labels parallel to the axis, and 0.25 for perpendicular ones.
Perpendicular labels no longer overlap, fixing bug PR#17384.
The default method of plot() gains new arguments xgap.axis = NA and ygap.axis = NA to be passed to the x– and y– axis(.., gap.axis=*) calls.
removeSource() now works not only for functions but also for some language objects.
as.call(), rep.int(), rep_len() and nchar() dispatch internally.
is(object, class2) looks for class2 in the calling namespace after looking in the namespace of class(object).
extendrange(.., f) with a length-2 f now extends separately to the left and the right.
lengths() dispatches internally to S4 methods.
download.file() on Windows now uses URLdecode() to determine the file extension, and uses binary transfer (mode = "wb") also for file extension ‘.rds’.
The help page for download.file() now contains the same information on all platforms.
Setting C locale for collation via environment variables LC_ALL and LC_COLLATE and via a call to Sys.setlocale() now takes precedence over environment variable R_ICU_LOCALE.
There is a new function, nullfile(), to give the file name of the null system device (e.g., ‘/dev/null’) on the current platform.
There are two new options, keep.parse.data and keep.parse.data.pkgs, which control whether parse data are included into sources when keep.source or keep.source.pkgs is TRUE. By default, keep.parse.data.pkgs is now FALSE, which changes previous behavior and significantly reduces space and time overhead when sources are kept when installing packages.
In rapply(x, ..), x can also be “list-like” and of length >= 2^{31}.
trimws() gets new optional whitespace argument, allowing more extensive definitions of “space”, such as including Unicode spaces (as wished in PR#17431).
weighted.mean() no longer coerces the weights to a double/numeric vector, since sum() now handles integer overflow. This makes weighted.mean() more polymorphic and endomorphic, but be aware that the results are no longer guaranteed to be a vector of type double.
When loading namespaces, S3 method registrations which overwrite previous registrations are now noted by default (using packageStartupMessage()).
compiler::cmpfile() gains a version argument, for use when the output file should be saved in serialization format 2.
The axis labeling in the default method of pairs() may now be toggled by new options horOdd and verOdd.
(Not Windows nor macOS.) Package tcltk now supports an environment variable R_DONT_USE_TK which if set disables Tk initialization. This is intended for use to circumvent errors in loading the package, e.g. with recent Linux running under an address sanitizer.
The numeric method of all.equal() gets optional arguments countEQ and formatFUN. If countEQ is true, the mean error is more sensible when many entries are equal.
outer(x,y, FUN = "*") is more efficient using tcrossprod(u,v) instead of u %*% t(v).
vcov(<mlm>) is more efficient via new optional arguments in summary.mlm().
The default method of summary() gets an option to choose the kind of quantile()s to use; wish of PR#17438.
Fitting multiple linear models via lm() does work with matrix offsets, as suggested in PR#17407.
The new functions mem.maxVSize() and mem.maxMSize() allow the maximal size of the vector heap and the maximal number of nodes allowed in the current R process to be queried and set.
news() gains support for ‘NEWS.md’ files.
An effort has been started to have our reference manuals, i.e., all help pages. show platform-independent information (rather than Windows or Unix-alike specifics visible only on that platform). Consequently, the Windows version of X11() / x11() got identical formal arguments to the Unix one.
sessionInfo()$running has been factored out in a new variable osVersion.
slice.index() now also works for multi-dimensional margins.
untar() used with an external tar command assumes this supports decompression including xz and automagically detecting the compression type. This has been true of all mainstream implementations since 2009 (for GNU tar, since version 1.22): older implementations are still supported via the new argument support_old_tars whose default is controlled by environment variable R_SUPPORT_OLD_TARS. (It looks like NetBSD and OpenBSD have ‘older’ tar commands for this purpose.)
The new function asplit() allow splitting an array or matrix by its margins.
New functions errorCondition() and warningCondition() provide a convenient way to create structured error and warning objects.
.Deprecated() now signals a warning of class "deprecatedWarning", and .Defunct() now signals an error of class "defunctError".
Many ‘package not found’ errors are now signaled as errors of class "packageNotFoundError".
As an experimental feature, when loadNamespace() fails because the requested package is not available the error is initially signaled with a retry_loadNamespace restart available. This allows a calling handler to try to install the package and continue.
S3method() directives in ‘NAMESPACE’ can now also be used to perform delayed S3 method registration.
Experimentally, setting environment variable _R_CHECK_LENGTH_1_LOGIC2_ will lead to warnings (or errors if the variable is set to a ‘true’ value) when && or || encounter and use arguments of length more than one.
Added "lines" and "chars" coordinate systems to grconvertX() and grconvertY().
getOption() is more efficient notably for the rare case when called with two arguments, from several contributors in PR#17394.
In .col(dim) and .row(dim), dim now may also be an integer-valued "double".
sQuote() and dQuote() get an explicit q argument with obvious default instead of using getOption("fancyQuotes") implicitly and unconditionally.
unzip() can list archives with comments and with spaces in file names even using an external unzip command.
Command line completion has a new setting rc.settings(dots = FALSE) to remove ... from the list of possible function arguments.
library() no longer checks packages with compiled code match R.version$platform. loadNamespace() never has, and increasingly the ‘canonical name’ does not reflect the important characteristics of compiled code.
The primitive functions drop() and unclass() now avoid duplicating their data for atomic vectors that are large enough, by returning ALTREP wrapper objects with adjusted attributes. R-level assignments to change attributes will also use wrapper objects to avoid duplicating data for larger atomic vectors. R functions like structure() and unname() will therefore not duplicate data in these settings. Generic vectors as produced by list() are not yet covered by this optimization but may be in due course.
In formals(), envir becomes an optional argument instead of being hardwired.
Instead of signalling an error for an invalid S4 object x, str(x) now gives a warning and subsequently still shows most parts of x, e.g., when slots are missing.
gamma(x) and lgamma(x) no longer warn when correctly returning Inf or underflowing to zero. This helps maximum likelihood and similar computations.
convertColor() is now vectorized, so a lot faster for converting many colours at once. The new argument vectorized to colorConverter() ensures that non-vectorized colour converters still work. (Thanks to Brodie Gaslam.)
download.file() and url() get new argument headers for custom HTTP headers, e.g., allowing to perform basic http authentication, thanks to a patch contributed by Gábor Csárdi.
File-based connection functions file(), gzfile(), bzfile() and xzfile() now signal an error when used on a directory.
For approx(), splinefun() etc, a new setting ties = c("ordered", <fun>) allows skipping the sorting and still treat ties.
format(x) gives a more user friendly error message in the case where no method is defined. A minimal method is provided in format.default(x) when isS4(x) is true.
which(x) now also works when x is a long vector, thanks to Suharto Anggono's PR#17201. NB: this may return a double result, breaking the previous guarantee of an integer result.
seq.default() is more careful to return an integer (as opposed to double) result when its arguments are large and/or classed objects; see comment #9 of Suharto Anggono's PR#17497.
The plot() method for lm and glm fits, plot.lm(), gains a new option iter.smooth with a default of 0 for binomial fits, no longer down-weighting when smoothing the residuals.
zip() passes its list of files via standard input to the external command when too long for the command line (on some platforms).
data() gains an overwrite argument.
t.test() now also returns the standard error (in list component stderr).
model.matrix(*, contrasts.arg = CC) now warns about invalid contrasts.args.
Performance of substr() and substring() has been improved.
stopifnot() has been simplified thanks to Suharto Anggono's proposals to become considerably faster for cheap expressions.
The default ‘user agent’ has been changed when accessing http:// and https:// sites using libcurl. (A site was found which caused libcurl to infinite-loop with the previous default.)
sessionInfo() now also contains RNGkind() and prints it when it differs from the default; based on a proposal and patch by Gabe Becker in PR#17535. Also, RNGversion(getRversion()) works directly.
library() and require() now allow more control over handling search path conflicts when packages are attached. The policy is controlled by the new conflicts.policy option.
barplot() gets a formula method, thanks to a patch proposal by Arni Magnusson in PR#17521.
pmax() and pmin(x) now also work for long vectors, thanks to Suharto Anggono's PR#17533.
bxp() now warns when omitting duplicated arguments.
New hcl.colors() function to provide wide range of HCL-based colour palettes with much better perceptual properties than the existing RGB/HSV-based palettes like rainbow().
Also a new hcl.pals() function to list available palette names for hcl.colors().
Contributed by Achim Zeileis.
The default colours for image() and filled.contour() are now based on hcl.colors().
The palette-generating functions rainbow(), gray.colors(), etc. get a new rev argument to facilitate reversing the order of colors.
New str2lang() and str2expression() as streamlined versions of parse(text=., keep.source=FALSE) allow to abstract typical call constructions, e.g., in formula manipulations. (Somewhat experimental)
Add update_PACKAGES() for incrementally updating a package repository index, instead of rebuilding the index from scratch. Thanks to Gabe Becker in PR#17544 for the patch, based on part of his switchr package.
INSTALLATION on a UNIX-ALIKE:
The options selected for the C++ compiler default to the C++11 standard if supported, otherwise to the C++98 standard.
Visibility macros such as C_VISIBILITY can now be user-set (including to empty), e.g. in ‘config.site’.
Macro FCLIBS, which has sometimes been needed on Solaris, has been renamed to FCLIBS_XTRA.
Macro F77 is always set to the value of FC, so the latter should be set to user-select the Fortran compiler for both fixed-form and free-form Fortran. In particular, gfortran is now the first choice for F77, not f95.
Macros FFLAGS and FCFLAGS remain distinct to allow for a compiler which needs a flag to select free- or fixed-form Fortran (most use the source-file extension to choose: ‘.f’ is fixed-form and ‘.f90’ and ‘.f95’ are free-form).
If only one of them is set, its value is used for both.
The special-casing of CFLAGS, CXXFLAGS and FFLAGS for Intel compilers on Linux has been removed: we do not have recent experience but the generic defaults now chosen are the same as those previously special-cased for x86_64.
If necessary, override the defaults on the configure command line or in file ‘config.site’.
Long-untested configure support for HP-UX and very old versions of Linux has been removed.
configure --with-blas (without specifying a value) includes OpenBLAS in its search (before ATLAS and a generic BLAS). This follows recent versions of the ax_blas autoconf macro.
The configure macro MAKEINFO has been updated to TEXI2ANY.
Support for make install-strip has been enhanced.
PACKAGE INSTALLATION:
Source package installation is by default ‘staged’: the package is installed into a temporary location under the final library directory and moved into place once the installation is complete. The benefit is that partially-installed packages are hidden from other R sessions.
The overall default is set by environment variable R_INSTALL_STAGED. R CMD INSTALL has new options --staged-install and --no-staged-install, and packages can use the StagedInstall field in their ‘DESCRIPTION’ file to opt out. (That opt-out is a temporary measure which may be withdrawn in future.)
Staged installation requires either --pkglock or --lock, one of which is used by default.
The interpretation of source code with extension ‘.f’ is changing. Previously this denoted FORTRAN 77 code, but current compilers no longer have a FORTRAN 77 mode and interpret it as ‘fixed-form’ Fortran 90 (or later where supported) code. Extensions ‘.f90’ and ‘.f95’ continue to indicate ‘free-form’ Fortran code.
Legal FORTRAN 77 code is also legal fixed-form Fortran 9x; however this change legitimizes the use of later features, in particular to replace features marked ‘obsolescent’ in Fortran 90 and ‘deleted’ in Fortran 2018 which gfortran 8.x and later warn about.
Packages containing files in the ‘src’ directory with extensions ‘.f90’ or ‘.f95’ are now linked using the C or C++ compiler rather than the Fortran 9x compiler. This is consistent with fixed-form Fortran code and allows mixing of C++ and free-form Fortran on most platforms.
Consequentially, a package which includes free-form Fortran 9x code which uses OpenMP should include SHLIB_OPENMP_CFLAGS (or the CXXFLAGS version if they also include C++ code) in PKG_LIBS rather than SHLIB_OPENMP_FCFLAGS — fortunately on almost all current platforms they are the same flag.
Macro PKG_FFLAGS will be used for the compilation of both fixed-form and free-form Fortran code unless PKG_FCFLAGS is also set (in ‘src/Makevars’ or ‘src/Makevars.win’).
The make macro F_VISIBILITY is now preferred for both fixed-form and free-form Fortran, for use in ‘src/Makevars’ and similar.
R CMD INSTALL gains a new option --strip which (where supported) strips installed shared object(s): this can also be achieved by setting the environment variable _R_SHLIB_STRIP_ to a true value.
The new option --strip-lib attempts stripping of static and shared libraries installed under ‘lib’.
These are most useful on platforms using GNU binutils (such as Linux) and compiling with -g flags.
There is more support for installing UTF-8-encoded packages in a strict Latin-1 locale (and probably for other Latin locales): non-ASCII comments in R code (and ‘NAMESPACE’ files) are worked around better.
UTILITIES:
R CMD check now optionally checks makefiles for correct and portable use of the SHLIB_OPENMP_*FLAGS macros.
R CMD check now evaluates Sexpr{} expressions (including those in macros) before checking the contents of ‘Rd’ files and so detects issues both in evaluating the expressions and in the expanded contents.
R CMD check now lists missing packages separated by commas and with regular quotes such as to be useful as argument in calling install.packages(c(..)); from a suggestion by Marcel Ramos.
tools::Rd2latex() now uses UTF-8 as its default output encoding.
R CMD check now checks line endings of files with extension ‘.hpp’ and those under ‘inst/include’. The check now includes that a non-empty file is terminated with a newline.
R CMD build will correct line endings in such files.
R CMD check now tries re-building all vignettes rather than stopping at the first error: whilst doing so it adds ‘bookmarks’ to the log. By default (see the ‘R Internals’ manual) it re-builds each vignette in a separate process.
It now checks for duplicated vignette titles (also known as ‘index entries’): they are used as hyperlinks on CRAN package pages and so do need to be unique.
R CMD check has more comprehensive checks on the ‘data’ directory and the functioning of data() in a package.
R CMD check now checks autoconf-generated ‘configure’ files have their corresponding source files, including optionally attempting to regenerate them on platforms with autoreconf.
R CMD build has a new option --compression to select the compression used for the tarball.
R CMD build now removes ‘src/*.mod’ files on all platforms.
C-LEVEL FACILITIES:
New pointer protection C functions R_PreserveInMSet and R_ReleaseFromMSet have been introduced to replace UNPROTECT_PTR, which is not safe to mix with UNPROTECT (and with PROTECT_WITH_INDEX). Intended for use in parsers only.
NAMEDMAX has been raised to 7 to allow further protection of intermediate results from (usually ill-advised) assignments in arguments to BUILTIN functions. Properly written package code should not be affected.
R_unif_index is now considered to be part of the C API.
R_GetCurrentEnv() allows C code to retrieve the current environment.
DEPRECATED AND DEFUNCT:
Argument compressed of untar() is deprecated — it is only used for external tar commands which increasingly for extraction auto-detect compression and ignore their zjJ flags.
var(f) and hence sd(f) now give an error for factor arguments; they gave a deprecation warning since R 3.2.3, PR#16564.
Package tools' vignetteDepends() has been deprecated (it called a function deprecated since Feb 2016), being partly replaced by newly exported vignetteInfo().
The f77_f2c script has been removed: it no longer sufficed to compile the ‘.f’ files in R.
The deprecated legacy support of make macros such as CXX1X has been removed: use the CXX11 forms instead.
Make macro F77_VISIBILITY is deprecated in favour of F_VISIBILITY.
Make macros F77, FCPIFCPLAGS and SHLIB_OPENMP_FCFLAGS are deprecated in favour of FC, FPICFLAGS and SHLIB_OPENMP_FFLAGS respectively.
$.data.frame had become an expensive version of the default method, so has been removed. (Thanks to Radford Neal for picking this up and to Duncan Murdoch for providing a patch.)
BUG FIXES:
replayPlot(r) now also works in the same R session when r has been “reproduced” from serialization, typically after saving to and reading from an RDS file.
substr() and substring() now signal an error when the input is invalid UTF-8.
file.copy() now works also when its argument to is of length greater than one.
mantelhaen.test() no longer suffers from integer overflow in largish cases, thanks to Ben Bolker's PR#17383.
Calling setGeneric("foo") in a package no longer fails when the enclosing environment of the implicit generic foo() is .GlobalEnv.
untar(file("<some>.tar.gz"), *) now gives a better error message, suggesting to use gzfile() instead.
Method dispatch uses more relevant environments when looking up class definitions.
The documentation for identify() incorrectly claimed that the indices of identified points were returned in the order that the points were selected. identify() now has a new argument order to allow the return value to include the order in which points were identified; the documentation has been updated. Reported by Richard Rowe and Samuel Granjeaud.
order(...., decreasing=c(TRUE, FALSE)) could fail in some cases. Reported from StackOverflow via Karl Nordström.
User macros in Rd files now accept empty and multi-line arguments.
Calling install.packages() with a length zero pkgs argument now is a no-op (PR#17422).
unlist(x) now returns a correct factor when x is a nested list with factor leaves, fixing PR#12572 and PR#17419.
The documentation help(family) gives more details about the aic component, thanks to Ben Bolker's prompting.
The documentation for attributes and `attributes<-` now gives x as name of the first and main argument which the implementation has been requiring, fixing PR#17434. For consistency, the first argument name is also changed from obj to x for `mostattributes<-`.
strwidth() now uses par("font") as default font face (PR#17352).
plot(<table>, log="x") no longer warns about log.
The print() method for "htest" objects now formats the test statistic and parameter directly and hence no longer rounds to units before the decimal point. Consequently, printing of t.test() results with a small number of digits now shows non-large df's to the full precision (PR#17444).
kruskal.test() and fligner.test() no longer erroneously insist on numeric g group arguments (PR#16719).
Printing a news db via the browser now does a much better job (PR#17433).
print.aov() missed column names in the multivariate case due to misspelling (reported by Chris Andrews).
axis() now creates valid at locations also for small subnormal number ranges in log scale plots.
format.POSIXlt() now also recycles the zone and gmtoff list components to full length when needed, and its internal C code detects have_zone in more cases. In some cases, this changes its output to become compatible with format.POSIXct().
On Windows, detectCores() in package parallel now detects processors in all processor groups, not just the group R is running in (impacts particularly systems with more than 64 logical processors). Reported by Arunkumar Srinivasan.
On Windows, socketSelect() would hang with more than 64 sockets, and hence parallel::clusterApplyLB() would hang with more than 64 workers. Reported by Arunkumar Srinivasan.
as(1L, "double") now does coerce (PR#17457).
lm.influence(), influence.measures(), rstudent() etc now work (more) correctly for multivariate models ("mlm"), thanks to (anonymous) stackoverflow remarks.
sample.int(2.9, *, replace=TRUE) again behaves as documented and as in R < 3.0.0, namely identically to sample.int(2, ..).
Fixes to convertColor() for chromatic adaptation; thanks to Brodie Gaslam PR#17473.
Using Sexpr[stage=install]{..} to create an ‘Rd’ section no longer gives a warning in R CMD check; problem originally posted by Gábor Csárdi, then reported as PR#17479 with a partial patch by Duncan Murdoch.
Parse data now include a special node for equal assignment.
split.default() no longer relies on [[<-(), so it behaves as expected when splitting an object by a factor with the empty string as one of its levels. Thanks to Brad Friedman for the report.
Line numbers in messages about ‘.Rd’ files are now more reliable, thanks to a patch from Duncan Murdoch.
In the numeric method for all.equal(), a numeric scale argument is now checked to be positive and allowed to be of length > 1. (The latter worked originally and with a warning in recent years).
Deferred string conversions now record the OutDec option setting when not equal to the default. Reported by Michael Sannella.
When y is numeric and f a factor, plot(y ~ f) nicely uses "y" and "f" as y- and x-labels. The more direct boxplot(y ~ f) now does too. The new argument ann = FALSE may be used to suppress these.
Subassignment to no/empty rows of a data frame is more consistent and typically a no-op in all cases instead of sometimes an error; part of Emil Bode's PR#17483.
Calls like formatC(*, zero.print = "< 0.001") no longer give an error and are further improved via new optional argument replace.zero. Reported by David Hugh-Jones.
methods::formalArgs("<fn>") now finds the same function as formals("<fn>"), fixing Emil Bode's PR#17499.
The methods package better handles duplicated class names across packages.
The default method of seq() now avoids integer overflow, thanks to the report and "cumsum" patch of Suharto Anggono's PR#17497.
sub() no longer loses encodings for non-ASCII replacements (PR#17509).
Fix for rotated raster image on X11 device. (Partial fix for PR#17148; thanks to Mikko Korpela).
formula(model.frame(frml, ..)) now returns frml in all cases, thanks to Bill Dunlap. The previous behavior is available as DF2formula(<model.frame>).
ar.ols() also returns scalar var.pred in univariate case (PR#17517).
normalizePath() now treats NA path as non-existent and normalizes it to NA. file.access() treats NA file name as non-existent. file.edit() and connection functions such as file() now treat NA file names as errors.
The internal regularize.values() auxiliary of approx(), splinefun() etc now warns again when there are ties and the caller did not specify ties. Further, it no longer duplicates x and y unnecessarily when x is already sorted (PR#17515).
strtoi("", base) now gives NA on all platforms, following its documentation. Reported by Michael Chirico.
In the definition of an S4 class, prototype elements are checked against the slots of the class, with giving a prototype for an undefined slot now being an error. (Reported by Bill Dunlap.)
From setClassUnion(), if environment variable _R_METHODS_SHOW_CHECKSUBCLASSES is set to true, the internal .checkSubclasses() utility prints debugging info to see where it is used.
max.col(m) with an m of zero columns now returns integer NA (instead of 1).
axTicks() no longer returns small “almost zero” numbers (in exponential format) instead of zero, fixing Ilario Gelmetti's PR#17534.
isSymmetric(matrix(0, dimnames=list("A","b"))) is FALSE again, as always documented.
The cairo_pdf graphics device (and other Cairo-based devices) now clip correctly to the right and bottom border.
There was an off-by-one-pixel bug, reported by Lee Kelvin.
as.roman(3) <= 2:4 and all other comparisons now work, as do group "Summary" function calls such as max(as.roman(sample(20))) and as.roman(NA). (Partly reported by Bill Dunlap in PR#17542.)
reformulate("x", response = "sin(y)") no longer produces extra back quotes, PR#17359, and gains new optional argument env.
When reading console input from ‘stdin’ with re-encoding (R --encoding=enc < input) the code on a Unix-alike now ensures that each converted input line is terminated with a newline even if re-encoding fails.
Class unions are unloaded when their namespace is unloaded (PR#17531, adapted from a patch by Brodie Gaslam).
selectMethod() is robust to ANY-truncation of method signatures (thanks to Herve Pages for the report).
as.matrix.data.frame() now produces better strings from logicals, thanks to PR#17548 from Gabe Becker.
The S4 generic signature of rowSums(), rowMeans(), colSums() and colMeans() is restricted to "x".
match(x, tab) now works for long character vectors x, thanks to PR#17552 by Andreas Kersting.
Changes in print.*(), thanks to Lionel Henry's patches in PR#17398:
Printing lists, pairlists or attributes containing calls with S3 class no longer evaluate those.
Printing S4 objects within lists and pairlists dispatches with show() rather than print(), as with auto-printing.
The indexing tags (names or [[<n>]]) of recursive data structures are now printed correctly in complex cases.
Arguments supplied to print() are now properly forwarded to methods when printing lists, pairlists or attributes containing S3 objects.
The print parameters are now preserved when printing S3 objects or deparsing symbols and calls. Previously, printing lists containing S3 objects or expressions would reset these parameters.
Printing lists, pairlists or attributes containing functions now uses srcref attributes if present.

New in R for Windows 3.5.3 (Mar 12, 2019)

New in R for Windows 3.5.2 (Dec 21, 2018)

PACKAGE INSTALLATION:
New macro CXX_VISIBILITY analogous to C_VISIBILITY (which several packages have been misusing for C++ code) for the default C++ compiler (but not necessarily one used for non-default C++ dialects like C++14).
TESTING:
The random number generator tests in tests/p-r-random-tests.R no longer fail occasionally as they now randomly sample from "certified" random seeds.
BUG FIXES:
The "glm" method of drop1() miscalculated the score test (test="Rao") when the model contained an offset.
Linear multiple empty models such as lm(y ~ 0) now have a correctly dimensioned empty coefficient matrix; reported by Brett Presnell.
vcov(<empty mlm>) and hence confint() now work (via a consistency change in summary.lm()).
confint(<multiple lm()>) now works correctly; reported on R-devel by Steven Pav.
quade.test() now also works correctly when its arguments are not yet sorted along groups, fixing PR#15842.
Installation on a Unix-alike tries harder to link to the pthread library where required (rather than relying on OpenMP to provide it: configuring with --disable-openmp was failing on some Linux systems).
The data.frame method for print(x) is fast now also for large data frames x and got an optional argument max, thanks to suggestions by Juan Telleria.
hist() no longer integer overflows in very rare cases, fixing PR#17450.
untar() ignored a character compressed argument: however many external tar programs ignore the flags which should have been set and automagically choose the compression type, and if appropriate gzip or bzip2 compression would have been chosen from the magic header of the tarball.
zapsmall(x) now works for more "number-like" objects.
The tools-internal function called from R CMD INSTALL now gets a warnOption = 1 argument and only sets options(warn = warnOption) when that increases the warning level (PR#17453).
Analogously, the tools-internal function called from R CMD check gets a warnOption = 1 argument and uses the larger of that and getOption("warn"), also allowing to be run with increased warning level.
Parse data now have deterministic parent nodes (PR#16041).
Calling match() with length one x and POSIXlt table gave a segfault (PR#17459).
Fork clusters could hang due to a race condition in cluster initialization (makeCluster()).
nextn(n) now also works for larger n and no longer loops infinitely for e.g, n <- 214e7.
cooks.distance() and rstandard() now work correctly for multiple linear models ("mlm").
polym() and corresponding lm() prediction now also work for a boundary "vector" case fixing PR#17474, reported by Alexandre Courtiol.
With a very large number of variables terms() could segfault (PR#17480).
cut(rep(0, 7)) now works, thanks to Joey Reid and Benjamin Tyner (PR#16802).
download.file(*, method = "curl", cacheOK = FALSE) should work now on Windows, thanks to Kevin Ushey's patch in PR#17323.
duplicated(<dataframe with 'f'>) now works, too, thanks to Andreas Kersting's PR#17485; ditto for anyDuplicated().
legend(*, cex = 1:2) now works less badly.
The print() method for POSIXct and POSIXlt now correctly obeys getOption("max.print"), fixing a long-standing typo, and it also gets a corresponding optional max argument.
Unserialization of raw vectors serialized in ASCII representation now works correctly.
<data frame>[TRUE, <new>] <- list(c1, c2) now works correctly, thanks to Suharto Anggono's PR#15362 and Emil Bode's patch in PR#17504.
seq.int(*, by=by, length=n) no longer wrongly "drops fractional parts" when by is integer, thanks to Suharto Anggono's report PR#17506.
Buffering is disabled for file() connections to non-regular files (like sockets), as well as fifo() and pipe() connections. Fixes PR#17470, reported by Chris Culnane.

New in R for Windows 3.5.0 (Apr 24, 2018)

SIGNIFICANT USER-VISIBLE CHANGES:
All packages are by default byte-compiled on installation. This makes the installed packages larger (usually marginally so) and may affect the format of messages and tracebacks (which often exclude .Call and similar).
NEW FEATURES:
factor() now uses order() to sort its levels, rather than sort.list(). This allows factor() to support custom vector-like objects if methods for the appropriate generics are defined. It has the side effect of making factor() succeed on empty or length-one non-atomic vector(-like) types (e.g., "list"), where it failed before.
diag() gets an optional names argument: this may require updates to packages defining S4 methods for it.
chooseCRANmirror() and chooseBioCmirror() no longer have a useHTTPS argument, not needed now all R builds support https:// downloads.
New summary() method for warnings() with a (somewhat experimental) print() method.
(methods package.) .self is now automatically registered as a global variable when registering a reference class method.
tempdir(check = TRUE) recreates the tempdir() directory if it is no longer valid (e.g. because some other process has cleaned up the /tmp directory).
New askYesNo() function and "askYesNo" option to ask the user binary response questions in a customizable but consistent way. (Suggestion of PR#17242.)
New low level utilities ...elt(n) and ...length() for working with ... parts inside a function.
isTRUE() is more tolerant and now true in x <- rlnorm(99) isTRUE(median(x) == quantile(x)["50%"])
New function isFALSE() defined analogously to isTRUE().
The default symbol table size has been increased from 4119 to 49157; this may improve the performance of symbol resolution when many packages are loaded. (Suggested by Jim Hester.)
line() gets a new option iter = 1.
Reading from connections in text mode is buffered, significantly improving the performance of readLines(), as well as scan() and read.table(), at least when specifying colClasses.
order() is smarter about picking a default sort method when its arguments are objects.
available.packages() has two new arguments which control if the values from the per-session repository cache are used (default true, as before) and if so how old cached values can be to be used (default one hour).
These arguments can be passed from install.packages(), update.packages() and functions calling that: to enable this available.packages(), packageStatus() and download.file() gain a ... argument.
packageStatus()'s upgrade() method no longer ignores its ... argument but passes it to install.packages().
installed.packages() gains a ... argument to allow arguments (including noCache) to be passed from new.packages(), old.packages(), update.packages() and packageStatus().
factor(x, levels, labels) now allows duplicated labels (not duplicated levels!). Hence you can map different values of x to the same level directly.
Attempting to use names<-() on an S4 derivative of a basic type no longer emits a warning.
The list method of within() gains an option keepAttrs = FALSE for some speed-up.
system() and system2() now allow the specification of a maximum elapsed time ('timeout').
debug() supports debugging of methods on any object of S4 class "genericFunction", including group generics.
Attempting to increase the length of a variable containing NULL using length()<- still has no effect on the target variable, but now triggers a warning.
type.convert() becomes a generic function, with additional methods that operate recursively over list and data.frame objects. Courtesy of Arni Magnusson (PR#17269).
lower.tri(x) and upper.tri(x) only needing dim(x) now work via new functions .row() and .col(), so no longer call as.matrix() by default in order to work efficiently for all kind of matrix-like objects.
print() methods for "xgettext" and "xngettext" now use encodeString() which keeps, e.g. "n", visible. (Wish of PR#17298.)
package.skeleton() gains an optional encoding argument.
approx(), spline(), splinefun() and approxfun() also work for long vectors.
deparse() and dump() are more useful for S4 objects, dput() now using the same internal C code instead of its previous imperfect workaround R code. S4 objects now typically deparse perfectly, i.e., can be recreated identically from deparsed code.
dput(), deparse() and dump() now print the names() information only once, using the more readable (tag = value) syntax, notably for list()s, i.e., including data frames.
These functions gain a new control option "niceNames" (see .deparseOpts()), which when set (as by default) also uses the (tag = value) syntax for atomic vectors. On the other hand, without deparse options "showAttributes" and "niceNames", names are no longer shown also for lists. as.character(list( c (one = 1))) now includes the name, as as.character(list(list(one = 1))) has always done.
m:n now also deparses nicely when m > n.
The "quoteExpressions" option, also part of "all", no longer quote()s formulas as that may not re-parse identically. (PR#17378)
If the option setWidthOnResize is set and TRUE, R run in a terminal using a recent readline library will set the width option when the terminal is resized. Suggested by Ralf Goertz.
If multiple on.exit() expressions are set using add = TRUE then all expressions will now be run even if one signals an error.
mclapply() gets an option affinity.list which allows more efficient execution with heterogeneous processors, thanks to Helena Kotthaus.
The character methods for as.Date() and as.POSIXlt() are more flexible _via_ new arguments tryFormats and optional: see their help pages.
on.exit() gains an optional argument after with default TRUE. Using after = FALSE with add = TRUE adds an exit expression before any existing ones. This way the expressions are run in a first-in last-out fashion. (From Lionel Henry.)
On Windows, file.rename() internally retries the operation in case of error to attempt to recover from possible anti-virus interference.
Command line completion on :: now also includes lazy-loaded data.
If the TZ environment variable is set when date-time functions are first used, it is recorded as the session default and so will be used rather than the default deduced from the OS if TZ is subsequently unset.
There is now a [ method for class "DLLInfoList".
glm() and glm.fit get the same singular.ok = TRUE argument that lm() has had forever. As a consequence, in glm(*, method = <your_own>), user specified methods need to accept a singular.ok argument as well.
aspell() gains a filter for Markdown (.md and .Rmd) files.
intToUtf8(multiple = FALSE) gains an argument to allow surrogate pairs to be interpreted.
The maximum number of DLLs that can be loaded into R e.g. _via_ dyn.load() has been increased up to 614 when the OS limit on the number of open files allows.
Sys.timezone() on a Unix-alike caches the value at first use in a session: _inter alia_ this means that setting TZ later in the session affects only the _current_ time zone and not the _system_ one.
Sys.timezone() is now used to find the system timezone to pass to the code used when R is configured with --with-internal-tzcode.
When tar() is used with an external command which is detected to be GNU tar or libarchive tar (aka bsdtar), a different command-line is generated to circumvent line-length limits in the shell.
system(*, intern = FALSE), system2() (when not capturing output), file.edit() and file.show() now issue a warning when the external command cannot be executed.
The "default" ("lm" etc) methods of vcov() have gained new optional argument complete = TRUE which makes the vcov() methods more consistent with the coef() methods in the case of singular designs. The former (back-compatible) behavior is given by vcov(*, complete = FALSE).
coef() methods (for lm etc) also gain a complete = TRUE optional argument for consistency with vcov(). For "aov", both coef() and vcov() methods remain back-compatibly consistent, using the _other_ default, complete = FALSE.
attach(*, pos = 1) is now an error instead of a warning.
New function getDefaultCluster() in package parallel to get the default cluster set via setDefaultCluster().
str(x) for atomic objects x now treats both cases of is.vector(x) similarly, and hence much less often prints "atomic". This is a slight non-back-compatible change producing typically both more informative and shorter output.
write.dcf() gets optional argument useBytes.
New, partly experimental packageDate() which tries to get a valid "Date" object from a package DESCRIPTION file, thanks to suggestions in PR#17324.
tools::resaveRdaFiles() gains a version argument, for use when packages should remain compatible with earlier versions of R.
ar.yw(x) and hence by default ar(x) now work when x has NAs, mostly thanks to a patch by Pavel Krivitsky in PR#17366. The ar.yw.default()'s AIC computations have become more efficient by using determinant().
New warnErrList() utility (from package nlme, improved).
By default the (arbitrary) signs of the loadings from princomp() are chosen so the first element is non-negative.
If --default-packages is not used, then Rscript now checks the environment variable R_SCRIPT_DEFAULT_PACKAGES. If this is set, then it takes precedence over R_DEFAULT_PACKAGES. If default packages are not specified on the command line or by one of these environment variables, then Rscript now uses the same default packages as R. For now, the previous behavior of not including methods can be restored by setting the environment variable R_SCRIPT_LEGACY to yes.
When a package is found more than once, the warning from find.package(*, verbose=TRUE) lists all library locations.
POSIXt objects can now also be rounded or truncated to month or year.
stopifnot() can be used alternatively via new argument exprs which is nicer and useful when testing several expressions in one call.
The environment variable R_MAX_VSIZE can now be used to specify the maximal vector heap size. On macOS, unless specified by this environment variable, the maximal vector heap size is set to the maximum of 16GB and the available physical memory. This is to avoid having the R process killed when macOS over-commits memory.
sum(x) and sum(x1,x2,..,x<N>) with many or long logical or integer vectors no longer overflows (and returns NA with a warning), but returns double numbers in such cases.
Single components of "POSIXlt" objects can now be extracted and replaced via [ indexing with 2 indices.
S3 method lookup now searches the namespace registry after the top level environment of the calling environment.
Arithmetic sequences created by 1:n, seq_along, and the like now use compact internal representations via the ALTREP framework. Coercing integer and numeric vectors to character also now uses the ALTREP framework to defer the actual conversion until first use.
Finalizers are now run with interrupts suspended.
merge() gains new option no.dups and by default suffixes the second of two duplicated column names, thanks to a proposal by Scott Ritchie (and Gabe Becker).
scale.default(x, center, scale) now also allows center or scale to be "numeric-alike", i.e., such that as.numeric(.) coerces them correctly. This also eliminates a wrong error message in such cases.
par*apply and par*applyLB gain an optional argument chunk.size which allows to specify the granularity of scheduling.
Some as.data.frame() methods, notably the matrix one, are now more careful in not accepting duplicated or NA row names, and by default produce unique non-NA row names. This is based on new function .rowNamesDF(x, make.names = *) <- rNms where the logical argument make.names allows to specify _how_ invalid row names rNms are handled. .rowNamesDF() is a "workaround" compatible default.
R has new serialization format (version 3) which supports custom serialization of ALTREP framework objects. These objects can still be serialized in format 2, but less efficiently. Serialization format 3 also records the current native encoding of unflagged strings and converts them when de-serialized in R running under different native encoding. Format 3 comes with new serialization magic numbers (RDA3, RDB3, RDX3). Format 3 can be selected by version = 3 in save(), serialize() and saveRDS(), but format 2 remains the default for all serialization and saving of the workspace. Serialized data in format 3 cannot be read by versions of R prior to version 3.5.0.
The "Date" and "date-time" classes "POSIXlt" and "POSIXct" now have a working `length<-` method, as wished in PR#17387.
optim(*, control = list(warn.1d.NelderMead = FALSE)) allows to turn off the warning when applying the default "Nelder-Mead" method to 1-dimensional problems.
matplot(.., panel.first = .) etc now work, as log becomes explicit argument and ... is passed to plot() unevaluated, as suggested by Sebastian Meyer in PR#17386.
Interrupts can be suspended while evaluating an expression using suspendInterrupts. Subexpression can be evaluated with interrupts enabled using allowInterrupts. These functions can be used to make sure cleanup handlers cannot be interrupted.
R 3.5.0 includes a framework that allows packages to provide alternate representations of basic R objects (ALTREP). The framework is still experimental and may undergo changes in future R releases as more experience is gained. For now, documentation is provided in <URL: https://svn.r-project.org/R/branches/ALTREP/ALTREP.html>.
UTILITIES:
install.packages() for source packages now has the possibility to set a 'timeout' (elapsed-time limit). For serial installs this uses the timeout argument of system2(): for parallel installs it requires the timeout utility command from GNU coreutils.
It is now possible to set 'timeouts' (elapsed-time limits) for most parts of R CMD check _via_ environment variables documented in the 'R Internals' manual.
The 'BioC extra' repository which was dropped from Bioconductor 3.6 and later has been removed from setRepositories(). This changes the mapping for 6-8 used by setRepositories(ind=).
R CMD check now also applies the settings of environment variables _R_CHECK_SUGGESTS_ONLY_ and _R_CHECK_DEPENDS_ONLY_ to the re-building of vignettes.
R CMD check with environment variable _R_CHECK_DEPENDS_ONLY_ set to a true value makes test-suite-management packages available and (for the time being) works around a common omission of rmarkdown from the VignetteBuilder field.
INSTALLATION on a UNIX-ALIKE:
Support for a system Java on macOS has been removed - install a fairly recent Oracle Java (see 'R Installation and Administration' SSC.3.2).
configure works harder to set additional flags in SAFE_FFLAGS only where necessary, and to use flags which have little or no effect on performance.
In rare circumstances it may be necessary to override the setting of SAFE_FFLAGS.
C99 functions expm1, hypot, log1p and nearbyint are now required.
configure sets a -std flag for the C++ compiler for all supported C++ standards (e.g., -std=gnu++11 for the C++11 compiler). Previously this was not done in a few cases where the default standard passed the tests made (e.g. clang 6.0.0 for C++11).
C-LEVEL FACILITIES:
'Writing R Extensions' documents macros MAYBE_REFERENCED, MAYBE_SHARED and MARK_NOT_MUTABLE that should be used by package C code instead NAMED or SET_NAMED.
The object header layout has been changed to support merging the ALTREP branch. This requires re-installing packages that use compiled code.
'Writing R Extensions' now documents the R_tryCatch, R_tryCatchError, and R_UnwindProtect functions.
NAMEDMAX has been raised to 3 to allow protection of intermediate results from (usually ill-advised) assignments in arguments to BUILTIN functions. Package C code using SET_NAMED may need to be revised.
DEPRECATED AND DEFUNCT:
Sys.timezone(location = FALSE) is defunct, and is ignored (with a warning).
methods:::bind_activation() is defunct now; it typically has been unneeded for years.
The undocumented 'hidden' objects .__H__.cbind and .__H__.rbind in package base are deprecated (in favour of cbind and rbind).
The declaration of pythag() in Rmath.h has been removed - the entry point has not been provided since R 2.14.0.
BUG FIXES:
printCoefmat() now also works without column names.
The S4 methods on Ops() for the "structure" class no longer cause infinite recursion when the structure is not an S4 object.
nlm(f, ..) for the case where f() has a "hessian" attribute now computes LL' = H + uI correctly. (PR#17249).
An S4 method that "rematches" to its generic and overrides the default value of a generic formal argument to NULL no longer drops the argument from its formals.
Rscript can now accept more than one argument given on the #! line of a script. Previously, one could only pass a single argument on the #! line in Linux.
Connections are now written correctly with encoding "UTF-16LE". (PR#16737).
Evaluation of ..0 now signals an error. When ..1 is used and ... is empty, the error message is more appropriate.
(Windows mainly.) Unicode code points which require surrogate pairs in UTF-16 are now handled. All systems should properly handle surrogate pairs, even those systems that do not need to make use of them. (PR#16098)
stopifnot(e, e2, ...) now evaluates the expressions sequentially and in case of an error or warning shows the relevant expression instead of the full stopifnot(..) call.
path.expand() on Windows now accepts paths specified as UTF-8-encoded character strings even if not representable in the current locale. (PR#17120)
line(x, y) now correctly computes the medians of the left and right group's x-values and in all cases reproduces straight lines.
Extending S4 classes with slots corresponding to special attributes like dim and dimnames now works.
Fix for legend() when fill has multiple values the first of which is NA (all colours used to default to par(fg)). (PR#17288)
installed.packages() did not remove the cached value for a library tree that had been emptied (but would not use the old value, just waste time checking it).
The documentation for installed.packages(noCache = TRUE) incorrectly claimed it would refresh the cache.
aggregate(<data.frame>) no longer uses spurious names in some cases. (PR#17283)
object.size() now also works for long vectors.
packageDescription() tries harder to solve re-encoding issues, notably seen in some Windows locales. This fixes the citation() issue in PR#17291.
poly(<matrix>, 3) now works, thanks to prompting by Marc Schwartz.
readLines() no longer segfaults on very large files with embedded '' (aka 'nul') characters. (PR#17311)
ns() (package splines) now also works for a single observation. interpSpline() gives a more friendly error message when the number of points is less than four.
dist(x, method = "canberra") now uses the correct definition; the result may only differ when x contains values of differing signs, e.g. not for 0-1 data.
methods:::cbind() and methods:::rbind() avoid deep recursion, thanks to Suharto Anggono via PR#17300.
Arithmetic with zero-column data frames now works more consistently; issue raised by Bill Dunlap.
Arithmetic with data frames gives a data frame for ^ (which previously gave a numeric matrix).
pretty(x, n) for large n or large diff(range(x)) now works better (though it was never meant for large n); internally it uses the same rounding fuzz (1e-10) as seq.default() - as it did up to 2010-02-03 when both were 1e-7.
Internal C-level R_check_class_and_super() and hence R_check_class_etc() now also consider non-direct super classes and hence return a match in more cases. This e.g., fixes behaviour of derived classes in package Matrix.
Reverted unintended change in behavior of return calls in on.exit expressions introduced by stack unwinding changes in R 3.3.0.
Attributes on symbols are now detected and prevented; attempt to add an attribute to a symbol results in an error.
fisher.test(*, workspace = <n>) now may also increase the internal stack size which allows larger problem to be solved, fixing PR#1662.
The methods package no longer directly copies slots (attributes) into a prototype that is of an "abnormal" (reference) type, like a symbol.
The methods package no longer attempts to call length<-() on NULL (during the bootstrap process).
The methods package correctly shows methods when there are multiple methods with the same signature for the same generic (still not fully supported, but at least the user can see them).
sys.on.exit() is now always evaluated in the right frame. (From Lionel Henry.)
seq.POSIXt(*, by = "<n> DSTdays") now should work correctly in all cases and is faster. (PR#17342)
.C() when returning a logical vector now always maps values other than FALSE and NA to TRUE (as documented).
Subassignment with zero length vectors now coerces as documented (PR#17344). Further, x <- numeric(); x[1] <- character() now signals an error 'replacement has length zero' (or a translation of that) instead of doing nothing.
(Package parallel.) mclapply(), pvec() and mcparallel() (when mccollect() is used to collect results) no longer leave zombie processes behind.
R CMD INSTALL <pkg> now produces the intended error message when, e.g., the LazyData field is invalid.
as.matrix(dd) now works when the data frame dd contains a column which is a data frame or matrix, including a 0-column matrix/d.f. .
mclapply(X, mc.cores) now follows its documentation and calls lapply() in case mc.cores = 1 also in the case mc.preschedule is false. (PR#17373)
aggregate(<data.frame>, drop=FALSE) no longer calls the function on <empty> parts but sets corresponding results to NA. (Thanks to Suharto Anggono's patches in PR#17280).
The duplicated() method for data frames is now based on the list method (instead of string coercion). Consequently unique() is better distinguishing data frame rows, fixing PR#17369 and PR#17381. The methods for matrices and arrays are changed accordingly.
Calling names() on an S4 object derived from "environment" behaves (by default) like calling names() on an ordinary environment.
read.table() with a non-default separator now supports quotes following a non-whitespace character, matching the behavior of scan().
parLapplyLB and parSapplyLB have been fixed to do load balancing (dynamic scheduling). This also means that results of computations depending on random number generators will now really be non-reproducible, as documented.
Indexing a list using dollar and empty string (l$"") returns NULL.
Using usage{ data(<name>, package="<pkg>") } no longer produces R CMD check warnings.
match.arg() more carefully chooses the environment for constructing default choices, fixing PR#17401 as proposed by Duncan Murdoch.
Deparsing of consecutive ! calls is now consistent with deparsing unary - and + calls and creates code that can be reparsed exactly; thanks to a patch by Lionel Henry in PR#17397. (As a side effect, this uses fewer parentheses in some other deparsing involving ! calls.)

New in R for Windows 3.4.4 (Mar 15, 2018)

New in R for Windows 3.4.3 (Dec 4, 2017)

New in R for Windows 3.4.2 (Dec 4, 2017)

NEW FEATURES:
Setting the LC_ALL category in Sys.setlocale() invalidates any cached locale-specific day/month names and the AM/PM indicator for strptime() (as setting LC_TIME has since R 3.1.0). The version of LAPACK included in the sources has been updated to 3.7.1, a bug-fix release. The default for tools::write_PACKAGES(rds_compress=) has been changed to "xz" to match the compression used by CRAN. c() and unlist() are now more efficient in constructing the names(.) of their return value, thanks to a proposal by Suharto Anggono. (PR#17284)
UTILITIES:
R CMD check checks for and R CMD build corrects CRLF line endings in shell scripts configure and cleanup (even on Windows).
INSTALLATION on a UNIX-ALIKE:
The order of selection of OpenMP flags has been changed: Oracle Developer Studio 12.5 accepts -fopenmp and -xopenmp but only the latter enables OpenMP so it is now tried first.
BUG FIXES:
within(List, rm(x1, x2)) works correctly again, including when List[["x2"]] is NULL.
regexec(pattern, text, *) now applies as.character(.) to its first two arguments, as documented.
write.table() and related functions, writeLines(), and perhaps other functions writing text to connections did not signal errors when the writes failed, e.g. due to a disk being full. Errors will now be signalled if detected during the write, warnings if detected when the connection is closed. (PR#17243)
rt() assumed the ncp parameter was a scalar. (PR#17306)
menu(choices) with more than 10 choices which easily fit into one getOption("width")-line no longer erroneously repeats choices. (PR#17312)
length()<- on a pairlist succeeds. (<URL: https://stat.ethz.ch/pipermail/r-devel/2017-July/074680.html>)
Language objects such as quote(("n")) or R functions are correctly printed again, where R 3.4.1 accidentally duplicated the backslashes.
Construction of names() for very large objects in c() and unlist() now works, thanks to Suharto Anggono's patch proposals in PR#17292.
Resource leaks (and similar) reported by Steve Grubb fixed. (PR#17314, PR#17316, PR#17317, PR#17318, PR#17319, PR#17320)
model.matrix(~1, mf) now gets the row names from mf also when they differ from 1:nrow(mf), fixing PR#14992
sigma(fm) now takes the correct denominator degrees of freedom for a fitted model with NA coefficients. (PR#17313)
hist(x, "FD") no longer "dies" with a somewhat cryptic error message when x has extreme outliers or IQR() zero: nclass.FD(x)
tries harder to find a robust bin width h in the latter case, and hist.default(*, breaks) now checks and corrects a too large breaks number. (PR#17274)
callNextMethod() works for ... methods.
qr.coef(qd, y) now has correct names also when qd is a complex QR or stems from qr(*, LAPACK=TRUE).
Setting options(device = *) to an invalid function no longer segfaults when plotting is initiated. (PR#15883)
encodeString(<very large string>) no longer segfaults. (PR#15885)
It is again possible to use configure --enable-maintainer-mode without having installed notangle (it was required in R 3.4.[01]).
S4 method dispatch on ... calls the method by name instead of Method (for consistency with default dispatch), and only attempts to pass non-missing arguments from the generic.
readRDS(textConnection(.)) works again. (PR#17325)
1:n)[-n] no longer segfaults for n <- 2.2e9 (on a platform with enough RAM).
x <- 1:2; tapply(x, list(x, x), function(x) "")[1,2] now correctly returns NA. (PR#17333)
Running of finalizers after explicit GC request moved from the R interface do_gc to the C interface R_gc. This helps with reclaiming inaccessible connections.
help.search(topic) and ??topic matching topics in vignettes with multiple file name extensions (e.g., *.md.rsp but not *.Rmd)
failed with an error when using options(help_type = "html").
The X11 device no longer uses the Xlib backing store (PR#16497).
array(character(), 1) now gives (a 1D array with) NA as has been documented for a long time as in the other cases of zero-length array initialization and also compatibly with matrix(character(),
As mentioned there, this also fixes PR#17333.
splineDesign(.., derivs = 4) no longer segfaults.
fisher.test(*, hybrid=TRUE) now (again) will use the hybrid
method when Cochran's conditions are met, fixing PR#16654.

New in R for Windows 3.4.1 (Jul 10, 2017)

New in R for Windows 3.4.0 (May 24, 2017)

SIGNIFICANT USER-VISIBLE CHANGES:
(Unix-alike) The default methods for download.file() and url() now choose "libcurl" except for file:// URLs. There will be small changes in the format and wording of messages, including in rare cases if an issue is a warning or an error. For example, when HTTP re-direction occurs, some messages refer to the final URL rather than the specified one.
Those who use proxies should check that their settings are compatible (see ?download.file: the most commonly used forms work for both "internal" and "libcurl").
table() has been amended to be more internally consistent and become back compatible to R <= 2.7.2 again. Consequently, table(1:2, exclude = NULL) no longer contains a zero count for <NA>, but useNA = "always" continues to do so.
summary.default() no longer rounds, but its print method does resulting in less extraneous rounding, notably of numbers in the ten thousands.
factor(x, exclude = L) behaves more rationally when x or L are character vectors. Further, exclude = <factor> now behaves as documented for long.
Arithmetic, logic (&, |) and comparison (aka 'relational', e.g., <, ==) operations with arrays now behave consistently, notably for arrays of length zero.
Arithmetic between length-1 arrays and longer non-arrays had silently dropped the array attributes and recycled. This now gives a warning and will signal an error in the future, as it has always for logic and comparison operations in these cases (e.g., compare matrix(1,1) + 2:3 and matrix(1,1) < 2:3).
The JIT ('Just In Time') byte-code compiler is now enabled by default at its level 3. This means functions will be compiled on first or second use and top-level loops will be compiled and then run. (Thanks to Tomas Kalibera for extensive work to make this possible.)
For now, the compiler will not compile code containing explicit calls to browser(): this is to support single stepping from the browser() call.
JIT compilation can be disabled for the rest of the session using compiler::enableJIT(0) or by setting environment variable R_ENABLE_JIT to 0.
xtabs() works more consistently with NAs, also in its result no longer setting them to 0. Further, a new logical option addNA allows to count NAs where appropriate. Additionally, for the case sparse = TRUE, the result's dimnames are identical to the default case's.
Matrix products now consistently bypass BLAS when the inputs have NaN/Inf values. Performance of the check of inputs has been improved. Performance when BLAS is used is improved for matrix/vector and vector/matrix multiplication (DGEMV is now used instead of DGEMM).
One can now choose from alternative matrix product implementations _via_ options(matprod = ). The "internal" implementation is not optimized for speed but consistent in precision with other summations in R (using long double accumulators where available). "blas" calls BLAS directly for best speed, but usually with undefined behavior for inputs with NaN/Inf.
factor() now uses order() to sort its levels, not sort.list(). This makes factor() support custom vector-like objects if methods for the appropriate generics are defined. This change has the side effect of making factor() succeed on empty or length-one non-atomic vector(-like) types (e.g., list), where it failed before.
NEW FEATURES:
User errors such as integrate(f, 0:1, 2) are now caught.
Add signature argument to debug(), debugonce(), undebug() and isdebugged() for more conveniently debugging S3 and S4 methods. (Based on a patch by Gabe Becker.)
Add utils::debugcall() and utils::undebugcall() for debugging the function that would be called by evaluating the given expression. When the call is to an S4 generic or standard S3 generic, debugcall() debugs the method that would be dispatched. A number of internal utilities were added to support this, most notably utils::isS3stdGeneric(). (Based on a patch by Gabe Becker.)
Add utils::strcapture(). Given a character vector and a regular expression containing capture expressions, strcapture() will extract the captured tokens into a tabular data structure, typically a data.frame.
str() and strOptions() get a new option drop.deparse.attr with improved but _changed_ default behaviour for expressions. For expression objects x, str(x) now may remove extraneous white space and truncate long lines.
str(<looooooooong_string>) is no longer very slow; inspired by Mikko Korpela's proposal in PR#16527.
str(x)'s default method is more "accurate" and hence somewhat more generous in displaying character vectors; this will occasionally change R outputs (and need changes to some *.Rout(.save) files).
For a classed integer vector such as x <- xtabs(~ c(1,9,9,9)), str(x) now shows both the class and "int", instead of only the latter.
isSymmetric(m) is much faster for large asymmetric matrices m _via_ pre-tests and a new option tol1 (with which strict back compatibility is possible but not the default).
The result of eigen() now is of class "eigen" in the default case when eigenvectors are computed.
Zero-length date and date-time objects (of classes "POSIX[cl]?t") now print() "recognizably".
xy.coords() and xyz.coords() get a new setLab option.
The method argument of sort.list(), order() and sort.int() gains an "auto" option (the default) which should behave the same as before when method was not supplied.
stopifnot(E, ..) now reports differences when E is a call to all.equal() and that is not true.
boxplot(<formula>, *) gain optional arguments drop, sep, and lex.order to pass to split.default() which itself gains an argument lex.order to pass to interaction() for more flexibility.
The plot() method for ppr() has enhanced default labels (xmin and main).
sample.int() gains an explicit useHash option (with a back compatible default).
identical() gains an ignore.srcref option which drops "srcref" and similar attributes when true (as by default).
diag(x, nrow = n) now preserves typeof(x), also for logical, integer and raw x (and as previously for complex and numeric).
smooth.spline() now allows direct specification of lambda, gets a hatvalues() method and keeps tol in the result, and optionally parts of the internal matrix computations.
addNA() is faster now, e.g. when applied twice. (Part of PR#16895.)
New option rstandard(<lm>, type = "predicted") provides the "PRESS"-related leave-one-out cross-validation errors for linear models.
After seven years of deprecation, duplicated factor levels now produce a warning when printed and an error in levels<- instead of a warning.
Invalid factors, e.g., with duplicated levels (invalid but constructable) now give a warning when printed, _via_ new function .valid.factor().
sessionInfo() has been updated for Apple's change in OS naming as from '10.12' ('macOS Sierra' _vs_ 'OS X El Capitan').
Its toLatex() method now includes the running component.
options(interrupt=) can be used to specify a default action for user interrupts. For now, if this option is not set and the error option is set, then an unhandled user interrupt invokes the error option. (This may be dropped in the future as interrupt conditions are not error conditions.)
In most cases user interrupt handlers will be called with a "resume" restart available. Handlers can invoke this restart to resume computation. At the browser prompt the r command will invoke a "resume" restart if one is available. Some read operations cannot be resumed properly when interrupted and do not provide a "resume" restart.
Radix sort is now chosen by method = "auto" for sort.int() for double vectors (and hence used for sort() for unclassed double vectors), excluding 'long' vectors.
sort.int(method = "radix") no longer rounds double vectors.
The default and data.frame methods for stack() preserve the names of empty elements in the levels of the ind column of the return value. Set the new drop argument to TRUE for the previous behavior.
Speedup in simplify2array() and hence sapply() and mapply() (for the case of names and common length > 1), thanks to Suharto Anggono's PR#17118.
table(x, exclude = NULL) now sets useNA = "ifany" (instead of "always"). Together with the bug fixes for this case, this recovers more consistent behaviour compatible to older versions of R. As a consequence, summary() for a logical vector no longer reports (zero) counts for NA when there are no NAs.
dump.frames() gets a new option include.GlobalEnv which allows to also dump the global environment, thanks to Andreas Kersting's proposal in PR#17116.
system.time() now uses message() instead of cat() when terminated early, such that suppressMessages() has an effect; suggested by Ben Bolker.
citation() supports inst/CITATION files from package source trees, with lib.loc pointing to the directory containing the package.
try() gains a new argument outFile with a default that can be modified _via_ options(try.outFile = .), useful notably for Sweave.
The unexported low-level functions in package parallel for passing serialized R objects to and from forked children now support long vectors on 64-bit platforms. This removes some limits on higher-level functions such as mclapply() (but returning gigabyte results from forked processes _via_ serialization should be avoided if at all possible).
Connections now print() without error even if invalid, e.g. after having been destroyed.
apropos() and find(simple.words = FALSE) no longer match object names starting with . which are known to be internal objects (such as .__S3MethodsTable__.).
Convenience function hasName() has been added; it is intended to replace the common idiom !is.null(x$name) without the usually unintended partial name matching.
strcapture() no longer fixes column names nor coerces strings to factors (suggested by Bill Dunlap).
strcapture() returns NA for non-matching values in x (suggested by Bill Dunlap).
source() gets new optional arguments, notably exprs; this is made use of in the new utility function withAutoprint().
sys.source() gets a new toplevel.env argument. This argument is useful for frameworks running package tests; contributed by Tomas Kalibera.
Sys.setFileTime() and file.copy(copy.date = TRUE) will set timestamps with fractions of seconds on platforms/filesystems which support this.
(Windows only.) file.info() now returns file timestamps including fractions of seconds; it has done so on other platforms since R 2.14.0. (NB: some filesystems do not record modification and access timestamps to sub-second resolution.)
The license check enabled by options(checkPackageLicense = TRUE) is now done when the package's namespace is first loaded.
ppr() and supsmu() get an optional trace argument, and ppr(.., sm.method = ..spline) is no longer limited to sample size n <= 2500.
The POSIXct method for print() gets optional tz and usetz arguments, thanks to a report from Jennifer S. Lyon.
New function check_packages_in_dir_details() in package tools for analyzing package-check log files to obtain check details.
Package tools now exports function CRAN_package_db() for obtaining information about current packages in the CRAN package repository, and several functions for obtaining the check status of these packages.
The (default) Stangle driver Rtangle allows annotate to be a function and gets a new drop.evalFALSE option.
The default method for quantile(x, prob) should now be monotone in prob, even in border cases, see PR#16672.
bug.report() now tries to extract an email address from a BugReports field, and if there is none, from a Contacts field.
The format() and print() methods for object.size() results get new options standard and digits; notably, standard = "IEC" and standard = "SI" allow more standard (but less common) abbreviations than the default ones, e.g. for kilobytes. (From contributions by Henrik Bengtsson.)
If a reference class has a validity method, validObject will be called automatically from the default initialization method for reference classes.
tapply() gets new option default = NA allowing to change the previously hardcoded value.
read.dcf() now consistently interprets any 'whitespace' to be stripped to include newlines.
The maximum number of DLLs that can be loaded into R e.g. _via_ dyn.load() can now be increased by setting the environment variable R_MAX_NUM_DLLS before starting R.
Assigning to an element of a vector beyond the current length now over-allocates by a small fraction. The new vector is marked internally as growable, and the true length of the new vector is stored in the truelength field. This makes building up a vector result by assigning to the next element beyond the current length more efficient, though pre-allocating is still preferred. The implementation is subject to change and not intended to be used in packages at this time.
Loading the parallel package namespace no longer sets or changes the .Random.seed, even if R_PARALLEL_PORT is unset.
NB: This can break reproducibility of output, and did for a CRAN package.
Methods "wget" and "curl" for download.file() now give an R error rather than a non-zero return value when the external command has a non-zero status.
Encoding name "utf8" is mapped to "UTF-8". Many implementations of iconv accept "utf8", but not GNU libiconv (including the late 2016 version 1.15).
sessionInfo() shows the full paths to the library or executable files providing the BLAS/LAPACK implementations currently in use (not available on Windows).
The binning algorithm used by bandwidth selectors bw.ucv(), bw.bcv() and bw.SJ() switches to a version linear in the input size n for n > nb/2. (The calculations are the same, but for larger n/nb it is worth doing the binning in advance.)
There is a new option PCRE_study which controls when grep(perl = TRUE) and friends 'study' the compiled pattern. Previously this was done for 11 or more input strings: it now defaults to 10 or more (but most examples need many more for the difference from studying to be noticeable).
grep(perl = TRUE) and friends can now make use of PCRE's Just-In-Time mechanism, for PCRE >= 8.20 on platforms where JIT is supported. It is used by default whenever the pattern is studied (see the previous item). (Based on a patch from Mikko Korpela.)
This is controlled by a new option PCRE_use_JIT.
Note that in general this makes little difference to the speed, and may take a little longer: its benefits are most evident on strings of thousands of characters. As a side effect it reduces the chances of C stack overflow in the PCRE library on very long strings (millions of characters, but see next item).
Warning: segfaults were seen using PCRE with JIT enabled on 64-bit Sparc builds.
There is a new option PCRE_limit_recursion for grep(perl = TRUE) and friends to set a recursion limit taking into account R's estimate of the remaining C stack space (or 10000 if that is not available). This reduces the chance of C stack overflow, but because it is conservative may report a non-match (with a warning) in examples that matched before. By default it is enabled if any input string has 1000 or more bytes. (PR#16757)
getGraphicsEvent() now works on X11(type = "cairo") devices. Thanks to Frederick Eaton (for reviving an earlier patch).
There is a new argument onIdle for getGraphicsEvent(), which allows an R function to be run whenever there are no pending graphics events. This is currently only supported on X11 devices. Thanks to Frederick Eaton.
The deriv() and similar functions now can compute derivatives of log1p(), sinpi() and similar one-argument functions, thanks to a contribution by Jerry Lewis.
median() gains a formal ... argument, so methods with extra arguments can be provided.
strwrap() reduces indent if it is more than half width rather than giving an error. (Suggested by Bill Dunlap.)
When the condition code in if(.) or while(.) is not of length one, an error instead of a warning may be triggered by setting an environment variable, see the help page.
Formatting and printing of bibliography entries (bibentry) is more flexible and better documented. Apart from setting options(citation.bibtex.max = 99) you can also use print(<citation>, bibtex=TRUE) (or format(..)) to get the BibTeX entries in the case of more than one entry. This also affects citation(). Contributions to enable style = "html+bibtex" are welcome.
C-LEVEL FACILITIES:
Entry points R_MakeExternalPtrFn and R_ExternalPtrFn are now declared in header Rinternals.h to facilitate creating and retrieving an R external pointer from a C function pointer without ISO C warnings about the conversion of function pointers.
There was an exception for the native Solaris C++ compiler to the dropping (in R 3.3.0) of legacy C++ headers from headers such as R.h and Rmath.h - this has now been removed. That compiler has strict C++98 compliance hence does not include extensions in its (non-legacy) C++ headers: some packages will need to request C++11 or replace non-C++98 calls such as lgamma: see SS1.6.4 of 'Writing R Extensions'.
Because it is needed by about 70 CRAN packages, headers R.h and Rmath.h still declare use namespace std;
when included on Solaris.
When included from C++, the R headers now use forms such as std::FILE directly rather than including the line using std::FILE;
C++ code including these headers might be relying on the latter.
Headers R_ext/BLAS.h and R_ext/Lapack.h have many improved declarations including const for double-precision complex routines. _Inter alia_ this avoids warnings when passing 'string literal' arguments from C++11 code.
Headers for Unix-only facilities R_ext/GetX11Image.h, R_ext/QuartzDevice.h and R_ext/eventloop.h are no longer installed on Windows.
No-longer-installed headers GraphicsBase.h, RGraphics.h, Rmodules/RX11.h and Rmodules/Rlapack.h which had a LGPL license no longer do so.
HAVE_UINTPTR_T is now defined where appropriate by Rconfig.h so that it can be included before Rinterface.h when CSTACK_DEFNS is defined and a C compiler (not C++) is in use. Rinterface.h now includes C header stdint.h or C++11 header cstdint where needed.
Package tools has a new function package_native_routine_registration_skeleton() to assist adding native-symbol registration to a package. See its help and SS5.4.1 of 'Writing R Extensions' for how to use it. (At the time it was added it successfully automated adding registration to over 90% of CRAN packages which lacked it. Many of the failures were newly-detected bugs in the packages, e.g. 50 packages called entry points with varying numbers of arguments and 65 packages called entry points not in the package.)
INSTALLATION on a UNIX-ALIKE:
readline headers (and not just the library) are required unless configuring with --with-readline=no.
configure now adds a compiler switch for C++11 code, even if the compiler supports C++11 by default. (This ensures that g++ 6.x uses C++11 mode and not its default mode of C++14 with 'GNU extensions'.)
The tests for C++11 compliance are now much more comprehensive. For gcc < 4.8, the tests from R 3.3.0 are used in order to maintain the same behaviour on Linux distributions with long-term support.
An alternative compiler for C++11 is now specified with CXX11, not CXX1X. Likewise C++11 flags are specified with CXX11FLAGS and the standard (e.g., -std=gnu++11 is specified with CXX11STD.
configure now tests for a C++14-compliant compiler by testing some basic features. This by default tries flags for the compiler specified by CXX11, but an alternative compiler, options and standard can be specified by variables CXX14, CXX14FLAGS and CXX14STD (e.g., -std=gnu++14).
There is a new macro CXXSTD to help specify the standard for C++ code, e.g. -std=c++98. This makes it easier to work with compilers which default to a later standard: for example, with CXX=g++6 CXXSTD=-std=c++98 configure will select commands for g++ 6.x which conform to C++11 and C++14 where specified but otherwise use C++98.
Support for the defunct IRIX and OSF/1 OSes and Alpha CPU has been removed.
configure checks that the compiler specified by $CXX $CXXFLAGS is able to compile C++ code.
configure checks for the required header sys/select.h (or sys/time.h on legacy systems) and system call select and aborts if they are not found.
If available, the POSIX 2008 system call utimensat will be used by Sys.setFileTime() and file.copy(copy.date = TRUE). This may result in slightly more accurate file times. (It is available on Linux and FreeBSD but not macOS.)
The minimum version requirement for libcurl has been reduced to 7.22.0, although at least 7.28.0 is preferred and earlier versions are little tested. (This is to support Debian 7 'Wheezy' LTS and Ubuntu 'Precise' 12.04 LTS, although the latter is close to end-of-life.)
configure tests for a C++17-compliant compiler. The tests are experimental and subject to change in the future.
INCLUDED SOFTWARE:
(Windows only) Tcl/Tk version 8.6.4 is now included in the binary builds. The tcltk*.chm help file is no longer included; please consult the online help at <URL: http://www.tcl.tk/man/> instead.
The version of LAPACK included in the sources has been updated to 3.7.0: no new routines have been added to R.
PACKAGE INSTALLATION:
There is support for compiling C++14 or C++17 code in packages on suitable platforms: see 'Writing R Extensions' for how to request this.
The order of flags when LinkingTo other packages has been changed so their include directories come earlier, before those specified in CPPFLAGS. This will only have an effect if non-system include directories are included with -I flags in CPPFLAGS (and so not the default -I/usr/local/include which is treated as a system include directory on most platforms).
Packages which register native routines for .C or .Fortran need to be re-installed for this version (unless installed with R-devel SVN revision r72375 or later).
Make variables with names containing CXX1X are deprecated in favour of those using CXX11, but for the time being are still made available _via_ file etc/Makeconf. Packages using them should be converted to the new forms and made dependent on R (>= 3.4.0).
UTILITIES:
Running R CMD check --as-cran with _R_CHECK_CRAN_INCOMING_REMOTE_ false now skips tests that require remote access. The remaining (local) tests typically run quickly compared to the remote tests.
R CMD build will now give priority to vignettes produced from files in the vignettes directory over those in the inst/doc directory, with a warning that the latter are being ignored.
R CMD config gains a --all option for printing names and values of all basic configure variables.
It now knows about all the variables used for the C++98, C++11 and C++14 standards.
R CMD check now checks that output files in inst/doc are newer than the source files in vignettes.
For consistency with other package subdirectories, files named *.r in the tests directory are now recognized as tests by R CMD check. (Wish of PR#17143.)
R CMD build and R CMD check now use the _union_ of R_LIBS and .libPaths(). They may not be equivalent, e.g., when the latter is determined by R_PROFILE.
R CMD build now preserves dates when it copies files in preparing the tarball. (Previously on Windows it changed the dates on all files; on Unix, it changed some dates when installing vignettes.)
The new option R CMD check --no-stop-on-test-error allows running the remaining tests (under tests/) even if one gave an error.
Check customization _via_ environment variables to detect side effects of .Call() and .External() calls which alter their arguments is described in SS8 of the 'R Internals' manual.
R CMD check now checks any BugReports field to be non-empty and a suitable single URL.
R CMD check --as-cran now NOTEs if the package does not register its native routines or does not declare its intentions on (native) symbol search. (This will become a WARNING in due course.)
DEPRECATED AND DEFUNCT:
(Windows only) Function setInternet2() is defunct.
Installation support for readline emulations based on editline (aka libedit) is deprecated.
Use of the C/C++ macro NO_C_HEADERS is defunct and silently ignored.
unix.time(), a traditional synonym for system.time(), has been deprecated.
structure(NULL, ..) is now deprecated as you cannot set attributes on NULL.
Header Rconfig.h no longer defines SUPPORT_OPENMP; instead use _OPENMP (as documented for a long time).
(C-level Native routine registration.) The deprecated styles member of the R_CMethodDef and R_FortranMethodDef structures has been removed. Packages using these will need to be re-installed for R 3.4.0.
The deprecated support for PCRE versions older than 8.20 will be removed in R 3.4.1. (Versions 8.20-8.31 will still be accepted but remain deprecated.)
BUG FIXES:
Getting or setting body() or formals() on non-functions for now signals a warning and may become an error for setting.
match(x, t), duplicated(x) and unique(x) work as documented for complex numbers with NAs or NaNs, where all those containing NA do match, whereas in the case of NaN's both real and imaginary parts must match, compatibly with how print() and format() work for complex numbers.
deparse(<complex>, options = "digits17") prints more nicely now, mostly thanks to a suggestion by Richie Cotton.
Rotated symbols in plotmath expressions are now positioned correctly on x11(type = "Xlib"). (PR#16948)
as<-() avoids an infinite loop when a virtual class is interposed between a subclass and an actual superclass.
Fix level propagation in unlist() when the list contains zero-length lists or factors.
Fix S3 dispatch on S4 objects when the methods package is not attached.
Internal S4 dispatch sets .Generic in the method frame for consistency with standardGeneric(). (PR#16929)
Fix order(x, decreasing = TRUE) when x is an integer vector containing MAX_INT. Ported from a fix Matt Dowle made to data.table.
Fix caching by callNextMethod(), resolves PR#16973 and PR#16974.
grouping() puts NAs last, to be consistent with the default behavior of order().
Point mass limit cases: qpois(-2, 0) now gives NaN with a warning and qgeom(1, 1) is 0. (PR#16972)
table() no longer drops an "NaN" factor level, and better obeys exclude = <chr>, thanks to Suharto Anggono's patch for PR#16936. Also, in the case of exclude = NULL and NAs, these are tabulated correctly (again).
Further, table(1:2, exclude = 1, useNA = "ifany") no longer erroneously reports <NA> counts.
Additionally, all cases of empty exclude are equivalent, and useNA is not overwritten when specified (as it was by exclude = NULL).
wilcox.test(x, conf.int=TRUE) no longer errors out in cases where the confidence interval is not available, such as for x = 0:2.
droplevels(f) now keeps <NA> levels when present.
In integer arithmetic, NULL is now treated as integer(0) whereas it was previously treated as double(0).
The radix sort considers NA_real_ and NaN to be equivalent in rank (like the other sort algorithms).
When index.return=TRUE is passed to sort.int(), the radix sort treats NAs like sort.list() does (like the other sort algorithms).
When in tabulate(bin, nbin) length(bin) is larger than the maximal integer, the result is now of type double and hence no longer silently overflows to wrong values. (PR#17140)
as.character.factor() respects S4 inheritance when checking the type of its argument. (PR#17141)
The factor method for print() no longer sets the class of the factor to NULL, which would violate a basic constraint of an S4 object.
formatC(x, flag = f) allows two new flags, and signals an error for invalid flags also in the case of character formatting.
Reading from file("stdin") now also closes the connection and hence no longer leaks memory when reading from a full pipe, thanks to G'abor Cs'ardi, see thread starting at <URL: https://stat.ethz.ch/pipermail/r-devel/2016-November/073360.html>.
Failure to create file in tempdir() for compressed pdf() graphics device no longer errors (then later segfaults). There is now a warning instead of error and compression is turned off for the device. Thanks to Alec Wysoker (PR#17191).
Asking for methods() on "|" returns only S3 methods. See <URL: https://stat.ethz.ch/pipermail/r-devel/2016-December/073476.html>.
dev.capture() using Quartz Cocoa device (macOS) returned invalid components if the back-end chose to use ARGB instead of RGBA image format. (Reported by Noam Ross.)
seq("2", "5") now works too, equivalently to "2":"5" and seq.int().
seq.int(to = 1, by = 1) is now correct, other cases are integer (instead of double) when seq() is integer too, and the "non-finite" error messages are consistent between seq.default() and seq.int(), no longer mentioning NaN etc.
rep(x, times) and rep.int(x, times) now work when times is larger than the largest value representable in an integer vector. (PR#16932)
download.file(method = "libcurl") does not check for URL existence before attempting downloads; this is more robust to servers that do not support HEAD or range-based retrieval, but may create empty or incomplete files for aborted download requests.
Bandwidth selectors bw.ucv(), bw.bcv() and bw.SJ() now avoid integer overflow for large sample sizes.
str() no longer shows "list output truncated", in cases that list was not shown at all. Thanks to Neal Fultz (PR#17219)
Fix for cairo_pdf() (and svg() and cairo_ps()) when replaying a saved display list that contains a mix of grid and graphics output. (Report by Yihui Xie.)
The str() and as.hclust() methods for "dendrogram" now also work for deeply nested dendrograms thanks to non-recursive implementations by Bradley Broom.
sample() now uses two uniforms for added precision when the uniform generator is Knuth-TAOCP, Knuth-TAOCP-2002, or a user-defined generator and the population size is 2^25 or greater.
If a vignette in the vignettes directory is listed in .Rbuildignore, R CMD build would not include it in the tarball, but would include it in the vignette database, leading to a check warning. (PR#17246)
tools::latexToUtf8() infinite looped on certain inputs. (PR#17138)
terms.formula() ignored argument names when determining whether two terms were identical. (PR#17235)
callNextMethod() was broken when called from a method that augments the formal arguments of a primitive generic.
Coercion of an S4 object to a vector during sub-assignment into a vector failed to dispatch through the as.vector() generic (often leading to a segfault).
Fix problems in command completion: Crash (PR#17222) and junk display in Windows, handling special characters in filenames on all systems.

New in R for Windows 3.3.2 (Feb 6, 2017)

NEW FEATURES:
extSoftVersion() now reports the version (if any) of the readline library in use.
The version of LAPACK included in the sources has been updated to 3.6.1, a bug-fix release including a speedup for the non-symmetric case of eigen().
Use options(deparse.max.lines=) to limit the number of lines recorded in .Traceback and other deparsing activities.
format(<AsIs>) looks more regular, also for non-character atomic matrices.
abbreviate() gains an option named = TRUE.
The online documentation for package methods is extensively rewritten. The goals are to simplify documentation for basic use, to note old features not recommended and to correct out-of-date information.
Calls to setMethod() no longer print a message when creating a generic function in those cases where that is natural: S3 generics and primitives.
INSTALLATION and INCLUDED SOFTWARE:
Versions of the readline library >= 6.3 had been changed so that terminal window resizes were not signalled to readline: code has been added using a explicit signal handler to work around that (when R is compiled against readline >= 6.3). (PR#16604)
configure works better with Oracle Developer Studio 12.5.
UTILITIES:
R CMD check reports more dubious flags in files src/Makevars[.in], including -w and -g.
R CMD check has been set up to filter important warnings from recent versions of gfortran with -Wall -pedantic: this now reports non-portable GNU extensions such as out-of-order declarations.
R CMD config works better with paths containing spaces, even those of home directories (as reported by Ken Beath).
DEPRECATED AND DEFUNCT:
Use of the C/C++ macro NO_C_HEADERS is deprecated (no C headers are included by R headers from C++ as from R 3.3.0, so it should no longer be needed).
BUG FIXES:
The check for non-portable flags in R CMD check could be stymied by src/Makevars files which contained targets.
(Windows only) When using certain desktop themes in Windows 7 or higher, Alt-Tab could cause Rterm to stop accepting input. (PR#14406; patch submitted by Jan Gleixner.)
pretty(d, ..) behaves better for date-time d (PR#16923).
When an S4 class name matches multiple classes in the S4 cache, perform a dynamic search in order to obey namespace imports. This should eliminate annoying messages about multiple hits in the class cache. Also, pass along the package from the ClassExtends object when looking up superclasses in the cache.
sample(NA_real_) now works.
Packages using non-ASCII encodings in their code did not install data properly on systems using different encodings.
merge(df1, df2) now also works for data frames with column names "na.last", "decreasing", or "method". (PR#17119)
contour() caused a segfault if the labels argument had length zero. (Reported by Bill Dunlap.)
unique(warnings()) works more correctly, thanks to a new duplicated.warnings() method.
findInterval(x, vec = numeric(), all.inside = TRUE) now returns 0s as documented. (Reported by Bill Dunlap.)
(Windows only) R CMD SHLIB failed when a symbol in the resulting library had the same name as a keyword in the .def file. (PR#17130)
pmax() and pmin() now work with (more ?) classed objects, such as "Matrix" from the Matrix package, as documented for a long time.
axis(side, x = D) and hence Axis() and plot() now work correctly for "Date" and time objects D, even when "time goes backward", e.g., with decreasing xlim. (Reported by William May.)
str(I(matrix(..))) now looks as always intended.
plot.ts(), the plot() method for time series, now respects cex, lwd and lty. (Reported by Greg Werbin.)
parallel::mccollect() now returns a named list (as documented) when called with wait = FALSE. (Reported by Michel Lang.)
If a package added a class to a class union in another package, loading the first package gave erroneous warnings about "undefined subclass".
c()'s argument use.names is documented now, as belonging to the (C internal) default method. In "parallel", argument recursive is also moved from the generic to the default method, such that the formal argument list of base generic c() is just (...).
rbeta(4, NA) and similarly rgamma() and rnbinom() now return NaN's with a warning, as other r<dist>(), and as documented. (PR#17155)
Using options(checkPackageLicense = TRUE) no longer requires acceptance of the licence for non-default standard packages such as compiler. (Reported by Mikko Korpela.)
split(<very_long>, *) now works even when the split off parts are long. (PR#17139)
min() and max() now also work correctly when the argument list starts with character(0). (PR#17160)
Subsetting very large matrices (prod(dim(.)) >= 2^31) now works thanks to Michael Schubmehl's PR#17158.
bartlett.test() used residual sums of squares instead of variances, when the argument was a list of lm objects. (Reported by Jens Ledet Jensen).
plot(<lm>, which = *) now correctly labels the contour lines for the standardized residuals for which = 6. It also takes the correct p in case of singularities (also for which = 5). (PR#17161)
xtabs(~ exclude) no longer fails from wrong scope, thanks to Suharto Anggono's PR#17147.
Reference class calls to methods() did not re-analyse previously defined methods, meaning that calls to methods defined later would fail. (Reported by Charles Tilford).
findInterval(x, vec, left.open = TRUE) misbehaved in some cases. (Reported by Dmitriy Chernykh.)

New in R for Windows 3.3.1 (Jun 25, 2016)

New in R for Windows 3.3.0 (May 9, 2016)

SIGNIFICANT USER-VISIBLE CHANGES:
nchar(x, *)'s argument keepNA governing how the result for NAs in x is determined, gets a new default keepNA = NA which returns NA where x is NA, except for type = "width" which still returns 2 the formatting / printing width of NA
All builds have support for https: URLs in the default methods for download.file(), url() and code making use of them
Unfortunately that cannot guarantee that any particular https URL can be accessed. For example, server and client have to successfully negotiate a cryptographic protocol (TLS/SSL, ...) and the server's identity has to be verifiable _via_ the available certificates. Different access methods may allow different protocols or use private certificate bundles: we encountered a https: CRAN mirror which could be accessed by one browser but not by another nor by download.file() on the same Linux machine
NEW FEATURES:
The print method for methods() gains a byclass argument.
New functions validEnc() and validUTF8() to give access to the validity checks for inputs used by grep() and friends.
Experimental new functionality for S3 method checking, notably isS3method().
Also, the names of the R 'language elements' are exported as character vector tools::langElts.
str(x) now displays "Time-Series" also for matrix (multivariate) time-series, i.e. when is.ts(x) is true.
Windows only) The GUI menu item to install local packages now accepts *.tar.gz files as well as *.zip files (but defaults to the latter).
New programmeR's utility function chkDots().
D() now signals an error when given invalid input, rather than silently returning NA. (Request of John Nash.)
formula objects are slightly more "first class": e.g., formula() or new("formula", y ~ x) are now valid. Similarly, for "table", "ordered" and "summary.table". Packages defining S4 classes with the above S3/S4 classes as slots should be reinstalled.
New function strrep() for repeating the elements of a character vector.
rapply() preserves attributes on the list when how = "replace".
New S3 generic function sigma() with methods for extracting the estimated standard deviation aka "residual standard deviation" from a fitted model.
news() now displays R and package news files within the HTML help system if it is available. If no news file is found, a visible NULL is returned to the console.
as.raster(x) now also accepts raw arrays x assuming values in 0:255.
Subscripting of matrix/array objects of type "expression" is now supported.
type.convert("i") now returns a factor instead of a complex value with zero real part and missing imaginary part.
Graphics devices cairo_pdf() and cairo_ps() now allow non-default values of the cairographics 'fallback resolution' to be set.
This now defaults to 300 on all platforms: that is the default documented by cairographics, but apparently was not used by all system installations.
file() gains an explicit method argument rather than implicitly using getOption("url.method", "default").
Thanks to a patch from Tomas Kalibera, x[x != 0] is now typically faster than x[which(x != 0)] (in the case where x has no NAs, the two are equivalent).
read.table() now always uses the names for a named colClasses argument (previously names were only used when colClasses was too short). (In part, wish of PR#16478.)
Windows only) download.file() with default method = "auto" and a ftps:// URL chooses "libcurl" if that is available.
The out-of-the box Bioconductor mirror has been changed to one using https://: use chooseBioCmirror() to choose a http:// mirror if required.
The data frame and formula methods for aggregate() gain a drop argument.
available.packages() gains a repos argument.
The undocumented switching of methods for url() on https: and ftps: URLs is confined to method = "default" (and documented).
smoothScatter() gains a ret.selection argument.
qr() no longer has a ... argument to pass additional arguments to methods.
has a method for class "table".
It is now possible (again) to replayPlot() a display list snapshot that was created by recordPlot() in a different R session.
It is still not a good idea to use snapshots as a persistent storage format for R plots, but it is now not completely silly to use a snapshot as a format for transferring an R plot between two R sessions.
The underlying changes mean that packages providing graphics devices (e.g., Cairo, RSvgDevice, cairoDevice, tikzDevice) will need to be reinstalled.
tools::undoc(dir = D) and codoc(dir = D) now also work when D is a directory whose normalizePath()ed version does not end in the package name, e.g. from a symlink.
abbreviate() has more support for multi-byte character sets - it no longer removes bytes within characters and knows about Latin vowels with accents. It is still only really suitable for (most) European languages, and still warns on non-ASCII input.
abbreviate(use.classes = FALSE) is now implemented, and that is more suitable for non-European languages.
match(x, table) is faster (sometimes by an order of magnitude) when x is of length one and incomparables is unchanged
More consistent, partly not back-compatible behavior of NA and NaN coercion to complex numbers, operations less often resulting in complex NA (NA_complex_).
lengths() considers methods for length and [[ on x, so it should work automatically on any objects for which appropriate methods on those generics are defined.
The logic for selecting the default screen device on OS X has been simplified: it is now quartz() if that is available even if environment variable DISPLAY has been set by the user.
The choice can easily be overridden _via_ environment variable R_INTERACTIVE_DEVICE.
On Unix-like platforms which support the getline C library function, system(*,intern = TRUE) no longer truncates (output) lines longer than 8192 characters
rank() gains a ties.method = "last" option, for convenience (and symmetry).
regmatches(invert = NA) can now be used to extract both non-matched and matched substrings.
data.frame() gains argument fix.empty.names; as.data.frame.list() gets new cut.names, col.names and fix.empty.names.
plot(x ~ x, *) now warns that it is the same as plot(x ~ 1, *).
recordPlot() has new arguments load and attach to allow package names to be stored as part of a recorded plot. replayPlot() has new argument reloadPkgs to load/attach any package names that were stored as part of a recorded plot.
S4 dispatch works within calls to .Internal(). This means explicit S4 generics are no longer needed for unlist() and as.vector().
Only font family names starting with "Hershey" (and not "Her" as before) are given special treatment by the graphics engine. S4 values are automatically coerced to vector (via as.vector) when subassigned into atomic vectors.
findInterval() gets a left.open option.
The version of LAPACK included in the sources has been updated to 3.6.0, including those 'deprecated' routines which were previously included. _Ca_ 40 double-complex routines have been added at the request of a package maintainer.
As before, the details of what is included are in src/modules/lapack/README and this now gives information on earlier additions.
tapply() has been made considerably more efficient without changing functionality
match.arg(arg) (the one-argument case) is faster; so is sort.int(). (PR#16640)
The format method for object_size objects now also accepts "binary" units such as "KiB" and e.g., "Tb". (Partly from PR#16649.)
Profiling now records calls of the form foo::bar and some similar cases directly rather than as calls to .
New string utilities startsWith(x, prefix) and endsWith(x, suffix). Also provide speedups for some grepl("^...",*) uses related to proposals in PR#16490).
Reference class finalizers run at exit, as well as on garbage collection.
Avoid parallel dependency on stats for port choice and random number seeds. (PR#16668)
The radix sort algorithm and implementation from data.table forder) replaces the previous radix (counting) sort and adds a new method for order(). The new algorithm supports logical, integer (even with large values), real, and character vectors. It outperforms all other methods, but there are some caveats (see ?sort). The order() function gains a method argument for choosing between "shell" and "radix".
New function grouping() returns a permutation that stably rearranges data so that identical values are adjacent. The return value includes extra partitioning information on the groups. The implementation came included with the new radix sort.
rhyper(nn, m, n, k) no longer returns NA when one of the three parameters exceeds the maximal integer.
switch() now warns when no alternatives are provided.
hist() for "Date" or "POSIXt" objects would sometimes give misleading labels on the breaks, as they were set to the day before the start of the period being displayed. The display format has been changed, and the shift of the start day has been made conditional on right = TRUE (the default). (PR#16679)
R now uses a new version of the logo (donated to the R Foundation by RStudio). It is defined in .svg format, so will resize without unnecessary degradation when displayed on HTML pages-there is also a vector PDF version.
New function .traceback() returns the stack trace which traceback() prints.
lengths() dispatches internally.
dotchart() gains a pt.cex argument to control the size of points separately from the size of plot labels.
as.roman(ch) now correctly deals with more diverse character vectors ch; also arithmetic with the resulting roman numbers works in more cases. (PR#16779)
prcomp() gains a new option rank. allowing to directly aim for less than min(n,p) PC's. The summary() and its print() method have been amended, notably for this case.
gzcon() gains a new option text, which marks the connection as text-oriented (so e.g. pushBack() works). It is still always opened in binary mode.
The import() namespace directive now accepts an argument except which names symbols to exclude from the imports. The except expression should evaluate to a character vector (after substituting symbols for strings). See Writing R Extensions.
New convenience function Rcmd() in package tools for invoking R CMD tools from within R.
New functions makevars_user() and makevars_site() in package tools to determine the location of the user and site specific Makevars files for customizing package compilation.
UTILITIES:
R CMD check has a new option --ignore-vignettes for use with non-Sweave vignettes whose VignetteBuilder package is not available
R CMD check now by default checks code usage (_via_ codetools) with only the base package attached. Functions from default packages other than base which are used in the package code but not imported are reported as undefined globals, with a suggested addition to the NAMESPACE file
R CMD check --as-cran now also checks DOIs in package CITATION and Rd files
R CMD Rdconv and R CMD Rd2pdf each have a new option
RdMacros=pkglist which allows Rd macros to be specified before processing
DEPRECATED AND DEFUNCT:
The previously included versions of zlib, bzip2, xz and PCRE have been removed, so suitable external (usually system) versions are required (see the 'R Installation and Administration' manual).
The unexported and undocumented Windows-only devices cairo_bmp(), cairo_png() and cairo_tiff() have been removed. (These devices should be used as e.g. bmp(type = "cairo").)
Windows only) Function setInternet2() has no effect and will be removed in due course. The choice between methods "internal" and "wininet" is now made by the method arguments of url() and download.file() and their defaults can be set _via_ options. The out-of-the-box default remains "wininet" (as it has been since R 3.2.2).
with an S4 value into a list currently embeds the S4 object into its own list such that the end result is roughly equivalent to using [[

New in R for Windows 3.2.5 (Apr 15, 2016)

New in R for Windows 3.2.4 (Apr 15, 2016)

NEW FEATURES:
install.packages() and related functions now give a more informative warning when an attempt is made to install a base package.
summary(x) now prints with less rounding when x contains infinite values. provideDimnames() gets an optional unique argument.
shQuote() gains type = "cmd2" for quoting in cmd.exe in Windows.
The data.frame method of rbind() gains an optional argument stringsAsFactors (instead of only depending on getOption("stringsAsFactors")).
smooth(x, *) now also works for long vectors.
tools::texi2dvi() has a workaround for problems with the texi2dvi script supplied by texinfo 6.1.
It extracts more error messages from the LaTeX logs when in emulation mode.
UTILITIES:
R CMD check will leave a log file build_vignettes.log from the re-building of vignettes in the .Rcheck directory if there is a problem, and always if environment variable _R_CHECK_ALWAYS_LOG_VIGNETTE_OUTPUT_ is set to a true value.
DEPRECATED AND DEFUNCT:
Use of SUPPORT_OPENMP from header Rconfig.h is deprecated in favour of the standard OpenMP define _OPENMP.
The make macro AWK which is long unused by R itself but recorded in file etc/Makeconf is deprecated and will be removed in R 3.3.0.
The C header file S.h is no longer documented: its use should be replaced by R.h.
BUG FIXES:
kmeans(x, centers = ) now works.
Vectorize() now checks for clashes in argument names.
file.copy(overwrite = FALSE) would signal a successful copy when none had taken place.
ngettext() now uses the same default domain as gettext().
array(.., dimnames = *) now warns about non-list dimnames and, from R 3.3.0, will signal the same error for invalid dimnames as matrix() has always done.
addmargins() now adds dimnames for the extended margins in all cases, as always documented.
heatmap() evaluated its add.expr argument in the wrong environment.
require() etc now give the correct entry of lib.loc in the warning about an old version of a package masking a newer required one.
The internal deparser did not add parentheses when necessary, e.g. before [] or [[]].
as.data.frame.vector(*, row.names=*) no longer produces â€˜corruptedâ€™ data frames from row names of incorrect length, but rather warns about them. This will become an error.
url connections with method = "libcurl" are destroyed properly.
withCallingHandler() now (again) handles warnings even during S4 generic's argument evaluation.
deparse(..., control = "quoteExpressions") incorrectly quoted empty expressions.
format()ting datetime objects ("POSIX[cl]?t") could segfault or recycle wrongly.
plot.ts(, las = 1) now does use las.
saveRDS(*, compress = "gzip") now works as documented.
(Windows only) The Rgui front end did not always initialize the console properly, and could cause R to crash. (PR#16998)
dummy.coef.lm() now works in more cases, thanks to a proposal by Werner Stahel . In addition, it now works for multivariate linear models ("mlm", manova) thanks to a proposal by Daniel Wollschlaeger.
The as.hclust() method for "dendrogram"s failed often when there were ties in the heights.
reorder() and midcache.dendrogram() now are non-recursive and hence applicable to somewhat deeply nested dendrograms, thanks to a proposal by Suharto Anggono in PR#16424.
cor.test() now calculates very small p values more accurately (affecting the result only in extreme not statistically relevant cases). (PR#16704)
smooth(*, do.ends=TRUE) did not always work correctly in R versions between 3.0.0 and 3.2.3.
pretty(D) for date-time objects D now also works well if range(D) is (much) smaller than a second. In the case of only one unique value in D, the pretty range now is more symmetric around that value than previously.
Similarly, pretty(dt) no longer returns a length 5 vector with duplicated entries for Date objects dt which span only a few days.
The figures in help pages such as ?points were accidentally damaged, and did not appear in R 3.2.3.
available.packages() sometimes deleted the wrong file when cleaning up temporary files.
The X11() device sometimes froze on Red Hat Enterprise Linux 6. It now waits for MapNotify events instead of Expose events, thanks to Siteshwar Vashisht. (PR#16497)
[dpqr]nbinom(*, size=Inf, mu=.) now works as limit case, for â€˜dpqâ€™ as the Poisson. (PR#16727) pnbinom() no longer loops infinitely in border cases.
approxfun(*, method="constant") and hence ecdf() which calls the former now correctly â€œpredictâ€ NaN values as NaN.
summary.data.frame() now displays NAs in Date columns in all cases.

New in R for Windows 3.2.3 (Jan 31, 2016)

NEW FEATURES:
Some recently-added Windows time zone names have been added to the conversion table used to convert these to Olson names Including those relating to changes for Russia in Oct 2014, as in PR#16503.)
Windows) Compatibility information has been added to the manifests for Rgui.exe, Rterm.exe and Rscript.exe. This should allow win.version() and Sys.info() to report the actual Windows version up to Windows 10
Windows "wininet" FTP first tries EPSV / PASV mode rather than only using active mode
which.min(x) and which.max(x) may be much faster for logical and integer x and now also work for long vectors
The 'emulation' part of tools::texi2dvi() has been somewhat enhanced, including supporting quiet = TRUE. It can be selected by texi2dvi = "emulation"
Windows) MiKTeX removed its texi2dvi.exe command in Sept 2015 tools::texi2dvi() tries texify.exe if it is not found
Windows only) Shortcuts for printing and saving have been added to menus in Rgui.exe. (Request of PR#16572.)
loess(..., iterTrace=TRUE) now provides diagnostics for robustness iterations, and the print() method for summary() shows slightly more
The included version of PCRE has been updated to 8.38, a bug-fix release
View() now displays nested data frames in a more friendly way
Request with patch in PR#15915.)
INSTALLATION and INCLUDED SOFTWARE:
The included configuration code for libintl has been updated to that from gettext version 0.19.5.1 - this should only affect how an external library is detected (and the only known instance is under OpenBSD). (Wish of PR#16464.)
configure has a new argument --disable-java to disable the checks for Java
The configure default for MAIN_LDFLAGS has been changed for the FreeBSD, NetBSD and Hurd OSes to one more likely to work with compilers other than gcc (FreeBSD 10 defaults to clang)
configure now supports the OpenMP flags -fopenmp=libomp (clang) and -qopenmp (Intel C)
Various macros can be set to override the default behaviour of configure when detecting OpenMP: see file config.site
Source installation on Windows has been modified to allow for MiKTeX installations without texi2dvi.exe. See file MkRules.dist
BUG FIXES:
regexpr(pat, x, perl = TRUE) with Python-style named capture did not work correctly when x contained NA strings. (PR#16484)
The description of dataset ToothGrowth has been improved/corrected. (PR#15953)
model.tables(type = "means") and hence TukeyHSD() now support "aov" fits without an intercept term. (PR#16437)
close() now reports the status of a pipe() connection opened with an explicit open argument. (PR#16481)
Coercing a list without names to a data frame is faster if the elements are very long. (PR#16467)
Unix-only) Under some rare circumstances piping the output from Rscript or R -f could result in attempting to close the input file twice, possibly crashing the process. (PR#16500)
Windows) Sys.info() was out of step with win.version() and did not report Windows 8
topenv(baseenv()) returns baseenv() again as in R 3.1.0 and earlier. This also fixes compilerJIT(3) when used in .Rprofile detach()ing the methods package keeps .isMethodsDispatchOn() true, as long as the methods namespace is not unloaded
Removed some spurious warnings from configure about the preprocessor not finding header files. (PR#15989)
rchisq(*, df=0, ncp=0) now returns 0 instead of NaN, and dchisq(*, df=0, ncp=*) also no longer returns NaN in limit cases where the limit is unique). (PR#16521)
pchisq(*, df=0, ncp > 0, log.p=TRUE) no longer underflows (for ncp > ~60)
nchar(x, "w") returned -1 for characters it did not know about e.g. zero-width spaces): it now assumes 1. It now knows about most zero-width characters and a few more double-width characters
Help for which.min() is now more precise about behavior with logical arguments. (PR#16532)
The print width of character strings marked as "latin1" or "bytes" was in some cases computed incorrectly
abbreviate() did not give names to the return value if minlength was zero, unlike when it was positive
Windows only) dir.create() did not always warn when it failed to create a directory. (PR#16537)
When operating in a non-UTF-8 multibyte locale (e.g. an East Asian locale on Windows), grep() and related functions did not handle UTF-8 strings properly. (PR#16264)
read.dcf() sometimes misread lines longer than 8191 characters
within(df, ..) no longer drops columns whose name start with a "."
The built-in HTTP server converted entire Content-Type to lowercase including parameters which can cause issues for multi-part form boundaries (PR#16541)
Modifying slots of S4 objects could fail when the methods package was not attached. (PR#16545)
splineDesign(*, outer.ok=TRUE) (splines) is better now PR#16549), and interpSpline() now allows sparse=TRUE for speedup with non-small sizes
If the expression in the traceback was too long, traceback() did not report the source line number.
The browser did not truncate the display of the function when exiting with options("deparse.max.lines") set. (PR#16581)
When bs(*, Boundary.knots=) had boundary knots inside the data range, extrapolation was somewhat off.
var() and hence sd() warn about factor arguments which are deprecated now. (PR#16564)
loess(*, weights = *) stored wrong weights and hence gave slightly wrong predictions for newdata. (PR#16587)
aperm(a, *) now preserves names(dim(a))
poly(x, ..) now works when either raw=TRUE or coef is specified (PR#16597)
data(package=*) is more careful in determining the path prettyNum(*, decimal.mark, big.mark): fixed bug introduced when fixing PR#16411

New in R for Windows 3.2.2 (Sep 8, 2015)

SIGNIFICANT USER-VISIBLE CHANGES:
It is now easier to use secure downloads from https:// URLs on builds which support them: no longer do non-default options need to be selected to do so. In particular, packages can be installed from repositories which offer https:// URLs, and those listed by setRepositories() now do so (for some of their mirrors)
Support for https:// URLs is available on Windows, and on other platforms if support for libcurl was compiled in and if that supports the https protocol (system installations can be expected to do). So https:// support can be expected except on rather old OSes (an example being OS X 'Snow Leopard', where a non-system version of libcurl can be used)
(Windows only) The default method for accessing URLs _via_ download.file() and url() has been changed to be "wininet" using Windows API calls. This changes the way proxies need to be set and security settings made: there have been some reports of sites being inaccessible under the new default method (but the previous methods remain available)
NEW FEATURES:
cmdscale() gets new option list. for increased flexibility when a list should be returned
configure now supports texinfo version 6.0, which (unlike the change from 4.x to 5.0) is a minor update. (Wish of PR#16456.)
(Windows only) setInternet2(TRUE) is now the default. The command-line option --internet2 and environment variable R_WIN_INTERNET2 are now ignored. Thus by default the "internal" method for download.file() and url() uses the "wininet" method: to revert to the previous default use setInternet2(FALSE). This means that https:// can be read by default by download.file() (they have been readable by file() and url() since R 3.2.0). There are implications for how proxies need to be set (see download.file): also, cacheOK = FALSE is not supported
chooseCRANmirror() and chooseBioCmirror() now offer HTTPS mirrors in preference to HTTP mirrors. This changes the interpretation of their ind arguments: see their help pages
capture.output() gets optional arguments type and split to pass to sink(), and hence can be used to capture messages
C-LEVEL FACILITIES:
Header Rconfig.h now defines HAVE_ALLOCA_H if the platform has the alloca.h header (it is needed to define alloca on Solaris and AIX, at least: see 'Writing R Extensions' for how to use it)
INSTALLATION and INCLUDED SOFTWARE:
The libtool script generated by configure has been modified to support FreeBSD >= 10 (PR#16410)
BUG FIXES:
The HTML help page links to demo code failed due to a change in R 3.2.0. (PR#16432)
If the na.action argument was used in model.frame(), the original data could be modified. (PR#16436)
getGraphicsEvent() could cause a crash if a graphics window was closed while it was in use. (PR#16438)
matrix(x, nr, nc, byrow = TRUE) failed if x was an object of type "expression"
strptime() could overflow the allocated storage on the C stack when the timezone had a non-standard format much longer than the standard formats. (Part of PR#16328.)
options(OutDec = s) now signals a warning (which will become an error in the future) when s is not a string with exactly one character, as that has been a documented requirement
prettyNum() gains a new option input.d.mark which together with other changes, e.g., the default for decimal.mark, fixes some format()ting variants with non-default getOption("OutDec") such as in PR#16411
download.packages() failed for type equal to either "both" or "binary".
The dendrogram method of labels() is much more efficient for large dendrograms, now using rapply(). (Comment #15 of PR#15215)
The "port" algorithm of nls() could give spurious errors
Reference classes that inherited from reference classes in another package could invalidate methods of the inherited class. Fixing this requires adding the ability for methods to be "external", with the object supplied explicitly as the first argument, named .self. See "Inter-Package Superclasses" in the documentation
readBin() could fail on the SPARC architecture due to alignment issues.
qt(*, df=Inf, ncp=.) now uses the natural qnorm() limit instead of returning NaN. (PR#16475)
Auto-printing of S3 and S4 values now searches for print() in the base namespace and show() in the methods namespace instead of searching the global environment
polym() gains a coefs = NULL argument and returns class "poly" just like poly() which gets a new simple=FALSE option. They now lead to correct predict()ions, e.g., on subsets of the original data
rhyper(nn, ) now works correctly. (PR#16489)
ttkimage() did not (and could not) work so was removed. Ditto for tkimage.cget() and tkimage.configure(). Added two Ttk widgets and missing subcommands for Tk's image command: ttkscale() ttkspinbox(), tkimage.delete(), tkimage.height(), tkimage.inuse(), tkimage.type(), tkimage.types(), tkimage.width(). (PR#15372, PR#16450)
getClass("foo") now also returns a class definition when it is found in the cache more than once

New in R for Windows 3.2.1 (Sep 8, 2015)

NEW FEATURES:
utf8ToInt() now checks that its input is valid UTF-8 and returns NA if it is not.
install.packages() now allows type = "both" with repos = NULL if it can infer the type of file.
nchar(x, *) and nzchar(x) gain a new argument keepNA which governs how the result for NAs in x is determined. For the R 3.2.x series, the default remains FALSE which is fully back compatible. From R 3.3.0, the default will change to keepNA = NA and you are advised to consider this for code portability.
news() more flexibly extracts dates from package NEWS.Rd files.
lengths(x) now also works (trivially) for atomic x and hence can be used more generally as an efficient replacement of sapply(x, length) and similar.
The included version of PCRE has been updated to 8.37, a bug-fix release.
diag() no longer duplicates a matrix when extracting its diagonal.
as.character.srcref() gains an argument to allow characters corresponding to a range of source references to be extracted.
BUG FIXES:
acf() and ccf() now guarantee values strictly in [-1,1] (instead of sometimes very slightly outside). PR#15832.
as.integer("111111111111") now gives NA (with a warning) as it does for the corresponding numeric or negative number coercions. Further, as.integer(M + 0.1) now gives M (instead of NA) when M is the maximal representable integer.
On some platforms nchar(x, "c") and nchar(x, "w") would return values (possibly NA) for inputs which were declared to be UTF-8 but were not, or for invalid strings without a marked encoding in a multi-byte locale, rather than give an error. Additional checks have been added to mitigate this.
apply(a, M, function(u) c(X = ., Y = .)) again has dimnames containing "X" and "Y" (as in R < 3.2.0).
(Windows only) In some cases, the --clean option to R CMD INSTALL could fail. (PR#16178)
(Windows only) choose.files() would occasionally include characters from the result of an earlier call in the result of a later one. (PR#16270)
A change in RSiteSearch() in R 3.2.0 caused it to submit invalid URLs. (PR#16329)
Rscript and command line R silently ignored incomplete statements at the end of a script; now they are reported as parse errors. PR#16350)
Parse data for very long strings was not stored. (PR#16354)
plotNode(), the workhorse of the plot method for "dendrogram"s is no longer recursive, and hence also works for deeply nested dendrograms. (PR#15215)
The parser could overflow internally when given numbers in scientific format with extremely large exponents. (PR#16358)
If the CRAN mirror was not set, install.packages(type = "both") and related functions could repeatedly query the user for it. (Part of PR#16362)
The low-level functions .rowSums() etc. did not check the length of their argument, so could segfault. (PR#16367)
The quietly argument of library() is now correctly propagated from .getRequiredPackages2().
Under some circumstances using the internal PCRE when building R from source would cause external libs such as -llzma to be omitted from the main link.
The .Primitive default methods of the logic operators, i.e., !, & and |, now give correct error messages when appropriate, e.g., for `&`(TRUE) or `!`(). (PR#16385)
cummax(x) now correctly propagates NAs also when x is of type integer and begins with an NA.
summaryRprof() could fail when the profile contained only two records. (PR#16395)
HTML vignettes opened using vignette() did not support links into the rest of the HTML help system. (Links worked properly when the vignette was opened using browseVignettes() or from within the help system.)
arima(*, xreg = .) (for d >= 1) computes estimated variances based on a the number of effective observations as in R version 3.0.1 and earlier. (PR#16278)
slotNames(.) is now correct for "signature" objects (mostly used internally in methods).
On some systems, the first string comparison after a locale change would result in NA.

New in R for Windows 3.2.0 (Apr 18, 2015)

NEW FEATURES:
anyNA() gains a recursive argument.
When x is missing and names is not false (including the default value), Sys.getenv(x, names) returns an object of class "Dlist" and hence prints tidily.
(Windows.) shell() no longer consults the environment variable SHELL: too many systems have been encountered where it was set incorrectly (usually to a path where software was compiled, not where it was installed). R_SHELL, the preferred way to select a non-default shell, can be used instead.
Some unusual arguments to embedFonts() can now be specified as character vectors, and the defaults have been changed accordingly.
Functions in the Summary group duplicate less. (PR#15798)
(Unix-alikes.) system(cmd, input = ) now uses 'shell-execution-environment' redirection, which will be more natural if cmd is not a single command (but requires a POSIX-compliant shell). (Wish of PR#15508)
read.fwf() and read.DIF() gain a fileEncoding argument, for convenience.
Graphics devices can add attributes to their description in .Device and .Devices. Several of those included with R use a "filepath" attribute.
pmatch() uses hashing in more cases and so is faster at the expense of using more memory. (PR#15697)
pairs() gains new arguments to select sets of variables to be plotted against each other.
file.info(, extra_cols = FALSE) allows a minimal set of columns to be computed on Unix-alikes: on some systems without properly-configured caching this can be significantly faster with large file lists.
New function dir.exists() in package base to test efficiently whether one or more paths exist and are directories.
dput() and friends gain new controls hexNumeric and digits17 which output double and complex quantities as, respectively, binary fractions (exactly, see sprintf("%a")) and as decimals with up to 17 significant digits.
save(), saveRDS() and serialize() now support ascii = NA which writes ASCII files using sprintf("%a") for double/complex quantities. This is read-compatible with ascii = TRUE but avoids binary->decimal->binary conversions with potential loss of precision. Unfortunately the Windows C runtime's lack of C99 compliance means that the format cannot be read correctly there in R before 3.1.2.
The default for formatC(decimal.mark =) has been changed to be getOption("OutDec"); this makes it more consistent with format() and suitable for use in print methods, e.g. those for classes "density", "ecdf", "stepfun" and "summary.lm".
getOption("OutDec") is now consulted by the print method for class "kmeans", by cut(), dendrogram(), plot.ts() and quantile() when constructing labels and for the report from legend(trace = TRUE). (In part, wish of PR#15819.)
printNum() and hence format() and formatC() give a warning if big.mark and decimal.mark are set to the same value (period and comma are not uncommonly used for each, and this is a check that conventions have not got mixed).
merge() can create a result which uses long vectors on 64-bit platforms.
dget() gains a new argument keep.source which defaults to FALSE for speed (dput() and dget() are most often used for data objects where this can make dget() many times faster).
Packages may now use a file of common macro definitions in their help files, and may import definitions from other packages.
A number of macros have been added in the new share/Rd directory for use in package overview help pages, and promptPackage() now makes use of them.
tools::parse_Rd() gains a new permissive argument which converts unrecognized macros into text. This is used by utils:::format.bibentry to allow LaTeX markup to be ignored.
options(OutDec =) can now specify a multi-byte character, e.g., options(OutDec = "\u00b7") in a UTF-8 locale.
is.recursive(x) is no longer true when x is an external pointer, a weak reference or byte code; the first enables all.equal(x, x) when x

New in R for Windows 3.1.3 (Apr 18, 2015)

NEW FEATURES:
The internal method of download.file() can now handle files larger than 2GB on 32-bit builds which support such files (tested on 32-bit R running on 64-bit Windows).
kruskal.test() warns on more types of suspicious input.
The as.dendrogram() method for "hclust" objects gains a check argument protecting against memory explosion for invalid inputs.
capabilities() has a new item long.double which indicates if the build uses a long double type which is longer than double.
nlm() no longer modifies the callback argument in place (a new vector is allocated for each invocation, which mimics the implicit duplication that occurred in R < 3.1.0); note that this is a change from the previously documented behavior. (PR#15958)
icuSetCollate() now accepts locale = "ASCII" which uses the basic C function strcmp and so collates strings byte-by-byte in numerical order.
sessionInfo() tries to report the OS version in use (not just that compiled under, and including details of Linux distributions).
model.frame() (used by lm() and many other modelling functions) now warns when it drops contrasts from factors. (Wish of PR#16119)
install.packages() and friends now accept the value type = "binary" as a synonym for the native binary type on the platform (if it has one).
Single source or binary files can be supplied for install.packages(type = "both") and the appropriate type and repos = NULL will be inferred.
New function pcre_config() to report on some of the configuration options of the version of PCRE in use. In particular, this reports if regular expressions using \p{xx} are supported.
(Windows.) download.file(cacheOK = FALSE) is now supported when internet2.dll is used.
browseURL() has been updated to work with Firefox 36.0 which has dropped support for the -remote interface.
INSTALLATION and INCLUDED SOFTWARE:
The included version of PCRE has been updated to 8.36.
configure accepts MAKEINFO=texi2any as another way to ensure texinfo 5.x is used when both 5.x and 4.x are installed.
UTILITIES:
R CMD check now checks the packages used in \donttest sections of the examples are specified in the DESCRIPTION file. (These are needed to run the examples interactively.)
R CMD check checks for the undeclared use of GNU extensions in Makefiles, and for Makefiles with a missing final linefeed.
R CMD build will correct line endings in all Makefiles, not just those in the src directory.
R CMD check notes uses of library() and require() in package code: see the section 'Suggested packages' of 'Writing R Extensions' for good practice.
DEPRECATED AND DEFUNCT:
The configure option --with-valgrind-instrumentation=3 is deprecated and will be removed in R 3.2.0.
BUG FIXES:
(Windows.) Rscript.exe was missing a manifest specifying the modern style for common controls (e.g., the download progress bar).
If a package had extra documentation files but no vignette, the HTML help system produced an empty index page.
The parser now gives an error if a null character is included in a string using Unicode escapes. (PR#16046)
qr.Q() failed on complex arguments due to pre-3.0(!) typo. (PR#16054)
abs() failed with named arguments when the argument was complex. (PR#16047)
"noquote" objects may now be used as columns in dataframes. (PR#15997)
Some values with extremely long names were printed incorrectly. (PR#15999)
Extremely large exponents on zero expressed in scientific notation (e.g. 0.0e50000) could give NaN. (PR#15976)
download.file() reported downloaded sizes as 0KB if less than 1MB, only for R 3.1.2 and only on big-endian platforms.
prompt() did not escape percent signs in the automatically generated usage section of help files.
drop.terms() dropped some of the attributes of the object it was working with. (PR#16029)
(Windows.) The command completion in Rgui.exe messed up the console. (PR#15791)
(Windows.) The choose.files() command returned a blank string when the user asked for a single file but cancelled the request. (PR#16074)
Math2 S4 group generics failed to correctly dispatch "structure"- and "nonStructure"-derived classes.
loadNamespace() imposed undocumented restrictions on the versionCheck parameter.
Rare over-runs detected by AddressSanitizer in substr() and its replacement version have been avoided.
_Inter alia_ that fix gives the documented behaviour for substr(x, 1, 2)

New in R for Windows 3.1.2 (Nov 10, 2014)

NEW FEATURES:
embedFonts() now defaults to format = "ps2write" for ‘.ps’ and ‘.eps’ files. This is available in Ghostscript 9.x (since 2010) whereas the previous default, format = "pswrite", was removed in Ghostscript 9.10.
For consistency with [dpqr]norm(), [dp]lnorm(sdlog = 0) model a point mass at exp(mulog) rather than return NaN (for an error).
capabilities() now reports if ICU is compiled in for use for collation (it is only actually used if a suitable locale is set for collation, and never for a C locale).
(OS X only.) Package tcltk checks when loaded if it is linked against the CRAN X11-based Tcl/Tk and if so that the Tcl/Tk component and the X11 libraries are installed. This allows more informative error messages to be given advising the installation of the missing component or of XQuartz.
The X11() device and X11-based versions of the data editor and viewer (invoked by edit() and View() for data frames and matrices from command-line R) check that the X11 libraries are installed and if not advises installing XQuartz.
icuSetCollate() allows locale = "default", and locale = "none" to use OS services rather than ICU for collation.
Environment variable R_ICU_LOCALE can be used to set the default ICU locale, in case the one derived from the OS locale is inappropriate (this is currently necessary on Windows).
New function icuGetCollate() to report on the ICU collation locale in use (if any).
utils::URLencode() was updated to use unreserved and reserved characters from RFC 3986, http://tools.ietf.org/html/rfc3986, instead of RFC 1738.
unique(warnings()) and c(warnings()) are now supported.
The Bioconductor ‘version’ used by setRepositories() now defaults to 3.0. (It can be set at runtime via environment variable R_BIOC_VERSION.)
INSTALLATION and INCLUDED SOFTWARE:
The configure script reports on the more important capabilities/options which will not be compiled in.
More types of external BLAS are recognized by name in that report.
When building R as a shared library, the -L${R_HOME}/lib${R_ARCH} flag is placed earlier in the link commands used during installation and when packages are installed: this helps ensure that the current build has priority if an R shared library has already been installed by e.g. install-libR in a library mentioned in LDFLAGS (and not in ‘your system's library directory’ as documented). (Wish of PR#15790.)
LaTeX package upquote is no longer required for R's use of inconsolata.
(Windows only) If both 32 and 64 bit versions of R are installed, the ‘bin/R.exe’ and ‘bin/Rscript.exe’ executables now run 64 bit R. (To run 32 bit R, overwrite these files with copies of ‘bin/i386/Rfe.exe’.)
UTILITIES:
Running R CMD check with _R_CHECK_DEPENDS_ONLY_ true now makes the VignetteBuilder packages available even if they are listed in Suggests, since they are needed to recognise and process non-Sweave vignettes.
R CMD check now reports empty importFrom declarations in a ‘NAMESPACE’ file, as these are common errors (writing importFrom(Pkg) where import(Pkg) was intended).
R CMD check now by default checks code usage directly on the package namespace without loading and attaching the package and its suggests and enhances. For good practice with packages in the Suggests field, see §1.1.3.1 of ‘Writing R Extensions’. For use of lazy-data objects in the package's own code, see ?data.
BUG FIXES:
dmultinom() did not handle non-finite probabilities correctly.
prettyNum(x, zero.print=*) now also works when x contains NAs.
A longstanding bug exhibited by nlminb() on Windows was traced to a compiler bug in gcc 4.6.3; a workaround has been put in place.
Rendering of \command in HTML versions of help pages has been improved: this is particularly evident on the help page for INSTALL.
as.hexmode(x) and as.octmode(x) now behave correctly for some numeric x, e.g., c(NA, 1) or c(1, pi).
drop1() failed if the scope argument had no variables to drop.
edit() (and hence fix()) failed if an object had a non-character attribute named "source" (an attribute that had been used in R prior to version 2.14.0).
callGeneric() could fail if the generic had ... as a formal argument.
Forking in package parallel called C entry point exit in the child. This was unsafe (_exit should have been called), and could flush stdin of the main R process (seen most often on Solaris).
As good practice, stdout is now flushed before forking a child.
R objects such as list(`a\b` = 1) now print correctly.
getAnywhere("C_pbinom") now returns correctly a single object (rather than unlisting it).
The confint() method for nls() fits failed it these has specified parameter limits despite using an algorithm other than "port".
Subclassing an S4 class failed if the class required arguments to the generator, through its initialize() method.
removeSource() did not properly handle expressions containing arguments that were supplied as missing, e.g. x[i,].
as.environment(list()) now works, and as.list() of such an environment is now the same as list().
Several tcltk functions failed when run in unusual environments.
options(list()) now works (trivially).
merge(, ..) now works correctly for two ‘independent’ dendrograms (PR#15648), and still compatibly via adjust = "auto" e.g. for two branches of an existing dendrogram.
The plot method for "hclust" objects gets an optional argument check; When that is true (the default) it checks more carefully for valid input.
(Windows only) If a user chose to install 64 bit R but not 32 bit R, the ‘bin/R’ and ‘bin/Rscript’ executables failed to run.
Various possible buffer overruns have been prevented, and missed memory protection added.
Rscript no longer passes --args to R when there are no extra (“user”) arguments.
objects like getClass("refClass")@prototype now print() and str() without error.
identical() now also looks at the S4 bit.
hist(x, breaks) is more robust in adding a small fuzz to few breaks when some are very large.
sub() and gsub() did not handle regular expressions like "\s{2,}" properly if the text contained NA or non-ascii elements in a UTF-8 locale. Part of this was due to a bug in the TRE library.
RShowDoc("NEWS") now displays the PDF version.
Matrices and arrays with last dimension zero did not print at all or incompletely.
plot.histogram() and hence hist() now respect the xaxs, yaxs and lab graphics parameters.
bw.SJ(x) and other bw.*() no longer segfault when x contains non-finite values.
R CMD Rd2pdf unintentionally ignored its --os option.
The internal method of download.file() was not reporting file sizes and progress correctly on files larger than 2GB (inherited from libxml2). This is corrected for 64-bit builds (32-bit platforms may not support such files, but where possible will be supported in future versions of R).
Work around a bug in OS X Yosemite where key environment variables may be duplicated causing issues in subprocesses. The duplicates are now removed on R startup (via Rprofile).
Adjust X11 auto-launch detection in DISPLAY on OS X to recognize latest XQuartz.

New in R for Windows 3.1.1 (Oct 10, 2014)

NEW FEATURES:
When attach() reports conflicts, it does so compatibly with library() by using message().
R CMD Sweave no longer cleans any files by default, compatibly with versions of R prior to 3.1.0. There are new options --clean, --clean=default and --clean=keepOuts.
tools::buildVignette() and tools::buildVignettes() with clean = FALSE no longer remove any created files. buildvignette() gains a keep argument for more cleaning customization.
The Bioconductor ‘version’ used by setRepositories() can now be set by environment variable R_BIOC_VERSION at runtime, not just when R is installed. (It has been stated that Bioconductor will switch from ‘version’ 2.14 to ‘version’ 3.0 during the lifetime of the R 3.1 series.)
Error messages from bugs in embedded Sexpr code in Sweave documents now report the source location.
type.convert(), read.table() and similar read.*() functions get a new numerals argument, specifying how numeric input is converted when its conversion to double precision loses accuracy. The default value, "allow.loss" allows accuracy loss, as in R versions before 3.1.0.
For some compilers, integer addition could overflow without a warning. R's internal code for both integer addition and subtraction is more robust now.
The function determining the default number of knots for smooth.spline() is now exported, as .nknots.smspl().
dbeta(, a,b), pbeta(), qbeta() and rbeta() are now defined also for a = 0, b = 0, or infinite a and b (where they typically returned NaN before).
Many package authors report that the RStudio graphics device does not work correctly with their package's use of dev.new(). The new option dev.new(noRStudioGD = TRUE) replaces the RStudio override by the default device as selected by R itself, still respecting environment variables R_INTERACTIVE_DEVICE and R_DEFAULT_DEVICE.
readRDS() now returns visibly.
Modifying internal logical scalar constants now results in an error instead of a warning.
install.packages(repos = NULL) now accepts http:// or ftp:// URLs of package archives as well as file paths, and will download as required. In most cases repos = NULL can be deduced from the extension of the URL.
The warning when using partial matching with the $ operator on data frames is now only given when options("warnPartialMatchDollar") is TRUE.
Package help requests like package?foo now try the package foo whether loaded or not.
General help requests now default to trying all loaded packages, not just those on the search path.
Added a new function promptImport(), to generate a help page for a function that was imported from another package (and presumably re-exported, or help would not be needed).
INSTALLATION and INCLUDED SOFTWARE:
configure option --with-internal-tzcode can now be used with variable rsharedir.
The included version of PCRE has been updated to 8.35.
There is a new target make uninstall-libR to remove an installed shared/static ‘libR’.
make install-libR now works if a sub-architecture is used, although the user will need to specify libdir differently for different sub-architectures.
There is more extensive advice on which LaTeX packages are required to install R or to make package manuals (as done by R CMD check) in the ‘Writing R Extensions’ manual.
Compilers/linkers were handling the visibility controls in ‘src/extra/xz’ inconsistently (and apparently in some cases incorrectly), so it has been simplified.
(Windows) There is updated support for the use of ICU for collation: see the ‘R Installation and Administration Manual’.
BUG FIXES:
dbinom(x, n), pbinom(), dpois(), etc, are slightly less restrictive in checking if n is integer-valued.
pchisq(x, df, ncp, log.p = TRUE) is more accurate and no longer underflows for small x and ncp < 80, e.g, for pchisq(1e-5, df = 100, ncp = 1, log = TRUE).
The s (“step into”) command in the debugger would cause R to step into expressions evaluated there, not just into functions being debugged.
The C code used by strptime() rejected time-zone offsets of more than +1200 (+1245, +1300 and +1400 can occur).
(Windows only.) png(type = "cairo", antialias = "gray") was not accepted.
Use of save(..., envir=) with named objects could fail.
Sweave() mis-parsed Sexpr expressions that contained backslashes.
The return value from options(foo = NULL) was not the previous value of the option.
enc2utf8() and enc2native() did not always mark the encoding of the return values when it was known.
dnbinom(x, size = , mu, log = TRUE) no longer underflows to -Inf for large mu, thanks to a suggestion from Alessandro Mammana (MPI MolGen, Berlin).
pbeta(x, a, b, log = TRUE) no longer behaves discontinuously (in a small x-region) because of denormalized numbers. Also, pbeta(1-1e-12, 1e30, 1.001, log=TRUE) now terminates “in real time”.
The "CRAN" filter (see available.packages()) no longer removes duplicates other than of packages on CRAN, and does not fail if there is no CRAN repository in getOption("repos").
The device listing from dev2bitmap() and bitmap() was truncated to 1000 characters: modern versions of GhostScript on most platforms have many more devices.
(Windows.) Commands such as Sys.which() and pipe() which needed to find the full path to a command could segfault if the ‘long’ path name was much longer than the ‘short’ path name (which Sys.which() returns), as the behaviour of the Windows API call had changed.
R CMD build will fail with an error if one of the packages specified in the VignetteBuilder field is not installed. (Without loading those packages it cannot be ascertained which files are intended to be vignettes. This means that the VignetteBuilder packages have to be installed for package checking too.)
Misguided attempts to use chull() with non-finite points now give an error
For a formula with exactly 32 variables the 32nd variable was aliased to the intercept in some C-level computations of terms, so that for example attempting to remove it would remove the intercept instead (and leave a corrupt internal structure).
anyDuplicated() silently returned wrong values when the first duplicate was at an index which was too large to be stored in an integer vector (although a lot of RAM and patience would have been needed to encounter this).
tools::Rd2ex(commentDontrun = FALSE) failed if the block had only one line.
Hexadecimal constants such as 0x110p-5L which were incorrectly qualified by L were parsed incorrectly since R 3.0.0, with a slightly garbled warning.
system() returned success on some platforms even if the system was unable to launch a process.
(Windows Rgui console.) Unbuffered output was sometimes not output immediately if the prompt was not on the last line of the console.
The built-in help server did not declare the encoding for the ‘DESCRIPTION’ or other text files to be the package encoding, so non-ASCII characters could be displayed incorrectly.
R is now trying harder to not cleanup child processes that were not spawned by mcparallel() on platforms that provide information about the source process of the SIGCHLD signal. This allows 3rd party libraries to manage the exit status of children that they spawn without R interfering.
mcmapply() was only parallelizing if the number of jobs was bigger than the number of cores. It now parallelizes if the number of jobs is more than one.
Auto-printing would re-evaluate its argument when trying to dispatch to a print method. This is now avoided when possible.
Unserializing (including load() and readRDS()) could silently return incorrect numeric values from ASCII saves if there was a read error.
getParseData() could return incorrect values for the parents of some elements.
Attempting to use data frames of 2^31 or more rows with merge() or to create a merged data frame of that size now gives a clearer error message.
parse() did not check its file argument was a connection if it was not a character string, so e.g. parse(FALSE) attempted to read from stdin.
Nor did dump() and dput().
The "help.try.all.packages" option was ignored when the shortcut syntax for help was used, e.g. ?foo.
A potential segfault in string allocation has been fixed.
Potential memory protection errors in sort() and D() have been fixed.
Fixed a lack of error checking in graphics event functions.
numericDeriv() sometimes miscalculated the gradient.

New in R for Windows 3.1.0 (Apr 14, 2014)

NEW FEATURES:
type.convert() (and hence by default read.table()) returns a character vector or factor when representing a numeric input as a double would lose accuracy. Similarly for complex inputs.
If a file contains numeric data with unrepresentable numbers of decimal places that are intended to be read as numeric, specify colClasses in read.table() to be "numeric".
tools::Rdiff(useDiff = FALSE) is closer to the POSIX definition of diff -b (as distinct from the description in the man pages of most systems).
New function anyNA(), a version of any(is.na(.)) which is fast for atomic vectors, based on a proposal by Tim Hesterberg. (Wish of PR#15239.)
arrayInd(*, useNames = TRUE) and, analogously, which(*, arr.ind = TRUE) now make use of names(.dimnames) when available.
is.unsorted() now also works for raw vectors.
The "table" method for as.data.frame() (also useful as as.data.frame.table()) now passes sep and base arguments to provideDimnames().
uniroot() gets new optional arguments, notably extendInt, allowing to auto-extend the search interval when needed. The return value has an extra component, init.it.
switch(f, ...) now warns when f is a factor, as this typically happens accidentally where the useR meant to pass a character string, but f is treated as integer (as always documented).
The parser has been modified to use less memory.
The way the unary operators (+ - !) handle attributes is now more consistent. If there is no coercion, all attributes (including class) are copied from the input to the result: otherwise only names, dims and dimnames are.
colorRamp() and colorRampPalette() now allow non-opaque colours and a ramp in opacity via the new argument alpha = TRUE. (Suggested by Alberto Krone-Martins, but optionally as there are existing uses which expect only RGB values.)
grid.show.layout() and grid.show.viewport() get an optional vp.ex argument.
There is a new function find_gs_cmd() in the tools package to locate a GhostScript executable. (This is an enhanced version of a previously internal function there.)
object.size() gains a format() method.
There is a new family, "ArialMT", for the pdf() and postscript() devices. This will only be rendered correctly on viewers which have access to Monotype TrueType fonts (which are sometimes requested by journals).
The text and PDF news files, including ‘NEWS’ and ‘NEWS.2’, have been moved to the ‘doc’ directory.
combn(x, simplify = TRUE) now gives a factor result for factor input x (previously user error). (Related to PR#15442.)
Added utils::fileSnapshot() and utils::changedFiles() functions to allow snapshots and comparison of directories of files.
make.names(names, unique=TRUE) now tries to preserve existing names. (Suggestion of PR#15452.)
New functions cospi(x), sinpi(x), and tanpi(x), for more accurate computation of cos(pi*x), etc, both in R and the C API. Using these gains accuracy in some cases, e.g., inside lgamma() or besselI(). (Suggested by Morten Welinder in PR#15529.)
print.table(x, zero.print = ".") now also has an effect when x is not integer-valued.
There is more support to explore the system's idea of time-zone names. Sys.timezone() tries to give the current system setting by name and OlsonNames() lists the names in the system's Olson database. Sys.timezone(location = FALSE) gives the previous behaviour.
Currently time_t is usually long and hence 64-bit on Unix-alike 64-bit platforms: however it several cases the time-zone database is 32-bit. On R for Windows it is 64-bit (for both architectures as from this version).
The "save.defaults" option can include a value for compression_level. (Wish of PR#15579.)
colSums() and friends now have support for arrays and data-frame columns with 2^31 or more elements.
as.factor() is faster when f is an unclassed integer vector (for example, when called from tapply()).
fft() now works with longer inputs, from the 12 million previously supported up to 2 billion. (PR#15593)
Complex svd() now uses LAPACK subroutine ZGESDD, the complex analogue of the routine used for the real case.
Sweave now outputs ‘.tex’ files in UTF-8 if the input encoding is declared to be UTF-8, regardless of the local encoding. The UTF-8 encoding may now be declared using a LaTeX comment containing the string %\SweaveUTF8 on a line by itself.
file.copy() gains a copy.date argument.
Printing of date-times will make use of the time-zone abbreviation in use at the time, if known. For example, for Paris pre-1940 this could be LMT, PMT, WET or WEST. To enable this, the "POSIXlt" class has an optional component "zone" recording the abbreviation for each element.
For platforms which support it, there is also a component "gmtoff" recording the offset from GMT where known.
The system C function strftime has been replaced by a more comprehensive version with closer conformance to the POSIX 2008 standard.
dnorm(x, log = FALSE) is more accurate (but somewhat slower) for |x| > 5; as suggested in PR#15620.
Some versions of the tiff() device have further compression options.
read.table(), readLines() and scan() have a new argument to influence the treatment of embedded nuls.
Avoid duplicating the right hand side values in complex assignments when possible. This reduces copying of replacement values in expressions such as Z$a

New in R for Windows 3.0.3 (Mar 11, 2014)

NEW FEATURES:
On Windows there is support for making ‘.texi’ manuals using texinfo 5.0 or later: the setting is in file ‘src/gnuwin32/MkRules.dist’.
A packaging of the Perl script and modules for texinfo 5.2 has been made available
write.table() now handles matrices of 2^31 or more elements, for those with large amounts of patience and disc space.
There is a new function, La_version(), to report the version of LAPACK in use.
The HTML version of ‘An Introduction to R’ now has links to PNG versions of the figures.
There is some support to produce manuals in ebook formats. (See ‘doc/manual/Makefile’. Suggested by Mauro Cavalcanti.)
On a Unix-alike Sys.timezone() returns NA if the environment variable TZ is unset, to distinguish it from an empty string which on some OSes means the UTC time zone.
The backtick may now be escaped in strings, to allow names containing them to be constructed, e.g. `\``. (PR#15621)
read.table(), readLines() and scan() now warn when an embedded nul is found in the input. (Related to PR#15625 which was puzzled by the behaviour in this unsupported case.)
(Windows only.) file.symlink() works around the undocumented restriction of the Windows system call to backslashes. (Wish of PR#15631.)
KalmanForecast(fast = FALSE) is now the default, and the help contains an example of how fast = TRUE can be used in this version. (The usage will change in 3.1.0.)
strptime() now checks the locale only when locale-specific formats are used and caches the locale in use: this can halve the time taken on OSes with slow system functions (e.g. OS X).
strptime() and the format() methods for classes "POSIXct", "POSIXlt" and "Date" recognize strings with marked encodings: this allows, for example, UTF-8 French month names to be read on (French) Windows.
iconv(to = "utf8") is now accepted on all platforms (some implementations did already, but GNU libiconv did not: however converted strings were not marked as being in UTF-8). The official name, "UTF-8" is still preferred.
available.packages() is better protected against corrupt metadata files. (A recurring problem with Debian package shogun-r: PR#14713.)
Finalizers are marked to be run at garbage collection, but run only at a somewhat safer later time (when interrupts are checked). This circumvents some problems with finalizers running arbitrary code during garbage collection (the known instances being running options() and (C-level) path.expand() re-entrantly).
INSTALLATION and INCLUDED SOFTWARE:
The included version of PCRE has been updated to 8.34. This fixes bugs and makes the behaviour closer to Perl 5.18. In particular, the concept of ‘space’ includes VT and hence agrees with POSIX's.
PACKAGE INSTALLATION:
The new field SysDataCompression in the ‘DESCRIPTION’ file allows user control over the compression used for ‘sysdata.rda’ objects in the lazy-load database.
install.packages(dependencies = value) for value = NA (the default) or value = TRUE omits packages only in LinkingTo for binary package installs.
C-LEVEL FACILITIES:
The long undocumented remapping of rround() to Rf_fround() in header ‘Rmath.h’ is now formally deprecated: use fround() directly.
Remapping of prec() and trunc() in the ‘Rmath.h’ header has been disabled in C++ code (it has caused breakage with libc++ headers).
BUG FIXES:
getParseData() truncated the imaginary part of complex number constants. (Reported by Yihui Xie.)
dbeta(x, a, b) with a or b within a factor of 2 of the largest representable number could infinite-loop. (Reported by Ioannis Kosmidis.)
provideDimnames() failed for arrays with a 0 dimension. (PR#15465)
rbind() and cbind() did not handle list objects correctly. (PR#15468)
replayPlot() now checks if it is replaying a plot from the same session.
rasterImage() and grid.raster() now give error on an empty (zero-length) raster. (Reported by Ben North.)
plot.lm() would sometimes scramble the labels in plot type 5. (PR#15458 and PR#14837)
min() did not handle NA_character_ values properly. (Reported by Magnus Thor Torfason.)
(Windows only.) readRegistry() would duplicate default values for keys. (PR#15455)
str(..., strict.width = "cut") did not handle it properly when more than one line needed to be cut. (Reported by Gerrit Eichner.)
Removing subclass back-references when S4 classes were removed or their namespace unloaded had several bugs (e.g., PR#15481).
aggregate() could fail when there were too many levels present in the by argument. (PR#15004)
namespaceImportFrom() needed to detect primitive functions when checking for duplicated imports (reported by Karl Forner).
getGraphicsEvent() did not exit when a user closed the graphics window. (PR#15208)
Errors in vignettes were not always captured and displayed properly. (PR#15495)
contour() could fail when dealing with extremely small z values. (PR#15454)
Several functions did not handle zero-length vectors properly, including browseEnv(), format(), gl(), relist() and summary.data.frame(). (E.g., PR#15499)
Sweave() did not restore the R output to the console if it was interrupted by a user in the middle of evaluating a code chunk. (Reported by Michael Sumner.)
Fake installs of packages with vignettes work again.
Illegal characters in the input caused parse() (and thus source()) to segfault. (PR#15518)
The nonsensical use of nmax = 1 in duplicated() or unique() is now silently ignored.
qcauchy(p, *) is now fully accurate even when p is very close to 1. (PR#15521)
The validmu() and valideta() functions in the standard glm() families now also report non-finite values, rather than failing.
Saved vignette results (in a ‘.Rout.save’ file) were not being compared to the new ones during R CMD check.
Double-clicking outside of the list box (e.g. on the scrollbar) of a Tk listbox widget generated by tk_select.list() no longer causes the window to close. (PR#15407)
Improved handling of edge cases in parallel::splitindices(). (PR#15552)
HTML display of results from help.search() and ?? sometimes contained badly constructed links.
c() and related functions such as unlist() converted raw vectors to invalid logical vectors. (PR#15535)
(Windows only) When a call to system2() specified one of stdin, stdout or stderr to be a file, but the command was not found (e.g. it contained its arguments, or the program was not on the PATH), it left the file open and unusable until R terminated. (Reported by Mathew McLean.)
The bmp() device was not recording res = NA correctly: it is now recorded as 72 ppi.
Several potential problems with compiler-specific behaviour have been identified using the ‘Undefined Behaviour Sanitizer’ in conjunction with the clang compiler.
hcl() now honours NA inputs (previously they were mapped to black).
Some translations in base packages were being looked up in the main catalog rather than that for the package.
As a result of the 3.0.2 change about ‘the last second before the epoch’, most conversions which should have given NA returned that time. (The platforms affected include Linux and OS X, but not Windows nor Solaris.)
rowsum() has more support for matrices and dataframes with 2^31 or more elements. (PR#15587)
predict(, interval = "confidence", scale = ) now works. (PR#15564)
The bug fix in 3.0.2 for PR#15411 was too aggressive, and sometimes removed spaces that should not have been removed. (PR#15583)
Running R code in a tcltk callback failed to set the busy flag, which will be needed to tell OS X not to ‘App Nap’.
The code for date-times before 1902 assumed that the offset from GMT in 1902 was a whole number of minutes: that was not true of Paris (as recorded on some platforms).
Using Sys.setlocale to set LC_NUMERIC to "C" (to restore the sane behavior) no longer gives a warning.
deparse() now deparses complex vectors in a way that re-parses to the original values. (PR#15534, patch based on code submitted by Alex Bertram.)
In some extreme cases (more than 10^15) integer inputs to dpqrxxx() functions might have been rounded up by one (with a warning about being non-integer). (PR#15624)
Plotting symbol pch = 14 had the triangle upside down on some devices (typically screen devices). The triangle is supposed to be point up. (Reported by Bill Venables.)
getSrcref() did not work on method definitions if rematchDefinition() had been used.
KalmanForecast(fast = FALSE) reported a (harmless) stack imbalance.
The count of observations used by KalmanRun() did not take missing values into account.
In locales where the abbreviated name of one month is a partial match for the full name of a later one, the %B format in strptime() could fail. An example was French on OS X, where juin is abbreviated to jui and partially matches juillet. Similarly for weekday names.
pbeta(x, a, b, log.p = TRUE) sometimes underflowed to zero for very small and very differently sized a, b. (PR#15641)
approx() and approxfun() now handle infinite values with the "constant" method. (PR#15655)
stripchart() again respects reversed limits in xlim and ylim. (PR#15664)

New in R for Windows 3.0.2 (Oct 1, 2013)

NEW FEATURES:
The ‘NEWS’ files have been re-organized.
This file contains news for R >= 3.0.0: news for the 0.x.y, 1.x.y and 2.x.y releases is in files ‘NEWS.0’, ‘NEWS.1’ and ‘NEWS.2’. The latter files are now installed when R is installed. An HTML version of news from 2.10.0 to 2.15.3 is available as ‘doc/html/NEWS.2.html’.
sum() for integer arguments now uses an integer accumulator of at least 64 bits and so will be more accurate in the very rare case that a cumulative sum exceeds 2^53 (necessarily summing more than 4 million elements).
The example() and tools::Rd2ex() functions now have parameters to allow them to ignore \dontrun markup in examples. (Suggested by Peter Solymos.)
str(x) is considerably faster for very large lists, or factors with 100,000 levels, the latter as in PR#15337.
col2rgb() now converts factors to character strings not integer codes (suggested by Bryan Hanson).
tail(warnings()) now works, via the new `[` method.
There is now support for the LaTeX style file ‘zi4.sty’ which has in some distributions replaced ‘inconsolata.sty’.
unlist(x) now typically returns all non-list xs unchanged, not just the “vector” ones. Consequently, format(lst) now also works when the list lst has non-vector elements.
The tools::getVignetteInfo() function has been added to give information about installed vignettes.
New assertCondition(), etc. utilities in tools, useful for testing.
Profiling now records non-inlined calls from byte-compiled code to BUILTIN functions.
Various functions in stats and elsewhere that use non-standard evaluation are now more careful to follow the namespace scoping rules. E.g. stats::lm() can now find stats::model.frame() even if stats is not on the search path or if some package defines a function of that name.
If an invalid/corrupt .Random.seed object is encountered in the workspace it is ignored with a warning rather than giving an error. (This allows R itself to rely on a working RNG, e.g. to choose a random port.)
seq() and seq.int() give more explicit error messages if called with invalid (e.g. NaN) inputs.
When parse() finds a syntax error, it now makes partial parse information available up to the location of the error. (Request of Reijo Sund.)
Methods invoked by NextMethod() had a different dynamic parent to the generic. This was causing trouble where S3 methods invoked via lazy evaluation could lose track of their generic. (PR#15267)
Code for the negative binomial distribution now treats the case size == 0 as a one-point distribution at zero.
abbreviate() handles without warning non-ASCII input strings which require no abbreviation.
read.dcf() no longer has a limit of 8191 bytes per line. (Wish of PR#15250.)
formatC(x) no longer copies the class of x to the result, to avoid misuse creating invalid objects as in PR#15303. A warning is given if a class is discarded.
Dataset npk has been copied from MASS to allow more tests to be run without recommended packages being installed.
The initialization of the regression coefficients for non-degenerate differenced models in arima() has been changed and in some examples avoids a local maximum. (PR#15396)
termplot() now has an argument transform.x to control the display of individual terms in the plot. (PR#15329)
format() now supports digits = 0, to display nsmall decimal places.
There is a new read-only par() parameter called "page", which returns a logical value indicating whether the next plot.new() call will start a new page.
Processing Sweave and Rd documents to PDF now renders backticks and single quotes better in several instances, including in \code and \samp expressions.
utils::modifyList() gets a new argument keep.null allowing NULL components in the replacement to be retained, instead of causing corresponding components to be deleted.
tools::pkgVignettes() gains argument check; if set to TRUE, it will warn when it appears a vignette requests a non-existent vignette engine.
UTILITIES
R CMD check --as-cran checks the line widths in usage and examples sections of the package Rd files.
R CMD check --as-cran now implies --timings.
R CMD check looks for command gfile if a suitable file is not found. (Although file is not from GNU, OpenCSW on Solaris installs it as gfile.)
R CMD build (with the internal tar) checks the permissions of ‘configure’ and ‘cleanup’ files and adds execute permission to the recorded permissions for these files if needed, with a warning. This is useful on OSes and file systems which do not support execute permissions (notably, on Windows).
R CMD build now weaves and tangles all vignettes, so suggested packages are not required during package installation if the source tarball was prepared with current R CMD build.
checkFF() (used by R CMD check) does a better job of detecting calls from other packages, including not reporting those where a function has been copied from another namespace (e.g. as a default method). It now reports calls where .NAME is a symbol registered in another package.
On Unix-alike systems, R CMD INSTALL now installs packages group writably whenever the library (lib.loc) is group writable. Hence, update.packages() works for other group members (suggested originally and from a patch by Dirk Eddelbuettel).
R CMD javareconf now supports the use of symbolic links for JAVA_HOME on platforms which have realpath. So it is now possible to use
R CMD javareconf JAVA_HOME=/usr/lib/jvm/java-1.7.0
on a Linux system and record that value rather than the frequently-changing full path such as ‘/usr/lib/jvm/java-1.7.0-openjdk-1.7.0.25.x86_64’.
(Windows only.) Rscript -e requires a non-empty argument for consistency with Unix versions of R. (Also Rterm -e and R -e.)
R CMD check does more thorough checking of declared packages and namespaces. It reports
packages declared in more than one of the Depends, Imports, Suggests and Enhances fields of the ‘DESCRIPTION’ file.
namespaces declared in Imports but not imported from, neither in the ‘NAMESPACE’ file nor using the :: nor ::: operators.
packages which are used in library() or requires() calls in the R code but were already put on the search path via Depends.
packages declared in Depends not imported via the ‘NAMESPACE’ file (except the standard packages). Objects used from Depends packages should be imported to avoid conflicts and to allow correct operation when the namespace is loaded but not attached.
objects imported via ::: calls where :: would do.
objects imported by :: which are not exported.
objects imported by ::: calls which do not exist.
See ‘Writing R Extensions’ for good practice.
R CMD check optionally checks for non-standard top-level files and directories (which are often mistakes): this is enabled for --as-cran.
LaTeX style file upquote.sty is no longer included (the version was several years old): it is no longer used in R. A much later version is commonly included in LaTeX distributions but does not play well with the ae fonts which are the default for Sweave vignettes.
R CMD build makes more use of the ‘build’ sub-directory of package sources, for example to record information about the vignettes.
INSTALLATION and INCLUDED SOFTWARE
The macros used for the texinfo manuals have been changed to work better with the incompatible changes made in texinfo 5.x.
The minimum version for a system xz library is now 5.0.3 (was 4.999). This is in part to avoid 5.0.2, which can compress in ways other versions cannot decompress.
The included version of PCRE has been updated to 8.33.
The included version of zlib has been updated to 1.2.8, a bug-fix release.
The included version of xz utils's liblzma has been updated to 5.0.5.
Since javareconf (see above) is used when R is installed, a stable link for JAVA_HOME can be supplied then.
Configuring with --disable-byte-compilation will override the ‘DESCRIPTION’ files of recommended packages, which typically require byte-compilation.
More of the installation and checking process will work even when TMPDIR is set to a path containing spaces, but this is not recommended and external software (such as texi2dvi) may fail.
PACKAGE INSTALLATION:
Installation is aborted immediately if a LinkingTo package is not installed.
R CMD INSTALL has a new option --no-byte-compile which will override a ByteCompile field in the package's ‘DESCRIPTION’ file.
License BSD is deprecated: use BSD_3_clause or BSD_2_clause instead.
License X11 is deprecated: use MIT or BSD_2_clause instead.
Version requirements for LinkingTo packages are now recognized: they are checked at installation. (Fields with version requirements were previously silently ignored.)
The limit of 500 S3method entries in a NAMESPACE file has been removed.
The default ‘version’ of Bioconductor for its packages has been changed to the upcoming 2.13, but this can be set by the environment variable R_BIOC_VERSION, e.g. in file ‘Renviron.site’.
C-LEVEL FACILITIES:
‘Rdefines.h’ has been tweaked so it can be included in C++ code after ‘R_ext/Boolean.h’ (which is included by ‘R.h’).
Note that ‘Rdefines.h’ is not kept up-to-date, and ‘Rinternals.h’ is preferred for new code.
eval and applyClosure are now protected against package code supplying an invalid rho.
DEPRECATED AND DEFUNCT:
The unused namespace argument to package.skeleton() is now formally deprecated and will be removed in R 3.1.0.
plclust() is deprecated: use the plot() method for class "hclust" instead.
Functions readNEWS() and checkNEWS() in package tools are deprecated (and they have not worked with current ‘NEWS’ files for a long time).
DOCUMENTATION:
‘An Introduction to R’ has a new chapter on using R as a scripting language including interacting with the OS.
BUG FIXES:
help.request() could not determine the current version of R on CRAN. (PR#15241)
On Windows, file.info() failed on root directories unless the path was terminated with an explicit ".". (PR#15302)
The regmatches 1. (Bill Dunlap, PR#15386).
stem() now discards infinite inputs rather than hanging. (PR#15376)
The parser now enforces C99 syntax for floating point hexadecimal constants (e.g. 0x1.1p0), rather than returning unintended values for malformed constants. (PR#15234)
model.matrix() now works with very long LHS names (more than 500 bytes). (PR#15377)
integrate() reverts to the pre-2.12.0 behaviour: from 2.12.0 to 3.0.1 it sometimes failed to achieve the requested tolerance and reported error estimates that were exceeded. (PR#15219)
strptime() now handles %W fields with value 0. (PR#15915)
R is now better protected against people trying to interact with the console in startup code. (PR#15325)
Subsetting 1D arrays often lost dimnames (PR#15301).
Unary + on a logical vector did not coerce to integer, although unary - did.
na.omit() and na.exclude() added a row to a zero-row data frame. (PR#15399)
All the (where necessary cut-down) vignettes are installed if R was configured with --without-recommended-packages.
source() did not display filenames when reporting syntax errors.
Syntax error reports misplaced the caret pointing out the bad token.
(Windows only) Starting R with R (instead of Rterm or Rgui) would lose any zero-length strings from the command line arguments. (PR#15406)
Errors in the encoding specified on the command line via --encoding=foo were not handled properly. (PR#15405)
If x is a symbol, is.vector(x, "name") now returns TRUE, since "name" and "symbol" should be synonyms. (Reported by Hervé Pagès.)
R CMD rtags works on platforms (such as OS X) with a XSI-conformant shell command echo. (PR#15231)
is.unsorted(NA) returns false as documented (rather than NA).
R CMD LINK did not know about sub-architectures.
system() and system2() are better protected against users who misguidedly have spaces in the temporary directory path.
file.show() and edit() are now more likely to work on file paths containing spaces. (Where external utilities are used, not the norm on Windows nor in R.app which should previously have worked.)
Packages using the methods package are more likely to work when they import it but it is not attached. (Several parts of its C code were looking for its R functions on the search path rather than in its namespace.)
lgamma(-x) is no longer NaN for very small x.
(Windows) system2() now respects specifying stdout and stderr as files if called from Rgui. (PR#15393)
Closing an x11() device whilst locator() or identify() is in progress no longer hangs R. (PR#15253)
list.dirs(full.names = FALSE) was not implemented. (PR#15170)
format() sometimes added unnecessary spaces. (PR#15411)
all.equal(check.names = FALSE) would ignore the request to ignore the names and would check them as attributes.
The symbol set by tools::Rd2txt_options(itemBullet=) was not respected in some locales. (PR#15435)
mcMap() was not exported by package parallel. (PR#15439)
plot() for TukeyHSD objects did not balance dev.hold() and dev.flush() calls on multi-page plots. (PR#15449)

New in R for Windows 3.0.1 (May 22, 2013)

NEW FEATURES:
chooseCRANmirror() and chooseBioCmirror() gain an ind argument (like setRepositories()).
mcparallel has a new argument mc.interactive which can modify the interactive flag in the child process. The new default is FALSE which makes child processes non-interactive by default (this prevents lock-ups due to children waiting for interactive input).
scan() now warns when end-of-file occurs within a quoted string.
count.fields() is now consistent with scan() in its handling of newlines in quoted strings. Instead of triggering an error, this results in the current line receiving NA as the field count, with the next line getting the total count of the two lines.
The default method of image() will plot axes of the class of xlim and ylim (and hence of x and y if there is a suitable range() method). Based on a suggestion of Michael Sumner.
load() now has a verbose argument for debugging support, to print the names of objects just before loading them.
When loading a serialized object encounters a reference to a namespace which cannot be loaded, this is replaced by a reference to the global environment, with a warning.
pairs() gains a line.main option for title placement.
The remaining instances in which serialization to a raw vector was limited to 2GB have been unlimited on a 64-bit platform, and in most cases serialization to a vector of more than 1GB will be substantially faster.
UTILITIES:
R CMD config now make use of personal ‘Makevars’ files under ‘~/.R’ and a site file ‘Makevars.site’, in the same way as R CMD SHLIB and R CMD INSTALL. This makes the utility more useful in package configure scripts.
On Windows finding the personal files may require the environment variable HOME set.
The old behaviour can be obtained with the new options --no-user-files and --no-site-files.
PACKAGE INSTALLATION:
Alternatives to the site and user customization files ‘Makevars.site’ and ‘~/.R/Makevars’ can be specified via the environment variables R_MAKEVARS_SITE and R_MAKEVARS_USER respectively. These can be used to suppress the use of the default files by setting an empty value (where possible) or a non-existent path.
BUG FIXES:
sys.source() did not report error locations when keep.source = TRUE.
as.POSIXct.numeric was coercing origin using the tz argument and not "GMT" as documented
The active binding to assign fields in reference classes has been cleaned up to reduce dependence on the class' package environment, also fixing bug in initializing read-only fields
str(d) no longer gives an error when names(d) contain illegal multibyte strings
Profiling of built-in functions with line.profiling= TRUE did not record the line from which they were called.
citation(pkg) dropped the header and footer specified in the ‘CITATION’ file
Quotes were handled differently when reading the first line and reading the rest, so read.table() misread some files that contained quote characters
cat() with sep a character vector of length greater than one and more than one argument was using separators inconsistently
On Windows in R 3.0.0, savePlot() failed because of an incorrect check on the argument count.
unzip(list = TRUE) returned Names as a factor and not a character vector (as documented) for the internal method.
contourLines() now checks more comprehensively for conformance of its x, y and z arguments (it was used incorrectly in package R2G2).
Saved graphics display lists are R version-specific. Attempting to load workspaces containing them (or some other version-specific objects) aborted the load in R 3.0.0 and earlier; now it does a partial load and generates a warning instead.
In R 3.0.0, identify() and locator() did not record information correctly, so replaying a graph (e.g. by copying it to another device) would fail.
Calling file.copy() or dirname() with the invalid input "" (which was being used in packages, despite not being a file path) could have caused a segfault.
dirname("") is now "" rather than "." (unless it segfaulted).
supsmu() could read/write outside its input vectors for very short inputs (seen in package rms for n = 4).
as.dendrogram()'s hclust method uses less memory and hence gets considerably faster for large (n ~ 1000) clusterings
The return value when all workers failed from parallel::mclapply(mc.presechedule = TRUE) was a list of strings and not of error objects.
In R 3.0.0, when help() found multiple pages with the same alias, the HTML display of all the selections was not produced.
splinefun(method="monoH.FC") now produces a function with first argument named x and allows deriv=3, as documented.
summaryRprof() would only read the first chunksize lines of an Rprof file produced with line.profiling=TRUE. By default, this is the first 100 seconds.
lsfit() produced an incorrect error message when argument x had more columns than rows or x had a different number of rows than y.
Binary operations on equal length vectors copied the class name from the second operand when the first had no class name, but did not set the object bit.
The trace() method for reference generator objects failed after those objects became function definitions.
write.table() did not check that factors were constructed correctly, and so caused a segment fault when writing bad ones.
The internal HTTP server no longer chokes on POST requests without body. It will also pass-through other request types for custom handlers (with the method stored in Request-Method header) instead of failing.

New in R for Windows 3.0.0 (Apr 12, 2013)

SIGNIFICANT USER-VISIBLE CHANGES:
Packages need to be (re-)installed under this version (3.0.0) of R.
There is a subtle change in behaviour for numeric index values 2^31 and larger. These never used to be legitimate and so were treated as NA, sometimes with a warning. They are now legal for long vectors so there is no longer a warning, and x[2^31] = 3.0.0.
Palette changes get recorded on the display list, so replaying plots (including when resizing screen devices and using dev.copy()) will work better when the palette is changed during a plot.
chol(pivot = TRUE) now defaults to LAPACK, not LINPACK.
The parse() function has a new parameter keep.source, which defaults to options("keep.source").
Profiling via Rprof() now optionally records information at the statement level, not just the function level.
The Rprof() function now quotes function names in in its output file on Windows, to be consistent with the quoting in Unix.
Profiling via Rprof() now optionally records information about time spent in GC
The HTML help page for a package now displays non-vignette documentation files in a more accessible format.
To support options(stringsAsFactors = FALSE), model.frame(), model.matrix() and replications() now automatically convert character vectors to factors without a warning.
The print method for objects of class "table" now detects tables with 0-extents and prints the results as, e.g., < table of extent 0 x 1 x 2 >. (Wish of PR#15198.)
Deparsing involving calls to anonymous functions and has been made closer to reversible by the addition of extra parentheses.
The function utils::packageName() has been added as a lightweight version of methods::g
find.package(lib.loc = NULL) now treats loaded namespaces preferentially in the same way as attached packages have been for a long time.
In Windows, the Change Directory dialog now defaults to the current working directory, rather than to the last directory chosen in that dialog.
available.packages() gains a "license/restricts_use" filter which retains only packages for which installation can proceed solely based on packages which are guaranteed not to restrict use.
New check_packages_in_dir() function in package tools for conveniently checking source packages along with their reverse dependencies.
R's completion mechanism has been improved to handle help requests (starting with a question mark). In particular, help prefixes are now supported, as well as quoted help topics. To support this, completion inside quotes are now handled by R by default on all platforms.
The memory manager now allows the strategy used to balance garbage collection and memory growth to be controlled by setting the environment variable R_GC_MEM_GROW. See ?Memory for more details.
For experts only’, as the introductory manual says.) The use of environment variables R_NSIZE and R_VSIZE to control the initial (= minimum) garbage collection trigger for number of cons cels and size of heap has been restored: they can be overridden by the command-line options --min-nsize and --min-vsize; see ?Memory.
On Windows, the device name for bitmap devices as reported by .Device and .Devices no longer includes the file name. This is for consistency with other platforms and was requested by the lattice maintainer.
win.metafile() still uses the file name: the exact form is used by package tkrplot.
set.seed(NULL) re-initializes .Random.seed as done at the beginning of the session if not already set. (Suggestion of Bill Dunlap.)
The breaks argument in hist.default() can now be a function that returns the breakpoints to be used (previously it could only return the suggested number of breakpoints).
File ‘share/licenses/licenses.db’ has some clarifications, especially as to which variants of ‘BSD’ and ‘MIT’ is intended and how to apply them to packages. The problematic licence ‘Artistic-1.0’ has been removed.
LONG VECTORS:
This section applies only to 64-bit platforms.
There is support for vectors longer than 2^31 - 1 elements. This applies to raw, logical, integer, double, complex and character vectors, as well as lists. (Elements of character vectors remain limited to 2^31 - 1 bytes.)
Most operations which can sensibly be done with long vectors work: others may return the error ‘long vectors not supported yet’. Most of these are because they explicitly work with integer indices (e.g. anyDuplicated() and match()) or because other limits (e.g. of character strings or matrix dimensions) would be exceeded or the operations would be extremely slow.
length() returns a double for long vectors, and lengths can be set to 2^31 or more by the replacement function with a double value.
Most aspects of indexing are available. Generally double-valued indices can be used to access elements beyond 2^31 - 1
There is some support for matrices and arrays with each dimension less than 2^31 but total number of elements more than that. Only some aspects of matrix algebra work for such matrices, often taking a very long time. In other cases the underlying Fortran code has an unstated restriction (as was found for complex svd()).
PERFORMANCE IMPROVEMENTS:
More use has been made of R objects representing registered entry points, which is more efficient as the address is provided by the loader once only when the package is loaded.
This has been done for packages base, methods, splines and tcltk: it was already in place for the other standard packages.
Since these entry points are always accessed by the R entry points they do not need to be in the load table which can be substantially smaller and hence searched faster. This does mean that .C / .Fortran / .Call calls copied from earlier versions of R may no longer work – but they were never part of the API.
Many .Call() calls in package base have been migrated to .Internal() calls.
solve() makes fewer copies, especially when b is a vector rather than a matrix.
eigen() makes fewer copies if the input has dimnames.
Most of the linear algebra functions make fewer copies when the input(s) are not double (e.g. integer or logical).
A foreign function call (.C() etc) in a package without a PACKAGE argument will only look in the first DLL specified in the ‘NAMESPACE’ file of the package rather than searching all loaded DLLs. A few packages needed PACKAGE arguments added.
The @

New in R for Windows 2.15.3 (Apr 12, 2013)

NEW FEATURES:
lgamma(x) for very small x (in the denormalized range) is no longer Inf with a warning.
image() now sorts an unsorted breaks vector, with a warning.
The internal methods for tar() and untar() do a slightly more general job for ‘ustar’-style handling of paths of more than 100 bytes.
Packages compiler and parallel have been added to the reference index (‘refman.pdf’).
untar(tar = "internal") has some support for pax headers as produced by e.g. gnutar --posix (which seems prevalent on OpenSUSE 12.2) or bsdtar --format pax, including long path and link names.
sQuote() and dQuote() now handle 0-length inputs. (Suggestion of Ben Bolker.)
summaryRprof() returns zero-row data frames rather than throw an error if no events are recorded, for consistency.
The included version of PCRE has been updated to 8.32.
The tcltk namespace can now be re-loaded after unloading.
The Tcl/Tk event loop is inhibited in a forked child from package parallel (as in e.g. mclapply()).
parallel::makeCluster() recognizes the value random for the environment variable R_PARALLEL_PORT: this chooses a random value for the port and reduces the chance of conflicts when multiple users start a cluster at the same time.
UTILITIES:
The default for TAR on Windows for R CMD build has been changed to be internal if no tar command is on the path.
This enables most packages to be built ‘out of the box’ without Rtools: the main exceptions are those which need to be installed to re-build vignettes and need Rtools for installation (usually because they contain compiled code).
C-LEVEL FACILITIES:
On a 64-bit Windows platform with enough RAM, R_alloc can now allocate up to just under 32GB like other 64-bit platforms.
DEPRECATED AND DEFUNCT:
Use of col2rgb(0) is deprecated (see the help page for its limitations).
The deprecated intensities component returned by hist() is no longer recognized by the plot() method and will be removed in R 3.0.0.
real(), as.real() and is.real() are now formally deprecated and give a warning.
This is formal notice that the non-API EISPACK entry points in R will be removed shortly.
INSTALLATION:
The configure tests for Objective C and Objective C++ now work on Mac OS 10.8 with Xcode 4.5.2 (PR#15107).
The cairo-based versions of X11() now work with current versions of cairographics (e.g. 1.12.10). (PR#15168)
A workaround for earlier versions of R is to use X11.options(type = "nbcairo").
Configuration and R CMD javareconf now come up with a smaller set of library paths for Java on Oracle-format JDK (including OpenJDK). This helps avoid conflicts between libraries (such as libjpeg) supplied in the JDK and system libraries. This can always be overridden if needed: see the ‘R Installation and Administration’ manual.
BUG FIXES:
beta(a, b) could overflow to infinity in its calculations when one of a and b was less than one. (PR#15075)
lbeta(a, b) no longer gives NaN if a or b is very small (in the denormalized range).
bquote() is now able to substitute default arguments in single-argument functions. (PR#15077)
browseEnv(html = FALSE) would segfault if called from R (not R.app) on a CRAN-style Mac OS X build of R.
for lists (generic vectors) needed to increment NAMED count when RHS is used more than once. (PR#15098)
On Windows, warnings about opening a file or pipe with a non-ASCII description were sometimes output in UTF-8 rather than in the current locale's character set.
The call() function did not duplicate its arguments. (PR#15115)
TukeyHSD() could give NA results with some na.action methods such as na.exclude(). (Hinted at on R-help by John Fox.)
The deprecated svd(X, LINPACK = TRUE) could alter X in R 2.15.[12]. (Reported by Bill Dunlap.)
Under Windows, file.link() and file.symlink() used the link name twice, so would always fail. (Reported by Rui Barradas/Oliver Soong).
summaryRprof(memory = "both") mixed up the units of Vcells and Ncells: it now works in bytes. (PR#15138)
tools::Rd2HTML() would sometimes delete text. (PR#15134)
plot() failed for "table" objects containing just one entry. (PR#15118)
embedFonts() needed to quote some filepaths. (PR#15149)
parallel::mccollect() handled NULL returns incorrectly (removing the element rather than setting it to NULL).
The full reference index (‘fullrefman.pdf’) was missing packages compiler and parallel.
The report for optim(method = "L-BFGS-B", control = list(trace = 1)) reported the last completed and not the current iteration, unlike other methods and trace levels. (PR#15103)
qt(1e-12, 1.2) no longer gives NaN.
dt(1e160, 1.2, log=TRUE) no longer gives -Inf.
On Windows the untar() function now quotes the directory name when using an external tar utility, so R CMD check will handle pathnames containing spaces.
The version for Windows 8 and Windows Server 2012 is now displayed by win.version(). (Reported by Gabor Grothendieck.)
The custom Windows installer target myR in the installer ‘Makefile’ did not work in 2.15.2. (Reported by Erich Neuwirth.)
aperm(matrix(1:6, 2, dimnames=list(A={}, B={})), "A") no longer segfaults.
Expressions involving user defined operators were not always deparsed faithfully. (PR#15179)
The enc2utf8() function converted NA_character_ to "NA" in non-UTF-8 locales. (PR#15201)
The exclude argument to xtabs() was ignored for "factor" arguments.
On Windows, work around an event-timing problem when the RGui console was closed from the ‘X’ control and the closure cancelled. (This would on some 64-bit systems crash R, typically those with a slow GPU relative to the CPU.)
On unix Rscript will pass the r_arch setting it was compiled with on to the R process so that the architecture of Rscript and that of R will match unless overridden.
On Windows, basename(), dirname() and file.choose() have more support for long non-ASCII file names with 260 or more bytes when expressed in UTF-8.

New in R for Windows 2.15.2 (Oct 30, 2012)

NEW FEATURES:
The X11() window gains an icon: the latter may be especially useful on Ubuntu's ‘Unity’ interface.
The WM_CLASS should be set in circumstances where the Window Manager failed to make use of X11 resource settings.
The "Date" and "POSIXt" methods for cut() will accept an unsorted breaks argument (as the default method does, although this was undocumented).
Reference class methods (in the methods package) that use other methods in an indirect way (e.g., by sapply()) must tell the code analysis to include that method. They can now do so by invoking $usingMethods().
More Polish translations are available: for the RGui menus and for several recommended packages.
Multistratum MANOVA works. In fact, it seems to have done so for years in spite of the help page claiming it did not.
qqline() has new optional arguments distribution, probs and qtype, following the example of lattice's panel.qqmathline().
The handling of single quotes in the en@quot pseudo-language has been slightly improved. Double quotes are no longer converted.
New functions checkPoFiles() and checkPoFile() have been added to the tools package to check for consistency of format strings in translation files.
model.matrix(~1, ...) now also contains the same rownames that less trivial formulae produce.
Misuse of rep() on undocumented types of objects (e.g. calls) is now reported as an error.
The included LAPACK has been updated to 3.4.1, with some patches from the current SVN sources.
file.copy(recursive = TRUE) has some additional checks on user error leading to attempted infinite recursion (and on some platforms to crashing R).
PCRE has been updated to version 8.31, a bug-fix release.
The included version of liblzma has been updated to version 5.0.4, a minor bug-fix release.
New function .bincode(), a ‘bare-bones’ version of cut.default(labels = FALSE) for use in packages with image() methods.
The HTML manuals now use directional single quotes.
maintainer() now converts embedded new lines to spaces. It no longer gives a non-obvious error for non-installed packages.
The X11() device has some protection against being used with forked processes via package parallel.
Setting the environment variable R_OSX_VALGRIND (to any value) allows R to be run under valgrind on Mac OS 10.6 and 10.7 (valgrind currently has very limited support for 10.8), provided system() is not used (directly or indirectly). This should not be needed for valgrind >= 3.8.1.
The model.frame() method for "lm" uses xlevels: this is safer if data was supplied or model = FALSE was used and the levels of factors used in the fit had been re-ordered since fitting.
Similarly, model.frame(fm, data=) copies across the variables used for safe prediction from the fit.
Functions such as parLapply() in package parallel can make use of a default cluster if one is set. (Reported by Martin Morgan.)
chol(pivot = TRUE, LINPACK = FALSE) is now available using LAPACK 3.2 subroutine DPSTRF.
The functions .C(), .Call(), .External() and .Fortran() now check that they are called with an unnamed first argument: the formal arguments were changed from name= to .NAME= in R 2.13.0, but some packages were still using the old name. This is currently a warning, but will be an error in future.
step() no longer tries to improve a model with AIC of -Inf (a perfect fit).
spline() and splinefun() gain a new method "hyman", an implementation of Hyman's method of constructing monotonic interpolation splines.
On Windows, the C stack size has been increased to 64MB (it has been 10MB since the days of 32MB RAM systems).
PERFORMANCE IMPROVEMENTS:
array() is now implemented in C code (for speed) when data is atomic or an unclassed list (so it is known that as.vector(data) will have no class to be used by rep()).
rep() is faster and uses less memory, substantially so in some common cases (e.g. if times is of length one or length.out is given, and each = 1).
findInterval(), tabulate(), cut(), hist() and image.default() all use .Call() and are more efficient.
duplicated(), unique() and similar now support vectors of lengths above 2^29 on 64-bit platforms.
Omitting PACKAGE in .C() etc calls was supposed to make use of the DLL from the namespace within which the enclosing function was defined. It was less successful in doing so than it might be, and gave no indication it had failed.
A new search strategy is very successful and gives a warning when it fails. In most cases this is because the entry point is not actually provided by that package (and so PACKAGE should be used to indicate which package is intended) but in some the namespace does not have a DLL specified by a useDynLib() directive so PACKAGE is required.
UTILITIES:
R CMD check now checks if a package can be loaded by library(pkgname, lib.loc = "somewhere") without being on the library search path (unless it is already installed in .Library, when it always will be).
R CMD check --as-cran notes ‘hidden’ files and directories (with names starting with a dot) that are not needed for the operation of R CMD INSTALL or R CMD build: such files should be excluded from the published tarball.
R CMD check (if checking subdirectories) checks that the R code in any demos is ASCII and can be parsed, and warns if not.
When R CMD Rd2pdf is used with ‘inputenx.sty’, it allows further characters (mainly for Eastern European languages) by including ‘ix-utf8enc.dfu’ (if available).
R CMD build now omits several types of hidden files/directories, including ‘inst/doc/.Rinstignore’, ‘vignettes/.Rinstignore’, (‘.Rinstignore’ should be at top level), ‘.deps’ under ‘src’, ‘.Renviron’, ‘.Rprofile’, ‘.Rproj.user’, ‘.backups’, ‘.cvsignore’, ‘.cproject’, ‘.directory’, ‘.dropbox’, ‘.exrc’, ‘.gdb.history’, ‘.gitattributes’, ‘.gitignore’, ‘.gitmodules’, ‘.hgignore’, ‘.hgtags’, ‘.htaccess’, ‘.latex2html-init’, ‘.project’, ‘.seed’, ‘.settings’, ‘.tm_properties’ and various leftovers.
R CMD check now checks for .C(), .Call(), .External() and .Fortran() calls in other packages, and gives a warning on those found from R itself (which are not part of the API and change without notice: many will changed for R 2.16.0).
C-LEVEL FACILITIES:
The limit for R_alloc on 64-bit platforms has been raised to just under 32GB (from just under 16GB).
The misuse of .C("name", ..., PACKAGE = foo) where foo is an arbitrary R object is now an error.
The misuse .C("name",..., PACKAGE = "") is now warned about in R CMD check, and will be an error in future.
DEPRECATED AND DEFUNCT:
Use of array() with a 0-length dim argument is deprecated with a warning (and was contrary to the documentation).
Use of tapply() with a 0-length INDEX list is deprecated with a warning.
Translation packages are deprecated.
Calling rep() or rep.int() on a pairlist is deprecated and will give a warning. In any case, rep() converted a pairlist to a list so you may as well do that explicitly.
Entry point rcont2 is no longer part of the API, and will move to package stats in R 2.16.0.
The ‘internal’ graphics device invoked by .Call("R_GD_nullDevice", package = "grDevices") is about to be removed: use pdf(file = NULL) instead.
eigen(EISPACK = TRUE), chol(pivot = FALSE, LINPACK = TRUE), chol2inv(LINPACK = TRUE), solve(LINPACK = TRUE) and svd(LINPACK = TRUE) are deprecated and give a warning.
They were provided for compatibility with R 1.7.0 (Mar 2003)!
The ‘internal function’ kappa.tri() has been renamed to .kappa_tri() so it is not inadvertently called as a method for class "tri".
Functions sessionData() and browseAll() in package methods are on a help page describing them as ‘deprecated’ and are now formally deprecated.
PACKAGE INSTALLATION:
For a Windows or Mac OS X binary package install, install.packages() will check if a source package is available on the same repositories, and report if it is a later version or there is a source package but no binary package available.
This check can be suppressed: see the help page.
install.packages(type = "both") has been enhanced. In interactive use it will ask whether to choose the source version of a package if the binary version is older and contains compiled code, and also asks if source packages with no binary version should be installed).
INSTALLATION:
There is a new configure option --with-libtiff (mainly in case the system installation needs to be avoided).
LAPACK 3.4.1 does use some Fortran 90 features, so g77 no longer suffices.
If an external LAPACK is used, it must be version 3.2 or later.
BUG FIXES:
On Windows, starting Rterm via R.exe caused Ctrl-C to misbehave.
The tools::latexToUtf8() function missed conversions that were contained within braces.
Long timezone specifications (such as a file name preceded by :) could crash as.POSIXlt.
R CMD build --resave-data could fail if there was no ‘data’ directory but there was an ‘R/sysdata.rda’ file.
is.na() misbehaved on a 0-column data frame.
anova.lmlist() failed if test was supplied.
It was unable to compute Cp tests for object of class "lm" (it assumed class "glm").
The formula method for sunflowerplot() now allows xlab and ylab to be set.
The "POSIXt" and "Date" methods for hist() could fail on Windows where adjustments to the right-hand boundary crossed a DST transition time.
On Windows, the code in as.POSIXct() to handle incorrectly specified isdst fields might have resulted in NA being returned.
aov() and manova() gave spurious warning about singular error model in the multiresponse case.
In ns() and bs(), specifying knots = NULL is now equivalent to omitting it, also when df is specified.
sprintf() did not accept numbered arguments ending in zero.
rWishart() could overflow the C stack and maybe crash the R process for dimensions of several hundreds or more.
Base package vignettes (e.g. vignette("Sweave")) were not fully installed in builds of R from the tarball.
lchoose() and choose() could overflow the C stack and crash R.
When given a 0-byte file and asked to keep source references, parse() read input from stdin() instead.
pdf(compress = TRUE) did not delete temporary files it created until the end of the R session.
logLik() did not detect the error of applying it to a multiple-response linear model.
file.copy(recursive = TRUE) did not always report FALSE for a failure two or more directories deep.
qgeom() could return -1 for extremely small q.
smooth.spline() used DUP = FALSE which allowed its compiled C code to change the function: this was masked by the default byte-compilation.
In Windows, the GUI preferences for foreground color were not always respected.
On OS X, the Quartz versions of the bitmap devices did not respect antialias = "none".
unique() and similar would infinite-loop if called on a vector of length > 2^29 (but reported that the vector was too long for 2^30 or more).
parallel::stopCluster() now works with MPI clusters without snow being on the search path.
terms.formula() could exhaust the stack, and the stack check did not always catch this before the segfault.
sort.list(method = "radix") could give incorrect results on certain compilers (seen with clang on Mac OS 10.7 and Xcode 4.4.1).
backsolve(T, b) gave incorrect results when nrows(b) > ncols(T) and b had more than one column.
It could segfault or give nonsense if k was specified as more than ncols(T).
smooth.spline() did not check that a specified numeric spar was of length 1, and gave corrupt results if it was of length 0.
Protection added to do_system.
Printing of vectors with names > 1000 characters now works correctly rather than truncating.
qr() for a complex matrix did not pivot the column names.
with-blas='-framework vecLib' now also works on OS X 10.8.
R CMD check no longer fails with an error if a ‘DESCRIPTION’ file incorrectly contains a blank line.
install.packages(type = "both") could call chooseCRANmirror() twice.
lm.wfit() could segfault in R 2.15.1 if all the weights were zero.
A malformed package name could cause R CMD INSTALL to write outside the target library.
Some of the quality control functions (e.g. tools::checkFF()) were wrongly identifying the source of S4 methods in a package and so not checking them.
The default type of display by browseEnv() when using R.app on Mac OS X has been incorrect for a long time.
The implementation of importMethodsFrom in a NAMESPACE file could be confused and fail to find generics when importing from multiple packages
The detection of the C stack direction is better protected against compiler optimization.
Long custom line types would sometimes segfault on the cairographics-based devices.
tools::checkPoFile() unprotected too early in its C code and so segfaulted from time to time.
The Fortran code underlying nlminb() could infinite-loop if any of the input functions returned NA or NaN. This is now an error for the gradient or Hessian, and a warning for the function (with the value replaced by Inf).
The code for creating coerce() methods could generate false notes about ambiguous selection; the notes have been suppressed for this function.
arima.sim() could give too long an output in some corner cases
anova.glm() with test = "Rao" didn't work when models included an offset.
as.data.frame.matrix() could return invalid data frame with no row.names attribute for 0-row matrix.
Compilation with the vecLib or Accelerate frameworks on OS X without using that also for LAPACK is more likely to be successful.

New in R for Windows 2.15.1 (Jul 2, 2012)

NEW FEATURES:
source() now uses withVisible() rather than .Internal(eval.with.vis). This sometimes alters tracebacks slightly.
install.packages("pkg_version.tgz") on Mac OS X now has sanity checks that this is actually a binary package (as people have tried it with incorrectly named source packages).
splineDesign() and spline.des() in package splines have a new option sparse which can be used for efficient construction of a sparse B-spline design matrix (via Matrix).
norm() now allows type = "2" (the ‘spectral’ or 2-norm) as well, mainly for didactical completeness.
pmin() and pmax()) now also work when one of the inputs is of length zero and others are not, returning a zero-length vector, analogously to, say, +.
colorRamp() (and hence colorRampPalette()) now also works for the boundary case of just one color when the ramp is flat.
qqline() has new optional arguments distribution, probs and qtype, following the example of lattice's panel.qqmathline().
C() gains some protection against the misuse of character vector arguments. (An all too common error is to pass character(N), which initializes the elements to "", and then attempt to edit the strings in-place, sometimes forgetting to terminate them.)
Calls to the new function globalVariables() in package utils declare that functions and other objects in a package should be treated as globally defined, so that CMD check will not note them.
print(packageDescription(*)) trims the Collate field by default.
The included copy of zlib has been updated to version 1.2.7.
A new option "show.error.locations" has been added. When set to TRUE, error messages will contain the location of the most recent call containing source reference information. (Other values are supported as well; see ?options.)
The NA warning messages from e.g. pchisq() now report the call to the closure and not that of the .Internal.
Added Polish translations
PERFORMANCE IMPROVEMENTS:
In package parallel, makeForkCluster() and the multicore-based functions use native byte-order for serialization (deferred from 2.15.0).
lm.fit(), lm.wfit(), glm.fit() and lsfit() do less copying of objects, mainly by using .Call() rather than .Fortran().
C() and .Fortran() do less copying: arguments which are raw, logical, integer, real or complex vectors and are unnamed are not copied before the call, and (named or not) are not copied after the call. Lists are no longer copied (they are supposed to be used read-only in the C code).
tabulate() makes use of .C(DUP = FALSE) and hence does not copy bin.
It also avoids making a copy of a factor argument bin.
Other functions (often or always) doing less copying include cut(), dist(), the complex case of eigen(), hclust(), image(), kmeans(), loess(), stl() and svd(LINPACK = TRUE).
There is less copying when using primitive replacement functions such as names(), attr() and attributes().
DEPRECATED AND DEFUNCT:
The converters for use with .C() (see ?getCConverterDescriptions) are deprecated: use the .Call() interface instead. There are no known examples (they were never fully documented).
UTILITIES:
For R CMD check, a few people have reported problems with junctions on Windows (although they were tested on Windows 7, XP and Server 2008 machines and it is unknown under what circumstances the problems occur). Setting the environment variable R_WIN_NO_JUNCTIONS to a non-empty value (e.g. in ‘~/.R/check.Renviron’) will force copies to be used instead.
INSTALLATION:
R CMD INSTALL with _R_CHECK_INSTALL_DEPENDS_ set to a true value (as done by R CMD check --as-cran) now restricts the packages available when lazy-loading as well as when test-loading (since packages such as ETLUtils and agsemisc had top-level calls to library() for undeclared packages).
This check is now also available on Windows.
C-LEVEL FACILITIES:
C entry points mkChar and mkCharCE now check that the length of the string they are passed does not exceed 2^31-1 bytes: they used to overflow with unpredictable consequences.
C entry points R_GetCurrentSrcref and R_GetSrcFilename have been added to the API to allow debuggers access to the source references on the stack.
WINDOWS-SPECIFIC CHANGES
Windows-specific changes will now be announced in this file (‘NEWS’). Changes up and including R 2.15.0 remain in the ‘CHANGES’ file.
There are two new environment variables which control the defaults for command-line options.
If R_WIN_INTERNET2 is set to a non-empty value, it is as if --internet2 was used.
If R_MAX_MEM_SIZE is set, it gives the default memory limit if --max-mem-size is not specified: invalid values being ignored.
BUG FIXES:
lsfit() lost the names from the residuals.
More cases in which merge() could create a data frame with duplicate column names now give warnings. Cases where names specified in by match multiple columns are errors.
Nonsense uses such as seq(1:50, by = 5) (from package plotrix) and seq.int(1:50, by = 5) are now errors.
The residuals in the 5-number summary printed by summary() on an "lm" object are now explicitly labelled as weighted residuals when non-constant weights are present. (Wish of PR#14840.)
tracemem() reported that all objects were copied by .C() or .Fortran() whereas only some object types were ever copied.
It also reported and marked as copies some transformations such as rexp(n, x): it no longer does so.
The plot() method for class "stepfun" only used the optional xval argument to compute xlim and not the points at which to plot (as documented). (PR#14864)
Names containing characters which need to be escaped were not deparsed properly. (PR#14846)
Trying to update (recommended) packages in ‘R_HOME/library’ without write access is now dealt with more gracefully. Further, such package updates may be skipped (with a warning), when a newer installed version is already going to be used from .libPaths(). (PR#14866)
hclust() is now fast again (as up to end of 2003), with a different fix for the "median"/"centroid" problem. (PR#4195).
get_all_vars() failed when the data came entirely from vectors in the global environment. (PR#14847)
R CMD check with _R_CHECK_NO_RECOMMENDED_ set to a true value (as done by the --as-cran option) could issue false errors if there was an indirect dependency on a recommended package.
formatC() uses the C entry point str_signif which could write beyond the length allocated for the output string.
Missing default argument added to implicit S4 generic for backsolve(). (PR#14883)
Some bugs have been fixed in handling load actions that could fail to export assigned items or generate spurious warnings in CMD check on loading.
For tiff(type = "windows"), the numbering of per-page files except the last was off by one.
On Windows, loading package stats (which is done for a default session) would switch line endings on ‘stdout’ and ‘stderr’ from CRLF to LF. This affected Rterm and R CMD BATCH.
On Windows, the compatibility function x11() had not kept up with changes to windows(), and issued warnings about bad parameters. (PR#14880)
On Windows, the Sys.glob() function did not handle UNC paths as it was designed to try to do. (PR#14884)
In package parallel, clusterApply() and similar failed to handle a (pretty pointless) length-1 argument. (PR#14898)
Quartz Cocoa display reacted asynchronously to dev.flush() which means that the redraw could be performed after the plot has been already modified by subsequent code. The redraw is now done synchronously in dev.flush() to allow animations without sleep cycles.
Source locations reported in traceback() were incorrect when byte-compiled code was on the stack.
plogis(x, lower = FALSE, log.p = TRUE) no longer underflows early for large x (e.g. 800).
Arithmetic's “1 ^ y and y ^ 0 are 1, always” now also applies for integer vectors y.
X11-based pixmap devices like png(type = "Xlib") were trying to set the cursor style, which triggered some warnings and hangs.
Code executed by the built-in HTTP server no longer allows other HTTP clients to re-enter R until the current worker evaluation finishes, to prevent cascades.
The plot() and Axis() methods for class "table" now respect graphical parameters such as cex.axis. (Reported by Martin Becker.)
Under some circumstances package.skeleton() would give out progress reports that could not be translated and so were displayed by question marks. Now they are always in English. (This was seen for CJK locales on Windows, but may have occurred elsewhere.)
The evaluator now keeps track of source references outside of functions, e.g. when source() executes a script.
The replacement method for window() now works correctly for multiple time series of class "mts". (PR#14925)
is.unsorted() gave incorrect results on non-atomic objects such as data frames. (Reported by Matthew Dowle.)
The value returned by tools::psnice() for invalid pid values was not always NA as documented.
Closing an X11() window while locator() was active could abort the R process.
getMethod(f, sig) produced an incorrect error message in some cases when f was not a string).
Using a string as a “call” in an error condition with options(showErrorCalls=TRUE) could cause a segfault. (PR#14931)
The string "infinity" allowed by C99 was not accepted as a numerical string value by e.g. scan() and as.character(). (PR#14933)
In legend(), setting some entries of lwd to NA was inconsistent (depending on the graphics device) in whether it would suppress those lines; now it consistently does so. (PR#14926)
by() failed for a zero-row data frame.
Yates correction in chisq.test() could be bigger than the terms it corrected, previously leading to an infinite test statistic in some corner cases which are now reported as NaN.
xgettext() and related functions sometimes returned items that were not strings for translation. (PR#14935)
plot(, which=5) now correctly labels the factor level combinations for the special case where all h[i,i] are the same. (PR#14837)

New in R for Windows 2.15.0 (Apr 4, 2012)

SIGNIFICANT USER-VISIBLE CHANGES:
The behavior of unlink(recursive = TRUE) for a symbolic link to a directory has changed: it now removes the link rather than the directory contents (just as rm -r does)
On Windows it no longer follows reparse points (including junctions and symbolic links)
NEW FEATURES:
Environment variable RD2DVI_INPUTENC has been renamed to RD2PDF_INPUTENC
Deprecated() becomes a bit more flexible, getting an old argument
Even data-only packages without R code need a namespace and so may need to be installed under R 2.14.0 or later
assignInNamespace() has further restrictions on use apart from at top-level, as its help page has warned. Expect it to be disabled from programmatic use in the future
system() and system2() when capturing output report a non-zero status in the new "status" attribute
kronecker() now has an S4 generic in package methods on which packages can set methods. It will be invoked by X %x% Y if either X or Y is an S4 object
pdf() accepts forms like file = "|lpr" in the same way as postscript()
pdf() accepts file = NULL. This means that the device does NOT create a PDF file (but it can still be queried, e.g., for font metric info)
format() (and hence print()) on "bibentry" objects now uses options("width") to set the output width
legend() gains a text.font argument. (Suggested by Tim Paine, PR#14719.)
nchar() and nzchar() no longer accept factors (as integer vectors). (Wish of PR#6899.)
summary() behaves slightly differently (or more precisely, its print() method does). For numeric inputs, the number of NAs is printed as an integer and not a real. For dates and datetimes, the number of NAs is included in the printed output (the latter being the wish of PR#14720)
The "data.frame" method is more consistent with the default method: in particular it now applies zapsmall() to numeric/complex summaries
The number of items retained with options(warn = 0) can be set by options(nwarnings=)
There is a new function assignInMyNamespace() which uses the namespace of the function it is called from
attach() allows the default name for an attached file to be overridden
bxp(), the work horse of boxplot(), now uses a more sensible default xlim in the case where at is specified differently from 1:n, see the discussion on R-devel, https://stat.ethz.ch/pipermail/r-devel/2011-November/062586.html
New function paste0(), an efficient version of paste(*, sep=""), to be used in many places for more concise (and slightly more efficient) code
Function setClass() in package methods now returns, invisibly, a generator function for the new class, slightly preferred to calling new(), as explained on the setClass help page
The "dendrogram" method of str() now takes its default for last.str from option str.dendrogram.last
New simple fitted() method for "kmeans" objects
The traceback() function can now be called with an integer argument, to display a current stack trace. (Wish of PR#14770.)
setGeneric() calls can be simplified when creating a new generic function by supplying the default method as the def argument. See ?setGeneric
serialize() has a new option xdr = FALSE which will use the native byte-order for binary serializations. In scenarios where only little-endian machines are involved (these days, close to universal) and (un)serialization takes an appreciable amount of time this may speed up noticeably transferring data between systems
The internal (un)serialization code is faster for long vectors, particularly with XDR on some platforms. (Based on a suggested patch by Michael Spiegel.)
For consistency, circles with zero radius are omitted by points() and grid.circle(). Previously this was device-dependent, but they were usually invisible
NROW(x) and NCOL(x) now work whenever dim(x) looks appropriate, e.g., also for more generalized matrices
PCRE has been updated to version 8.30
The internal R_Srcref variable is now updated before the browser stops on entering a function. (Suggestion of PR#14818.)
There are ‘bare-bones’ functions .colSums(), .rowSums(), .colMeans() and .rowMeans() for use in programming where ultimate speed is required
The formerly internal function .package_dependencies() from package tools for calculating (recursive) (reverse) dependencies on package databases has been renamed to package_dependencies() and is now exported
There is a new function optimHess() to compute the (approximate) Hessian for an optim() solution if hessian = TRUE was forgotten
filled.contour() is a ‘bare-bones’ function to add a filled-contour rectangular plot to an already prepared plot region
The stepping in debugging and single-step browsing modes has changed slightly: now left braces at the start of the body are stepped over for if statements as well as for for and while statements. (Wish of PR#14814.)
library() no longer warns about a conflict with a function from package:base if the function has the same code as the base one but with a different environment. (An example is Matrix::det().)
When deparsing very large language objects, as.character() now inserts newlines after each line of approximately 500 bytes, rather than truncating to the first line
New function rWishart() generates Wishart-distributed random matrices
Packages may now specify actions to be taken when the package is loaded (setLoadActions())
options(max.print = Inf) and similar now give an error (instead of warnings later)
The "difftime" replacement method of units tries harder to preserve other attributes of the argument. (Wish of PR#14839.)
poly(raw = TRUE) no longer requires more unique points than the degree.
PACKAGE parallel:
There is a new function mcmapply(), a parallel version of mapply(), and a wrapper mcMap(), a parallel version of Map()
A default cluster can be registered by the new function setDefaultCluster(): this will be used by default in functions such as parLapply()
clusterMap() has a new argument .scheduling to allow the use of load-balancing
There are new load-balancing functions parLapplyLB() and parSapplyLB()
makePSOCKCluster() has a new option useXDR = FALSE which can be used to avoid byte-shuffling for serialization when all the nodes are known to be little-endian (or all big-endian)
PACKAGE INSTALLATION:
Non-ASCII vignettes without a declared encoding are no longer accepted
C/C++ code in packages is now compiled with -NDEBUG to mitigate against the C/C++ function assert being called in production use. Developers can turn this off during package development with PKG_CPPFLAGS = -UNDEBUG
R CMD INSTALL has a new option --dsym which on Mac OS X (Darwin) dumps the symbols alongside the ‘.so’ file: this is helpful when debugging with valgrind (and especially when installing packages into ‘R.framework’). [This can also be enabled by setting the undocumented environment variable PKG_MAKE_DSYM, since R 2.12.0.]
R CMD INSTALL will test loading under all installed sub-architectures even for packages without compiled code, unless the flag --no-multiarch is used. (Pure R packages can do things which are architecture-dependent: in the case which prompted this, looking for an icon in a Windows R executable.)
There is a new option install.packages(type = "both") which tries source packages if binary packages are not available, on those platforms where the latter is the default
The meaning of install.packages(dependencies = TRUE) has changed: it now means to install the essential dependencies of the named packages plus the Suggests, but only the essential dependencies of dependencies. To get the previous behaviour, specify dependencies as a character vector
R CMD INSTALL --merge-multiarch is now supported on OS X and other Unix-alikes using multiple sub-architectures
R CMD INSTALL --libs-only now by default does a test load on Unix-alikes as well as on Windows: suppress with --no-test-load
UTILITIES:
R CMD check now gives a warning rather than a note if it finds inefficiently compressed datasets. With bzip2 and xz compression having been available since R 2.10.0, it only exceptionally makes sense to not use them
The environment variable _R_CHECK_COMPACT_DATA2_ is no longer consulted: the check is always done if _R_CHECK_COMPACT_DATA_ has a true value (its default)
Where multiple sub-architectures are to be tested, R CMD check now runs the examples and tests for all the sub-architectures even if one fails
R CMD check can optionally report timings on various parts of the check: this is controlled by environment variable _R_CHECK_TIMINGS_ documented in ‘Writing R Extensions’. Timings (in the style of R CMD BATCH) are given at the foot of the output files from running each test and the R code in each vignette
There are new options for more rigorous testing by R CMD check selected by environment variables – see the ‘Writing R Extensions’ manual
R CMD check now warns (rather than notes) on undeclared use of other packages in examples and tests: increasingly people are using the metadata in the ‘DESCRIPTION’ file to compute information about packages, for example reverse dependencies
The defaults for some of the options in R CMD check (described in the ‘R Internals’ manual) have changed: checks for unsafe and .Internal() calls and for partial matching of arguments in R function calls are now done by default
R CMD check has more comprehensive facilities for checking compiled code and so gives fewer reports on entry points linked into ‘.so’/‘.dll’ files from libraries (including C++ and Fortran runtimes)
Checking compiled code is now done on FreeBSD (as well as the existing supported platforms of Linux, Mac OS X, Solaris and Windows)
R CMD build has more options for --compact-vignettes: see R CMD build --help
R CMD build has a new option --md5 to add an ‘MD5’ file (as done by CRAN): this is used by R CMD INSTALL to check the integrity of the distribution
If this option is not specified, any existing (and probably stale) ‘MD5’ file is removed
DEPRECATED AND DEFUNCT:
R CMD Rd2dvi is now defunct: use R CMD Rd2pdf
Options such --max-nsize, --max-vsize and the function mem.limits() are now defunct. (Options --min-nsize and --min-vsize remain available.)
Use of library.dynam() without specifying all the first three arguments is now disallowed
Use of an argument chname in library.dynam() including the extension ‘.so’ or ‘.dll’ (which was never allowed according to the help page) is defunct. This also applies to library.dynam.unload() and to useDynLib directives in ‘NAMESPACE’ files
The internal functions .readRDS() and .saveRDS() are now defunct
The off-line help() types "postscript" and "ps" are defunct
Sys.putenv(), replaced and deprecated in R 2.5.0, is finally removed
Some functions/objects which have been defunct for five or more years have been removed completely. These include .Alias(), La.chol(), La.chol2inv(), La.eigen(), Machine(), Platform(), Version, codes(), delay(), format.char(), getenv(), httpclient(), loadURL(), machine(), parse.dcf(), printNoClass(), provide(), read.table.url(), restart(), scan.url(), symbol.C(), symbol.For() and unix()
The ENCODING argument to .C() is deprecated. It was intended to smooth the transition to multi-byte character strings, but can be replaced by the use of iconv() in the rare cases where it is still needed
INSTALLATION:
Building with a positive value of --with-valgrind-instrumentation now also instruments logical, complex and raw vectors
There is experimental support for link-time optimization with gcc 4.5.0 or later on platforms which support it
C-LEVEL FACILITIES
Passing R objects other than atomic vectors, functions, lists and environments to .C() is now deprecated and will give a warning. Most cases (especially NULL) are actually coding errors. NULL will be disallowed in future
C() now passes a pairlist as a SEXP to the compiled code. This is as was documented, but pairlists were in reality handled differently as a legacy from the early days of R
call_R and call_S are deprecated. They still exist in the headers and as entry points, but are no longer documented and should not be used for new code
BUG FIXES:
str(x, width) now obeys its width argument also for function headers and other objects x where deparse() is applied
The convention for x %/% 0L for integer-mode x has been changed from 0L to NA_integer_. (PR#14754)
The exportMethods directive in a ‘NAMESPACE’ file now exports S4 generics as necessary, as the extensions manual said it does. The manual has also been updated to be a little more informative on this point
It is now required that there is an S4 generic (imported or created in the package) when methods are to be exported
Reference methods cannot safely use non-exported entries in the namespace. We now do not do so, and warn in the documentation
The namespace import code was warning when identical S4 generic functions were imported more than once, but should not merge() is no longer allowed (in some ways) to create a data frame with duplicate column names (which confused PR#14786)
Fixes for rendering raster images on X11 and Windows devices when the x-axis or y-axis scale is reversed
getAnywhere() found S3 methods as seen from the utils namespace and not from the environment from which it was called
selectMethod(f, sig) would not return inherited group methods when caching was off (as it is by default)
dev.copy2pdf(out.type = "cairo") gave an error. (PR#14827)
Virtual classes (e.g., class unions) had a NULL prototype even if that was not a legal subclass. See ?setClassUnion
The C prototypes for zdotc and zdotu in ‘R_ext/BLAS.h’ have been changed to the more modern style rather than that used by f2c.
isGeneric() produced an error for primitives that can not have methods
C() or .Fortran() had a lack-of-protection error if the registration information resulted in an argument being coerced to another type
boxplot(x=x, at=at) with non finite elements in x and non integer at could not generate a warning but failed
heatmap(x, symm=TRUE, RowSideColors=*) no longer draws the colors in reversed order predict() was incorrect in the multivariate case, for p >= 2
print(x, max=m) is now consistent when x is a "Date"; also the “reached ... max.print ..” messages are now consistently using single brackets
Closed the tag in pages generated by Rd2HTML(). (PR#14841.)
Axis tick marks could go out of range when a log scale was used. (PR#14833.)
Signature objects in methods were not allocated as S4 objects

New in R for Windows 2.14.2 (Mar 6, 2012)

New in R for Windows 2.14.1 (Mar 6, 2012)

New in R for Windows 2.14.0 (Nov 24, 2011)

All packages must have a namespace, and one is created on installation if not supplied in the sources. This means that any package without a namespace must be re-installed under this version of R (but previously-installed data-only packages without R code can still be used).
The yLineBias of the X11() and windows() families of devices has been changed from 0.1 to 0.2: this changes slightly the vertical positioning of text in the margins (including axis annotations). This is mainly for consistency with other devices such as quartz() and pdf(). (Wish of PR#14538.)
There is a new graphics parameter "ylbias" which allows the y-line bias of the graphics device to be tweaked, including to reproduce output from earlier versions of R.
Labeling of the p-values in various anova tables has been rationalized to be either "Pr(>F)" or "Pr(>Chi)" (i.e. the "Pr(F)", "Pr(Chi)" and "P(>|Chi|)" variants have been eliminated). Code which extracts the p value via indexing by name may need adjustment.
:: can now be used for datasets made available for lazy-loading in packages with namespaces (which makes it consistent with its use for data-only packages without namespaces in earlier versions of R).
There is a new package parallel.
It incorporates (slightly revised) copies of packages multicore and snow (excluding MPI, PVM and NWS clusters). Code written to use the higher-level API functions in those packages should work unchanged (apart from changing any references to their namespaces to a reference to parallel, and links explicitly to multicore or snow on help pages).
It also contains support for multiple RNG streams following L'Ecuyer et al (2002), with support for both mclapply and snow clusters. This replaces functions like clusterSetupRNG() from snow (which are not in parallel).
The version released for R 2.14.0 contains base functionality: higher-level convenience functions are planned (and some are already available in the ‘R-devel’ version of R).
Building PDF manuals (for R itself or packages, e.g. via R CMD check) by default requires the LaTeX package inconsolata: see the section on ‘Making the manuals’ in the ‘R Installation and Administration Manual’.
axTicks(*, log=TRUE) has changed in some cases to satisfy the documented behavior and be consistent.

New in R for Windows 2.13.2 (Oct 6, 2011)

NEW FEATURES:
mem.limits() now reports values larger than the maximum integer (previously documented to be reported as NA), and allows larger values to be set, including Inf to remove the limit.
The print() methods for classes "Date", "POSIXct" and "POSIXlt" respect the option "max.print" and so are much faster for very long datetime vectors. (Suggestion of Yohan Chalabi.)
untar2() now works around errors generated with tar files that use more than the standard 6 digits for the checksum. (https://bugs.R-project.org/bugzilla3/show_bug.cgi?id=14654PR#14654)
install.packages() with Ncpus > 1 guards against simultaneous installation of indirect dependencies as well as direct ones.
Sweave now knows about a few more Windows' encodings (including cp1250 and cp1257) and some inputenx encodings such as koi8-r.
postscript(colormodel = "rgb-nogray") no longer sets the sRGB colorspace for each colour and so some viewers may render its files much faster than the default colormodel ="rgb".
The default for pdf(maxRasters=) has been increased from 64 to 1000.
readBin() now warns if signed = FALSE is used inappropriately (rather than being silently ignored).
It enforces the documented limit of 2^31-1 bytes in a single call.
PCRE has been updated to version 8.13, a bug-fix release with updated Unicode tables (version 6.0.0). An additional patch (r611 from PCRE 8.20-to-be) has been added to fix a collation symbol recognition issue.
INSTALLATION:
It is possible to build in ‘src/extra/xdr’ on more platforms. (Needed since glibc 2.14 hides its RPC implementation.)
configure will find the Sun TI-RPC implementation of xdr (in ‘libtirpc’) provided its header files are in the search path: see the ‘R Installation and Administration Manual’.
PACKAGE INSTALLATION
Using a broad exportPattern directive in a ‘NAMESPACE’ file is no longer allowed to export internal objects such as .onLoad and .__S3MethodsTable__. .
These are also excluded from imports, along with .First.lib.
BUG FIXES:
fisher.test() had a buglet: If arguments were factors with unused levels, levels were dropped and you would get an error saying that there should be at least two levels, inconsistently with pre-tabulated data. (Reported by Michael Fay).
package.skeleton() will no longer dump S4 objects supplied directly rather than in a code file. These cannot be restored correctly from the dumped version.
Build-time expressions in help files did not have access to functions in the package being built (with R CMD build).
Because quote() did not mark its result as being in use, modification of the result could in some circumstances modify the original call.
Plotting pch = '.' now guarantees at least a one-pixel dot if cex > 0.
The very-rarely-used command-line option --max-vsize was incorrectly interpreted as a number of Vcells and not in bytes as documented.
The HTML generated by Rd2HTML() comes closer to being standards compliant.
filter(x, recursive = TRUE) gave incorrect results on a series containing NAs. (Spotted by Bill Dunlap.)
Profiling stats::mle() fits with a fixed parameter was not supported. (https://bugs.R-project.org/bugzilla3/show_bug.cgi?id=14646PR#14646)
retracemem() was still using positional matching. (https://bugs.R-project.org/bugzilla3/show_bug.cgi?id=14650PR#14650)
The quantile method for "ecdf" objects now works and is documented.
xtabs(~ .., ..., sparse=TRUE) now also works together with an exclude = .. specification.
decompose() computed an incorrect seasonal component for time series with odd frequencies.
The pdf() device only includes the definition of the sRGB colorspace in the output file for the "rgb" colormodel (and not for "gray" nor "cmyk"): this saves ca 9KB in the output file.
.hasSlot() wrongly gave FALSE in some cases.
Sweave() with keep.source=TRUE could generate spurious NA lines when a chunk reference appeared last in a code chunk.
\Sexpr[results=rd] in an ‘.Rd’ file now first tries parse_Rd(fragment=FALSE) to allow Rd section-level macros to be inserted.
The print() method for class "summary.aov" did not pass on arguments such as signif.stars when summary() was called on a single object. (https://bugs.R-project.org/bugzilla3/show_bug.cgi?id=14684PR#14684)
In rare cases ks.test() could return a p-value very slightly less than 0 by rounding error. (https://bugs.R-project.org/bugzilla3/show_bug.cgi?id=14671PR#14671)
If trunc() was called on a "POSIXlt" vector and the result was subsetted, all but the first element was converted to NA. (https://bugs.R-project.org/bugzilla3/show_bug.cgi?id=14679PR#14679)
cbind() and rbind() could cause memory corruption when used on a combination of raw and logical/integer vectors.