Julia Language Changelog

What's new in Julia Language 1.11.0 Alpha 1

Mar 3, 2024

New language features:
public is a new keyword. Symbols marked with public are considered public API. Symbols marked with export are now also treated as public API. The difference between public and export is that public names do not become available when using a package/module (#50105).
ScopedValue implements dynamic scope with inheritance across tasks (#50958).
The new macro Base.Cartesian.@ncallkw is analogous to Base.Cartesian.@ncall, but allows to add keyword arguments to the function call (#51501).
Support for Unicode 15.1 (#51799).
Three new types around the idea of text with "annotations" (Pair{Symbol, Any} entries, e.g. :lang => "en" or :face => :magenta). These annotations are preserved across operations (e.g. string concatenation with *) when possible.
AnnotatedString is a new AbstractString type. It wraps an underlying string and allows for annotations to be attached to regions of the string. This type is used extensively in the new StyledStrings standard library to hold styling information.
AnnotatedChar is a new AbstractChar type. It wraps another char and holds a list of annotations that apply to it.
AnnotatedIOBuffer is a new IO type that mimics an IOBuffer, but has specialised read/write methods for annotated content. This can be thought of both as a "string builder" of sorts and also as glue between annotated and unannotated content.
Manifest.toml files can now be renamed in the format Manifest-v{major}.{minor}.toml to be preferentially picked up by the given julia version. i.e. in the same folder, a Manifest-v1.11.toml would be used by v1.11 and Manifest.toml by every other julia version. This makes managing environments for multiple julia versions at the same time easier (#43845).
@time now reports a count of any lock conflicts where a ReentrantLock had to wait, plus a new macro @lock_conflicts which returns that count (#52883).
Language changes:
During precompilation, the atexit hooks now run before saving the output file. This allows users to safely tear down background state (such as closing Timers and sending disconnect notifications to heartbeat tasks) and cleanup other resources when the program wants to begin exiting.
Code coverage and malloc tracking is no longer generated during the package precompilation stage. Further, during these modes pkgimage caches are now used for packages that are not being tracked. This means that coverage testing (the default for julia-actions/julia-runtest) will by default use pkgimage caches for all other packages than the package being tested, likely meaning faster test execution. (#52123)
Specifying a path in JULIA_DEPOT_PATH now results in the expansion of empty strings to omit the default user depot (#51448).
Compiler/Runtime improvements:
Updated GC heuristics to count allocated pages instead of individual objects (#50144).
A new LazyLibrary type is exported from Libdl for use in building chained lazy library loads, primarily to be used within JLLs (#50074).
Added support for annotating Base.@assume_effects on code blocks (#52400).
The libuv library has been updated from a base of v1.44.2 to v1.48.0 (#49937).
Command-line option changes:
The entry point for Julia has been standardized to Main.main(ARGS). This must be explicitly opted into using the @main macro (see the docstring for further details). When opted-in, and julia is invoked to run a script or expression (i.e. using julia script.jl or julia -e expr), julia will subsequently run the Main.main function automatically. This is intended to unify script and compilation workflows, where code loading may happen in the compiler and execution of Main.main may happen in the resulting executable. For interactive use, there is no semantic difference between defining a main function and executing the code directly at the end of the script ([50974]).
The --compiled-modules and --pkgimages flags can now be set to existing, which will cause Julia to consider loading existing cache files, but not to create new ones (#50586 and #52573).
Multi-threading changes:
Threads.@threads now supports the :greedy scheduler, intended for non-uniform workloads (#52096).
A new exported struct Lockable{T, L<:AbstractLock} makes it easy to bundle a resource and its lock together (#52898).
Build system changes:
New library functions:
in!(x, s::AbstractSet) will return whether x is in s, and insert x in s if not.
The new Libc.mkfifo function wraps the mkfifo C function on Unix platforms (#34587).
copyuntil(out, io, delim) and copyline(out, io) copy data into an out::IO stream (#48273).
eachrsplit(string, pattern) iterates split substrings right to left.
Sys.username() can be used to return the current user's username (#51897).
wrap(Array, m::Union{MemoryRef{T}, Memory{T}}, dims) is the safe counterpart to unsafe_wrap (#52049).
GC.logging_enabled() can be used to test whether GC logging has been enabled via GC.enable_logging (#51647).
IdSet is now exported from Base and considered public (#53262).
New library features:
invmod(n, T) where T is a native integer type now computes the modular inverse of n in the modular integer ring that T defines (#52180).
invmod(n) is an abbreviation for invmod(n, typeof(n)) for native integer types (#52180).
replace(string, pattern...) now supports an optional IO argument to write the output to a stream rather than returning a string (#48625).
New methods allequal(f, itr) and allunique(f, itr) taking a predicate function (#47679).
sizehint!(s, n) now supports an optional shrink argument to disable shrinking (#51929).
New function Docs.hasdoc(module, symbol) tells whether a name has a docstring (#52139).
New function Docs.undocumented_names(module) returns a module's undocumented public names (#52413).
Passing an IOBuffer as a stdout argument for Process spawn now works as expected, synchronized with wait or success, so a Base.BufferStream is no longer required there for correctness to avoid data races (#52461).
After a process exits, closewrite will no longer be automatically called on the stream passed to it. Call wait on the process instead to ensure the content is fully written, then call closewrite manually to avoid data-races. Or use the callback form of open to have all that handled automatically.
@timed now additionally returns the elapsed compilation and recompilation time (#52889)
filter can now act on a NamedTuple (#50795).
Iterators.cycle(iter, n) runs over iter a fixed number of times, instead of forever (#47354)
zero(::AbstractArray) now applies recursively, so zero([[1,2],[3,4,5]]) now produces the additive identity [[0,0],[0,0,0]] rather than erroring (#38064).
Standard library changes:
StyledStrings:
A new standard library for handling styling in a more comprehensive and structured way (#49586).
The new Faces struct serves as a container for text styling information (think typeface, as well as color and decoration), and comes with a framework to provide a convenient, extensible (via addface!), and customisable (with a user's Faces.toml and loadfaces!) approach to styled content (#49586).
The new @styled_str string macro provides a convenient way of creating a AnnotatedString with various faces or other attributes applied (#49586).
JuliaSyntaxHighlighting:
A new standard library for applying syntax highlighting to Julia code, this uses JuliaSyntax and StyledStrings to implement a highlight function that creates an AnnotatedString with syntax highlighting applied.
Package Manager:
LinearAlgebra:
cbrt(::AbstractMatrix{<:Real}) is now defined and returns real-valued matrix cube roots of real-valued matrices (#50661).
eigvals/eigen(A, bunchkaufman(B)) and eigvals/eigen(A, lu(B)), which utilize the Bunchkaufman (LDL) and LU decomposition of B, respectively, now efficiently compute the generalized eigenvalues (eigen: and eigenvectors) of A and B. Note: The second argument is the output of bunchkaufman or lu (#50471).
There is now a specialized dispatch for eigvals/eigen(::Hermitian{<:Tridiagonal}) which performs a similarity transformation to create a real symmetrix triagonal matrix, and solve that using the LAPACK routines (#49546).
Structured matrices now retain either the axes of the parent (for Symmetric/Hermitian/AbstractTriangular/UpperHessenberg), or that of the principal diagonal (for banded matrices) (#52480).
bunchkaufman and bunchkaufman! now work for any AbstractFloat, Rational and their complex variants. bunchkaufman now supports Integer types, by making an internal conversion to Rational{BigInt}. Added new function inertia that computes the inertia of the diagonal factor given by the BunchKaufman factorization object of a real symmetric or Hermitian matrix. For complex symmetric matrices, inertia only computes the number of zero eigenvalues of the diagonal factor (#51487).
Packages that specialize matrix-matrix mul! with a method signature of the form mul!(::AbstractMatrix, ::MyMatrix, ::AbstractMatrix, ::Number, ::Number) no longer encounter method ambiguities when interacting with LinearAlgebra. Previously, ambiguities used to arise when multiplying a MyMatrix with a structured matrix type provided by LinearAlgebra, such as AbstractTriangular, which used to necessitate additional methods to resolve such ambiguities. Similar sources of ambiguities have also been removed for matrix-vector mul! operations (#52837).
lu and issuccess(::LU) now accept an allowsingular keyword argument. When set to true, a valid factorization with rank-deficient U factor will be treated as success instead of throwing an error. Such factorizations are now shown by printing the factors together with a "rank-deficient" note rather than printing a "Failed Factorization" message (#52957).
Logging
Printf
Profile
Random
rand now supports sampling over Tuple types (#35856, #50251).
rand now supports sampling over Pair types (#28705).
When seeding RNGs provided by Random, negative integer seeds can now be used (#51416).
Seedable random number generators from Random can now be seeded by a string, e.g. seed!(rng, "a random seed") (#51527).
REPL:
Tab complete hints now show in lighter text while typing in the repl. To disable set Base.active_repl.options.hint_tab_completes = false interactively, or in startup.jl:
if VERSION >= v"1.11.0-0"
atreplinit() do repl
repl.options.hint_tab_completes = false
end
end
``` ([#51229]).
Meta-M with an empty prompt now toggles the contextual module between the previous non-Main contextual module and Main so that switching back and forth is simple. (#51616, #52670)
SuiteSparse
SparseArrays
Test
Dates
The undocumented function adjust is no longer exported but is now documented
Statistics:
Statistics is now an upgradeable standard library (#46501).
Distributed:
pmap now defaults to using a CachingPool (#33892).
Unicode
DelimitedFiles
InteractiveUtils
Deprecated or removed:
Base.map, Iterators.map, and foreach lost their single-argument methods (#52631).
External dependencies:
tput is no longer called to check terminal capabilities, it has been replaced with a pure-Julia terminfo parser (#50797).
Tooling Improvements:
CI now performs limited automatic typo detection on all PRs. If you merge a PR with a failing typo CI check, then the reported typos will be automatically ignored in future CI runs on PRs that edit those same files (#51704).

New in Julia Language 1.10.0 (Dec 28, 2023)

New language features:
JuliaSynta.jl is now used as the default parser, providing better diagnostics and faster parsing. Set environment variable JULIA_USE_FLISP_PARSER to 1 to switch back to the old parser if necessary (and if you find this necessary, please file an issue) (#46372).
(U+297A, leftarrowsubset) and ? (U+2977, leftarrowless) may now be used as binary operators with arrow precedence (#45962).
Language changes:
When a task forks a child, the parent task's task-local RNG (random number generator) is no longer affected. The seeding of child based on the parent task also takes a more disciplined approach to collision resistance, using a design based on the SplitMi and DotMi splittable RNG schemes (#49110).
A new more-specific rule for methods resolves ambiguities containing Union{} in favor of the method defined eplicitly to handle the Union{} argument. This makes it possible to define methods to eplicitly handle Union{} without the ambiguities that commonly would result previously. This also lets the runtime optimize certain method lookups in a way that significantly improves load and inference times for heavily overloaded methods that dispatch on Types (such as traits and constructors).
The "h bar" ? (hslash U+210F) character is now treated as equivalent to h (hbar U+0127).
The @simd macro now has more limited and clearer semantics: it only enables reordering and contraction of floating-point operations, instead of turning on all "fastmath" optimizations. If you observe performance regressions due to this change, you can recover previous behavior with @fastmath @simd, if you are OK with all the optimizations enabled by the @fastmath macro (#49405).
When a method with keyword arguments is displayed in the stack trace view, the tetual representation of the keyword arguments' type is simplified using the new @Kwargs{key1::Type1, ...} macro synta (#49959).
Compiler/Runtime improvements:
Updated GC heuristics to count allocated pages instead of object sizes (#50144). This should help some programs that consumed ecessive memory before.
The mark phase of the garbage collector is now multi-threaded (#48600).
JITLink is enabled by default on Linu aarch64 when Julia is linked to LLVM 15 or later versions (#49745). This should resolve many segmentation faults previously observed on this platform.
The precompilation process now uses pidfile locks and orchestrates multiple julia processes to only have one process spend effort precompiling while the others wait. Previously all would do the work and race to overwrite the cache files. ([#49052])
Command-line option changes:
New option --gcthreads to set how many threads will be used by the garbage collector (#48600). The default is N/2 where N is the number of worker threads (--threads) used by Julia.
Build system changes:
SparseArrays and SuiteSparse are no longer included in the default system image, so the core language no longer contains GPL libraries. However, these libraries are still included alongside the language in the standard binary distribution (#44247, #48979, #49266).
New library functions:
tanpi is now defined. It computes tan(p*) more accurately than tan(pi*) (#48575).
fourthroot() is now defined in Base.Math and can be used to compute the fourth root of . It can also be accessed using the unicode character ?, which can be typed by fourthroot<tab> (#48899).
Libc.memmove, Libc.memset, and Libc.memcpy are now defined, whose functionality matches that of their respective C calls.
Base.isprecompiled(pkg::PkgId) has been added, to identify whether a package has already been precompiled (#50218).
New library features:
binomial(, k) now supports non-integer (#48124).
A CartesianInde is now treated as a "scalar" for broadcasting (#47044).
printstyled now supports italic output (#45164).
parent and parentindices support SubStrings.
replace(string, pattern...) now supports an optional IO argument to write the output to a stream rather than returning a string (#48625).
startswith now supports seekable IO streams (#43055).
Standard library changes:
The initialized=true keyword assignment for sortperm! and partialsortperm! is now a no-op (#47979). It previously eposed unsafe behavior (#47977).
Printing integral Rationals will skip the denominator in Rational-typed IO contet (e.g. in arrays) (#45396).
Package Manager:
Pkg.precompile now accepts timing as a keyword argument which displays per package timing information for precompilation (e.g. Pkg.precompile(timing=true)).
LinearAlgebra:
AbstractQ no longer subtypes AbstractMatri. Moreover, adjoint(Q::AbstractQ) no longer wraps Q in an Adjoint type, but instead in an AdjointQ, that itself subtypes AbstractQ. This change accounts for the fact that typically AbstractQ instances behave like function-based, matri-backed linear operators, and hence don't allow for efficient indeing. Also, many AbstractQ types can act on vectors/matrices of different size, acting like a matri with contet-dependent size. With this change, AbstractQ has a well-defined API that is described in detail in the Julia documentation (#46196).
Adjoints and transposes of Factorization objects are no longer wrapped in Adjoint and Transpose wrappers, respectively. Instead, they are wrapped in AdjointFactorization and TransposeFactorization types, which themselves subtype Factorization (#46874).
New functions hermitianpart and hermitianpart! for etracting the Hermitian (real symmetric) part of a matri (#31836).
The norm of the adjoint or transpose of an AbstractMatri now returns the norm of the parent matri by default, matching the current behaviour for AbstractVectors (#49020).
eigen(A, B) and eigvals(A, B), where one of A or B is symmetric or Hermitian, are now fully supported (#49533).
eigvals/eigen(A, cholesky(B)) now computes the generalized eigenvalues (eigen: and eigenvectors) of A and B via Cholesky decomposition for positive definite B. Note: The second argument is the output of cholesky.
Printf:
Format specifiers now support dynamic width and precision, e.g. %*s and %*.*g (#40105).
REPL:
When stack traces are printed, the printed depth of types in function signatures will be limited to avoid overly verbose output (#49795).
Test:
The @test_broken macro (or @test with broken=true) now complains if the test epression returns a non-boolean value in the same way as a non-broken test (#47804).
When a call to @test fails or errors inside a function, a larger stacktrace is now printed such that the location of the test within a @testset can be retrieved (#49451).
InteractiveUtils:
code_native and @code_native now default to intel synta instead of AT&T.
@time_imports now shows the timing of any module __init__()s that are run (#49529).
Deprecated or removed:
The @pure macro is now deprecated. Use Base.@assume_effects :foldable instead (#48682).

New in Julia Language 1.10.0 Alpha 1 (Jul 7, 2023)

New language features:
JuliaSyntax.jl is now used as the default parser, providing better diagnostics and faster parsing. Set environment variable JULIA_USE_NEW_PARSER to 0 to switch back to the old parser if necessary (and if you find this necessary, please file an issue) ([#46372]).
(U+297A, leftarrowsubset) and ? (U+2977, leftarrowless) may now be used as binary operators with arrow precedence. ([#45962])
Language changes:
When a task forks a child, the parent task's task-local RNG (random number generator) is no longer affected. The seeding of child based on the parent task also takes a more disciplined approach to collision resistance, using a design based on the SplitMix and DotMix splittable RNG schemes ([#49110]).
A new more-specific rule for methods resolves ambiguities containing Union{} in favor of the method defined explicitly to handle the Union{} argument. This makes it possible to define methods to explicitly handle Union{} without the ambiguities that commonly would result previously. This also lets the runtime optimize certain method lookups in a way that significantly improves load and inference times for heavily overloaded methods that dispatch on Types (such as traits and constructors).
The "h bar" ? (hslash U+210F) character is now treated as equivalent to h (hbar U+0127).
The @simd macro now has a more limited and clearer semantics, it only enables reordering and contraction of floating-point operations, instead of turning on all "fastmath" optimizations. If you observe performance regressions due to this change, you can recover previous behavior with @fastmath @simd, if you are OK with all the optimizations enabled by the @fastmath macro. ([#49405])
When a method with keyword arguments is displayed in the stack trace view, the textual representation of the keyword arguments' types is simplified using the new @Kwargs{key1::Type1, ...} macro syntax ([#49959]).
Compiler/Runtime improvements:
The @pure macro is now deprecated. Use Base.@assume_effects :foldable instead ([#48682]).
The mark phase of the Garbage Collector is now multi-threaded ([#48600]).
JITLink is enabled by default on Linux aarch64 when Julia is linked to LLVM 15 or later versions ([#49745]). This should resolve many segmentation faults previously observed on this platform.
Command-line option changes:
New option --gcthreads to set how many threads will be used by the Garbage Collector ([#48600]). The default is set to N/2 where N is the amount of worker threads (--threads) used by Julia.
New library functions:
tanpi is now defined. It computes tan(px) more accurately than tan(pi*x) ([#48575]).
fourthroot(x) is now defined in Base.Math and can be used to compute the fourth root of x. It can also be accessed using the unicode character ?, which can be typed by fourthroot<tab> ([#48899]).
Libc.memmove, Libc.memset, and Libc.memcpy are now defined, whose functionality matches that of their respective C calls.
Base.isprecompiled(pkg::PkgId) to identify whether a package has already been precompiled ([#50218]).
New library features:
The initialized=true keyword assignment for sortperm! and partialsortperm! is now a no-op ([#47979]). It previously exposed unsafe behavior ([#47977]).
binomial(x, k) now supports non-integer x ([#48124]).
A CartesianIndex is now treated as a "scalar" for broadcasting ([#47044]).
printstyled now supports italic output ([#45164]).
parent and parentindices support SubStrings
Standard library changes:
startswith now supports seekable IO streams ([#43055])
printing integral Rationals will skip the denominator in Rational-typed IO context (e.g. in Arrays) ([#45396])
Package Manager:
Pkg.precompile now accepts timing as a keyword argument which displays per package timing information for precompilation (e.g. Pkg.precompile(timing=true))
LinearAlgebra:
AbstractQ no longer subtypes to AbstractMatrix. Moreover, adjoint(Q::AbstractQ) no longer wraps Q in an Adjoint type, but instead in an AdjointQ, that itself subtypes AbstractQ. This change accounts for the fact that typically AbstractQ instances behave like function-based, matrix-backed linear operators, and hence don't allow for efficient indexing. Also, many AbstractQ types can act on vectors/matrices of different size, acting like a matrix with context-dependent size. With this change, AbstractQ has a well-defined API that is described in detail in the Julia documentation ([#46196]).
Adjoints and transposes of Factorization objects are no longer wrapped in Adjoint and Transpose wrappers, respectively. Instead, they are wrapped in AdjointFactorization and TranposeFactorization types, which themselves subtype Factorization ([#46874]).
New functions hermitianpart and hermitianpart! for extracting the Hermitian (real symmetric) part of a matrix ([#31836]).
The norm of the adjoint or transpose of an AbstractMatrix now returns the norm of the parent matrix by default, matching the current behaviour for AbstractVectors ([#49020]).
eigen(A, B) and eigvals(A, B), where one of A or B is symmetric or Hermitian, are now fully supported ([#49533])
eigvals/eigen(A, cholesky(B)) now computes the generalized eigenvalues (eigen: and eigenvectors) of A and B via Cholesky decomposition for positive definite B. Note: The second argument is the output of cholesky.
Printf:
Format specifiers now support dynamic width and precision, e.g. %*s and %*.*g ([#40105]).
REPL:
When stack traces are printed, the printed depth of types in function signatures will be limited to avoid overly verbose output ([#49795]).
Test:
The @test_broken macro (or @test with broken=true) now complains if the test expression returns a non-boolean value in the same way as a non-broken test. ([#47804])
When a call to @test fails or errors inside a function, a larger stacktrace is now printed such that the location of the test within a @testset can be retrieved ([#49451])
InteractiveUtils:
code_native and @code_native now default to intel syntax instead of AT&T.
@time_imports now shows the timing of any module __init__()s that are run ([#49529])

New in Julia Language 1.9.0 (May 11, 2023)

New language features:
It is now possible to assign to bindings in another module using setproperty!(::Module, ::Symbol, x) (#44137).
Slurping in assignments is now also allowed in non-final position. This is handled via Base.split_rest (#42902).
Character literals now support the same syntax allowed in string literals; i.e. the syntax can represent invalid UTF-8 sequences as allowed by the Char type (#44989).
Support for Unicode 15 (#47392).
Nested combinations of tuples and named tuples of symbols are now allowed as type parameters (#46300).
New builtins getglobal(::Module, ::Symbol[, order]) and setglobal!(::Module, ::Symbol, x[, order]) for reading from and writing to globals. getglobal should now be preferred for accessing globals over getfield (#44137).
Language changes:
The @invoke macro introduced in 1.7 is now exported. Additionally, it now uses Core.Typeof(x) rather than Any when a type annotation is omitted for an argument x so that types passed as arguments are handled correctly (#45807).
The invokelatest function and @invokelatest macro introduced in 1.7 are now exported (#45831).
Compiler/Runtime improvements:
Time to first execution (TTFX, sometimes called time to first plot) is greatly reduced. Package precompilation now saves native code into a "pkgimage", meaning that code generated during the precompilation process will not require compilation after package load. Use of pkgimages can be disabled via --pkgimages=no ([#44527]) ([#47184]).
The known quadratic behavior of type inference is now fixed and inference uses less memory in general. Certain edge cases with auto-generated long functions (e.g. ModelingToolkit.jl with partial differential equations and large causal models) should see significant compile-time improvements (#45276, #45404).
Non-concrete call sites can now be union-split to be inlined or statically resolved even if there are multiple dispatch candidates. This may improve runtime performance in certain situations where object types are not fully known statically, by statically resolving @nospecialize-d call sites and avoiding excessive compilation (#44512).
All uses of the @pure macro in Base have been replaced with the now-preferred Base.@assume_effects (#44776).
invoke(f, invokesig, args...) calls to a less-specific method than would normally be chosen for f(args...) are no longer spuriously invalidated when loading package precompile files (#46010).
Command-line option changes:
In Linux and Windows, --threads=auto now tries to infer the usable number of CPUs from the process affinity which is set typically in HPC and cloud environments (#42340).
--math-mode=fast is now a no-op (#41638). Users are encouraged to use the @fastmath macro instead, which has more well-defined semantics.
The --threads command-line option now accepts auto|N[,auto|M] where M specifies the number of interactive threads to create (auto currently means 1) (#42302).
New option --heap-size-hint=<size> suggests a size limit to invoke garbage collection more eagerly. The size may be specified in bytes, kilobytes (1000k), megabytes (300M), or gigabytes (1.5G) (#45369).
Multi-threading changes:
Threads.@spawn now accepts an optional first argument: :default or :interactive. An interactive task desires low latency and implicitly agrees to be short duration or to yield frequently. Interactive tasks will run on interactive threads, if any are specified when Julia is started (#42302).
Threads started outside the Julia runtime (e.g. from C or Java) can now become able to call into Julia code by calling jl_adopt_thread. This is done automatically when entering Julia code via cfunction or a @ccallable entry point. As a consequence, the number of threads can now change during execution (#46609).
Build system changes:
New library functions:
New function Iterators.flatmap (#44792).
New pkgversion(m::Module) function to get the version of the package that loaded a given module, similar to pkgdir(m::Module) (#45607).
New function stack(x) which generalises reduce(hcat, x::Vector{<:Vector}) to any dimensionality, and allows any iterator of iterators. Method stack(f, x) generalises mapreduce(f, hcat, x) and is more efficient (#43334).
New macro @allocations which is similar to @allocated except reporting the total number of allocations rather than the total size of memory allocated (#47367).
New library features:
RoundFromZero now works for non-BigFloat types (#41246).
Dict can be now shrunk manually by sizehint! (#45004).
@time now separates out % time spent recompiling invalidated methods (#45015).
Standard library changes:
A known concurrency issue in iterate methods on Dict and other derived objects such as keys(::Dict), values(::Dict), and Set is fixed. These methods of iterate can now be called on a dictionary or set shared by arbitrary tasks provided that there are no tasks mutating the dictionary or set (#44534).
Predicate function negation !f now returns a composed function (!) ∘ f instead of an anonymous function (#44752).
eachslice now works over multiple dimensions; eachslice, eachrow and eachcol return a Slices object, which allows dispatching to provide more efficient methods (#32310).
@kwdef is now exported and added to the public API (#46273).
An issue with order of operations in fld1 is now fixed (#28973).
Sorting is now always stable by default, as QuickSort was stabilized (#45222).
Base.splat is now exported. The return value is now a Base.Splat instead of an anonymous function, which allows for pretty printing (#42717).
Package Manager:
LinearAlgebra:
The methods a / b and b a with a a scalar and b a vector, which were equivalent to a * pinv(b), have been removed due to the risk of confusion with elementwise division (#44358).
We are now wholly reliant on libblastrampoline (LBT) for calling BLAS and LAPACK. OpenBLAS is shipped by default, but building the system image with other BLAS/LAPACK libraries is not supported. Instead, it is recommended that the LBT mechanism be used for swapping BLAS/LAPACK with vendor provided ones (#44360).
lu supports a new pivoting strategy RowNonZero() that chooses the first non-zero pivot element, for use with new arithmetic types and for pedagogy (#44571).
normalize(x, p=2) now supports any normed vector space x, including scalars (#44925).
The default number of BLAS threads is now set to the number of CPU threads on ARM CPUs, and half the number of CPU threads on other architectures (#45412, #46085).
Printf:
Error messages for bad format strings have been improved, to make it clearer what and where in the format string is wrong (#45366).
Profile:
New function Profile.take_heap_snapshot(file) that writes a file in Chrome's JSON-based .heapsnapshot format (#46862).
Random:
randn and randexp now work for any AbstractFloat type defining rand (#44714).
REPL:
Alt-e now opens the current input in an editor. The content (if modified) will be executed upon exiting the editor (#33759).
The contextual module which is active in the REPL can be changed (it is Main by default), via the REPL.activate(::Module) function or via typing the module in the REPL and pressing the keybinding Alt-m (#33872).
A "numbered prompt" mode which prints numbers for each input and output and stores evaluated results in Out can be activated with REPL.numbered_prompt!(). See the manual for how to enable this at startup (#46474).
Tab completion displays available keyword arguments ([#43536])
SuiteSparse:
Code for the SuiteSparse solver wrappers has been moved to SparseArrays.jl. Solvers are now re-exported by SuiteSparse.jl.
SparseArrays:
SuiteSparse solvers are now available as submodules of SparseArrays (JuliaSparse/SparseArrays.jl#95).
UMFPACK (JuliaSparse/SparseArrays.jl#179) and CHOLMOD (JuliaSparse/SparseArrays.jl#206) thread safety are improved by avoiding globals and using locks. Multithreaded ldiv! of UMFPACK objects may now be performed safely.
An experimental function SparseArrays.allowscalar(::Bool) allows scalar indexing of sparse arrays to be disabled or enabled. This function is intended to help find accidental scalar indexing of SparseMatrixCSC objects, which is a common source of performance issues (JuliaSparse/SparseArrays.jl#200).
Test:
New fail-fast mode for testsets that will terminate the test run early if a failure or error occurs. Set either via the @testset kwarg failfast=true or by setting env var JULIA_TEST_FAILFAST to "true" i.e. in CI runs to request the job failure be posted eagerly when issues occur (#45317)
Dates:
Empty strings are no longer incorrectly parsed as valid DateTimes, Dates or Times and instead throw an ArgumentError in constructors and parse, while nothing is returned by tryparse ([#47117]).
Distributed:
The package environment (active project, LOAD_PATH, DEPOT_PATH) is now propagated when adding local workers (e.g. with addprocs(N::Int) or through the --procs=N command line flag) (#43270).
addprocs for local workers now accepts the env keyword argument for passing environment variables to worker processes. This was already supported for remote workers (#43270).
Unicode:
graphemes(s, m:n) returns a substring of the m-th to n-th graphemes in s (#44266).
DelimitedFiles:
DelimitedFiles has been moved out as a separate package.
Deprecated or removed:
External dependencies:
On Linux, now autodetects the system libstdc++ version, and automatically loads the system library if it is newer. The old behavior of loading the bundled libstdc++ regardless of the system version can be restored by setting the environment variable JULIA_PROBE_LIBSTDCXX=0 (#46976).
Removed RPATH from the julia binary. On Linux this may break libraries that have failed to set RUNPATH.
Tooling Improvements:
Printing of MethodError and methods (such as from methods(my_func)) is now prettified and colored consistently with printing of methods in stacktraces (#45069).

New in Julia Language 1.8.0 (Aug 18, 2022)

New language features:
Mutable struct fields may now be annotated as const to prevent changing them after construction, providing for greater clarity and optimization ability of these objects (#43305.
Type annotations can now be added to global variables to make accessing them type stable (#43671.
Empty n-dimensional arrays can now be created using multiple semicolons inside square brackets, e.g. [;;;] creates a 0×0×0 Array (#41618.
try-blocks can now optionally have an else-block which is executed right after the main body only if no errors were thrown (#42211.
@inline and @noinline can now be placed within a function body, allowing one to annotate anonymous function (#41312.
@inline and @noinline can now be applied to a function at callsite or block to enforce the involved function calls to be (or not to be inlined (#41328.
, ?, and ? are now allowed as identifier characters (#42314.
Support for Unicode 14.0.0 (#43443.
Module(:name, false, false can be used to create a module that contains no names (it does not import Base or Core and does not contain a reference to itself (#40110, #42154.
Language changes:
Newly-created Task objects (@spawn, @async, etc. now adopt the world age for methods from their parent Task upon creation, instead of using the global latest world at start. This is done to enable inference to eventually optimize these calls. Places that wish for the old behavior may use Base.invokelatest (#41449.
Unbalanced Unicode bidirectional formatting directives are now disallowed within strings and comments, to mitigate the "trojan source" vulnerability (#42918.
Base.ifelse is now defined as a generic function rather than a builtin one, allowing packages to extend its definition (#37343.
Every assignment to a global variable now first goes through a call to convert(Any, x (or convert(T, x respectively if a type T has been declared for the global. This means great care should be taken to ensure the invariant convert(Any, x === x always holds, as this change could otherwise lead to unexpected behavior (#43671.
Builtin functions are now a bit more like generic functions, and can be enumerated with methods (#43865.
Compiler/Runtime improvements:
Bootstrapping time has been improved by about 25% (#41794.
The LLVM-based compiler has been separated from the run-time library into a new library, libjulia-codegen. It is loaded by default, so normal usage should see no changes. In deployments that do not need the compiler (e.g. system images where all needed code is precompiled, this library (and its LLVM dependency can simply be excluded (#41936.
Conditional type constraints are now be forwarded interprocedurally (i.e. propagated from caller to callee. This allows inference to understand e.g. Base.ifelse(isa(x, Int, x, 0 returns ::Int-value even if the type of x is not known (#42529.
Julia-level SROA (Scalar Replacement of Aggregates has been improved: allowing elimination of getfield calls with constant global fields (#42355, enabling elimination of mutable structs with uninitialized fields (#43208, improving performance (#43232, and handling more nested getfield calls (#43239.
Abstract call sites can now be inlined or statically resolved as long as the call site has a single matching method (#43113.
Inference now tracks various effects such as side-effectful-ness and nothrow-ness on a per-specialization basis. Code heavily dependent on constant propagation should see significant compile-time performance improvements and certain cases (e.g. calls to uninlinable functions that are nevertheless effect free should see runtime performance improvements. Effects may be overwritten manually with the Base.@assume_effects macro (#43852.
Command-line option changes:
The default behavior of observing @inbounds declarations is now an option via auto in --check-bounds=yes|no|auto (#41551.
New option --strip-metadata to remove docstrings, source location information, and local variable names when building a system image (#42513.
New option --strip-ir to remove the compiler's IR (intermediate representation of source code when building a system image. The resulting image will only work if --compile=all is used, or if all needed code is precompiled (#42925.
When the program file is - the code to be executed is read from standard in (#43191.
Multi-threading changes:
Threads.@threads now defaults to a new :dynamic schedule option which is similar to the previous behavior except that iterations will be scheduled dynamically to available worker threads rather than pinned to each thread. This behavior is more composable with (possibly nested @spawn and @threads loops (#43919, #44136.
Build system changes:
New library functions:
New function eachsplit(str for iteratively performing split(str (#39245.
New function allequal(itr for testing if all elements in an iterator are equal (#43354.
hardlink(src, dst can be used to create hard links (#41639.
setcpuaffinity(cmd, cpus can be used to set CPU affinity of sub-processes (#42469.
diskstat(path=pwd( can be used to return statistics about the disk (#42248.
New @showtime macro to show both the line being evaluated and the @time report (#42431.
The LazyString and the lazy"str" macro were added to support delayed construction of error messages in error paths (#33711.
New library features:
@time and @timev now take an optional description to allow annotating the source of time reports, e.g. @time "Evaluating foo" foo( (#42431.
range accepts either stop or length as a sole keyword argument (#39241.
precision and setprecision now accept a base keyword argument (#42428.
TCP socket objects now expose closewrite functionality and support half-open mode usage (#40783.
extrema now accepts an init keyword argument (#36265, #43604.
Iterators.countfrom now accepts any type that defines + (#37747.
Standard library changes:
Keys with value nothing are now removed from the environment in addenv (#43271.
Iterators.reverse (and hence last now supports eachline iterators (#42225.
The length function on certain ranges of certain element types no longer checks for integer overflow in most cases. The new function checked_length is now available, which will try to use checked arithmetic to error if the result may be wrapping. Or use a package such as SaferIntegers.jl when constructing the range (#40382.
Intersect returns a result with the eltype of the type-promoted eltypes of the two inputs (#41769.
Iterating an Iterators.Reverse now falls back on reversing the eachindex iterator, if possible (#43110.
InteractiveUtils:
New macro @time_imports for reporting any time spent importing packages and their dependencies, highlighting compilation and recompilation time as percentages per import (#41612,[#45064].
LinearAlgebra:
The BLAS submodule now supports the level-2 BLAS subroutine spr! (#42830.
cholesky[!] now supports LinearAlgebra.PivotingStrategy (singleton type values as its optional pivot argument: the default is cholesky(A, NoPivot( (vs. cholesky(A, RowMaximum(; the former Val{true/false}-based calls are deprecated (#41640.
The standard library LinearAlgebra.jl is now completely independent of SparseArrays.jl, both in terms of the source code as well as unit testing (#43127. As a consequence, sparse arrays are no longer (silently returned by methods from LinearAlgebra applied to Base or LinearAlgebra objects. Specifically, this results in the following breaking changes:
Concatenations involving special "sparse" matrices (*diagonal now return dense matrices; As a consequence, the D1 and D2 fields of SVD objects, constructed upon getproperty calls are now dense matrices.
arg similar(::SpecialSparseMatrix, ::Type, ::Dims returns a dense zero matrix. As a consequence, products of bi-, tri- and symmetric tridiagonal matrices with each other result in dense output. Moreover, constructing 3-arg similar matrices of special "sparse" matrices of (nonstatic matrices now fails for the lack of zero(::Type{Matrix{T}}.
Printf:
Now uses textwidth for formatting %s and %c widths (#41085.
Profile:
CPU profiling now records sample metadata including thread and task. Profile.print( has a new groupby kwarg that allows grouping by thread, task, or nested thread/task, task/thread, and threads and tasks kwargs to allow filtering. Further, percent utilization is now reported as a total or per-thread, based on whether the thread is idle or not at each sample. Profile.fetch( includes the new metadata by default. For backwards compatibility with external profiling data consumers, it can be excluded by passing include_meta=false (#41742.
The new Profile.Allocs module allows memory allocations to be profiled. The stack trace, type, and size of each allocation is recorded, and a sample_rate argument allows a tunable amount of allocations to be skipped, reducing performance overhead (#42768.
A fixed duration cpu profile can now be triggered by the user during running tasks without Profile being loaded first and the report will show during execution. On MacOS & FreeBSD press ctrl-t or raise a SIGINFO. For other platforms raise a SIGUSR1 i.e. % kill -USR1 $julia_pid. Not currently available on windows (#43179.
REPL:
RadioMenu now supports optional keybindings to directly select options (#41576.
(x, y followed by TAB displays all methods that can be called with arguments x, y, .... (The space at the beginning prevents entering help-mode. MyModule.?(x, y limits the search to MyModule. TAB requires that at least one argument have a type more specific than Any; use SHIFT-TAB instead of TAB to allow any compatible methods (#38791.
New err global variable in Main set when an expression throws an exception, akin to ans. Typing err reprints the exception information (#40642.
SparseArrays:
The code for SparseArrays has been moved from the Julia repo to the external repo at https://github.com/JuliaSparse/SparseArrays.jl. This is only a code movement and does not impact any usage (#43813.
New sparse concatenation functions sparse_hcat, sparse_vcat, and sparse_hvcat return SparseMatrixCSC output independent from the types of the input arguments. They make concatenation behavior available, in which the presence of some special "sparse" matrix argument resulted in sparse output by multiple dispatch. This is no longer possible after making LinearAlgebra.jl independent from SparseArrays.jl (#43127.
Logging:
The standard log levels BelowMinLevel, Debug, Info, Warn, Error, and AboveMaxLevel are now exported from the Logging stdlib (#40980.
Unicode:
Added function isequal_normalized to check for Unicode equivalence without explicitly constructing normalized strings (#42493.
The Unicode.normalize function now accepts a chartransform keyword that can be used to supply custom character mappings, and a Unicode.julia_chartransform function is provided to reproduce the mapping used in identifier normalization by the Julia parser (#42561.
Test:
@test_throws "some message" triggers_error( can now be used to check whether the displayed error text contains "some message" regardless of the specific exception type. Regular expressions, lists of strings, and matching functions are also supported (#41888.
@testset foo( can now be used to create a test set from a given function. The name of the test set is the name of the called function. The called function can contain @test and other @testset definitions, including to other function calls, while recording all intermediate test results (#42518.
TestLogger and LogRecord are now exported from the Test stdlib (#44080.
Distributed:
SSHManager now supports workers with csh/tcsh login shell, via addprocs( option shell=:csh (#41485.

New in Julia Language 1.8.0 Beta 1 (Feb 27, 2022)

New language features:
Mutable struct fields may now be annotated as const to prevent changing them after construction, providing for greater clarity and optimization ability of these objects (#43305).
Type annotations can now be added to global variables to make accessing them type stable (#43671).
Empty n-dimensional arrays can now be created using multiple semicolons inside square brackets, e.g. [;;;] creates a 0×0×0 Array (#41618).
try-blocks can now optionally have an else-block which is executed right after the main body only if no errors were thrown (#42211).
@inline and @noinline annotations can now be placed within a function body (#41312).
@inline and @noinline annotations can now be applied to a function call site or block to enforce the involved function calls to be (or not to be) inlined (#41312).
∀, ∃, and ∄ are now allowed as identifier characters (#42314).
Support for Unicode 14.0.0 (#43443).
Module(:name, false, false) can be used to create a module that contains no names (it does not import Base or Core and does not contain a reference to itself) ([#40110, #42154]).
Language changes:
Newly-created Task objects (@spawn, @async, etc.) now adopt the world age for methods from their parent Task upon creation, instead of using the global latest world at start. This is done to enable inference to eventually optimize these calls. Places that wish for the old behavior may use Base.invokelatest (#41449).
Unbalanced Unicode bidirectional formatting directives are now disallowed within strings and comments, to mitigate the "trojan source" vulnerability (#42918).
Base.ifelse is now defined as a generic function rather than a builtin one, allowing packages to extend its definition (#37343).
Every assignment to a global variable now first goes through a call to convert(Any, x) (or convert(T, x) respectively if a type T has been declared for the global). This means great care should be taken to ensure the invariant convert(Any, x) === x always holds, as this change could otherwise lead to unexpected behavior (#43671).
Builtin functions are now a bit more like generic functions, and can be enumerated with methods (#43865).
Compiler/Runtime improvements:
Bootstrapping time has been improved by about 25% (#41794).
The LLVM-based compiler has been separated from the run-time library into a new library, libjulia-codegen. It is loaded by default, so normal usage should see no changes. In deployments that do not need the compiler (e.g. system images where all needed code is precompiled), this library (and its LLVM dependency) can simply be excluded (#41936).
Conditional type constraints can now be forwarded interprocedurally (i.e. propagated from caller to callee) (#42529).
Julia-level SROA (Scalar Replacement of Aggregates) has been improved: allowing elimination of getfield calls with constant global fields (#42355), enabling elimination of mutable structs with uninitialized fields (#43208), improving performance (#43232), and handling more nested getfield calls (#43239).
Abstract call sites can now be inlined or statically resolved as long as the call site has a single matching method (#43113).
Inference now tracks various effects such as side-effectful-ness and nothrow-ness on a per-specialization basis. Code heavily dependent on constant propagation should see significant compile-time performance improvements and certain cases (e.g. calls to uninlinable functions that are nevertheless effect free) should see runtime performance improvements. Effects may be overwritten manually with the @Base.assume_effects macro (#43852).
Command-line option changes:
The default behavior of observing @inbounds declarations is now an option via auto in --check-bounds=yes|no|auto (#41551).
New option --strip-metadata to remove docstrings, source location information, and local variable names when building a system image (#42513).
New option --strip-ir to remove the compiler's IR (intermediate representation) of source code when building a system image. The resulting image will only work if --compile=all is used, or if all needed code is precompiled (#42925).
When the program file is - the code to be executed is read from standard in (#43191).
Multi-threading changes:
Threads.@threads now defaults to a new :dynamic schedule option which is similar to the previous behavior except that iterations will be scheduled dynamically to available worker threads rather than pinned to each thread. This behavior is more composable with (possibly nested) @spawn and @threads loops (#43919, #44136).
Build system changes:
New library functions:
New function eachsplit(str) for iteratively performing split(str) (#39245).
New function allequal(itr) for testing if all elements in an iterator are equal (#43354).
hardlink(src, dst) can be used to create hard links (#41639).
setcpuaffinity(cmd, cpus) can be used to set CPU affinity of sub-processes (#42469).
diskstat(path=pwd()) can be used to return statistics about the disk (#42248).
New @showtime macro to show both the line being evaluated and the @time report (#42431).
The LazyString and the lazy"str" macro were added to support delayed construction of error messages in error paths (#33711).
New library features:
@time and @timev now take an optional description to allow annotating the source of time reports, e.g. @time "Evaluating foo" foo() (#42431).
range accepts either stop or length as a sole keyword argument (#39241).
precision and setprecision now accept a base keyword argument (#42428).
TCP socket objects now expose closewrite functionality and support half-open mode usage (#40783).
extrema now accepts an init keyword argument (#36265, #43604).
Iterators.countfrom now accepts any type that defines + (#37747).
Standard library changes:
Keys with value nothing are now removed from the environment in addenv (#43271).
Iterators.reverse (and hence last) now supports eachline iterators (#42225).
The length function on certain ranges of certain element types no longer checks for integer overflow in most cases. The new function checked_length is now available, which will try to use checked arithmetic to error if the result may be wrapping. Or use a package such as SaferIntegers.jl when constructing the range (#40382).
Intersect returns a result with the eltype of the type-promoted eltypes of the two inputs (#41769).
Iterating an Iterators.Reverse now falls back on reversing the eachindex iterator, if possible (#43110).
InteractiveUtils:
New macro @time_imports for reporting any time spent importing packages and their dependencies (#41612).
LinearAlgebra:
The BLAS submodule now supports the level-2 BLAS subroutine spr! (#42830).
cholesky[!] now supports LinearAlgebra.PivotingStrategy (singleton type) values as its optional pivot argument: the default is cholesky(A, NoPivot()) (vs. cholesky(A, RowMaximum())); the former Val{true/false}-based calls are deprecated (#41640).
The standard library LinearAlgebra.jl is now completely independent of SparseArrays.jl, both in terms of the source code as well as unit testing (#43127). As a consequence, sparse arrays are no longer (silently) returned by methods from LinearAlgebra applied to Base or LinearAlgebra objects. Specifically, this results in the following breaking changes:
Concatenations involving special "sparse" matrices (*diagonal) now return dense matrices; As a consequence, the D1 and D2 fields of SVD objects, constructed upon getproperty calls are now dense matrices.
3-arg similar(::SpecialSparseMatrix, ::Type, ::Dims) returns a dense zero matrix. As a consequence, products of bi-, tri- and symmetric tridiagonal matrices with each other result in dense output. Moreover, constructing 3-arg similar matrices of special "sparse" matrices of (nonstatic) matrices now fails for the lack of zero(::Type{Matrix{T}}).
Printf:
Now uses textwidth for formatting %s and %c widths (#41085).
Profile:
CPU profiling now records sample metadata including thread and task. Profile.print() has a new groupby kwarg that allows grouping by thread, task, or nested thread/task, task/thread, and threads and tasks kwargs to allow filtering. Further, percent utilization is now reported as a total or per-thread, based on whether the thread is idle or not at each sample. Profile.fetch() includes the new metadata by default. For backwards compatibility with external profiling data consumers, it can be excluded by passing include_meta=false (#41742).
The new Profile.Allocs module allows memory allocations to be profiled. The stack trace, type, and size of each allocation is recorded, and a sample_rate argument allows a tunable amount of allocations to be skipped, reducing performance overhead (#42768).
A fixed duration cpu profile can now be triggered by the user during running tasks without Profile being loaded first and the report will show during execution. On MacOS & FreeBSD press ctrl-t or raise a SIGINFO. For other platforms raise a SIGUSR1 i.e. % kill -USR1 $julia_pid. Not currently available on windows (#43179).
REPL:
RadioMenu now supports optional keybindings to directly select options (#41576).
?(x, y followed by TAB displays all methods that can be called with arguments x, y, .... (The space at the beginning prevents entering help-mode.) MyModule.?(x, y limits the search to MyModule. TAB requires that at least one argument have a type more specific than Any; use SHIFT-TAB instead of TAB to allow any compatible methods (#38791).
New err global variable in Main set when an expression throws an exception, akin to ans. Typing err reprints the exception information (#40642).
SparseArrays:
The code for SparseArrays has been moved from the Julia repo to the external repo at https://github.com/JuliaSparse/SparseArrays.jl. This is only a code movement and does not impact any usage (#43813).
New sparse concatenation functions sparse_hcat, sparse_vcat, and sparse_hvcat return SparseMatrixCSC output independent from the types of the input arguments. They make concatenation behavior available, in which the presence of some special "sparse" matrix argument resulted in sparse output by multiple dispatch. This is no longer possible after making LinearAlgebra.jl independent from SparseArrays.jl (#43127).
Logging:
The standard log levels BelowMinLevel, Debug, Info, Warn, Error, and AboveMaxLevel are now exported from the Logging stdlib (#40980).
Unicode:
Added function isequal_normalized to check for Unicode equivalence without explicitly constructing normalized strings (#42493).
The Unicode.normalize function now accepts a chartransform keyword that can be used to supply custom character mappings, and a Unicode.julia_chartransform function is provided to reproduce the mapping used in identifier normalization by the Julia parser (#42561).
Test:
@test_throws "some message" triggers_error() can now be used to check whether the displayed error text contains "some message" regardless of the specific exception type. Regular expressions, lists of strings, and matching functions are also supported (#41888).
@testset foo() can now be used to create a test set from a given function. The name of the test set is the name of the called function. The called function can contain @test and other @testset definitions, including to other function calls, while recording all intermediate test results (#42518).
TestLogger and LogRecord are now exported from the Test stdlib (#44080).
Tooling Improvements:
GC.enable_logging(true) can be used to log each garbage collection, with the time it took and the amount of memory that was collected (#43511).

New in Julia Language 1.7.0 (Dec 1, 2021)

New language features:
(; a, b) = x can now be used to destructure properties a and b of x. This syntax is equivalent to a = getproperty(x, :a); b = getproperty(x, :b) (#39285).
Implicit multiplication by juxtaposition is now allowed for radical symbols (e.g. x√y and x∛y) (#40173).
The short-circuiting operators && and || can now be dotted to participate in broadcast fusion as .&& and .|| (#39594).
⫪ (U+2AEA, Top, downvDash) and ⫫ (U+2AEB, Bot, upvDash, indep) may now be used as binary operators with comparison precedence (#39403).
Repeated semicolons can now be used inside array concatenation expressions to separate dimensions of an array, with the number of semicolons specifying the dimension. Just as a single semicolon in [A; B] has always described concatenating in the first dimension (vertically), now two semicolons [A;; B] do so in the second dimension (horizontally), three semicolons ;;; in the third, and so on (#33697).
A backslash () before a newline inside a string literal now removes the newline while also respecting indentation. This can be used to split up long strings without newlines into multiple lines of code (#40753).
A backslash before a newline in command literals now always removes the newline, similar to standard string literals, whereas the result was not well-defined before (#40753).
Language changes:
macroexpand, @macroexpand, and @macroexpand1 no longer wrap errors in a LoadError. To reduce breakage, @test_throws has been modified so that many affected tests will still pass (#38379).
The middle dot · (cdotp U+00b7) and the Greek interpunct · (U+0387) are now treated as equivalent to the dot operator ⋅ (cdot U+22c5) (#25157).
The minus sign − (minus U+2212) is now treated as equivalent to the hyphen-minus sign - (U+002d) (#40948).
Destructuring will no longer mutate values on the left-hand side while iterating through values on the right-hand side. In the example of an array x, x[2], x[1] = x will now swap the first and second elements of x, whereas it used to fill both entries with x[1] because x[2] was mutated during the iteration of x (#40737).
The default random number generator has changed, so all random numbers will be different (even with the same seed) unless an explicit RNG object is used. See the section on the Random standard library below (#40546).
Iterators.peel(itr) now returns nothing when itr is empty instead of throwing a BoundsError (#39607).
Multiple successive semicolons in an array expresion were previously ignored (e.g., [1 ;; 2] == [1 ; 2]). This syntax is now used to separate dimensions (see New language features).
Compiler/Runtime improvements:
Command-line option changes:
The Julia --project option and the JULIA_PROJECT environment variable now support selecting shared environments like .julia/environments/myenv the same way the package management console does: use julia --project=@myenv resp. export JULIA_PROJECT="@myenv" (#40025).
Multi-threading changes:
Intrinsics for atomic pointer operations are now defined for certain byte sizes (#37847).
Support for declaring and using individual fields of a mutable struct as atomic has been added; see the new @atomic macro (#37847).
If the JULIA_NUM_THREADS environment variable is set to auto, then the number of threads will be set to the number of CPU threads (#38952).
Every Task object has a local random number generator state, providing reproducible (schedule-independent) execution of parallel simulation code by default. The default generator is also significantly faster in parallel than in previous versions (#40546).
Tasks can now migrate among threads when they are re-scheduled. Previously, a Task would always run on whichever thread executed it first (#40715).
Build system changes:
New library functions:
Two argument methods findmax(f, domain), argmax(f, domain) and the corresponding min versions (#35316).
isunordered(x) returns true if x is a value that is normally unordered, such as NaN or missing (#35316).
New keepat!(vector, inds) function which is the inplace equivalent of vector[inds] for a list inds of integers (#36229).
Two arguments method lock(f, lck) now accepts a Channel as the second argument (#39312).
New functor Returns(value), which returns value for any arguments (#39794).
New macros @something and @coalesce which are short-circuiting versions of something and coalesce, respectively (#40729).
New function redirect_stdio for redirecting stdin, stdout and stderr (#37978).
New macro Base.@invoke f(arg1::T1, arg2::T2; kwargs...) provides an easier syntax to call invoke(f, Tuple{T1,T2}, arg1, arg2; kwargs...) (#38438).
New macro Base.@invokelatest f(args...; kwargs...) providing a convenient way to call Base.invokelatest(f, args...; kwargs...) (#37971).
New library features:
The optional keyword argument context of sprint can now be set to a tuple of :key => value pairs to specify multiple attributes (#39381).
bytes2hex and hex2bytes are no longer limited to arguments of type Union{String,AbstractVector{UInt8}} and now only require that they're iterable and have a length (#39710).
stat(file) now has a more detailed and user-friendly show method (#39463).
Standard library changes:
count and findall now accept an AbstractChar argument to search for a character in a string (#38675).
New methods range(start, stop) and range(start, stop, length) (#39228).
range now supports start as an optional keyword argument (#38041).
Some operations on ranges will return a StepRangeLen instead of a StepRange, to allow the resulting step to be zero. Previously, λ .* (1:9) gave an error when λ = 0 (#40320).
islowercase and isuppercase are now compliant with the Unicode lower/uppercase categories (#38574).
iseven and isodd functions now support non-Integer numeric types (#38976).
escape_string now accepts a collection of characters via the keyword keep that are to be kept as they are (#38597).
getindex for NamedTuples now accepts a tuple of symbols in order to index multiple values (#38878).
Subtypes of AbstractRange now correctly follow the general array indexing behavior when indexed by Bools, erroring for scalar Bools and treating arrays (including ranges) of Bool as logical indices (#31829).
keys(::RegexMatch) is now defined to return the capture's keys, by name if named, or by index if not (#37299).
keys(::Generator) is now defined to return the iterator's keys (#34678).
RegexMatch is now iterable, giving the captured substrings (#34355).
lpad/rpad are now defined in terms of textwidth (#39044).
Test.@test now accepts broken and skip boolean keyword arguments, which mimic Test.@test_broken and Test.@test_skip behavior, but allows skipping tests failing only under certain conditions.
@lock is now exported from Base (#39588).
The experimental function Base.catch_stack() has been renamed to current_exceptions(), exported from Base and given a more specific return type (#29901).
Some degree trigonometric functions, sind, cosd, tand, asind, acosd, asecd, acscd, acotd, atand now accept a square matrix (#39758).
replace(::String) now accepts multiple patterns, which will be applied left-to-right simultaneously, so only one pattern will be applied to any character, and the patterns will only be applied to the input text, not the replacements (#40484).
New replace methods to replace elements of a Tuple (#38216).
Package Manager:
If a package is using or imported from the julia> prompt that isn't found but is available from a registry, a pkg> add prompt now offers to install the package into the current environment, precompile it, and continue to load it (#39026).
A new Manifest.toml format is now used that captures extensible metadata fields, including the julia version that generated the manifest. Old format manifests are still supported and will be maintained in their original format, unless the user runs Pkg.upgrade_manifest() to upgrade the format of the current environment's manifest without re-resolving (#40765).
pkg> precompile will now precompile new versions of packages that are already loaded, rather than postponing to the next session (the ?-marked dependencies) (#40345).
pkg> rm, pin, and free now accept the --all argument to call the action on all packages.
Registries downloaded from the Pkg Server (not git) are no longer uncompressed into files but instead read directly from the compressed tarball into memory. This improves performance on filesystems which do not handle a large number of files well. To turn this feature off, set the environment variable JULIA_PKG_UNPACK_REGISTRY=true.
It is now possible to use an external git executable instead of the default libgit2 library for the downloads that happen via the Git protocol by setting the environment variable JULIA_PKG_USE_CLI_GIT=true.
Registries downloaded from the Pkg Server (not git) is now assumed to be immutable. Manual changes to their files might not be picked up by a running Pkg session.
Adding packages by directory name in the REPL mode now requires prepending ./ to the name if the package is in the current directory; e.g. add ./Package is required instead of add Package. This is to avoid confusion between the package name Package and the local directory Package.
The mode keyword for PackageSpec has been removed.
LinearAlgebra:
Use Libblastrampoline to pick a BLAS and LAPACK at runtime. By default it forwards to OpenBLAS in the Julia distribution. The forwarding mechanism can be used by packages to replace the BLAS and LAPACK with user preferences (#39455).
On aarch64, OpenBLAS now uses an ILP64 BLAS like all other 64-bit platforms (#39436).
OpenBLAS is updated to 0.3.13 (#39216).
SuiteSparse is updated to 5.8.1 (#39455).
The shape of an UpperHessenberg matrix is preserved under certain arithmetic operations, e.g. when multiplying or dividing by an UpperTriangular matrix (#40039).
Real quasitriangular Schur factorizations S can now be efficiently converted to complex upper-triangular form with Schur{Complex}(S) (#40573).
cis(A) now supports matrix arguments (#40194).
dot now supports UniformScaling with AbstractMatrix (#40250).
qr[!] and lu[!] now support LinearAlgebra.PivotingStrategy (singleton type) values as their optional pivot argument: defaults are qr(A, NoPivot()) (vs. qr(A, ColumnNorm()) for pivoting) and lu(A, RowMaximum()) (vs. lu(A, NoPivot()) without pivoting); the former Val{true/false}-based calls are deprecated (#40623).
det(M::AbstractMatrix{BigInt}) now calls det_bareiss(M), which uses the Bareiss algorithm to calculate precise values (#40868).
Markdown:
Printf:
Random:
The default random number generator has been changed from Mersenne Twister to Xoshiro256++. The new generator has smaller state, better performance, and superior statistical properties. This generator is the one used for reproducible Task-local randomness (#40546).
REPL:
Long strings are now elided using the syntax "head" ⋯ 12345 bytes ⋯ "tail" when displayed in the REPL (#40736).
Pasting repl examples into the repl (prompt pasting) now supports all repl modes (julia, pkg, shell, help?) and switches mode automatically (#40604).
help?> for modules without docstrings now returns a list of exported names and prints the contents of an associated README.md if found (#39093).
SparseArrays:
new sizehint!(::SparseMatrixCSC, ::Integer) method (#30676).
cholesky() now fully preserves the user-specified permutation (#40560).
issparse now applies consistently to all wrapper arrays, including nested, by checking issparse on the wrapped parent array (#37644).
Dates:
The Dates.periods function can be used to get the Vector of Periods that comprise a CompoundPeriod (#39169).
Downloads:
If a cookie header is set in a redirected request, the cookie will now be sent in following requests (https://github.com/JuliaLang/Downloads.jl/pull/98).
If a ~/.netrc file exists, it is used to get passwords for authenticated websites (https://github.com/JuliaLang/Downloads.jl/pull/98).
Server Name Indication is now sent with all TLS connections, even when the server's identity is not verified (see NetworkOptions; https://github.com/JuliaLang/Downloads.jl/pull/114).
When verifying TLS connections on Windows, if the certificate revocation server cannot be reached, the connection is allowed; this matches what other applications do and how revocation is performed on macOS (https://github.com/JuliaLang/Downloads.jl/pull/115).
There is now a 30-second connection timeout and a 20-second timeout if no data is sent; in combination, this guarantees that connections must make some progress or they will timeout in under a minute (https://github.com/JuliaLang/Downloads.jl/pull/126).
Statistics:
Sockets:
Tar:
Tar.extract now ignores the exact permission mode in a tarball and normalizes modes in the same way that Tar.create does, which is, in turn the same way that git normalizes them (https://github.com/JuliaIO/Tar.jl/pull/99).
Functions that consume tarballs now handle hard links: the link target must be a previously seen file; Tar.list lists the entry with :hardlink type and .link field giving the path to the target; other functions — Tar.extract, Tar.rewrite, Tar.tree_hash — treat a hard link as a copy of the target file (https://github.com/JuliaIO/Tar.jl/pull/102).
The standard format generated by Tar.create and Tar.rewrite now includes entries for non-empty directories; this shouldn't be neccessary, but some tools that consume tarballs (including docker) are confused by the absence of these directory entries (https://github.com/JuliaIO/Tar.jl/pull/106).
Tar now accepts tarballs with leading spaces in octal integer header fields: this is technically not a valid format according to the POSIX spec, but old Solaris tar commands produced tarballs like this so this format does occur in the wild, and it seems harmless to accept it (https://github.com/JuliaIO/Tar.jl/pull/116).
Tar.extract now takes a set_permissions keyword argument, which defaults to true; if false is passed instead, the permissions of extracted files are not modified on extraction (https://github.com/JuliaIO/Tar.jl/pull/113).
Distributed:
UUIDs:
Mmap:
mmap is now exported (#39816).
DelimitedFiles:
readdlm now defaults to use_mmap=false on all OSes for consistent reliability in abnormal filesystem situations (#40415).

New in Julia Language 1.6.0 (Mar 26, 2021)

New language features:
Types written with where syntax can now be used to define constructors, e.g. (Foo{T} where T)(x) = ....
<-- and <--> are now available as infix operators, with the same precedence and associativity as other arrow-like operators (#36666).
Compilation and type inference can now be enabled or disabled at the module level using the experimental macro Base.Experimental.@compiler_options (#37041).
The library name passed to ccall or @ccall can now be an expression involving global variables and function calls. The expression will be evaluated the first time the ccall executes (#36458).
? (U+A71B), ? (U+A71C) and ? (U+A71D) can now also be used as operator suffixes. They can be tab-completed from ^uparrow, ^downarrow and ^! in the REPL (#37542).
Standalone "dotted" operators now get parsed as Expr(:., :op), which gets lowered to Base.BroadcastFunction(op). This means .op is functionally equivalent to (x...) -> (op).(x...), which can be useful for passing the broadcasted version of an operator to higher-order functions, for example map(.*, A, B) for an elementwise product of two arrays of arrays (#37583).
The syntax import A as B (plus import A: x as y, import A.x as y, and using A: x as y) can now be used to rename imported modules and identifiers (#1255).
Unsigned literals (starting with 0x) which are too big to fit in a UInt128 object are now interpreted as BigInt (#23546).
It is now possible to use ... on the left-hand side of assignments for taking any number of items from the front of an iterable collection, while also collecting the rest, for example a, b... = [1, 2, 3]. This syntax is implemented using Base.rest, which can be overloaded to customize its behavior for different collection types (#37410).
Language changes:
The postfix conjugate transpose operator ' now accepts Unicode modifiers as suffixes, so e.g. a'? is parsed as var"'?"(a), which can be defined by the user. a'? parsed as a' * ? before, so this is a minor breaking change (#37247).
Macros that return :quote expressions (e.g. via Expr(:quote, ...)) were previously able to work without escaping (esc(...)) their output when needed. This has been corrected, and now esc must be used in these macros as it is in other macros (#37540).
The --> operator now lowers to a :call expression, so it can be defined as a function like other operators. The dotted version .--> is now parsed as well. For backwards compatibility, --> still parses using its own expression head instead of :call.
The a[begin, k] syntax now calls firstindex(a, 1) rather than first(axes(a, 1)) (#35779), but the former now defaults to the latter for any a (#38742).
? (U+233F) and ¦ (U+00A6) are now infix operators with times-like and plus-like precedence, respectively. Previously they were parsed as identifier characters (#37973).
Compiler/Runtime improvements:
All platforms can now use @executable_path within jl_load_dynamic_library(). This allows executable-relative paths to be embedded within executables on all platforms, not just MacOS, which the syntax is borrowed from (#35627).
Constant propagation now occurs through keyword arguments (#35976).
The precompilation cache is now created atomically (#36416). Invoking n Julia processes simultaneously may create n temporary caches.
Command-line option changes:
There is no longer a concept of "home project": starting julia --project=dir is now exactly equivalent to starting julia and then doing pkg> activate $dir and julia --project is exactly equivalent to doing that where dir = Base.current_project(). In particular, this means that if you do pkg> activate after starting julia with the --project option (or with JULIA_PROJECT set) it will take you to the default active project, which is @v1.6 unless you have modified LOAD_PATH (#36434).
Multi-threading changes:
Locks now automatically inhibit finalizers from running, to avoid deadlock (#38487).
New function Base.Threads.foreach(f, channel::Channel) for multithreaded Channel consumption (#34543).
Build system changes:
Windows Installer now has the option to 'Add Julia to Path'. To unselect this option from the commandline simply remove the tasks you do not want to be installed: e.g. ./julia-installer.exe /TASKS="desktopicon,startmenu,addtopath", adds a desktop icon, a startmenu group icon, and adds Julia to system PATH.
New library functions:
New function Base.kron! and corresponding overloads for various matrix types for performing Kronecker product in-place (#31069).
New function Base.readeach(io, T) for iteratively performing read(io, T) (#36150).
Iterators.map is added. It provides another syntax Iterators.map(f, iterators...) for writing (f(args...) for args in zip(iterators...)), i.e. a lazy map (#34352).
New function sincospi for simultaneously computing sinpi(x) and cospi(x) more efficiently (#35816).
New function cispi(x) for more accurately computing cis(pi * x) (#38449).
New function addenv for adding environment mappings into a Cmd object, returning the new Cmd object.
New function insorted for determining whether an element is in a sorted collection or not (#37490).
New function Base.rest for taking the rest of a collection, starting from a specific iteration state, in a generic way (#37410).
New library features:
The redirect_* functions now accept devnull to discard all output redirected to it, and as an empty input (#36146).
The redirect_* functions can now be called on IOContext objects (#36688).
findfirst, findnext, findlast, and findall now support AbstractVector{<:Union{Int8,UInt8}} (pattern, array) arguments (#37283).
New constructor NamedTuple(iterator) that constructs a named tuple from a key-value pair iterator.
A new reinterpret(reshape, T, a::AbstractArray{S}) reinterprets a to have eltype T while potentially inserting or consuming the first dimension depending on the ratio of sizeof(T) and sizeof(S).
New append!(vector, collections...) and prepend!(vector, collections...) methods accept multiple collections to be appended or prepended (#36227).
keys(io::IO) has been added, which returns all keys of io if io is an IOContext and an empty Base.KeySet otherwise (#37753).
count now accepts an optional init argument to control the accumulation type (#37461).
New method occursin(haystack) that returns a function that checks whether its argument occurs in haystack (#38475).
New methods ?(collection), ?(item), and ?(item) returning corresponding containment-testing functions (#38475).
The nextprod function now accepts tuples and other array types for its first argument (#35791).
The reverse(A; dims) function for multidimensional A can now reverse multiple dimensions at once by passing a tuple for dims, and defaults to reversing all dimensions; there is also a multidimensional in-place reverse!(A; dims) (#37367).
The function isapprox(x,y) now accepts the norm keyword argument also for numeric (i.e., non-array) arguments x and y (#35883).
ispow2(x) now supports non-Integer arguments x (#37635).
view, @view, and @views now work on AbstractStrings, returning a SubString when appropriate (#35879).
All AbstractUnitRange{<:Integer}s now work with SubString, view, @view and @views on strings (#35879).
sum, prod, maximum, and minimum now support init keyword argument (#36188, #35839).
unique(f, itr; seen=Set{T}()) now allows you to declare the container type used for keeping track of values returned by f on elements of itr (#36280).
first and last functions now accept an integer as second argument to get that many leading or trailing elements of any iterable (#34868).
CartesianIndices now supports step different from 1. It can also be constructed from three CartesianIndexes I, S, J using I:S:J. step for CartesianIndices now returns a CartesianIndex (#37829).
RegexMatch objects can now be probed for whether a named capture group exists within it through haskey() (#36717).
For consistency haskey(r::RegexMatch, i::Integer) has also been added and returns if the capture group for i exists (#37300).
Standard library changes:
A new standard library TOML has been added for parsing and printing TOML files (#37034).
A new standard library Downloads has been added, which replaces the old Base.download function with Downloads.download, providing cross-platform, multi-protocol, in-process download functionality implemented with libcurl (#37340).
Libdl has been moved to Base.Libc.Libdl, however it is still accessible as an stdlib (#35628).
To download artifacts lazily, LazyArtifacts now must be explicitly listed as a dependency, to avoid needing the support machinery to be available when it is not commonly needed (#37844).
It is no longer possible to create a LinRange, StepRange, or StepRangeLen with a <: Integer eltype but non-integer step (#32439).
intersect on CartesianIndices now returns CartesianIndices instead of Vector{<:CartesianIndex} (#36643).
push!(c::Channel, v) now returns channel c. Previously, it returned the pushed value v (#34202).
The composition operator ° now returns a Base.ComposedFunction instead of an anonymous function (#37517).
Logging (such as @warn) no longer catches exceptions in the logger itself (#36600).
@time now reports if the time presented included any compilation time, which is shown as a percentage (#37678).
@varinfo can now report non-exported objects within modules, look recursively into submodules, and return a sorted results table (#38042).
@testset now supports the option verbose to show the test result summary of the children even if they all pass (#33755).
In LinearIndices(::Tuple) and CartesianIndices(::Tuple), integers (as opposed to ranges of integers) in the argument tuple now consistently describe 1-based ranges, e.g, CartesianIndices((3, 1:3)) is equivalent to CartesianIndices((1:3, 1:3)). This is how tuples of integers have always been documented to work, but a bug had caused erroneous behaviors with heterogeneous tuples containing both integers and ranges (#37829, [#37928]).
Package Manager:
pkg> precompile is now parallelized through depth-first precompilation of dependencies. Errors will only throw for direct dependencies listed in the Project.toml.
pkg> precompile is now automatically triggered whenever Pkg changes the active manifest. Auto-precompilation will remember if a package has errored within the given environment and will not retry until it changes. Auto-precompilation can be gracefully interrupted with a ctrl-c and disabled by setting the environment variable JULIA_PKG_PRECOMPILE_AUTO=0.
The Pkg.BinaryPlatforms module has been moved into Base as Base.BinaryPlatforms and heavily reworked. Applications that want to be compatible with the old API should continue to import Pkg.BinaryPlatforms, however new users should use Base.BinaryPlatforms directly (#37320).
The Pkg.Artifacts module has been imported as a separate standard library. It is still available as Pkg.Artifacts, however starting from Julia v1.6+, packages may import simply Artifacts without importing all of Pkg alongside (#37320).
LinearAlgebra:
New method LinearAlgebra.issuccess(::CholeskyPivoted) for checking whether pivoted Cholesky factorization was successful (#36002).
UniformScaling can now be indexed into using ranges to return dense matrices and vectors (#24359).
New function LinearAlgebra.BLAS.get_num_threads() for getting the number of BLAS threads (#36360).
(+)(::UniformScaling) is now defined, making +I a valid unary operation (#36784).
Instances of UniformScaling are no longer isequal to matrices. Previous behaviour violated the rule that isequal(x, y) implies hash(x) == hash(y).
Transposing *Triangular matrices now returns matrices of the opposite triangular type, consistently with adjoint!(::*Triangular) and transpose!(::*Triangular). Packages containing methods with, e.g., Adjoint{<:Any,<:LowerTriangular{<:Any,<:OwnMatrixType}} should replace that by UpperTriangular{<:Any,<:Adjoint{<:Any,<:OwnMatrixType}} in the method signature (#38168).
Markdown:
Printf:
Complete overhaul of internal code to use the ryu float printing algorithms (from Julia 1.4); leads to consistent 2-5x performance improvements.
New Printf.tofloat function allowing custom float types to more easily integrate with Printf formatting by converting their type to Float16, Float32, Float64, or BigFloat.
New Printf.format"..." and Printf.Format(...) functions that allow creating Printf.Format objects that can be passed to Printf.format for easier dynamic printf formatting.
Printf.format(f::Printf.Format, args...) as a non-macro function that applies a printf format f to provided args.
Random:
REPL:
The AbstractMenu extension interface of REPL.TerminalMenus has been extensively overhauled. The new interface does not rely on global configuration variables, is more consistent in delegating printing of the navigation/selection markers, and provides improved support for dynamic menus. These changes are compatible with the previous (deprecated) interface, so are non-breaking.
The new API offers several enhancements:
Menus are configured in their constructors via keyword arguments.
For custom menu types, the new Config and MultiSelectConfig replace the global CONFIG Dict.
request(menu; cursor=1) allows you to control the initial cursor position in the menu (defaults to first item).
MultiSelectMenu allows you to pass a list of initially-selected items with the selected keyword argument.
writeLine was deprecated to writeline, and writeline methods are not expected to print the cursor indicator. The old writeLine continues to work, and any of its method extensions should print the cursor indicator as before.
printMenu has been deprecated to printmenu, and it both accepts a state input and returns a state output that controls the number of terminal lines erased when the menu is next refreshed. This plus related changes makes printmenu work properly when the number of menu items might change depending on user choices.
numoptions, returning the number of items in the menu, has been added as an alternative to implementing options.
suppress_output (primarily a testing option) has been added as a keyword argument to request, rather than a configuration option.
Tab completion now supports runs of consecutive sub/superscript characters, e.g. ^(3) tab-completes to ?³? (#38649).
Windows REPL now supports 24-bit colors, by correctly interpreting virtual terminal escapes.
SparseArrays:
Display large sparse matrices with a Unicode "spy" plot of their nonzero patterns, and display small sparse matrices by an Matrix-like 2d layout of their contents (#33821).
New convenient spdiagm([m, n,] v::AbstractVector) methods which call spdiagm([m, n,] 0 => v), consistently with their dense diagm counterparts (#37684).
Dates:
Quarter period is defined (#35519).
canonicalize can now take Period as an input (#37391).
Zero-valued FixedPeriods and OtherPeriods now compare equal, e.g., Year(0) == Day(0). The behavior of non-zero Periods is not changed (#37486).
Statistics:
Sockets:
Distributed:
Now supports invoking Windows workers via ssh (via new keyword argument shell=:wincmd in addprocs) (#30614).
Other new keyword arguments in addprocs: ssh to specify the ssh client path, env to pass environment variables to workers, and cmdline_cookie to work around an ssh problem with Windows workers that run older (pre-ConPTY) versions of Windows, Julia or OpenSSH (#30614).
UUIDs:
Change uuid1 and uuid4 to use Random.RandomDevice() as default random number generator (#35872).
Added parse(::Type{UUID}, ::AbstractString) method.
Mmap:
On Unix systems, the Mmap.madvise! function (along with OS-specific Mmap.MADV_* constants) has been added to give advice on handling of memory-mapped arrays (#37369).
Deprecated or removed:
The Base.download function has been deprecated (silently, by default) in favor of the new Downloads.download standard library function (#37340).
The Base.Grisu code has been officially removed (float printing was switched to the ryu algorithm code in 1.4). The code is available from JuliaAttic if needed.

New in Julia Language 1.6.0 RC 1 (Feb 8, 2021)

New language features:
Types written with where syntax can now be used to define constructors, e.g. (Foo{T} where T)(x) = ....
<-- and <--> are now available as infix operators, with the same precedence and associativity as other arrow-like operators (#36666).
Compilation and type inference can now be enabled or disabled at the module level using the experimental macro Base.Experimental.@compiler_options (#37041).
The library name passed to ccall or @ccall can now be an expression involving global variables and function calls. The expression will be evaluated the first time the ccall executes (#36458).
? (U+A71B), ? (U+A71C) and ? (U+A71D) can now also be used as operator suffixes. They can be tab-completed from ^uparrow, ^downarrow and ^! in the REPL (#37542).
Standalone "dotted" operators now get parsed as Expr(:., :op), which gets lowered to Base.BroadcastFunction(op). This means .op is functionally equivalent to (x...) -> (op).(x...), which can be useful for passing the broadcasted version of an operator to higher-order functions, for example map(.*, A, B) for an elementwise product of two arrays of arrays (#37583).
The syntax import A as B (plus import A: x as y, import A.x as y, and using A: x as y) can now be used to rename imported modules and identifiers (#1255).
Unsigned literals (starting with 0x) which are too big to fit in a UInt128 object are now interpreted as BigInt (#23546).
It is now possible to use ... on the left-hand side of assignments for taking any number of items from the front of an iterable collection, while also collecting the rest, for example a, b... = [1, 2, 3]. This syntax is implemented using Base.rest, which can be overloaded to customize its behavior for different collection types (#37410).
Language changes:
The postfix conjugate transpose operator ' now accepts Unicode modifiers as suffixes, so e.g. a'? is parsed as var"'?"(a), which can be defined by the user. a'? parsed as a' * ? before, so this is a minor breaking change (#37247).
Macros that return :quote expressions (e.g. via Expr(:quote, ...)) were previously able to work without escaping (esc(...)) their output when needed. This has been corrected, and now esc must be used in these macros as it is in other macros (#37540).
The --> operator now lowers to a :call expression, so it can be defined as a function like other operators. The dotted version .--> is now parsed as well. For backwards compatibility, --> still parses using its own expression head instead of :call.
The a[begin, k] syntax now calls firstindex(a, 1) rather than first(axes(a, 1)) (#35779), but the former now defaults to the latter for any a (#38742).
? (U+233F) and ¦ (U+00A6) are now infix operators with times-like and plus-like precedence, respectively. Previously they were parsed as identifier characters (#37973).
Compiler/Runtime improvements:
All platforms can now use @executable_path within jl_load_dynamic_library(). This allows executable-relative paths to be embedded within executables on all platforms, not just MacOS, which the syntax is borrowed from (#35627).
Constant propagation now occurs through keyword arguments (#35976).
The precompilation cache is now created atomically (#36416). Invoking n Julia processes simultaneously may create n temporary caches.
Command-line option changes:
There is no longer a concept of "home project": starting julia --project=dir is now exactly equivalent to starting julia and then doing pkg> activate $dir and julia --project is exactly equivalent to doing that where dir = Base.current_project(). In particular, this means that if you do pkg> activate after starting julia with the --project option (or with JULIA_PROJECT set) it will take you to the default active project, which is @v1.6 unless you have modified LOAD_PATH (#36434).
Multi-threading changes:
Locks now automatically inhibit finalizers from running, to avoid deadlock (#38487).
New function Base.Threads.foreach(f, channel::Channel) for multithreaded Channel consumption (#34543).
Build system changes:
Windows Installer now has the option to 'Add Julia to Path'. To unselect this option from the commandline simply remove the tasks you do not want to be installed: e.g. ./julia-installer.exe /TASKS="desktopicon,startmenu,addtopath", adds a desktop icon, a startmenu group icon, and adds Julia to system PATH.
New library functions:
New function Base.kron! and corresponding overloads for various matrix types for performing Kronecker product in-place (#31069).
New function Base.readeach(io, T) for iteratively performing read(io, T) (#36150).
Iterators.map is added. It provides another syntax Iterators.map(f, iterators...) for writing (f(args...) for args in zip(iterators...)), i.e. a lazy map (#34352).
New function sincospi for simultaneously computing sinpi(x) and cospi(x) more efficiently (#35816).
New function cispi(x) for more accurately computing cis(pi * x) (#38449).
New function addenv for adding environment mappings into a Cmd object, returning the new Cmd object.
New function insorted for determining whether an element is in a sorted collection or not (#37490).
New function Base.rest for taking the rest of a collection, starting from a specific iteration state, in a generic way (#37410).
New library features:
The redirect_* functions now accept devnull to discard all output redirected to it, and as an empty input (#36146).
The redirect_* functions can now be called on IOContext objects (#36688).
findfirst, findnext, findlast, and findall now support AbstractVector{<:Union{Int8,UInt8}} (pattern, array) arguments (#37283).
New constructor NamedTuple(iterator) that constructs a named tuple from a key-value pair iterator.
A new reinterpret(reshape, T, a::AbstractArray{S}) reinterprets a to have eltype T while potentially inserting or consuming the first dimension depending on the ratio of sizeof(T) and sizeof(S).
New append!(vector, collections...) and prepend!(vector, collections...) methods accept multiple collections to be appended or prepended (#36227).
The postfix operator '? can now be used as an alias for transpose (#38062).
keys(io::IO) has been added, which returns all keys of io if io is an IOContext and an empty Base.KeySet otherwise (#37753).
count now accepts an optional init argument to control the accumulation type (#37461).
New method occursin(haystack) that returns a function that checks whether its argument occurs in haystack (#38475).
New methods ?(collection), ?(item), and ?(item) returning corresponding containment-testing functions (#38475).
The nextprod function now accepts tuples and other array types for its first argument (#35791).
The reverse(A; dims) function for multidimensional A can now reverse multiple dimensions at once by passing a tuple for dims, and defaults to reversing all dimensions; there is also a multidimensional in-place reverse!(A; dims) (#37367).
The function isapprox(x,y) now accepts the norm keyword argument also for numeric (i.e., non-array) arguments x and y (#35883).
ispow2(x) now supports non-Integer arguments x (#37635).
view, @view, and @views now work on AbstractStrings, returning a SubString when appropriate (#35879).
All AbstractUnitRange{<:Integer}s now work with SubString, view, @view and @views on strings (#35879).
sum, prod, maximum, and minimum now support init keyword argument (#36188, #35839).
unique(f, itr; seen=Set{T}()) now allows you to declare the container type used for keeping track of values returned by f on elements of itr (#36280).
first and last functions now accept an integer as second argument to get that many leading or trailing elements of any iterable (#34868).
CartesianIndices now supports step different from 1. It can also be constructed from three CartesianIndexes I, S, J using I:S:J. step for CartesianIndices now returns a CartesianIndex (#37829).
RegexMatch objects can now be probed for whether a named capture group exists within it through haskey() (#36717).
For consistency haskey(r::RegexMatch, i::Integer) has also been added and returns if the capture group for i exists (#37300).
Standard library changes:
A new standard library TOML has been added for parsing and printing TOML files (#37034).
A new standard library Downloads has been added, which replaces the old Base.download function with Downloads.download, providing cross-platform, multi-protocol, in-process download functionality implemented with libcurl (#37340).
Libdl has been moved to Base.Libc.Libdl, however it is still accessible as an stdlib (#35628).
To download artifacts lazily, LazyArtifacts now must be explicitly listed as a dependency, to avoid needing the support machinery to be available when it is not commonly needed (#37844).
It is no longer possible to create a LinRange, StepRange, or StepRangeLen with a <: Integer eltype but non-integer step (#32439).
intersect on CartesianIndices now returns CartesianIndices instead of Vector{<:CartesianIndex} (#36643).
push!(c::Channel, v) now returns channel c. Previously, it returned the pushed value v (#34202).
The composition operator ° now returns a Base.ComposedFunction instead of an anonymous function (#37517).
Logging (such as @warn) no longer catches exceptions in the logger itself (#36600).
@time now reports if the time presented included any compilation time, which is shown as a percentage (#37678).
@varinfo can now report non-exported objects within modules, look recursively into submodules, and return a sorted results table (#38042).
@testset now supports the option verbose to show the test result summary of the children even if they all pass (#33755).
Package Manager:
pkg> precompile is now parallelized through depth-first precompilation of dependencies. Errors will only throw for direct dependencies listed in the Project.toml.
pkg> precompile is now automatically triggered whenever Pkg changes the active manifest. Auto-precompilation will remember if a package has errored within the given environment and will not retry until it changes. Auto-precompilation can be gracefully interrupted with a ctrl-c and disabled by setting the environment variable JULIA_PKG_PRECOMPILE_AUTO=0.
The Pkg.BinaryPlatforms module has been moved into Base as Base.BinaryPlatforms and heavily reworked. Applications that want to be compatible with the old API should continue to import Pkg.BinaryPlatforms, however new users should use Base.BinaryPlatforms directly (#37320).
The Pkg.Artifacts module has been imported as a separate standard library. It is still available as Pkg.Artifacts, however starting from Julia v1.6+, packages may import simply Artifacts without importing all of Pkg alongside (#37320).
LinearAlgebra:
New method LinearAlgebra.issuccess(::CholeskyPivoted) for checking whether pivoted Cholesky factorization was successful (#36002).
UniformScaling can now be indexed into using ranges to return dense matrices and vectors (#24359).
New function LinearAlgebra.BLAS.get_num_threads() for getting the number of BLAS threads (#36360).
(+)(::UniformScaling) is now defined, making +I a valid unary operation (#36784).
Instances of UniformScaling are no longer isequal to matrices. Previous behaviour violated the rule that isequal(x, y) implies hash(x) == hash(y).
Transposing *Triangular matrices now returns matrices of the opposite triangular type, consistently with adjoint!(::*Triangular) and transpose!(::*Triangular). Packages containing methods with, e.g., Adjoint{<:Any,<:LowerTriangular{<:Any,<:OwnMatrixType}} should replace that by UpperTriangular{<:Any,<:Adjoint{<:Any,<:OwnMatrixType}} in the method signature (#38168).
Markdown:
Printf:
Complete overhaul of internal code to use the ryu float printing algorithms (from Julia 1.4); leads to consistent 2-5x performance improvements.
New Printf.tofloat function allowing custom float types to more easily integrate with Printf formatting by converting their type to Float16, Float32, Float64, or BigFloat.
New Printf.format"..." and Printf.Format(...) functions that allow creating Printf.Format objects that can be passed to Printf.format for easier dynamic printf formatting.
Printf.format(f::Printf.Format, args...) as a non-macro function that applies a printf format f to provided args.
Random:
REPL:
The AbstractMenu extension interface of REPL.TerminalMenus has been extensively overhauled. The new interface does not rely on global configuration variables, is more consistent in delegating printing of the navigation/selection markers, and provides improved support for dynamic menus. These changes are compatible with the previous (deprecated) interface, so are non-breaking.
The new API offers several enhancements:
Menus are configured in their constructors via keyword arguments.
For custom menu types, the new Config and MultiSelectConfig replace the global CONFIG Dict.
request(menu; cursor=1) allows you to control the initial cursor position in the menu (defaults to first item).
MultiSelectMenu allows you to pass a list of initially-selected items with the selected keyword argument.
writeLine was deprecated to writeline, and writeline methods are not expected to print the cursor indicator. The old writeLine continues to work, and any of its method extensions should print the cursor indicator as before.
printMenu has been deprecated to printmenu, and it both accepts a state input and returns a state output that controls the number of terminal lines erased when the menu is next refreshed. This plus related changes makes printmenu work properly when the number of menu items might change depending on user choices.
numoptions, returning the number of items in the menu, has been added as an alternative to implementing options.
suppress_output (primarily a testing option) has been added as a keyword argument to request, rather than a configuration option.
Tab completion now supports runs of consecutive sub/superscript characters, e.g. ^(3) tab-completes to ?³? (#38649).
Windows REPL now supports 24-bit colors, by correctly interpreting virtual terminal escapes.
SparseArrays:
Display large sparse matrices with a Unicode "spy" plot of their nonzero patterns, and display small sparse matrices by an Matrix-like 2d layout of their contents (#33821).
New convenient spdiagm([m, n,] v::AbstractVector) methods which call spdiagm([m, n,] 0 => v), consistently with their dense diagm counterparts (#37684).
Dates:
Quarter period is defined (#35519).
canonicalize can now take Period as an input (#37391).
Zero-valued FixedPeriods and OtherPeriods now compare equal, e.g., Year(0) == Day(0). The behavior of non-zero Periods is not changed (#37486).
Statistics:
Sockets:
Distributed:
Now supports invoking Windows workers via ssh (via new keyword argument shell=:wincmd in addprocs) (#30614).
Other new keyword arguments in addprocs: ssh to specify the ssh client path, env to pass environment variables to workers, and cmdline_cookie to work around an ssh problem with Windows workers that run older (pre-ConPTY) versions of Windows, Julia or OpenSSH (#30614).
UUIDs:
Change uuid1 and uuid4 to use Random.RandomDevice() as default random number generator (#35872).
Added parse(::Type{UUID}, ::AbstractString) method.
Mmap:
On Unix systems, the Mmap.madvise! function (along with OS-specific Mmap.MADV_* constants) has been added to give advice on handling of memory-mapped arrays (#37369).
Deprecated or removed:
The Base.download function has been deprecated (silently, by default) in favor of the new Downloads.download standard library function (#37340).
The Base.Grisu code has been officially removed (float printing was switched to the ryu algorithm code in 1.4). The code is available from JuliaAttic if needed.

New in Julia Language 1.5.1 (Aug 27, 2020)

New language features:
Macro calls @foo {...} can now also be written @foo{...} (without the space) (#34498).
⨟ is now parsed as a binary operator with times precedence. It can be entered in the REPL with bbsemi followed by TAB (#34722).
± and ∓ are now unary operators as well, like + or -. Attention has to be paid in macros and matrix constructors, which are whitespace sensitive, because expressions like [a ±b] now get parsed as [a ±(b)] instead of [±(a, b)] (#34200).
Passing an identifier x by itself as a keyword argument or named tuple element is equivalent to x=x, implicitly using the name of the variable as the keyword or named tuple field name. Similarly, passing an a.b expression uses b as the keyword or field name (#29333).
Support for Unicode 13.0.0 (via utf8proc 2.5) (#35282).
The compiler optimization level can now be set per-module using the experimental macro Base.Experimental.@optlevel n. For code that is not performance-critical, setting this to 0 or 1 can provide significant latency improvements (#34896).
Language changes:
The interactive REPL now uses "soft scope" for top-level expressions: an assignment inside a scope block such as a for loop automatically assigns to a global variable if one has been defined already. This matches the behavior of Julia versions 0.6 and prior, as well as IJulia. Note that this only affects expressions interactively typed or pasted directly into the default REPL (#28789, #33864).
Outside of the REPL (e.g. in a file), assigning to a variable within a top-level scope block is considered ambiguous if a global variable with the same name exists. A warning is given if that happens, to alert you that the code will work differently than in the REPL. A new command line option --warn-scope controls this warning (#33864).
Converting arbitrary tuples to NTuple, e.g. convert(NTuple, (1, "")) now gives an error, where it used to be incorrectly allowed. This is because NTuple refers only to homogeneous tuples (this meaning has not changed) (#34272).
The syntax (;) (which was deprecated in v1.4) now creates an empty named tuple (#30115).
@inline macro can now be applied to short-form anonymous functions (#34953).
In triple-quoted string literals, whitespace stripping is now done before processing escape sequences instead of after. For example, the syntax
"""
an b"""
used to yield the string " anb", since the single space before b set the indent level. Now the result is "an b", since the space before b is no longer considered to occur at the start of a line. The old behavior is considered a bug (#35001).
<: and >: can now be broadcasted over arrays with .<: and .>: ([#35085])
The line number of function definitions is now added by the parser as an additional LineNumberNode at the start of each function body (#35138).
Statements of the form a' now get lowered to var"'"(a) instead of Base.adjoint(a). This allows for shadowing this function in local scopes, although this is generally discouraged. By default, Base exports var"'" as an alias of Base.adjoint, so custom types should still extend Base.adjoint (#34634).
Compiler/Runtime improvements:
Immutable structs (including tuples) that contain references can now be allocated on the stack, and allocated inline within arrays and other structs (#33886). This significantly reduces the number of heap allocations in some workloads. Code that requires assumptions about object layout and addresses (usually for interoperability with C or other languages) might need to be updated; for example any object that needs a stable address should be a mutable struct. As a result, Array views no longer allocate ([#34126]).
Command-line option changes:
Deprecation warnings are no longer shown by default. i.e. if the --depwarn=... flag is not passed it defaults to --depwarn=no. The warnings are printed from tests run by Pkg.test() (#35362).
Color now defaults to on when stdout and stderr are TTYs (#34347).
-t N, --threads N starts Julia with N threads. This option takes precedence over JULIA_NUM_THREADS. The specified number of threads also propagates to worker processes spawned using the -p/--procs or --machine-file command line arguments. In order to set number of threads for worker processes spawned with addprocs use the exeflags keyword argument, e.g. addprocs(...; exeflags=`--threads 4`) (#35108).
Multi-threading changes:
Parts of the multi-threading API are now considered stable, with caveats. This includes all documented identifiers from Base.Threads except the atomic_ operations.
@threads now allows an optional schedule argument. Use @threads :static ... to ensure that the same schedule will be used as in past versions; the default schedule is likely to change in the future.
Build system changes:
The build system now contains a pure-make caching system for expanding expensive operations at the latest possible moment, while still expanding it only once (#35626).
New library functions:
Packages can now provide custom hints to help users resolve errors by using the experimental Base.Experimental.register_error_hint function. Packages that define custom exception types can support hints by calling the Base.Experimental.show_error_hints from their showerror method (#35094).
The @ccall macro has been added to Base. It is a near drop-in replacement for ccall with more Julia-like syntax. It also wraps the new foreigncall API for varargs of different types, though it lacks the capability to specify an LLVM calling convention (#32748).
New functions mergewith and mergewith! supersede merge and merge! with combine argument. They don't have the restriction for combine to be a Function and also provide one-argument method that returns a closure. The old methods of merge and merge! are still available for backward compatibility (#34296).
The new isdisjoint function indicates whether two collections are disjoint (#34427).
Add function ismutable and deprecate isimmutable to check whether something is mutable (#34652).
include now accepts an optional mapexpr first argument to transform the parsed expressions before they are evaluated (#34595).
New function bitreverse for reversing the order of bits in a fixed-width integer (#34791).
New function bitrotate(x, k) for rotating the bits in a fixed-width integer (#33937).
New function contains(haystack, needle) and its one argument partially applied form have been added, it acts like occursin(needle, haystack) (#35132).
New function Base.exit_on_sigint is added to control if InterruptException is thrown by Ctrl-C (#29411).
New library features:
Function composition now works also on one argument ∘(f) = f (#34251).
One argument methods startswith(x) and endswith(x) have been added, returning partially-applied versions of the functions, similar to existing methods like isequal(x) (#33193).
isapprox (or ≈) now has a one-argument "curried" method isapprox(x) which returns a function, like isequal (or ==) (#32305).
@NamedTuple{key1::Type1, ...} macro for convenient NamedTuple declarations (#34548).
Ref{NTuple{N,T}} can be passed to Ptr{T}/Ref{T} ccall signatures (#34199).
x::Signed % Unsigned and x::Unsigned % Signed are supported for integer bitstypes.
signed(unsigned_type) is supported for integer bitstypes, unsigned(signed_type) has been supported.
accumulate, cumsum, and cumprod now support Tuple (#34654) and arbitrary iterators (#34656).
pop!(collection, key, [default]) now has a method for Vector to remove an element at an arbitrary index (#35513).
In splice! with no replacement, values to be removed can now be specified with an arbitrary iterable (instead of a UnitRange) (#34524).
The @view and @views macros now support the a[begin] syntax that was introduced in Julia 1.4 (#35289).
open for files now accepts a keyword argument lock controlling whether file operations will acquire locks for safe multi-threaded access. Setting it to false provides better performance when only one thread will access the file (#35426).
The introspection macros (@which, @code_typed, etc.) now work with do-block syntax (#35283) and with dot syntax (#35522).
count now accepts the dims keyword.
new in-place count! function similar to sum!.
peek is now exported and accepts a type to peek from a stream ([#28811]).
Standard library changes:
Empty ranges now compare equal, regardless of their startpoint and step (#32348).
A 1-d Zip iterator (where Base.IteratorSize is Base.HasShape{1}()) with defined length of n has now also size of (n,) (instead of throwing an error with truncated iterators) (#29927).
The @timed macro now returns a NamedTuple (#34149).
New supertypes(T) function returns a tuple of all supertypes of T (#34419).
Views of builtin ranges are now recomputed ranges (like indexing returns) instead of SubArrays (#26872).
Sorting-related functions such as sort that take the keyword arguments lt, rev, order and by now do not discard order if by or lt are passed. In the former case, the order from order is used to compare the values of by(element). In the latter case, any order different from Forward or Reverse will raise an error about the ambiguity.
close on a file (IOStream) can now throw an exception if an error occurs when trying to flush buffered data to disk (#35303).
The large StridedArray Union now has special printing to avoid printing out its entire contents (#31149).
LinearAlgebra:
The BLAS submodule now supports the level-2 BLAS subroutine hpmv! (#34211).
normalize now supports multidimensional arrays (#34239).
lq factorizations can now be used to compute the minimum-norm solution to under-determined systems (#34350).
sqrt(::Hermitian) now treats slightly negative eigenvalues as zero for nearly semidefinite matrices, and accepts a new rtol keyword argument for this tolerance ([#35057]).
The BLAS submodule now supports the level-2 BLAS subroutine spmv! (#34320).
The BLAS submodule now supports the level-1 BLAS subroutine rot! (#35124).
New generic rotate!(x, y, c, s) and reflect!(x, y, c, s) functions (#35124).
Markdown:
In docstrings, a level-1 markdown header "Extended help" is now interpreted as a marker dividing "brief help" from "extended help". The REPL help mode only shows the brief help (the content before the "Extended help" header) by default; prepend the expression with '?' (in addition to the one that enters the help mode) to see the full docstring (#25930).
Random:
randn!(::MersenneTwister, ::Array{Float64}) is faster, and as a result, for a given state of the RNG, the corresponding generated numbers have changed (#35078).
rand!(::MersenneTwister, ::Array{Bool}) is faster, and as a result, for a given state of the RNG, the corresponding generated numbers have changed (#33721).
A new faster algorithm ("nearly division less") is used for generating random numbers within a range (#29240). As a result, the streams of generated numbers are changed (for ranges, like in rand(1:9), and for collections in general, like in rand([1, 2, 3])). Also, for performance, the undocumented property that, given a seed and a, b of type Int, rand(a:b) produces the same stream on 32 and 64 bits architectures, is dropped.
SparseArrays:
lu! accepts UmfpackLU as an argument to make use of its symbolic factorization.
The trim keyword argument for the functions fkeep!, tril!, triu!, droptol!,dropzeros! and dropzeros has been removed in favour of always trimming. Calling these with trim=false could result in invalid sparse arrays.
Dates:
The eps function now accepts TimeType types (#31487).
The zero function now accepts TimeType types (#35554).
Statistics:
Sockets:
Joining and leaving UDP multicast groups on a UDPSocket is now supported through join_multicast_group() and leave_multicast_group() (#35521).
Distributed:
launch_on_machine now supports and parses ipv6 square-bracket notation (#34430).

New in Julia Language 1.5.0 (Aug 3, 2020)

New language features:
Macro calls @foo {...} can now also be written @foo{...} (without the space) (#34498).
? is now parsed as a binary operator with times precedence. It can be entered in the REPL with bbsemi followed by TAB (#34722).
± and ± are now unary operators as well, like + or -. Attention has to be paid in macros and matrix constructors, which are whitespace sensitive, because expressions like [a ±b] now get parsed as [a ±(b)] instead of [±(a, b)] (#34200).
Passing an identifier x by itself as a keyword argument or named tuple element is equivalent to x=x, implicitly using the name of the variable as the keyword or named tuple field name. Similarly, passing an a.b expression uses b as the keyword or field name (#29333).
Support for Unicode 13.0.0 (via utf8proc 2.5) (#35282).
The compiler optimization level can now be set per-module using the experimental macro Base.Experimental.@optlevel n. For code that is not performance-critical, setting this to 0 or 1 can provide significant latency improvements (#34896).
Language changes:
The interactive REPL now uses "soft scope" for top-level expressions: an assignment inside a scope block such as a for loop automatically assigns to a global variable if one has been defined already. This matches the behavior of Julia versions 0.6 and prior, as well as IJulia. Note that this only affects expressions interactively typed or pasted directly into the default REPL (#28789, #33864).
Outside of the REPL (e.g. in a file), assigning to a variable within a top-level scope block is considered ambiguous if a global variable with the same name exists. A warning is given if that happens, to alert you that the code will work differently than in the REPL. A new command line option --warn-scope controls this warning (#33864).
Converting arbitrary tuples to NTuple, e.g. convert(NTuple, (1, "")) now gives an error, where it used to be incorrectly allowed. This is because NTuple refers only to homogeneous tuples (this meaning has not changed) (#34272).
The syntax (;) (which was deprecated in v1.4) now creates an empty named tuple (#30115).
@inline macro can now be applied to short-form anonymous functions (#34953).
In triple-quoted string literals, whitespace stripping is now done before processing escape sequences instead of after. For example, the syntax
"""
an b"""
used to yield the string " anb", since the single space before b set the indent level. Now the result is "an b", since the space before b is no longer considered to occur at the start of a line. The old behavior is considered a bug (#35001).
<: and >: can now be broadcasted over arrays with .<: and .>: ([#35085])
The line number of function definitions is now added by the parser as an additional LineNumberNode at the start of each function body (#35138).
Statements of the form a' now get lowered to var"'"(a) instead of Base.adjoint(a). This allows for shadowing this function in local scopes, although this is generally discouraged. By default, Base exports var"'" as an alias of Base.adjoint, so custom types should still extend Base.adjoint (#34634).
Compiler/Runtime improvements:
Immutable structs (including tuples) that contain references can now be allocated on the stack, and allocated inline within arrays and other structs (#33886). This significantly reduces the number of heap allocations in some workloads. Code that requires assumptions about object layout and addresses (usually for interoperability with C or other languages) might need to be updated; for example any object that needs a stable address should be a mutable struct. As a result, Array views no longer allocate ([#34126]).
Command-line option changes:
Deprecation warnings are no longer shown by default. i.e. if the --depwarn=... flag is not passed it defaults to --depwarn=no. The warnings are printed from tests run by Pkg.test() (#35362).
Color now defaults to on when stdout and stderr are TTYs (#34347).
t N, --threads N starts Julia with N threads. This option takes precedence over JULIA_NUM_THREADS. The specified number of threads also propagates to worker processes spawned using the -p/--procs or --machine-file command line arguments. In order to set number of threads for worker processes spawned with addprocs use the exeflags keyword argument, e.g. addprocs(...; exeflags=`--threads 4`) (#35108).
Multi-threading changes:
Parts of the multi-threading API are now considered stable, with caveats. This includes all documented identifiers from Base.Threads except the atomic_ operations.
@threads now allows an optional schedule argument. Use @threads :static ... to ensure that the same schedule will be used as in past versions; the default schedule is likely to change in the future.
Build system changes:
The build system now contains a pure-make caching system for expanding expensive operations at the latest possible moment, while still expanding it only once (#35626).
New library functions:
Packages can now provide custom hints to help users resolve errors by using the experimental Base.Experimental.register_error_hint function. Packages that define custom exception types can support hints by calling the Base.Experimental.show_error_hints from their showerror method (#35094).
The @ccall macro has been added to Base. It is a near drop-in replacement for ccall with more Julia-like syntax. It also wraps the new foreigncall API for varargs of different types, though it lacks the capability to specify an LLVM calling convention (#32748).
New functions mergewith and mergewith! supersede merge and merge! with combine argument. They don't have the restriction for combine to be a Function and also provide one-argument method that returns a closure. The old methods of merge and merge! are still available for backward compatibility (#34296).
The new isdisjoint function indicates whether two collections are disjoint (#34427).
Add function ismutable and deprecate isimmutable to check whether something is mutable (#34652).
include now accepts an optional mapexpr first argument to transform the parsed expressions before they are evaluated (#34595).
New function bitreverse for reversing the order of bits in a fixed-width integer (#34791).
New function bitrotate(x, k) for rotating the bits in a fixed-width integer (#33937).
New function contains(haystack, needle) and its one argument partially applied form have been added, it acts like occursin(needle, haystack) (#35132).
New function Base.exit_on_sigint is added to control if InterruptException is thrown by Ctrl-C (#29411).
New library features:
Function composition now works also on one argument °(f) = f (#34251).
One argument methods startswith(x) and endswith(x) have been added, returning partially-applied versions of the functions, similar to existing methods like isequal(x) (#33193).
isapprox (or ˜) now has a one-argument "curried" method isapprox(x) which returns a function, like isequal (or ==) (#32305).
@NamedTuple{key1::Type1, ...} macro for convenient NamedTuple declarations (#34548).
Ref{NTuple{N,T}} can be passed to Ptr{T}/Ref{T} ccall signatures (#34199).
x::Signed % Unsigned and x::Unsigned % Signed are supported for integer bitstypes.
signed(unsigned_type) is supported for integer bitstypes, unsigned(signed_type) has been supported.
accumulate, cumsum, and cumprod now support Tuple (#34654) and arbitrary iterators (#34656).
pop!(collection, key, [default]) now has a method for Vector to remove an element at an arbitrary index (#35513).
In splice! with no replacement, values to be removed can now be specified with an arbitrary iterable (instead of a UnitRange) (#34524).
The @view and @views macros now support the a[begin] syntax that was introduced in Julia 1.4 (#35289).
open for files now accepts a keyword argument lock controlling whether file operations will acquire locks for safe multi-threaded access. Setting it to false provides better performance when only one thread will access the file (#35426).
The introspection macros (@which, @code_typed, etc.) now work with do-block syntax (#35283) and with dot syntax (#35522).
count now accepts the dims keyword.
new in-place count! function similar to sum!.
peek is now exported and accepts a type to peek from a stream ([#28811]).
Standard library changes:
Empty ranges now compare equal, regardless of their startpoint and step (#32348).
A 1-d Zip iterator (where Base.IteratorSize is Base.HasShape{1}()) with defined length of n has now also size of (n,) (instead of throwing an error with truncated iterators) (#29927).
The @timed macro now returns a NamedTuple (#34149).
New supertypes(T) function returns a tuple of all supertypes of T (#34419).
Views of builtin ranges are now recomputed ranges (like indexing returns) instead of SubArrays (#26872).
Sorting-related functions such as sort that take the keyword arguments lt, rev, order and by now do not discard order if by or lt are passed. In the former case, the order from order is used to compare the values of by(element). In the latter case, any order different from Forward or Reverse will raise an error about the ambiguity.
close on a file (IOStream) can now throw an exception if an error occurs when trying to flush buffered data to disk (#35303).
The large StridedArray Union now has special printing to avoid printing out its entire contents (#31149).
LinearAlgebra:
The BLAS submodule now supports the level-2 BLAS subroutine hpmv! (#34211).
normalize now supports multidimensional arrays (#34239).
lq factorizations can now be used to compute the minimum-norm solution to under-determined systems (#34350).
sqrt(::Hermitian) now treats slightly negative eigenvalues as zero for nearly semidefinite matrices, and accepts a new rtol keyword argument for this tolerance ([#35057]).
The BLAS submodule now supports the level-2 BLAS subroutine spmv! (#34320).
The BLAS submodule now supports the level-1 BLAS subroutine rot! (#35124).
New generic rotate!(x, y, c, s) and reflect!(x, y, c, s) functions (#35124).
Markdown:
In docstrings, a level-1 markdown header "Extended help" is now interpreted as a marker dividing "brief help" from "extended help". The REPL help mode only shows the brief help (the content before the "Extended help" header) by default; prepend the expression with '?' (in addition to the one that enters the help mode) to see the full docstring (#25930).
Random:
randn!(::MersenneTwister, ::Array{Float64}) is faster, and as a result, for a given state of the RNG, the corresponding generated numbers have changed (#35078).
rand!(::MersenneTwister, ::Array{Bool}) is faster, and as a result, for a given state of the RNG, the corresponding generated numbers have changed (#33721).
A new faster algorithm ("nearly division less") is used for generating random numbers within a range (#29240). As a result, the streams of generated numbers are changed (for ranges, like in rand(1:9), and for collections in general, like in rand([1, 2, 3])). Also, for performance, the undocumented property that, given a seed and a, b of type Int, rand(a:b) produces the same stream on 32 and 64 bits architectures, is dropped.
REPL:
SparseArrays
lu! accepts UmfpackLU as an argument to make use of its symbolic factorization.
The trim keyword argument for the functions fkeep!, tril!, triu!, droptol!,dropzeros! and dropzeros has been removed in favour of always trimming. Calling these with trim=false could result in invalid sparse arrays.
Dates:
The eps function now accepts TimeType types (#31487).
The zero function now accepts TimeType types (#35554).
Statistics
Sockets
Joining and leaving UDP multicast groups on a UDPSocket is now supported through join_multicast_group() and leave_multicast_group() (#35521).
Distributed:
launch_on_machine now supports and parses ipv6 square-bracket notation (#34430).
Deprecated or removed
External dependencies
OpenBLAS has been updated to v0.3.9 ([#35113]).
Tooling Improvements

New in Julia Language 1.4.0 (Mar 22, 2020)

New language features:
Structs with all isbits and isbitsunion fields are now stored inline in arrays (#32448).
import now allows quoted symbols, e.g. import Base.:+ (#33158).
a[begin] can now be used to address the first element of an integer-indexed collection a. The index is computed by firstindex(a) (#33946).
Language changes:
The syntax (;), which used to parse as an empty block expression, is deprecated. In the future it will indicate an empty named tuple (#30115).
Multi-threading changes:
Values can now be interpolated into @async and @spawn via $, which copies the value directly into the constructed underlying closure (#33119).
Build system changes:
Windows build installer has switched to Inno Setup. Installer command line parameters have thus changed. For example, to extract the installer to a specific directory, the command line parameter is now /DIR=x:dirname. Use julia-installer.exe /? to list all new command line parameters.
New library functions:
The new only(x) function returns the one-and-only element of a collection x, and throws an ArgumentError if x contains zero or multiple elements (#33129).
takewhile and dropwhile have been added to the Iterators submodule (#33437).
accumulate has been added to the Iterators submodule (#34033).
There is a now an evalpoly function meant to take the role of the @evalpoly macro. The function is just as efficient as the macro while giving added flexibility, so it should be preferred over @evalpoly. evalpoly takes a list of coefficients as a tuple, so where one might write @evalpoly(x, p1, p2, p3) one would instead write evalpoly(x, (p1, p2, p3)).
New library features:
Function composition now supports multiple functions: ∘(f, g, h) = f ∘ g ∘ h and splatting ∘(fs...) for composing an iterable collection of functions (#33568).
Functions gcd, lcm, and gcdx now support Rational arguments (#33910).
The splitpath function now accepts any AbstractString whereas previously it only accepted paths of type String (#33012).
filter can now act on a Tuple (#32968).
The tempname function now takes an optional parent::AbstractString argument to give it a directory in which to attempt to produce a temporary path name (#33090).
The tempname function now takes a cleanup::Bool keyword argument defaulting to true, which causes the process to try to ensure that any file or directory at the path returned by tempname is deleted upon process exit (#33090).
The readdir function now takes a join::Bool keyword argument defaulting to false, which when set causes readdir to join its directory argument with each listed name (#33113).
div now accepts a rounding mode as the third argument, consistent with the corresponding argument to rem. Support for rounding division, by passing one of the RoundNearest modes to this function, was added. For future compatibility, library authors should now extend this function, rather than extending the two-argument fld/cld/div directly (#33040).
methods now accepts a module (or a list thereof) to filter methods defined in it (#33403).
Standard library changes:
Calling show or repr on an undef/UndefInitializer() array initializer now shows valid Julia code (#33211).
Calling show or repr on a 0-dimensional AbstractArray now shows valid code for creating an equivalent 0-dimensional array, instead of only showing the contained value (#33206).
readdir output is now guaranteed to be sorted. The sort keyword allows opting out of sorting to get names in OS-native order (#33542).
The methods of mktemp and mktempdir that take a function to pass temporary paths to no longer throw errors if the path is already deleted when the function returns (#33091).
Verbose display of Char (text/plain output) now shows the codepoint value in standard-conforming "U+XXXX" format (#33291).
Iterators.partition now uses views (or smartly re-computed ranges) for partitions of all AbstractArrays (#33533).
Sets are now displayed less compactly in the REPL, as a column of elements, like vectors and dictionaries (#33300).
delete! on WeakKeyDicts now returns the WeakKeyDict itself instead of the underlying Dict used for implementation
LinearAlgebra:
qr and qr! functions support blocksize keyword argument (#33053).
dot now admits a 3-argument method dot(x, A, y) to compute generalized dot products dot(x, A*y), but without computing and storing the intermediate result A*y (#32739).
ldlt and non-pivoted lu now throw a new ZeroPivotException type (#33372).
cond(A, p) with p=1 or p=Inf now computes the exact condition number instead of an estimate (#33547).
UniformScaling objects may now be exponentiated such that (a*I)^x = a^x * I.
Markdown:
Tables now have the align attribute set when shown as HTML (#33849).
Random:
AbstractRNGs now behave like scalars when used in broadcasting (#33213).
The performance of rand(::Tuple) is improved in some cases (#32208). As a consequence, the stream of generated values produced for a given seed has changed.
REPL:
The attributes of the implicit IOContext used by the REPL to display objects can be modified by the user (experimental feature) (#29249).
SparseArrays:
The return value of zero(x::AbstractSparseArray) has no stored zeros anymore (#31835). Previously, it would have stored zeros wherever x had them. This makes the operation constant time instead of O(<number of stored values>).
Products involving sparse arrays now allow more general sparse eltypes, such as StaticArrays (#33205)

New in Julia Language 1.4.0 RC 2 (Feb 27, 2020)

New language features:
Structs with all isbits and isbitsunion fields are now stored inline in arrays (#32448).
import now allows quoted symbols, e.g. import Base.:+ (#33158).
a[begin] can now be used to address the first element of an integer-indexed collection a. The index is computed by firstindex(a) (#33946).
Language changes:
The syntax (;), which used to parse as an empty block expression, is deprecated. In the future it will indicate an empty named tuple (#30115).
Multi-threading changes:
Values can now be interpolated into @async and @spawn via $, which copies the value directly into the constructed underlying closure (#33119).
Build system changes:
Windows build installer has switched to Inno Setup. Installer command line parameters have thus changed. For example, to extract the installer to a specific directory, the command line parameter is now /DIR=x:dirname. Use julia-installer.exe /? to list all new command line parameters.
New library functions:
The new only(x) function returns the one-and-only element of a collection x, and throws an ArgumentError if x contains zero or multiple elements (#33129).
takewhile and dropwhile have been added to the Iterators submodule (#33437).
accumulate has been added to the Iterators submodule (#34033).
There is a now an evalpoly function meant to take the role of the @evalpoly macro. The function is just as efficient as the macro while giving added flexibility, so it should be preferred over @evalpoly. evalpoly takes a list of coefficients as a tuple, so where one might write @evalpoly(x, p1, p2, p3) one would instead write evalpoly(x, (p1, p2, p3)).
New library features:
Function composition now supports multiple functions: ∘(f, g, h) = f ∘ g ∘ h and splatting ∘(fs...) for composing an iterable collection of functions (#33568).
Functions gcd, lcm, and gcdx now support Rational arguments (#33910).
The splitpath function now accepts any AbstractString whereas previously it only accepted paths of type String (#33012).
filter can now act on a Tuple (#32968).
The tempname function now takes an optional parent::AbstractString argument to give it a directory in which to attempt to produce a temporary path name (#33090).
The tempname function now takes a cleanup::Bool keyword argument defaulting to true, which causes the process to try to ensure that any file or directory at the path returned by tempname is deleted upon process exit (#33090).
The readdir function now takes a join::Bool keyword argument defaulting to false, which when set causes readdir to join its directory argument with each listed name (#33113).
div now accepts a rounding mode as the third argument, consistent with the corresponding argument to rem. Support for rounding division, by passing one of the RoundNearest modes to this function, was added. For future compatibility, library authors should now extend this function, rather than extending the two-argument fld/cld/div directly (#33040).
methods now accepts a module (or a list thereof) to filter methods defined in it (#33403).
Standard library changes:
Calling show or repr on an undef/UndefInitializer() array initializer now shows valid Julia code (#33211).
Calling show or repr on a 0-dimensional AbstractArray now shows valid code for creating an equivalent 0-dimensional array, instead of only showing the contained value (#33206).
readdir output is now guaranteed to be sorted. The sort keyword allows opting out of sorting to get names in OS-native order (#33542).
The methods of mktemp and mktempdir that take a function to pass temporary paths to no longer throw errors if the path is already deleted when the function returns (#33091).
Verbose display of Char (text/plain output) now shows the codepoint value in standard-conforming "U+XXXX" format (#33291).
Iterators.partition now uses views (or smartly re-computed ranges) for partitions of all AbstractArrays (#33533).
Sets are now displayed less compactly in the REPL, as a column of elements, like vectors and dictionaries (#33300).
delete! on WeakKeyDicts now returns the WeakKeyDict itself instead of the underlying Dict used for implementation
LinearAlgebra:
qr and qr! functions support blocksize keyword argument (#33053).
dot now admits a 3-argument method dot(x, A, y) to compute generalized dot products dot(x, A*y), but without computing and storing the intermediate result A*y (#32739).
ldlt and non-pivoted lu now throw a new ZeroPivotException type (#33372).
cond(A, p) with p=1 or p=Inf now computes the exact condition number instead of an estimate (#33547).
UniformScaling objects may now be exponentiated such that (a*I)^x = a^x * I.
Markdown:
Tables now have the align attribute set when shown as HTML (#33849).
Random:
AbstractRNGs now behave like scalars when used in broadcasting (#33213).
The performance of rand(::Tuple) is improved in some cases (#32208). As a consequence, the stream of generated values produced for a given seed has changed.
REPL:
The attributes of the implicit IOContext used by the REPL to display objects can be modified by the user (experimental feature) (#29249).
SparseArrays:
The return value of zero(x::AbstractSparseArray) has no stored zeros anymore (#31835). Previously, it would have stored zeros wherever x had them. This makes the operation constant time instead of O(<number of stored values>).
Products involving sparse arrays now allow more general sparse eltypes, such as StaticArrays (#33205)

New in Julia Language 1.3.0 (Nov 27, 2019)

New language features:
Support for Unicode 12.1.0 (#32002).
Methods can now be added to an abstract type (#31916).
Support for unicode bold digits and double-struck digits 0 through 9 as valid identifiers (#32838).
Added the syntax var"#str#" for printing and parsing non-standard variable names (#32408).
Multi-threading changes:
New experimental Threads.@spawn macro that runs a task on any available thread (#32600).
All system-level I/O operations (e.g. files and sockets) are now thread-safe. This does not include subtypes of IO that are entirely in-memory, such as IOBuffer, although it specifically does include BufferStream. (#32309, #32174, #31981, #32421).
The global random number generator (GLOBAL_RNG) is now thread-safe (and thread-local) (#32407).
New Channel(f::Function, spawn=true) keyword argument to schedule the created Task on any available thread, matching the behavior of Threads.@spawn (#32872).
Simplified the Channel constructor, which is now easier to read and more idiomatic julia. Use of the keyword arguments csize and ctype is now discouraged (#30855, #32818).
New library functions:
findfirst, findlast, findnext and findprev now accept a character as first argument to search for that character in a string passed as the second argument (#31664).
New findall(pattern, string) method where pattern is a string or regex (#31834).
count(pattern, string) gives the number of things findall would match (#32849).
istaskfailed is now documented and exported, like its siblings istaskdone and istaskstarted (#32300).
RefArray and RefValue objects now accept index CartesianIndex() in getindex and setindex! (#32653)
Added sincosd(x) to simultaneously compute the sine and cosine of x, where x is in degrees (#30134).
The function nonmissingtype, which removes Missing from type unions, is now exported (#31562).
Standard library changes:
Pkg won't clobber pre-compilation files as often when switching environments ([#32651])
Pkg can now download and install binary artifacts through the Pkg.Artifacts submodule and supporting functions. ([#32918])
When wait (or @sync, or fetch) is called on a failing Task, the exception is propagated as a TaskFailedException wrapping the task. This makes it possible to see the location of the original failure inside the task (as well as the location of the wait call, as before) (#32814).
Regex can now be multiplied (*) and exponentiated (^), like strings (#23422).
Cmd interpolation (`$(x::Cmd) a b c` where) now propagates x's process flags (environment, flags, working directory, etc) if x is the first interpolant and errors otherwise (#24353).
Zero-dimensional arrays are now consistently preserved in the return values of mathematical functions that operate on the array(s) as a whole (and are not explicitly broadcasted across their elements). Previously, the functions +, -, *, /, conj, real and imag returned the unwrapped element when operating over zero-dimensional arrays (#32122).
IPAddr subtypes now behave like scalars when used in broadcasting (#32133).
Pair is now treated as a scalar for broadcasting (#32209).
clamp can now handle missing values (#31066).
empty now accepts a NamedTuple (#32534).
mod now accepts a unit range as the second argument to easily perform offset modular arithmetic to ensure the result is inside the range (#32628).
nothing can now be printed, and interpolated into strings etc. as the string "nothing". It is still not permitted to be interpolated into Cmds (i.e. echo `$(nothing)` will still error without running anything.) (#32148)
When open is called with a function, command, and keyword argument (e.g. open(`ls`, read=true) do f ...) it now correctly throws a ProcessFailedException like other similar calls (#32193).
mktemp and mktempdir now try, by default, to remove temporary paths they create before the process exits (#32851).
Added argument keep to unescape_string (#27125).
Libdl:
dlopen() can now be invoked in do-block syntax, similar to open().
LinearAlgebra:
The BLAS submodule no longer exports dot, which conflicts with that in LinearAlgebra (#31838).
diagm and spdiagm now accept optional m,n initial arguments to specify a size (#31654).
Hessenberg factorizations H now support efficient shifted solves (H+µI) b and determinants, and use a specialized tridiagonal factorization for Hermitian matrices. There is also a new UpperHessenberg matrix type (#31853).
Added keyword argument alg to svd and svd! that allows one to switch between different SVD algorithms (#31057).
Five-argument mul!(C, A, B, α, β) now implements inplace multiplication fused with addition C = A B α + C β (#23919).
SparseArrays:
SparseMatrixCSC(m,n,colptr,rowval,nzval) perform consistency checks for arguments: colptr must be properly populated and lengths of colptr, rowval, and nzval must be compatible with m, n, and eltype(colptr).
sparse(I, J, V, m, n) verifies lengths of I, J, V are equal and compatible with eltype(I) and m, n.
Dates:
DateTime and Time formatting/parsing now supports 12-hour clocks with AM/PM via I and p codes, similar to strftime (#32308).
Fixed repr such that it displays Time as it would be entered in Julia (#32103).
Statistics:
mean now accepts both a function argument and a dims keyword (#31576).
Sockets:
Sockets.recvfrom now returns both host and port as an InetAddr (#32729).
Added InetAddr constructor from AbstractString, representing IP address, and Integer, representing port number (#31459).
Miscellaneous:
foldr and mapfoldr now work on any iterator that supports Iterators.reverse, not just arrays (#31781).
Deprecated or removed:
@spawn expr from the Distributed standard library should be replaced with @spawnat :any expr (#32600).
Threads.Mutex and Threads.RecursiveSpinLock have been removed; use ReentrantLock (preferred) or Threads.SpinLock instead (#32875).
Tooling Improvements:
The ClangSA.jl static analysis package has been imported, which makes use of the clang static analyzer to validate GC invariants in Julia's C code. The analysis may be run using make -C src analyzegc.

New in Julia Language 1.3.0 RC2 (Sep 17, 2019)

New language features:
Support for Unicode 12.1.0 (#32002).
Methods can now be added to an abstract type (#31916).
Support for unicode bold digits and double-struck digits 0 through 9 as valid identifiers (#32838).
Added the syntax var"#str#" for printing and parsing non-standard variable names (#32408).
Language changes:
Multi-threading changes:
New experimental Threads.@spawn macro that runs a task on any available thread (#32600).
All system-level I/O operations (e.g. files and sockets) are now thread-safe. This does not include subtypes of IO that are entirely in-memory, such as IOBuffer, although it specifically does include BufferStream. (#32309, #32174, #31981, #32421).
The global random number generator (GLOBAL_RNG) is now thread-safe (and thread-local) (#32407).
New Channel(f::Function, spawn=true) keyword argument to schedule the created Task on any available thread, matching the behavior of Threads.@spawn (#32872).
Simplified the Channel constructor, which is now easier to read and more idiomatic julia. Use of the keyword arguments csize and ctype is now discouraged (#30855, #32818).
Build system changes:
New library functions:
findfirst, findlast, findnext and findprev now accept a character as first argument to search for that character in a string passed as the second argument (#31664).
New findall(pattern, string) method where pattern is a string or regex (#31834).
count(pattern, string) gives the number of things findall would match (#32849).
istaskfailed is now documented and exported, like its siblings istaskdone and istaskstarted (#32300).
RefArray and RefValue objects now accept index CartesianIndex() in getindex and setindex! (#32653)
Added sincosd(x) to simultaneously compute the sine and cosine of x, where x is in degrees (#30134).
The function nonmissingtype, which removes Missing from type unions, is now exported (#31562).
Standard library changes:
Pkg can now download and install binary artifacts through the Pkg.Artifacts submodule and supporting functions. ([#32918])
When wait (or @sync, or fetch) is called on a failing Task, the exception is propagated as a TaskFailedException wrapping the task. This makes it possible to see the location of the original failure inside the task (as well as the location of the wait call, as before) (#32814).
Regex can now be multiplied (*) and exponentiated (^), like strings (#23422).
Cmd interpolation (`$(x::Cmd) a b c` where) now propagates x's process flags (environment, flags, working directory, etc) if x is the first interpolant and errors otherwise (#24353).
Zero-dimensional arrays are now consistently preserved in the return values of mathematical functions that operate on the array(s) as a whole (and are not explicitly broadcasted across their elements). Previously, the functions +, -, *, /, conj, real and imag returned the unwrapped element when operating over zero-dimensional arrays (#32122).
IPAddr subtypes now behave like scalars when used in broadcasting (#32133).
Pair is now treated as a scalar for broadcasting (#32209).
clamp can now handle missing values (#31066).
empty now accepts a NamedTuple (#32534).
mod now accepts a unit range as the second argument to easily perform offset modular arithmetic to ensure the result is inside the range (#32628).
nothing can now be printed, and interpolated into strings etc. as the string "nothing". It is still not permitted to be interpolated into Cmds (i.e. echo `$(nothing)` will still error without running anything.) (#32148)
When open is called with a function, command, and keyword argument (e.g. open(`ls`, read=true) do f ...) it now correctly throws a ProcessFailedException like other similar calls (#32193).
mktemp and mktempdir now try, by default, to remove temporary paths they create before the process exits (#32851).
Added argument keep to unescape_string (#27125).
Libdl:
dlopen() can now be invoked in do-block syntax, similar to open().
LinearAlgebra:
The BLAS submodule no longer exports dot, which conflicts with that in LinearAlgebra (#31838).
diagm and spdiagm now accept optional m,n initial arguments to specify a size (#31654).
Hessenberg factorizations H now support efficient shifted solves (H+µI) b and determinants, and use a specialized tridiagonal factorization for Hermitian matrices. There is also a new UpperHessenberg matrix type (#31853).
Added keyword argument alg to svd and svd! that allows one to switch between different SVD algorithms (#31057).
Five-argument mul!(C, A, B, a, ß) now implements inplace multiplication fused with addition C = A B a + C ß (#23919).
SparseArrays:
SparseMatrixCSC(m,n,colptr,rowval,nzval) perform consistency checks for arguments: colptr must be properly populated and lengths of colptr, rowval, and nzval must be compatible with m, n, and eltype(colptr).
sparse(I, J, V, m, n) verifies lengths of I, J, V are equal and compatible with eltype(I) and m, n.
Dates:
DateTime and Time formatting/parsing now supports 12-hour clocks with AM/PM via I and p codes, similar to strftime (#32308).
Fixed repr such that it displays Time as it would be entered in Julia (#32103).
Statistics:
mean now accepts both a function argument and a dims keyword (#31576).
Sockets:
Sockets.recvfrom now returns both host and port as an InetAddr (#32729).
Added InetAddr constructor from AbstractString, representing IP address, and Integer, representing port number (#31459).
Miscellaneous:
foldr and mapfoldr now work on any iterator that supports Iterators.reverse, not just arrays (#31781).
Deprecated or removed:
@spawn expr from the Distributed standard library should be replaced with @spawnat :any expr (#32600).
Threads.Mutex and Threads.RecursiveSpinLock have been removed; use ReentrantLock (preferred) or Threads.SpinLock instead (#32875).

New in Julia Language 1.3.0 Alpha (Jul 24, 2019)

New in Julia Language 1.1.0 (Jan 22, 2019)

New language features:
An exception stack is maintained on each task to make exception handling more robust and enable root cause analysis. The stack may be accessed using the experimental function Base.catch_stack (#28878).
The experimental macro Base.@locals returns a dictionary of current local variable names and values (#29733).
Binary ~ can now be dotted, as in x .~ y (#30341).
Language changes:
Parser inputs ending with a comma are now consistently treated as incomplete. Previously they were sometimes parsed as tuples, depending on whitespace (#28506).
Spaces were accidentally allowed in broadcast call syntax, e.g. f. (x). They are now disallowed, consistent with normal function call syntax (#29781).
Big integer literals and command syntax (backticks) are now parsed with the name of the macro (@int128_str, @uint128_str, @big_str, @cmd) qualified to refer to the Core module (#29968).
Using the same name for both a local variable and a static parameter is now an error instead of a warning (#29429).
findall(in(b), a) now returns a CartesianIndex when a is a matrix or a higher-dimensional array, for consistency with other findall methods. Use LinearIndices(a)[findall(in(b), a)] to get the old behavior, or CartesianIndices(a)[findall(in(b), a)] to get the new behavior on previous Julia versions (#30226).
findmin(::BitArray) and findmax(::BitArray) now return a CartesianIndex when a is a matrix or a higher-dimensional array, for consistency with other array types. Use LinearIndices(a)[findmin(a)[2]] to get the old behavior, or CartesianIndices(a)[findmin(a)[2]] to get the new behavior on previous Julia versions (#30102).
Method signatures such as f(::Type{T}, ::T) where {T <: X} and f(::Type{X}, ::Any) are now considered ambiguous. Previously a bug caused the first one to be considered more specific in some cases (#30160).
Command-line option changes:
When a script run in interactive mode (-i) throws an error, the REPL now starts after the error is displayed. Previously the REPL only started if the script completed without error (#21233).
New library functions:
splitpath(p::String) function, which is the opposite of joinpath(parts...): it splits a filepath into its components (#28156).
isnothing(::Any) predicate, to check whether the argument is nothing. (#29679).
getpid(::Process) method (#24064).
eachrow, eachcol and eachslice functions provide efficient iterators over slices of arrays (#29749).
fieldtypes(T::Type) which returns the declared types of the field in type T (#29600).
uuid5 has been added to the UUIDs standard library (#28761).
Predicates Sys.isfreebsd, Sys.isopenbsd, Sys.isnetbsd, and Sys.isdragonfly for detecting BSD systems have been added (#30249).
Internal Base.disable_library_threading that sets libraries to use one thread. It executes function hooks that have been registered with Base.at_disable_library_threading (#30004).
Standard library changes:
CartesianIndices can now be constructed from two CartesianIndexes I and J with I:J (#29440).
CartesianIndices support broadcasting arithmetic (+ and -) with a CartesianIndex (#29890).
copy! support for arrays, dicts, and sets has been moved to Base from the Future package (#29173).
Channels now convert inserted values (like containers) instead of requiring types to match (#29092).
range can accept the stop value as a positional argument, e.g. range(1,10,step=2) (#28708).
diff now supports arrays of arbitrary dimensionality and can operate over any dimension (#29827).
The constructor BigFloat(::BigFloat) now respects the global precision setting and always returns a BigFloat with precision equal to precision(BigFloat) (#29127). The optional precision argument to override the global setting is now a keyword instead of positional argument (#29157).
The use of scientific notation when printing BigFloat values is now consistent with other floating point types (#29211).
Regex now behaves like a scalar when used in broadcasting (#29913).
Char now behaves like a read-only 0-dimensional array (#29819).
parse now allows strings representing integer 0 and 1 for type Bool (#29980).
Base.tail now works on named tuples (#29595).
The process id is appended to malloc log files in order to track memory allocations of multiple processes (#29969).
Base.julia_cmd now propagates the --inline=(yes|no) flag (#29858).
Base.@kwdef can now be used for parametric structs, and for structs with supertypes (#29316).
merge(::NamedTuple, ::NamedTuple...) can now be used with more than 2 NamedTuples (#29259).
New ncodeunits(c::Char) method as a fast equivalent to ncodeunits(string(c)) (#29153).
New sort!(::AbstractArray; dims) method that can sort the array along the dims dimension (#28902).
range now accepts stop as a positional argument (#28708).
get(A::AbstractArray, (), default) now returns A[] instead of an empty array (#30270).
parse(Bool, str) is now supported (#29997).
copyto!(::AbstractMatrix, ::UniformScaling) now supports rectangular matrices (#28790).
current_project() now searches the parent directories of a Git repository for a Project.toml file. This also affects the behavior of the --project command line option when using the default --project=@. (#29108).
The spawn API is now more flexible and supports taking IOBuffer directly as an I/O stream, converting to a system pipe as needed (#30278).
Dates:
New DateTime(::Date, ::Time) constructor (#29754).
TimeZone now behaves like a scalar when used in broadcasting (#30159).
InteractiveUtils:
edit can now be called on a module to edit the file that defines it (#29636).
All compiler-reflection tools (i.e. the code_ class of functions and macros) now print accurate line number and inlining information in a common style, and take an optional parameter (debuginfo=:default) to control the verbosity of the metadata shown (#29893).
LinearAlgebra:
isdiag and isposdef for Diagonal and UniformScaling (#29638).
mul!, rmul! and lmul! methods for UniformScaling (#29506).
Symmetric and Hermitian matrices now preserve the wrapper when scaled with a number (#29469).
Exponentiation operator ^ now supports raising an Irrational to an AbstractMatrix power (#29782).
Added keyword arguments rtol, atol to rank (#29926).
Random:
randperm and randcycle now use the type of their argument to determine the element type of the returned array (#29670).
A new method rand(::Tuple) implements sampling from the values of a tuple (#25278).
serialize and deserialize now accept a filename argument, like write and read (#30151).
SparseArrays:
sprandn now supports specifying the output element type (#30083).
Statistics:
mean and var now handle more kinds of empty inputs (#29033).
External dependencies:
7zip (bundled with Julia on Windows) has been upgraded from version 16.04 to 18.05 (#30035).
Busybox is no longer bundled with Julia on Windows (#30022).
OpenBLAS has been upgraded from 0.3.2 to 0.3.3 (#29845).
The source code for Pkg is no longer included in JuliaLang/julia. Pkg is instead downloaded during the build process (#29615).
LLVM has been upgraded to 6.0.1 and support for LLVM < 6.0 has been dropped (#28745, #28696).
Pkg has been upgraded to version 1.1 (#30342).
Deprecated or removed:
one(i::CartesianIndex) should be replaced with oneunit(i::CartesianIndex) (#29442).
The internal array Base.Grisu.DIGITS is deprecated; new code should use Base.Grisu.getbuf() to get an appropriate task-local buffer and pass it to grisu() instead (#29907).
The internal function Base._default_type(T) has been removed. Calls to it should be replaced with just the argument T (#29739).
peakflops has been scheduled to move from InteractiveUtils to LinearAlgebra but is already now available as LinearAlgebra.peakflops (#29978).

New in Julia Language 1.1.0 RC2 (Jan 15, 2019)

New in Julia Language 1.1.0 RC1 (Jan 3, 2019)

New language features:
An exception stack is maintained on each task to make exception handling more robust and enable root cause analysis. The stack may be accessed using the experimental function Base.catch_stack (#28878).
The experimental macro Base.@locals returns a dictionary of current local variable names and values (#29733).
Binary ~ can now be dotted, as in x .~ y (#30341).
Language changes:
Parser inputs ending with a comma are now consistently treated as incomplete. Previously they were sometimes parsed as tuples, depending on whitespace (#28506).
Spaces were accidentally allowed in broadcast call syntax, e.g. f. (x). They are now disallowed, consistent with normal function call syntax (#29781).
Big integer literals and command syntax (backticks) are now parsed with the name of the macro (@int128_str, @uint128_str, @big_str, @cmd) qualified to refer to the Core module (#29968).
Using the same name for both a local variable and a static parameter is now an error instead of a warning (#29429).
findall(in(b), a) now returns a CartesianIndex when a is a matrix or a higher-dimensional array, for consistency with other findall methods. Use LinearIndices(a)[findall(in(b), a)] to get the old behavior, or CartesianIndices(a)[findall(in(b), a)] to get the new behavior on previous Julia versions (#30226).
findmin(::BitArray) and findmax(::BitArray) now return a CartesianIndex when a is a matrix or a higher-dimensional array, for consistency with other array types. Use LinearIndices(a)[findmin(a)[2]] to get the old behavior, or CartesianIndices(a)[findmin(a)[2]] to get the new behavior on previous Julia versions (#30102).
Method signatures such as f(::Type{T}, ::T) where {T <: X} and f(::Type{X}, ::Any) are now considered ambiguous. Previously a bug caused the first one to be considered more specific in some cases (#30160).
Command-line option changes:
When a script run in interactive mode (-i) throws an error, the REPL now starts after the error is displayed. Previously the REPL only started if the script completed without error (#21233).
New library functions:
splitpath(p::String) function, which is the opposite of joinpath(parts...): it splits a filepath into its components (#28156).
isnothing(::Any) predicate, to check whether the argument is nothing. (#29679).
getpid(::Process) method (#24064).
eachrow, eachcol and eachslice functions provide efficient iterators over slices of arrays (#29749).
fieldtypes(T::Type) which returns the declared types of the field in type T (#29600).
uuid5 has been added to the UUIDs standard library (#28761).
Predicates Sys.isfreebsd, Sys.isopenbsd, Sys.isnetbsd, and Sys.isdragonfly for detecting BSD systems have been added (#30249).
Internal Base.disable_library_threading that sets libraries to use one thread. It executes function hooks that have been registered with Base.at_disable_library_threading (#30004).
Standard library changes:
CartesianIndices can now be constructed from two CartesianIndexes I and J with I:J (#29440).
CartesianIndices support broadcasting arithmetic (+ and -) with a CartesianIndex (#29890).
copy! support for arrays, dicts, and sets has been moved to Base from the Future package (#29173).
Channels now convert inserted values (like containers) instead of requiring types to match (#29092).
range can accept the stop value as a positional argument, e.g. range(1,10,step=2) (#28708).
diff now supports arrays of arbitrary dimensionality and can operate over any dimension (#29827).
The constructor BigFloat(::BigFloat) now respects the global precision setting and always returns a BigFloat with precision equal to precision(BigFloat) (#29127). The optional precision argument to override the global setting is now a keyword instead of positional argument (#29157).
The use of scientific notation when printing BigFloat values is now consistent with other floating point types (#29211).
Regex now behaves like a scalar when used in broadcasting (#29913).
Char now behaves like a read-only 0-dimensional array (#29819).
parse now allows strings representing integer 0 and 1 for type Bool (#29980).
Base.tail now works on named tuples (#29595).
The process id is appended to malloc log files in order to track memory allocations of multiple processes (#29969).
Base.julia_cmd now propagates the --inline=(yes|no) flag (#29858).
Base.@kwdef can now be used for parametric structs, and for structs with supertypes (#29316).
merge(::NamedTuple, ::NamedTuple...) can now be used with more than 2 NamedTuples (#29259).
New ncodeunits(c::Char) method as a fast equivalent to ncodeunits(string(c)) (#29153).
New sort!(::AbstractArray; dims) method that can sort the array along the dims dimension (#28902).
range now accepts stop as a positional argument (#28708).
get(A::AbstractArray, (), default) now returns A[] instead of an empty array (#30270).
parse(Bool, str) is now supported (#29997).
copyto!(::AbstractMatrix, ::UniformScaling) now supports rectangular matrices (#28790).
current_project() now searches the parent directories of a Git repository for a Project.toml file. This also affects the behavior of the --project command line option when using the default --project=@. (#29108).
The spawn API is now more flexible and supports taking IOBuffer directly as an I/O stream, converting to a system pipe as needed (#30278).
Dates:
New DateTime(::Date, ::Time) constructor (#29754).
TimeZone now behaves like a scalar when used in broadcasting (#30159).
InteractiveUtils:
edit can now be called on a module to edit the file that defines it (#29636).
All compiler-reflection tools (i.e. the code_ class of functions and macros) now print accurate line number and inlining information in a common style, and take an optional parameter (debuginfo=:default) to control the verbosity of the metadata shown (#29893).
LinearAlgebra:
isdiag and isposdef for Diagonal and UniformScaling (#29638).
mul!, rmul! and lmul! methods for UniformScaling (#29506).
Symmetric and Hermitian matrices now preserve the wrapper when scaled with a number (#29469).
Exponentiation operator ^ now supports raising an Irrational to an AbstractMatrix power (#29782).
Added keyword arguments rtol, atol to rank (#29926).
Random:
randperm and randcycle now use the type of their argument to determine the element type of the returned array (#29670).
A new method rand(::Tuple) implements sampling from the values of a tuple (#25278).
serialize and deserialize now accept a filename argument, like write and read (#30151).
SparseArrays:
sprandn now supports specifying the output element type (#30083).
Statistics:
mean and var now handle more kinds of empty inputs (#29033).
External dependencies:
7zip (bundled with Julia on Windows) has been upgraded from version 16.04 to 18.05 (#30035).
Busybox is no longer bundled with Julia on Windows (#30022).
OpenBLAS has been upgraded from 0.3.2 to 0.3.3 (#29845).
The source code for Pkg is no longer included in JuliaLang/julia. Pkg is instead downloaded during the build process (#29615).
LLVM has been upgraded to 6.0.1 and support for LLVM < 6.0 has been dropped (#28745, #28696).
Pkg has been upgraded to version 1.1 (#30342).
Deprecated or removed:
one(i::CartesianIndex) should be replaced with oneunit(i::CartesianIndex) (#29442).
The internal array Base.Grisu.DIGITS is deprecated; new code should use Base.Grisu.getbuf() to get an appropriate task-local buffer and pass it to grisu() instead (#29907).
The internal function Base._default_type(T) has been removed. Calls to it should be replaced with just the argument T (#29739).
peakflops has been scheduled to move from InteractiveUtils to LinearAlgebra but is already now available as LinearAlgebra.peakflops (#29978).

New in Julia Language 1.0.1 (Oct 22, 2018)

New language features:
Local variables can be tested for being defined using the new @isdefined variable macro (#22281).
Destructuring in function arguments: when an expression such as (x, y) is used as a function argument name, the argument is unpacked into local variables x and y as in the assignment (x, y) = arg (#6614).
Named tuples, with the syntax (a=1, b=2). These behave very similarly to tuples, except components can also be accessed by name using dot syntax t.a (#22194).
Keyword argument containers (kw in f(; kw...)) are now based on named tuples. Dictionary functions like haskey and indexing can be used on them, and name-value pairs can be iterated using pairs(kw). kw can no longer contain multiple entries for the same argument name (#4916).
Custom infix operators can now be defined by appending Unicode combining marks, primes, and sub/superscripts to other operators. For example, +̂ₐ″ is parsed as an infix operator with the same precedence as + (#22089).
The macro call syntax @macroname[args] is now available and is parsed as @macroname([args]) (#23519).
The construct if @generated ...; else ...; end can be used to provide both @generated and normal implementations of part of a function. Surrounding code will be common to both versions (#23168).
Added ⟂ (perp) operator with comparison precedence (#24404).
The missing singleton object (of type Missing) has been added to represent missing values (#24653). It propagates through standard operators and mathematical functions, and implements three-valued logic, similar to SQLs NULL and R's NA.
Field access via dot-syntax can now be overloaded by adding methods to Base.getproperty and Base.setproperty! (#1974), optionally along with a corresponding Base.propertynames method for reflection (#25311).
Values for Enums can now be specified inside of a begin block when using the @enum macro (#25424).
Keyword arguments can be required: if a default value is omitted, then an exception is thrown if the caller does not assign the keyword a value (#25830).
The pair operator => is now broadcastable as .=> which was previously a parsing error (#27447)
Language changes:
The syntax for parametric methods, function f{T}(x::T), has been changed to function f(x::T) where {T} (#11310).
The fallback constructor that calls convert is deprecated. Instead, new types should prefer to define constructors, and add convert methods that call those constructors only as necessary (#15120).
The syntax 1.+2 is deprecated, since it is ambiguous: it could mean either 1 .+ 2 (the current meaning) or 1. + 2 (#19089).
Mutable structs with no fields are no longer singletons; it is now possible to make multiple instances of them that can be distinguished by === (#25854). Zero-size immutable structs are still singletons.
In string and character literals, backslash may no longer precede unrecognized escape characters (#22800).
Juxtaposing binary, octal, and hexadecimal literals is deprecated, since it can lead to confusing code such as 0xapi == 0xa * pi (#16356).
Numeric literal juxtaposition now has slighty lower precedence than unary operators, so for example √2x parses as (√2) * x (#27641).
Declaring arguments as x::ANY to avoid specialization has been replaced by @nospecialize x. (#22666).
This can also be used in global scope, to apply to all subsequent method definitions in the module (until @specialize). (#28065)
Keyword argument default values are now evaluated in successive scopes --- the scope for each expression includes only previous keyword arguments, in left-to-right order (#17240).
The parsing of 1<<2*3 as 1<<(2*3) is deprecated, and will change to (1<<2)*3 in a future version (#13079).
The parsing of <| is now right associative. |> remains left associative (#24153).
: now parses like other operators, as a call to a function named :, instead of calling colon (#25947).
{ } expressions now use braces and bracescat as expression heads instead of cell1d and cell2d, and parse similarly to vect and vcat (#8470).
Nested if expressions that arise from the keyword elseif now use elseif as their expression head instead of if (#21774).
let blocks now parse the same as for loops; the first argument is either an assignment or block of assignments, and the second argument is a block of statements (#21774).
do syntax now parses to an expression with head :do, instead of as a function call (#21774).
Parsed and lowered forms of type definitions have been synchronized with their new keywords (#23157). Expression heads are renamed as follows:
type => struct
bitstype => primitive (order of arguments is also reversed, to match syntax)
composite_type => struct_type
bits_type => primitive_type
The global keyword now only introduces a new binding if one doesn't already exist in the module. This means that assignment to a global (global sin = 3) may now throw the error: "cannot assign variable Base.sin from module Main", rather than emitting a warning. Additionally, the new bindings are now created before the statement is executed. For example, f() = (global sin = "gluttony"; nothing) will now resolve which module contains sin eagerly, rather than delaying that decision until f is run. (#22984).
global const declarations may no longer appear inside functions (#12010).
Uninitialized BitArray constructors of the form BitArray[{N}](shape...) have been deprecated in favor of equivalents accepting undef (an alias for UndefInitializer()) as their first argument, as in BitArray[{N}](undef, shape...). For example, BitVector(3) is now BitVector(undef, 3), BitMatrix((2, 4)) is now BitMatrix(undef, (2, 4)), and BitArray{3}(11, 13, 17) is now BitArray{3}(undef, 11, 14, 17) (#24785).
Dispatch rules have been simplified: method matching is now determined exclusively by subtyping; the rule that method type parameters must also be captured has been removed. Instead, attempting to access the unconstrained parameters will throw an UndefVarError. Linting in package tests is recommended to confirm that the set of methods which might throw UndefVarError when accessing the static parameters (need_to_handle_undef_sparam = Set{Any}(m.sig for m in Test.detect_unbound_args(Base, recursive=true))) is equal (==) to some known set (expected = Set()). (#23117)
const declarations on local variables were previously ignored. They now give a warning, so that this syntax can be disallowed or given a new meaning in a future version (#5148).
Placing an expression after catch, as in catch f(x), is deprecated. Use catch; f(x) instead (#19987).
In for i = ..., if a local variable i already existed it would be overwritten during the loop. This behavior is deprecated, and in the future for loop variables will always be new variables local to the loop (#22314). The old behavior of overwriting an existing variable is available via for outer i = ....
In for i in x, x used to be evaluated in a new scope enclosing the for loop. Now it is evaluated in the scope outside the for loop.
In for i in x, j in y, all variables now have fresh bindings on each iteration of the innermost loop. For example, an assignment to i will not be visible on the next j loop iteration (#330).
Variable bindings local to while loop bodies are now freshly allocated on each loop iteration, matching the behavior of for loops.
Prefix & for by-reference arguments to ccall has been deprecated in favor of Ref argument types (#6080).
The constructor Ref(x::T) now always returns a Ref{T} (#21527).
All line numbers in ASTs are represented by LineNumberNodes; the :line expression head is no longer used. QuoteNodes are also consistently used for quoted symbols instead of the :quote expression head (though :quote Exprs are still used for quoted expressions) (#23885).
The + and - methods for Number and UniformScaling are not ambiguous anymore since + and - no longer do automatic broadcasting. Hence, the methods for UniformScaling and Number are no longer deprecated (#23923).
The keyword importall is deprecated. Use using and/or individual import statements instead (#22789).
reduce(+, [...]) and reduce(*, [...]) no longer widen the iterated over arguments to system word size. sum and prod still preserve this behavior. (#22825)
Like _, variable names consisting only of underscores can be assigned, but accessing their values is deprecated (#24221).
Raw string literal escaping rules have been changed to make it possible to write all strings. The rule is that backslashes escape both quotes and other backslashes, but only when a sequence of backslashes precedes a quote character. Thus, 2n backslashes followed by a quote encodes n backslashes and the end of the literal while 2n+1 backslashes followed by a quote encodes n backslashes followed by a quote character (#22926).
reprmime(mime, x) has been renamed to repr(mime, x), and along with repr(x) and sprint it now accepts an optional context keyword for IOContext attributes. stringmime has been moved to the Base64 stdlib package (#25990).
The syntax (x...) for constructing a tuple is deprecated; use (x...,) instead (#24452).
Non-parenthesized interpolated variables in strings, e.g. "$x", must be followed by a character that will never be an allowed identifier character (currently operators, space/control characters, or common punctuation characters) (#25231).
The syntax using A.B can now only be used when A.B is a module, and the syntax using A: B can only be used for adding single bindings (#8000).
=> now has its own precedence level, giving it strictly higher precedence than = and , (#25391).
The conditions under which unary operators followed by ( are parsed as prefix function calls have changed (#26154).
begin is disallowed inside indexing expressions, in order to enable the syntax a[begin] (for selecting the first element) in the future (#23354).
Underscores for _italics_ and __bold__ are now supported by the Base Markdown parser. (#25564)
… (dots) and ⁝ (tricolon) are now parsed as binary operators (#26262).
Assignment syntax (a=b) inside square bracket expressions (e.g. A[...], [x, y]) is deprecated. It will likely be reclaimed in a later version for passing keyword arguments. Note this does not affect updating operators like += (#25631).
try blocks without catch or finally are no longer allowed. An explicit empty catch block should be written instead (#27554).
AbstractArray types that use unconventional (not 1-based) indexing can now support size, length, and @inbounds. To optionally enforce conventional indices, you can @assert !has_offset_axes(A).
Module pre-compilation is now the default for code loading. Adding a __precompile__() declaration is no longer necessary, although __precompile__(false) can still be used to opt-out ([#26991]).
Breaking changes:
This section lists changes that do not have deprecation warnings.:
The package manager Pkg has been replaced with a new one. See the manual entries on "Code Loading" and "Pkg" for documentation.
replace(s::AbstractString, pat=>repl) for function repl arguments formerly passed a substring to repl in all cases. It now passes substrings for string patterns pat, but a Char for character patterns (when pat is a Char, collection of Char, or a character predicate) (#25815).
readuntil now does not include the delimiter in its result, matching the behavior of readline. Pass keep=true to get the old behavior (#25633).
lu methods now return decomposition objects such as LU rather than tuples of arrays or tuples of numbers (#26997, #27159, #27212).
schur methods now return decomposition objects such as Schur and GeneralizedSchur rather than tuples of arrays (#26997, #27159, #27212).
lq methods now return decomposition objects such as LQ rather than tuples of arrays (#26997, #27159, #27212).
qr methods now return decomposition objects such as QR, QRPivoted, and QRCompactWY rather than tuples of arrays (#26997, #27159, #27212).
svd methods now return decomposition objects such as SVD and GeneralizedSVD rather than tuples of arrays or tuples of numbers (#26997, #27159, #27212).
countlines now always counts the last non-empty line even if it does not end with EOL, matching the behavior of eachline and readlines (#25845).
getindex(s::String, r::UnitRange{Int}) now throws StringIndexError if last(r) is not a valid index into s (#22572).
ntuple(f, n::Integer) throws ArgumentError if n is negative. Previously an empty tuple was returned (#21697).
⋮, ⋱, ⋰, and ⋯ are now parsed as binary operators, not ordinary identifiers. ≔, ≕, and ⩴ now parse with assignment rather than comparison precedence (#26262).
Juxtaposing string literals (e.g. "x"y) is now a syntax error (#20575).
finalizer(function, object) now returns object rather than nothing (#24679).
The constructor of SubString now checks if the requested view range is defined by valid indices in the parent AbstractString (#22511).
Macro calls with for expressions are now parsed as generators inside function argument lists (#18650).
Examples:
sum(@inbounds a[i] for i = 1:n) used to give a syntax error, but is now parsed as sum(@inbounds(a[i]) for i = 1:n).
sum(@m x for i = 1:n end) used to parse the argument to sum as a 2-argument call to macro @m, but now parses it as a generator plus a syntax error for the dangling end.
@__DIR__ returns the current working directory rather than nothing when not run from a file (#21759).
@__FILE__ and @__DIR__ return information relative to the file that it was parsed from, rather than from the task-local SOURCE_PATH global when it was expanded.
All macros receive an extra argument __source__::LineNumberNode which describes the parser location in the source file for the @ of the macro call. It can be accessed as a normal argument variable in the body of the macro. This is implemented by inserting an extra leading argument into the Expr(:macrocall, :@name, LineNumberNode(...), args...) surface syntax. (#21746)
Passing the same keyword argument multiple times is now a syntax error (#16937).
getsockname on a TCPSocket now returns the locally bound address and port of the socket. Previously the address of the remote endpoint was being returned (#21825).
The ~/.juliarc.jl file has been moved to ~/.julia/config/startup.jl and /etc/julia/juliarc.jl file has been renamed to /etc/julia/startup.jl (#26161).
Using ARGS within startup.jl files or within a .jl file loaded with --load will no longer contain the script name as the first argument. Instead, the script name will be assigned to PROGRAM_FILE. (#22092)
The format for a ClusterManager specifying the cookie on the command line is now --worker=<cookie>. --worker <cookie> will not work as it is now an optional argument.
The representation of CartesianRange has changed to a tuple-of-AbstractUnitRanges; the start and stop fields are no longer present. Use first(R) and last(R) to obtain start/stop. (#20974)
The Diagonal, Bidiagonal, Tridiagonal and SymTridiagonal type definitions have changed from Diagonal{T}, Bidiagonal{T}, Tridiagonal{T} and SymTridiagonal{T} to Diagonal{T,V<:AbstractVector{T}}, Bidiagonal{T,V<:AbstractVector{T}}, Tridiagonal{T,V<:AbstractVector{T}} and SymTridiagonal{T,V<:AbstractVector{T}} respectively (#22718, #22925, #23035, #23154).
The immediate supertype of BitArray is now simply AbstractArray. BitArray is no longer considered a subtype of DenseArray and StridedArray (#25858).
When called with an argument that contains NaN elements, findmin and findmax now return the first NaN found and its corresponding index. Previously, NaN elements were ignored. The new behavior matches that of min, max, minimum, and maximum.
isapprox(x,y) now tests norm(x-y) <= max(atol, rtol*max(norm(x), norm(y))) rather than norm(x-y) <= atol + ..., and rtol defaults to zero if an atol > 0 is specified (#22742).
Spaces are no longer allowed between @ and the name of a macro in a macro call (#22868).
Juxtaposition of a non-literal with a macro call (x@macro) is no longer valid syntax (#22868).
On a cluster, all files are now loaded from the local file system rather than node 1 (#22588). To load the same file everywhere from node 1, one possible alternative is to broadcast a call to include_string: @everywhere include_string(Main, $(read("filename", String)), "filename"). Improving upon this API is left as an opportunity for packages.
randperm(n) and randcycle(n) now always return a Vector{Int} (independent of the type of n). Use the corresponding mutating functions randperm! and randcycle! to control the array type (#22723).
Hermitian now ignores any imaginary components in the diagonal instead of checking the diagonal. (#17367)
Worker-worker connections are setup lazily for an :all_to_all topology. Use keyword arg lazy=false to force all connections to be setup during a addprocs call. (#22814)
In joinpath(a, b) on Windows, if the drive specifications of a and b do not match, joinpath now returns b instead of throwing an ArgumentError. joinpath(path...) is defined to be left associative, so if any argument has a drive path which does not match the drive of the join of the preceding paths, the prior ones are dropped. (#20912)
^(A::AbstractMatrix{<:Integer}, p::Integer) now throws a DomainError if p < 0, unless A == one(A) or A == -one(A) (same as for ^(A::Integer, p::Integer)) (#23366).
^(A::AbstractMatrix{<:Integer}, p::Integer) now promotes the element type in the same way as ^(A::Integer, p::Integer). This means, for instance, that [1 1; 0 1]^big(1) will return a Matrix{BigInt} instead of a Matrix{Int} (#23366).
The element type of the input is now preserved in unique. Previously the element type of the output was shrunk to fit the union of the type of each element in the input. (#22696)
The promote function now raises an error if its arguments are of different types and if attempting to convert them to a common type fails to change any of their types. This avoids stack overflows in the common case of definitions like f(x, y) = f(promote(x, y)...) (#22801).
indmin and indmax have been renamed to argmin and argmax, respectively (#25654).
findmin, findmax, argmin, and argmax used to always return linear indices. They now return CartesianIndexes for all but 1-d arrays, and in general return the keys of indexed collections (e.g. dictionaries) (#22907).
The openspecfun library is no longer built and shipped with Julia, as it is no longer used internally (#22390).
All loaded packages used to have bindings in Main (e.g. Main.Package). This is no longer the case; now bindings will only exist for packages brought into scope by typing using Package or import Package (#17997).
The rules for mixed-signedness integer arithmetic (e.g. Int32(1) + UInt64(1)) have been simplified: if the arguments have different sizes (in bits), then the type of the larger argument is used. If the arguments have the same size, the unsigned type is used (#9292).
All command line arguments passed via -e, -E, and -L will be executed in the order given on the command line (#23665).
I now yields UniformScaling{Bool}(true) rather than UniformScaling{Int64}(1) to better preserve types in operations involving I (#24396).
The return type of reinterpret has changed to ReinterpretArray. reinterpret on sparse arrays has been discontinued.
Base.find_in_path is now Base.find_package or Base.find_source_file (#24320).
finalizer now takes functions or pointers as its first argument, and the object being finalized as its second (rather than the reverse). For the majority of use cases deprecation warnings will be triggered. However, deprecation warnings will not trigger where (1) the callable argument is not a subtype of Function; or (2) both arguments are Functions or Ptr{Cvoid}s (#24605).
The kill function now throws errors on user error (e.g. on permission errors), but returns successfully if the process had previously exited. Its return value has been removed. Use the process_running function to determine if a process has already exited.
The logging system has been redesigned - info and warn are deprecated and replaced with the logging macros @info, @warn, @debug and @error. The logging function is also deprecated and replaced with AbstractLogger and the functions from the new standard Logging library. (#24490)
The RevString type has been removed from the language; reverse(::String) returns a String with code points (or fragments thereof) in reverse order. In general, reverse(s) should return a string of the same type and encoding as s with code points in reverse order; any string type overrides reverse to return a different type of string must also override reverseind to compute reversed indices correctly.
eachindex(A, B...) now requires that all inputs have the same number of elements. When the chosen indexing is Cartesian, they must have the same axes.
AbstractRange objects are now considered as equal to other AbstractArray objects by == and isequal if all of their elements are equal (#16401). This has required changing the hashing algorithm: ranges now use an O(N) fallback instead of a O(1) specialized method unless they define the Base.RangeStepStyle trait; see its documentation for details. Types which support subtraction (operator -) must now implement widen for hashing to work inside heterogeneous arrays.
findn(x::AbstractArray) has been deprecated in favor of findall(!iszero, x), which now returns cartesian indices for multidimensional arrays (see below, #25532).
Broadcasting operations are no longer fused into a single operation by Julia's parser. Instead, a lazy Broadcasted object is created to represent the fused expression and then realized with copy(bc::Broadcasted) or copyto!(dest, bc::Broadcasted) to evaluate the wrapper. Consequently, package authors generally need to specialize copy and copyto! methods rather than broadcast and broadcast!. This also allows for more customization and control of fused broadcasts. See the Interfaces chapter for more information.
find has been renamed to findall. findall, findfirst, findlast, findnext now take and/or return the same type of indices as keys/pairs for AbstractArray, AbstractDict, AbstractString, Tuple and NamedTuple objects (#24774, #25545). In particular, this means that they use CartesianIndex objects for matrices and higher-dimensional arrays instead of linear indices as was previously the case. Use LinearIndices(a)[findall(f, a)] and similar constructs to compute linear indices.
The find* functions, i.e. findnext, findprev, findfirst, and findlast, as well as indexin, now return nothing when no match is found rather than 0 or 0:-1 (#25472, #25662, #26149)
The Base.HasShape iterator trait has gained a type parameter N indicating the number of dimensions, which must correspond to the length of the tuple returned by size (#25655).
AbstractSet objects are now considered equal by == and isequal if all of their elements are equal (#25368). This has required changing the hashing algorithm for BitSet.
the default behavior of titlecase is changed in two ways (#23393):
characters not starting a word are converted to lowercase; a new keyword argument strict is added which allows to get the old behavior when it's false.
any non-letter character is considered as a word separator; to get the old behavior (only "space" characters are considered as word separators), use the keyword wordsep=isspace.
writedlm in the standard library module DelimitedFiles now writes numeric values using print rather than print_shortest (#25745).
The tempname function used to create a file on Windows but not on other platforms. It now never creates a file (#9053).
The fieldnames and propertynames functions now return a tuple rather than an array (#25725).
indexin now returns the first rather than the last matching index (#25998).
parse(::Type, ::Char) now uses a default base of 10, like other number parsing methods, instead of 36 (#26576).
isequal for Ptrs now compares element types; == still compares only addresses (#26858).
widen on 8- and 16-bit integer types now widens to 16- and 32-bit types, respectively. (#28045).
mv,cp, touch, mkdir, mkpath, chmod and chown now return the path that was created/modified rather than nothing (#27071).
Regular expressions now default to UCP mode. Escape sequences such as w will now match based on unicode character properties, e.g. r"w+" will match café (not just caf). Add the a modifier (e.g. r"w+"a) to restore the previous behavior (#27189).
@sync now waits only for lexically enclosed (i.e. visible directly in the source text of its argument) @async expressions. If you need to wait for a task created by a called function f, have f return the task and put @async wait(f(...)) within the @sync block. This change makes @schedule redundant with @async, so @schedule has been deprecated (#27164).
norm(A::AbstractMatrix, p=2) computes no longer the operator/matrix norm but the norm of A as for other iterables, i.e. as if it were a vector. Especially, norm(A::AbstractMatrix) is the Frobenius norm. To compute the operator/matrix norm, use the new function opnorm (#27401).
dot(u, v) now acts recursively. Instead of sum(u[i]' * v[i] for i in ...), it computes sum(dot(u[i], v[i]) for i in ...), similarly to vecdot before (#27401).
Sys.CPU_CORES has been renamed to Sys.CPU_THREADS; it still gives the number of "logical cores" (including hyperthreading) rather than the number of physical cores present on the CPU. Similarly, the environment variable JULIA_CPU_CORES is deprecated in favor of JULIA_CPU_THREADS (#27856).
WeakKeyDict does not convert keys on insertion anymore (#24941).
Library improvements:
The function thisind(s::AbstractString, i::Integer) returns the largest valid index less or equal than i in the string s or 0 if no such index exists (#24414).
Support for Unicode 11 (#28266).
Char is now a subtype of AbstractChar, and most of the functions that take character arguments now accept any AbstractChar (#26286).
pathof(module) returns the path a module was imported from ([#28310]).
bytes2hex now accepts an optional io argument to output to a hexadecimal stream without allocating a String first (#27121).
String(array) now accepts an arbitrary AbstractVector{UInt8}. For Vector inputs, it "steals" the memory buffer, leaving them with an empty buffer which is guaranteed not to be shared with the String object. For other types of vectors (in particular immutable vectors), a copy is made and the input is not truncated (#26093).
Irrational is now a subtype of AbstractIrrational (#24245).
Introduced the empty function, the functional pair to empty! which returns a new, empty container (#24390).
Jump to first/last history entries in the REPL via "Alt-<" and "Alt->" (#22829).
REPL LaTeX-like tab completions have been simplified for several Unicode characters, e.g. 𝔸 is now bbA rather than BbbA (#25980).
The function chop now accepts two arguments head and tail allowing to specify number of characters to remove from the head and tail of the string (#24126).
get(io, :color, false) can now be used to query whether a stream io supports ANSI color codes (#25067), rather than using the undocumented Base.have_color global flag.
print_with_color has been deprecated in favor of printstyled([io], xs...; bold=false, color=:normal) for printing styled text (#25522).
Functions first and last now accept nchar argument for AbstractString. If this argument is used they return a string consisting of first/last nchar characters from the original string (#23960).
Expressions x^-n where n is an integer literal now correspond to inv(x)^n. For example, x^-1 is now essentially a synonym for inv(x), and works in a type-stable way even if typeof(x) != typeof(inv(x)) (#24240).
New Iterators.reverse(itr) for reverse-order iteration (#24187). Iterator types T can implement start etc. for Iterators.Reverse{T} to support this.
The functions nextind and prevind now accept nchar argument that indicates the number of characters to move (#23805).
The functions strip, lstrip and rstrip now return SubString (#22496).
The functions strwidth and charwidth have been merged into textwidth(#20816).
The functions base and digits digits now accept a negative base (like ndigits did) (#21692).
The function randn now accepts complex arguments (Complex{T <: AbstractFloat}) (#21973).
parse(Complex{T}, string) can parse complex numbers in some common formats (#24713).
The function rand can now pick up random elements from strings, associatives and sets (#22228, #21960, #18155, #22224).
It's now possible to specify the characters to pick from in the randstring function (#22222).
Allow multidimensional arrays in shuffle and shuffle! functions (#22226).
Method lists are now printed as a numbered list. In addition, the source code of a method can be opened in an editor by entering the corresponding number in the REPL and pressing ^Q (#22007).
getpeername on a TCPSocket returns the address and port of the remote endpoint of the TCP connection (#21825).
resize! and sizehint! methods no longer over-reserve memory when the requested array size is more than double of its current size (#22038).
The crc32c function for CRC-32c checksums is now exported (#22274).
eye(::Type{Diagonal{T}}, m::Integer) has been deprecated in favor of Diagonal{T}(I, m) (#24415).
The output of versioninfo is now controlled with keyword arguments (#21974).
The function LibGit2.set_remote_url now always sets both the fetch and push URLs for a git repo. Additionally, the argument order was changed to be consistent with the git command line tool (#22062).
Added unique! which is an inplace version of unique (#20549).
@test isequal(x, y) and @test isapprox(x, y) now prints an evaluated expression when the test fails (#22296).
Uses of Val{c} in Base has been replaced with Val{c}(), which is now easily accessible via the efficient constructor Val(c). Functions are defined as f(::Val{c}) = ... and called by f(Val(c)). Notable affected functions include: ntuple, Base.literal_pow, sqrtm, lufact, lufact!, qrfact, qrfact!, cholfact, cholfact!, _broadcast!, reshape, cat and cat_t.
A new @macroexpand1 macro for non recursive macro expansion (#21662).
Chars can now be concatenated with Strings and/or other Chars using * (#22532).
Diagonal, Bidiagonal, Tridiagonal and SymTridiagonal are now parameterized on the type of the wrapped vectors, allowing Diagonal, Bidiagonal, Tridiagonal and SymTridiagonal matrices with arbitrary AbstractVectors (#22718, #22925, #23035, #23154).
Mutating versions of randperm and randcycle have been added: randperm! and randcycle! (#22723).
BigFloat random numbers can now be generated (#22720).
The efficiency of random generation for MersenneTwister RNGs has been improved for integers, Float64 and ranges; as a result, given a seed, the produced stream of numbers has changed (#27560, #25277, #25197, #25058, #25047).
REPL Undo via Ctrl-/ and Ctrl-_
diagm now accepts several diagonal index/vector Pairs (#24047).
isequal, ==, and in have one argument "curried" forms. For example isequal(x) returns a function that compares its argument to x using isequal (#26436).
reinterpret now works on any AbstractArray using the new ReinterpretArray type. This supersedes the old behavior of reinterpret on Arrays. As a result, reinterpreting arrays with different alignment requirements (removed in 0.6) is once again allowed (#23750).
The keys of an Associative are now an AbstractSet. Base.KeyIterator{<:Associative} has been changed to KeySet{K, <:Associative{K}} <: AbstractSet{K} (#24580).
New function ncodeunits(s::AbstractString) gives the number of code units in a string. The generic definition is constant time but calls lastindex(s) which may be inefficient. Therefore custom string types may want to define direct ncodeunits methods.
reverseind(s::AbstractString, i::Integer) now has an efficient generic fallback, so custom string types do not need to provide their own efficient definitions. The generic definition relies on ncodeunits however, so for optimal performance you may need to define a custom method for that function.
The global RNG is being re-seeded with its own seed at the beginning of each @testset, and have its original state restored at the end (#24445). This is breaking for testsets relying implicitly on the global RNG being in a specific state.
permutedims(m::AbstractMatrix) is now short for permutedims(m, (2,1)), and is now a more convenient way of making a "shallow transpose" of a 2D array. This is the recommended approach for manipulating arrays of data, rather than the recursively defined, linear-algebra function transpose. Similarly, permutedims(v::AbstractVector) will create a row matrix (#24839).
A new replace(A, old=>new) function is introduced to replace old by new in collection A. There is also another method with a different API, and a mutating variant, replace! (#22324, #25697, #26206, #27944).
Adding integers to CartesianIndex objects is now deprecated. Instead of i::Int + x::CartesianIndex, use i*one(x) + x (#26284).
CartesianRange changes (#24715):
Inherits from AbstractArray, and linear indexing can be used to provide linear-to-cartesian conversion (#24715)
It has a new constructor taking an array
several missing set-like operations have been added (#23528): union, intersect, symdiff, setdiff are now implemented for all collections with arbitrary many arguments, as well as the mutating counterparts (union! etc.). The performance is also much better in many cases. Note that this change is slightly breaking: all the non-mutating functions always return a new object even if only one argument is passed.
Moreover the semantics of intersect and symdiff is changed for vectors:
intersect doesn't preserve the multiplicity anymore (use filter for the old behavior)
symdiff has been made consistent with the corresponding methods for other containers, by taking the multiplicity of the arguments into account. Use unique to get the old behavior.
The linearindices function has been deprecated in favor of the new LinearIndices type, which additionally provides conversion from cartesian indices to linear indices using the normal indexing operation. (#24715, #26775).
IdDict{K,V} replaces ObjectIdDict. It has type parameters like other AbstractDict subtypes and its constructors mirror the ones of Dict. (#25210)
IOBuffer can take the sizehint keyword argument to suggest a capacity of the buffer (#25944).
lstrip and rstrip now accept a predicate function that defaults to isspace (#27309).
trunc, floor, ceil, and round specify digits, sigdigits and base using keyword arguments. (#26156, #26670)
Sys.which() provides a cross-platform method to find executable files, similar to the Unix which command. (#26559)
Added an optimized method of vecdot for taking the Frobenius inner product of sparse matrices. (#27470)
Added an optimized method of kron for taking the tensor product of two Diagonal matrices. ([27581])
An official API for extending rand is now defined (#23964, #25002).
The constructor MersenneTwister() is re-enabled, producing a randomly initialized RNG (similar to Random.seed!(MersenneTwister(0))) (#21909).
BitSet can now store any Int (instead of only positive ones) (#25029).
The initial element v0 in reduce(op, v0, itr) has been replaced with an init optional keyword argument, as in reduce(op, itr; init=v0). Similarly for foldl, foldr, mapreduce, mapfoldl, mapfoldr, accumulate and accumulate!. (#27711, #27859)
Compiler/Runtime improvements:
The inlining heuristic now models the approximate runtime cost of a method (using some strongly-simplifying assumptions). Functions are inlined unless their estimated runtime cost substantially exceeds the cost of setting up and issuing a subroutine call. (#22210, #22732)
Inference recursion-detection heuristics are now more precise, allowing them to be triggered less often, but being more aggressive when they are triggered to drive the inference computation to a solution (#23912).
Inference now propagates constants inter-procedurally, and can compute various constants expressions at compile-time (#24362).
The LLVM SLP Vectorizer optimization pass is now enabled at the default optimization level.
Deprecated or removed:
The JULIA_HOME environment variable has been renamed to JULIA_BINDIR and Base.JULIA_HOME has been moved to Sys.BINDIR (#20899).
The keyword immutable is fully deprecated to struct, and type is fully deprecated to mutable struct (#19157, #20418).
lufact, schurfact, lqfact, qrfact, ldltfact, svdfact, bkfact, hessfact, eigfact, and cholfact have respectively been deprecated to lu, schur, lq, qr, ldlt, svd, bunchkaufman, hessenberg, eigen, and cholesky (#26997, #27159, #27212).
lufact!, schurfact!, lqfact!, qrfact!, ldltfact!, svdfact!, bkfact!, hessfact!, and eigfact! have respectively been deprecated to lu!, schur!, lq!, qr!, ldlt!, svd!, bunchkaufman!, hessenberg!, and eigen! (#26997, #27159, #27212).
eig(A[, args...]) has been deprecated in favor of eigen(A[, args...]). Whereas the former returns a tuple of arrays, the latter returns an Eigen object. So for a direct replacement, use (eigen(A[, args...])...,). But going forward, consider using the direct result of eigen(A[, args...]) instead, either destructured into its components (vals, vecs = eigen(A[, args...])) or as an Eigen object (X = eigen(A[, args...])) (#26997, #27159, #27212).
eig(A::AbstractMatrix, B::AbstractMatrix) and eig(A::Number, B::Number) have been deprecated in favor of eigen(A, B). Whereas the former each return a tuple of arrays, the latter returns a GeneralizedEigen object. So for a direct replacement, use (eigen(A, B)...,). But going forward, consider using the direct result of eigen(A, B) instead, either destructured into its components (vals, vecs = eigen(A, B)), or as a GeneralizedEigen object (X = eigen(A, B)) (#26997, #27159, #27212).
ordschur(T::StridedMatrix{Ty}, Z::StridedMatrix{Ty}, select::Union{Vector{Bool},BitVector}) and ordschur(S::StridedMatrix{Ty}, T::StridedMatrix{Ty}, Q::StridedMatrix{Ty}, Z::StridedMatrix{Ty}, select::Union{Vector{Bool},BitVector}) and their respective inplace versions have been deprecated. Use ordschur(schur::Schur, select::Union{Vector{Bool},BitVector}) and ordschur(gschur::GeneralizedSchur, select::Union{Vector{Bool},BitVector}) instead (#28155).
Indexing into multidimensional arrays with more than one index but fewer indices than there are dimensions is no longer permitted when those trailing dimensions have lengths greater than 1. Instead, reshape the array or add trailing indices so the dimensionality and number of indices match (#14770, #23628).
The use of a positional dimension argument has largely been deprecated in favor of a dims keyword argument. This includes the functions sum, prod, maximum, minimum, all, any, findmax, findmin, mean, varm, std, var, cov, cor, median, mapreducedim, reducedim, sort, accumulate, accumulate!, cumsum, cumsum!, cumprod, cumprod!, flipdim, dropdims, and cat (#25501, #26660, #27100).
indices(a) and indices(a,d) have been deprecated in favor of axes(a) and axes(a, d) (#25057).
EnvHash has been renamed to EnvDict (#24167).
Uninitialized Array constructors of the form Array[{T,N}](shape...) have been deprecated in favor of equivalents accepting undef (an alias for UndefInitializer()) as their first argument, as in Array[{T,N}](undef, shape...). For example, Vector(3) is now Vector(undef, 3), Matrix{Int}((2, 4)) is now, Matrix{Int}(undef, (2, 4)), and Array{Float32,3}(11, 13, 17) is now Array{Float32,3}(undef, 11, 13, 17) (#24781).
Previously setindex!(A, x, I...) (and the syntax A[I...] = x) supported two different modes of operation when supplied with a set of non-scalar indices I (e.g., at least one index is an AbstractArray) depending upon the value of x on the right hand side. If x is an AbstractArray, its contents are copied elementwise into the locations in A selected by I and it must have the same number of elements as I selects locations. Otherwise, if x is not an AbstractArray, then its value is implicitly broadcast to all locations to all locations in A selected by I. This latter behavior—implicitly broadcasting "scalar"-like values across many locations—is now deprecated in favor of explicitly using the broadcasted assignment syntax A[I...] .= x or fill!(view(A, I...), x) (#26347).
broadcast_getindex(A, I...) and broadcast_setindex!(A, v, I...) are deprecated in favor of getindex.((A,), I...) and setindex!.((A,), v, I...), respectively (#27075).
LinAlg.fillslots! has been renamed LinAlg.fillstored! (#25030).
fill!(A::Diagonal, x) and fill!(A::AbstractTriangular, x) have been deprecated in favor of Base.LinAlg.fillstored!(A, x) (#24413).
eye has been deprecated in favor of I and Matrix constructors. Please see the deprecation warnings for replacement details (#24438).
zeros(D::Diagonal[, opts...]) has been deprecated (#24654).
Using Bool values directly as indices is now deprecated and will be an error in the future. Convert them to Int before indexing if you intend to access index 1 for true and 0 for false.
slicedim(A, d, i) has been deprecated in favor of copy(selectdim(A, d, i)). The new selectdim function now always returns a view into A; in many cases the copy is not necessary. Previously, slicedim on a vector V over dimension d=1 and scalar index i would return the just selected element (unless V was a BitVector). This has now been made consistent: selectdim now always returns a view into the original array, with a zero-dimensional view in this specific case (#26009).
whos has been renamed varinfo, and now returns a markdown table instead of printing output (#12131).
Uninitialized RowVector constructors of the form RowVector{T}(shape...) have been deprecated in favor of equivalents accepting undef (an alias for UndefInitializer()) as their first argument, as in RowVector{T}(undef, shape...). For example, RowVector{Int}(3) is now RowVector{Int}(undef, 3), and RowVector{Float32}((1, 4)) is now RowVector{Float32}(undef, (1, 4)) (#24786).
writecsv(io, a; opts...) has been deprecated in favor of writedlm(io, a, ','; opts...) (#23529).
The method srand(rng, filename, n=4) has been deprecated (#21359).
readcsv(io[, T::Type]; opts...) has been deprecated in favor of readdlm(io, ','[, T]; opts...) (#23530).
sparse(s::UniformScaling, m::Integer) has been deprecated in favor of the three-argument equivalent sparse(s::UniformScaling, m, n) (#24472).
The cholfact/cholfact! methods that accepted an uplo symbol have been deprecated in favor of using Hermitian (or Symmetric) views (#22187, #22188).
The thin keyword argument for orthogonal decomposition methods has been deprecated in favor of full, which has the opposite meaning: thin == true if and only if full == false (#24279).
isposdef(A::AbstractMatrix, UL::Symbol) and isposdef!(A::AbstractMatrix, UL::Symbol) have been deprecated in favor of isposdef(Hermitian(A, UL)) and isposdef!(Hermitian(A, UL)) respectively (#22245).
The bkfact/bkfact! methods that accepted uplo and issymmetric symbols have been deprecated in favor of using Hermitian (or Symmetric) views (#22605).
The function current_module is deprecated and replaced with @__MODULE__. This caused the deprecation of some reflection methods (such as macroexpand and isconst), which now require a module argument. And it caused the bugfix of other default arguments to use the Main module (including whos, which) (#22064).
expand(ex) and expand(module, ex) have been deprecated in favor of Meta.lower(module, ex) (#22064, #24278).
ones(A::AbstractArray[, opts...]) and zeros(A::AbstractArray[, opts...]) methods have been deprecated. For zeros(A), consider zero(A). For ones(A) or zeros(A), consider ones(size(A)), zeros(size(A)), fill(v, size(A)) for v an appropriate one or zero, fill!(copy(A), {1|0}), fill!(similar(A), {1|0}), or any of the preceding with different element type and/or shape depending on opts.... Where strictly necessary, consider fill!(similar(A[, opts...]), {one(eltype(A)) | zero(eltype(A))}). For an algebraic multiplicative identity, consider one(A) (#24656).
The similar(dims->f(..., dims...), [T], axes...) method to add offset array support to a function f that would otherwise create a non-offset array has been deprecated. Instead, call f(..., axes...) directly and, if needed, the offset array implementation should add offset axis support to the function f directly (#26733).
The functions ones and zeros used to accept any objects as dimensional arguments, implicitly converting them to Ints. This is now deprecated; only Integers or AbstractUnitRanges are accepted as arguments. Instead, convert the arguments before calling ones or zeros (#26733).
The variadic size(A, dim1, dim2, dims...) method to return a tuple of multiple dimension lengths of A has been deprecated (#26862).
The Operators module is deprecated. Instead, import required operators explicitly from Base, e.g. import Base: +, -, *, / (#22251).
Bindings to the FFTW library have been removed from Base. The DFT framework for building FFT implementations is now in AbstractFFTs.jl, the bindings to the FFTW library are in FFTW.jl, and the Base signal processing functions which used FFTs are now in DSP.jl (#21956).
The corrected positional argument to cov has been deprecated in favor of a keyword argument with the same name (#21709).
Omitting spaces around the ? and the : tokens in a ternary expression has been deprecated. Ternaries must now include some amount of whitespace, e.g. x ? a : b rather than x?a:b (#22523 and #22712).
? can no longer be used as an identifier name (#22712)
The method replace(s::AbstractString, pat, r, [count]) is deprecated in favor of replace(s::AbstractString, pat => r; [count]) (#25165). Moreover, count cannot be negative anymore (use typemax(Int) instead (#22325).
read(io, type, dims) is deprecated to read!(io, Array{type}(undef, dims)) (#21450).
read(::IO, ::Ref) is now a method of read!, since it mutates its Ref argument (#21592).
nb_available is now bytesavailable (#25634).
skipchars(io::IO, predicate; linecomment=nothing) is deprecated in favor of skipchars(predicate, io::IO; linecomment=nothing) (#25667).
Bidiagonal constructors now use a Symbol (:U or :L) for the upper/lower argument, instead of a Bool or a Char (#22703).
Bidiagonal, Tridiagonal and SymTridiagonal constructors that automatically converted the input vectors to the same type are deprecated in favor of explicit conversion (#22925, #23035, #23154.
Calling nfields on a type to find out how many fields its instances have is deprecated. Use fieldcount instead. Use nfields only to get the number of fields in a specific object (#22350).
fieldnames now operates only on types. To get the names of fields in an object, use fieldnames(typeof(x)) (#22350).
InexactError, DomainError, and OverflowError now take arguments. InexactError(func::Symbol, type, -3) now prints as "ERROR: InexactError: func(type, -3)", DomainError(val, [msg]) prints as "ERROR: DomainError with val:nmsg", and OverflowError(msg) prints as "ERROR: OverflowError: msg". (#20005, #22751, #22761)
The operating system identification functions: is_linux, is_bsd, is_apple, is_unix, and is_windows, have been deprecated in favor of Sys.islinux, Sys.isbsd, Sys.isapple, Sys.isunix, and Sys.iswindows, respectively (#22182).
The forms of read, readstring, and eachline that accepted both a Cmd object and an input stream are deprecated. Use e.g. read(pipeline(stdin, cmd)) instead (#22762).
The unexported type AbstractIOBuffer has been renamed to GenericIOBuffer (#17360 #22796).
IOBuffer(data::AbstractVector{UInt8}, read::Bool, write::Bool, maxsize::Integer), IOBuffer(read::Bool, write::Bool), and IOBuffer(maxsize::Integer) are deprecated in favor of constructors taking keyword arguments (#25872).
Display has been renamed to AbstractDisplay (#24831).
Remaining vectorized methods over SparseVectors, particularly floor, ceil, trunc, round, and most common transcendental functions such as exp, log, and sin variants, have been deprecated in favor of dot-syntax (#22961).
The method String(io::IOBuffer) is deprecated to String(take!(copy(io))) (#21438).
The function readstring is deprecated in favor of read(io, String) (#22793)
The function showall is deprecated. Showing entire values is the default, unless an IOContext specifying :limit=>true is in use (#22847).
issubtype has been deprecated in favor of <: (which used to be an alias for issubtype).
Calling write on non-isbits arrays is deprecated in favor of explicit loops or serialize (#6466).
The default startup.jl file on Windows has been removed. Now must explicitly include the full path if you need access to executables or libraries in the Sys.BINDIR directory, e.g. joinpath(Sys.BINDIR, "7z.exe") for 7z.exe (#21540).
sqrtm has been deprecated in favor of sqrt (#23504).
expm has been deprecated in favor of exp (#23233).
logm has been deprecated in favor of log (#23505).
full has been deprecated in favor of more specific, better defined alternatives. On structured matrices A, consider instead Matrix(A), Array(A), SparseMatrixCSC(A), or sparse(A). On sparse arrays S, consider instead Vector(S), Matrix(S), or Array(S) as appropriate. On factorizations F, consider instead Matrix(F), Array(F), AbstractMatrix(F), or AbstractArray(F). On implicit orthogonal factors Q, consider instead Matrix(Q) or Array(Q); for implicit orthogonal factors that can be recovered in square or truncated form, see the deprecation message for square recovery instructions. On Symmetric, Hermitian, or AbstractTriangular matrices A, consider instead Matrix(S), Array(S), SparseMatrixCSC(S), or sparse(S). On Symmetric matrices A particularly, consider instead LinAlg.copytri!(copy(parent(A)), A.uplo). On Hermitian matrices A particularly, consider instead LinAlg.copytri!(copy(parent(A)), A.uplo, true). On UpperTriangular matrices A particularly, consider instead triu!(copy(parent(A))). On LowerTriangular matrices A particularly, consider instead tril!(copy(parent(A))) (#24250).
speye has been deprecated in favor of I, sparse, and SparseMatrixCSC constructor methods (#24356).
Calling union with no arguments is deprecated; construct an empty set with an appropriate element type using Set{T}() instead (#23144).
Vectorized DateTime, Date, and format methods have been deprecated in favor of dot-syntax (#23207).
Base.cpad has been removed; use an appropriate combination of rpad and lpad instead (#23187).
ctranspose and ctranspose! have been deprecated in favor of adjoint and adjoint!, respectively (#23235).
filter and filter! on dictionaries now pass a single key=>value pair to the argument function, instead of two arguments (#17886).
rol, rol!, ror, and ror! have been deprecated in favor of specialized methods for circshift/circshift! (#23404).
Base.SparseArrays.SpDiagIterator has been removed (#23261).
The function cfunction, has been deprecated in favor of a macro form @cfunction. Most existing uses can be upgraded simply by adding a @. The new syntax now additionally supports allocating closures at runtime, for dealing with C APIs that don't provide a separate void* env-type callback argument. (#26486)
diagm(v::AbstractVector, k::Integer=0) has been deprecated in favor of diagm(k => v) (#24047).
diagm(x::Number) has been deprecated in favor of fill(x, 1, 1) (#24047).
diagm(A::SparseMatrixCSC) has been deprecated in favor of spdiagm(sparsevec(A)) (#23341).
diagm(A::BitMatrix) has been deprecated, use diagm(0 => vec(A)) or BitMatrix(Diagonal(vec(A))) instead (#23373, #24047).
ℯ (written as mscre<TAB> or euler<TAB>) is now the only (by default) exported name for Euler's number, and the type has changed from Irrational{:e} to Irrational{:ℯ} (#23427).
The mathematical constants π, pi, ℯ, e, γ, eulergamma, catalan, φ and golden have been moved from Base to a new module; Base.MathConstants. Only π, pi and ℯ are now exported by default from Base (#23427).
eu (previously an alias for ℯ) has been deprecated in favor of ℯ (or MathConstants.e) (#23427).
GMP.gmp_version(), GMP.GMP_VERSION, GMP.gmp_bits_per_limb(), and GMP.GMP_BITS_PER_LIMB have been renamed to GMP.version(), GMP.VERSION, GMP.bits_per_limb(), and GMP.BITS_PER_LIMB, respectively. Similarly, MPFR.get_version(), has been renamed to MPFR.version() (#23323). Also, LinAlg.LAPACK.laver() has been renamed to LinAlg.LAPACK.version() and now returns a VersionNumber.
select, select!, selectperm and selectperm! have been renamed respectively to partialsort, partialsort!, partialsortperm and partialsortperm! (#23051).
The Range abstract type has been renamed to AbstractRange (#23570).
map on dictionaries previously operated on key=>value pairs. This behavior is deprecated, and in the future map will operate only on values (#5794).
map on sets previously returned a Set, possibly changing the order or number of elements. This behavior is deprecated and in the future map will preserve order and number of elements (#26980).
Previously, broadcast defaulted to treating its arguments as scalars if they were not arrays. This behavior is deprecated, and in the future broadcast will default to iterating over all its arguments. Wrap arguments you wish to be treated as scalars with Ref() or a 1-tuple. Package developers can choose to allow a non-iterable type T to always behave as a scalar by implementing broadcastable(x::T) = Ref(x) (#26212).
Automatically broadcasted + and - for array + scalar, scalar - array, and so-on have been deprecated due to inconsistency with linear algebra. Use .+ and .- for these operations instead (#22880, #22932).
flipbits!(B) is deprecated in favor of using in-place broadcast to negate each element: B .= .!B (#27067).
isleaftype is deprecated in favor of the simpler predicates isconcretetype and isdispatchtuple. Concrete types are those that might equal typeof(x) for some x; isleaftype included some types for which this is not true. Those are now categorized more precisely as "dispatch tuple types" and "!has_free_typevars" (not exported). (#17086, #25496)
contains(eq, itr, item) is deprecated in favor of any with a predicate (#23716).
spdiagm(x::AbstractVector) has been deprecated in favor of sparse(Diagonal(x)) alternatively spdiagm(0 => x) (#23757).
spdiagm(x::AbstractVector, d::Integer) and spdiagm(x::Tuple{<:AbstractVector}, d::Tuple{<:Integer}) have been deprecated in favor of spdiagm(d => x) and spdiagm(d[1] => x[1], d[2] => x[2], ...) respectively. The new spdiagm implementation now always returns a square matrix (#23757).
spones(A::AbstractSparseArray) has been deprecated in favor of LinAlg.fillstored!(copy(A), 1) (#25037).
Constructors for LibGit2.UserPasswordCredentials and LibGit2.SSHCredentials which take a prompt_if_incorrect argument are deprecated. Instead, prompting behavior is controlled using the allow_prompt keyword in the LibGit2.CredentialPayload constructor (#23690).
gradient is deprecated and will be removed in the next release (#23816).
The timing functions tic, toc, and toq are deprecated in favor of @time and @elapsed (#17046).
Methods of findfirst, findnext, findlast, and findprev that accept a value to search for are deprecated in favor of passing a predicate (#19186, #10593).
find functions now operate only on booleans by default. To look for non-zeros, use x->x!=0 or !iszero (#23120).
The ability of reinterpret to yield Arrays of different type than the underlying storage has been removed. The reinterpret function is still available, but now returns a ReinterpretArray. The three argument form of reinterpret that implicitly reshapes has been deprecated (#23750).
bits has been deprecated in favor of bitstring (#24281, #24263).
num2hex and hex2num have been deprecated in favor of reinterpret combined with parse/hex (#22088).
copy! is deprecated for AbstractSet and AbstractDict, with the intention to re-enable it with a cleaner meaning in a future version (#24844).
copy! (resp. unsafe_copy!) is deprecated for AbstractArray and is renamed copyto! (resp. unsafe_copyto!); it will be re-introduced with a different meaning in a future version (#24808).
a:b is deprecated for constructing a StepRange when a and b have physical units (Dates and Times). Use a:s:b, where s = Dates.Day(1) or s = Dates.Second(1).
trues(A::AbstractArray) and falses(A::AbstractArray) are deprecated in favor of trues(size(A)) and falses(size(A)) respectively (#24595).
workspace is discontinued, check out Revise.jl for an alternative workflow (#25046).
cumsum, cumprod, accumulate, their mutating versions, and diff all now require a dim argument instead of defaulting to using the first dimension unless there is only one dimension (#24684, #25457).
The sum_kbn and cumsum_kbn functions have been moved to the KahanSummation package (#24869).
isnumber has been renamed to isnumeric (#25021).
isalpha has been renamed to isletter (#26932).
is_assigned_char and normalize_string have been renamed to isassigned and normalize, and moved to the new Unicode standard library module. graphemes has also been moved to that module (#25021).
Sparse array functionality has moved to the SparseArrays standard library module (#25249).
Linear algebra functionality, and specifically the LinAlg module has moved to the LinearAlgebra standard library module (#25571).
@printf and @sprintf have been moved to the Printf standard library (#23929,#25056).
The Libdl module has moved to the Libdl standard library module (#25459).
The aliases Complex32, Complex64 and Complex128 have been deprecated in favor of ComplexF16, ComplexF32 and ComplexF64 respectively (#24647).
Base.parentindexes and SharedArrays.localindexes have been renamed to parentindices and localindices, respectively. Similarly, the indexes field in the SubArray type has been renamed to indices without deprecation (#25088).
Associative has been deprecated in favor of AbstractDict (#25012).
Void has been renamed back to Nothing with an alias Cvoid for use when calling C with a return type of Cvoid or a return or argument type of Ptr{Cvoid} (#25162).
Nullable{T} has been deprecated and moved to the Nullables package (#23642). Use Union{T, Nothing} instead, or Union{Some{T}, Nothing} if nothing is a possible value (i.e. Nothing <: T). isnull(x) can be replaced with x === nothing and unsafe_get/get can be dropped or replaced with coalesce. NullException has been removed.
unshift! and shift! have been renamed to pushfirst! and popfirst! (#23902)
ipermute! has been deprecated in favor of invpermute! (#25168).
CartesianRange has been renamed CartesianIndices (#24715).
sub2ind and ind2sub are deprecated in favor of using CartesianIndices and LinearIndices (#24715).
getindex(F::Factorization, s::Symbol) (usually seen as e.g. F[:Q]) is deprecated in favor of dot overloading (getproperty) so factors should now be accessed as e.g. F.Q instead of F[:Q] (#25184).
search and rsearch have been deprecated in favor of findfirst/findnext and findlast/findprev respectively, in combination with curried isequal and in predicates for some methods (#24673).
search(buf::IOBuffer, delim::UInt8) has been deprecated in favor of either occursin(delim, buf) (to test containment) or readuntil(buf, delim) (to read data up to delim) (#26600).
ismatch(regex, str) has been deprecated in favor of occursin(regex, str) (#26283).
matchall has been deprecated in favor of collect(m.match for m in eachmatch(r, s)) (#26071).
similar(::Associative) has been deprecated in favor of empty(::Associative), and similar(::Associative, ::Pair{K, V}) has been deprecated in favour of empty(::Associative, K, V) (#24390).
findin(a, b) has been deprecated in favor of findall(in(b), a) (#24673).
module_name has been deprecated in favor of a new, general nameof function. Similarly, the unexported Base.function_name and Base.datatype_name have been deprecated in favor of nameof methods (#25622).
The module Random.dSFMT is renamed Random.DSFMT (#25567).
Random.RandomDevice(unlimited::Bool) (on non-Windows systems) is deprecated in favor of Random.RandomDevice(; unlimited=unlimited) (#25668).
The generic implementations of strides(::AbstractArray) and stride(::AbstractArray, ::Int) have been deprecated. Subtypes of AbstractArray that implement the newly introduced strided array interface should define their own strides method (#25321).
module_parent, Base.datatype_module, and Base.function_module have been deprecated in favor of parentmodule ([#TODO]).
rand(t::Tuple{Vararg{Int}}) is deprecated in favor of rand(Float64, t) or rand(t...); rand(::Tuple) will have another meaning in the future (#25429, #25278).
randjump, which produced an array, is deprecated in favor of the scalar version Future.randjump used with accumulate (#27746).
The assert function (and @assert macro) have been documented that they are not guaranteed to run under various optimization levels and should therefore not be used to e.g. verify passwords.
ObjectIdDict has been deprecated in favor of IdDict{Any,Any} (#25210).
gc and gc_enable have been deprecated in favor of GC.gc and GC.enable (#25616).
Base.@gc_preserve has been deprecated in favor of GC.@preserve (#25616).
print_shortest has been discontinued, but is still available in the Base.Grisu submodule (#25745).
scale! has been deprecated in favor of mul!, lmul!, and rmul! (#25701, #25812).
The remove_destination keyword argument to cp, mv, and the unexported cptree has been renamed to force (#25979).
contains has been deprecated in favor of a more general occursin function, which takes its arguments in reverse order from contains (#26283).
Regex objects are no longer callable. Use occursin instead (#26283).
The methods of range based on positional arguments have been deprecated in favor of keyword arguments (#25896).
linspace has been deprecated in favor of range with stop and length keyword arguments (#25896).
LinSpace has been renamed to LinRange (#25896).
logspace has been deprecated to its definition (#25896).
endof(a) has been renamed to lastindex(a), and the end keyword in indexing expressions now lowers to either lastindex(a) (in the case with only one index) or lastindex(a, d) (in cases where there is more than one index and end appears at dimension d) (#23554, #25763).
DateTime(), Date(), and Time() have been deprecated, instead use DateTime(1), Date(1) and Time(0) respectively (#23724).
The fallback method ^(x, p::Integer) is deprecated. If your type relied on this definition, add a method such as ^(x::MyType, p::Integer) = Base.power_by_squaring(x, p) (#23332).
DevNull, STDIN, STDOUT, and STDERR have been renamed to devnull, stdin, stdout, and stderr, respectively (#25786).
wait and fetch on Task now resemble the interface of Future.
showcompact(io, x...) has been deprecated in favor of show(IOContext(io, :compact => true), x...) (#26080). Use sprint(show, x..., context=:compact => true) instead of sprint(showcompact, x...).
isupper, islower, ucfirst and lcfirst have been deprecated in favor of isuppercase, islowercase, uppercasefirst and lowercasefirst, respectively (#26442).
signif has been deprecated in favor of the sigdigits keyword argument to round.
Base.IntSet has been deprecated in favor of Base.BitSet (#24282).
setrounding has been deprecated for Float32 and Float64, as the behaviour was too unreliable (#26935).
gamma, lgamma, beta, lbeta and lfact have been moved to SpecialFunctions.jl (#27459, #27473).
atan2 is now a 2-argument method of atan (#27248).
The functions eigs and svds have been moved to the Arpack.jl package (#27616).
vecdot and vecnorm are deprecated in favor of dot and norm, respectively (#27401).
clipboard has been moved to the InteractiveUtils standard library package (along with other utilities mostly used at the interactive prompt, such as edit and less) (#27635).
ndigits(n, b, [pad]) is deprecated in favor of ndigits(n, base=b, pad=pad) (#27908).
squeeze is deprecated in favor of dropdims.
srand is deprecated in favor of the unexported Random.seed! (#27726).
realmin/realmax are deprecated in favor of floatmin/floatmax (#28302).
sortrows/sortcols have been deprecated in favor of the more general sortslices.
nextpow2/prevpow2 have been deprecated in favor of the more general nextpow/prevpow functions.
Command-line option changes:
New option --warn-overwrite={yes|no} to control the warning for overwriting method definitions. The default is no (#23002).
New option --banner={yes,no} allows suppressing or forcing the printing of the startup banner, overriding the def

New in Julia Language 0.7.0 (Aug 8, 2018)

New language features:
Local variables can be tested for being defined using the new @isdefined variable macro (#22281).
Destructuring in function arguments: when an expression such as (x, y) is used as a function argument name, the argument is unpacked into local variables x and y as in the assignment (x, y) = arg (#6614).
Named tuples, with the syntax (a=1, b=2). These behave very similarly to tuples, except components can also be accessed by name using dot syntax t.a (#22194).
Keyword argument containers (kw in f(; kw...)) are now based on named tuples. Dictionary functions like haskey and indexing can be used on them, and name-value pairs can be iterated using pairs(kw). kw can no longer contain multiple entries for the same argument name (#4916).
Custom infix operators can now be defined by appending Unicode combining marks, primes, and sub/superscripts to other operators. For example, +̂ₐ″ is parsed as an infix operator with the same precedence as + (#22089).
The macro call syntax @macroname[args] is now available and is parsed as @macroname([args]) (#23519).
The construct if @generated ...; else ...; end can be used to provide both @generated and normal implementations of part of a function. Surrounding code will be common to both versions (#23168).
Added ⟂ (perp) operator with comparison precedence (#24404).
The missing singleton object (of type Missing) has been added to represent missing values (#24653). It propagates through standard operators and mathematical functions, and implements three-valued logic, similar to SQLs NULL and R's NA.
Field access via dot-syntax can now be overloaded by adding methods to Base.getproperty and Base.setproperty! (#1974), optionally along with a corresponding Base.propertynames method for reflection (#25311).
Values for Enums can now be specified inside of a begin block when using the @enum macro (#25424).
Keyword arguments can be required: if a default value is omitted, then an exception is thrown if the caller does not assign the keyword a value (#25830).
The pair operator => is now broadcastable as .=> which was previously a parsing error (#27447)
Language changes:
The syntax for parametric methods, function f{T}(x::T), has been changed to function f(x::T) where {T} (#11310).
The fallback constructor that calls convert is deprecated. Instead, new types should prefer to define constructors, and add convert methods that call those constructors only as necessary (#15120).
The syntax 1.+2 is deprecated, since it is ambiguous: it could mean either 1 .+ 2 (the current meaning) or 1. + 2 (#19089).
Mutable structs with no fields are no longer singletons; it is now possible to make multiple instances of them that can be distinguished by === (#25854). Zero-size immutable structs are still singletons.
In string and character literals, backslash may no longer precede unrecognized escape characters (#22800).
Juxtaposing binary, octal, and hexadecimal literals is deprecated, since it can lead to confusing code such as 0xapi == 0xa * pi (#16356).
Numeric literal juxtaposition now has slighty lower precedence than unary operators, so for example √2x parses as (√2) * x (#27641).
Declaring arguments as x::ANY to avoid specialization has been replaced by @nospecialize x. (#22666).
This can also be used in global scope, to apply to all subsequent method definitions in the module (until @specialize). (#28065)
Keyword argument default values are now evaluated in successive scopes --- the scope for each expression includes only previous keyword arguments, in left-to-right order (#17240).
The parsing of 1<<2*3 as 1<<(2*3) is deprecated, and will change to (1<<2)*3 in a future version (#13079).
The parsing of <| is now right associative. |> remains left associative (#24153).
: now parses like other operators, as a call to a function named :, instead of calling colon (#25947).
{ } expressions now use braces and bracescat as expression heads instead of cell1d and cell2d, and parse similarly to vect and vcat (#8470).
Nested if expressions that arise from the keyword elseif now use elseif as their expression head instead of if (#21774).
let blocks now parse the same as for loops; the first argument is either an assignment or block of assignments, and the second argument is a block of statements (#21774).
do syntax now parses to an expression with head :do, instead of as a function call (#21774).
Parsed and lowered forms of type definitions have been synchronized with their new keywords (#23157)
The global keyword now only introduces a new binding if one doesn't already exist in the module. This means that assignment to a global (global sin = 3) may now throw the error: "cannot assign variable Base.sin from module Main", rather than emitting a warning. Additionally, the new bindings are now created before the statement is executed. For example, f() = (global sin = "gluttony"; nothing) will now resolve which module contains sin eagerly, rather than delaying that decision until f is run. (#22984).
global const declarations may no longer appear inside functions (#12010).
Uninitialized BitArray constructors of the form BitArray[{N}](shape...) have been deprecated in favor of equivalents accepting undef (an alias for UndefInitializer()) as their first argument, as in BitArray[{N}](undef, shape...). For example, BitVector(3) is now BitVector(undef, 3), BitMatrix((2, 4)) is now BitMatrix(undef, (2, 4)), and BitArray{3}(11, 13, 17) is now BitArray{3}(undef, 11, 14, 17) (#24785).
Dispatch rules have been simplified: method matching is now determined exclusively by subtyping; the rule that method type parameters must also be captured has been removed. Instead, attempting to access the unconstrained parameters will throw an UndefVarError. Linting in package tests is recommended to confirm that the set of methods which might throw UndefVarError when accessing the static parameters (need_to_handle_undef_sparam = Set{Any}(m.sig for m in Test.detect_unbound_args(Base, recursive=true))) is equal (==) to some known set (expected = Set()). (#23117)
const declarations on local variables were previously ignored. They now give a warning, so that this syntax can be disallowed or given a new meaning in a future version (#5148).
Placing an expression after catch, as in catch f(x), is deprecated. Use catch; f(x) instead (#19987).
In for i = ..., if a local variable i already existed it would be overwritten during the loop. This behavior is deprecated, and in the future for loop variables will always be new variables local to the loop (#22314). The old behavior of overwriting an existing variable is available via for outer i = ....
In for i in x, x used to be evaluated in a new scope enclosing the for loop. Now it is evaluated in the scope outside the for loop.
In for i in x, j in y, all variables now have fresh bindings on each iteration of the innermost loop. For example, an assignment to i will not be visible on the next j loop iteration (#330).
Variable bindings local to while loop bodies are now freshly allocated on each loop iteration, matching the behavior of for loops.
Prefix & for by-reference arguments to ccall has been deprecated in favor of Ref argument types (#6080).
The constructor Ref(x::T) now always returns a Ref{T} (#21527).
All line numbers in ASTs are represented by LineNumberNodes; the :line expression head is no longer used. QuoteNodes are also consistently used for quoted symbols instead of the :quote expression head (though :quote Exprs are still used for quoted expressions) (#23885).
The + and - methods for Number and UniformScaling are not ambiguous anymore since + and - no longer do automatic broadcasting. Hence, the methods for UniformScaling and Number are no longer deprecated (#23923).
The keyword importall is deprecated. Use using and/or individual import statements instead (#22789).
reduce(+, [...]) and reduce(*, [...]) no longer widen the iterated over arguments to system word size. sum and prod still preserve this behavior. (#22825)
Like _, variable names consisting only of underscores can be assigned, but accessing their values is deprecated (#24221).
Raw string literal escaping rules have been changed to make it possible to write all strings. The rule is that backslashes escape both quotes and other backslashes, but only when a sequence of backslashes precedes a quote character. Thus, 2n backslashes followed by a quote encodes n backslashes and the end of the literal while 2n+1 backslashes followed by a quote encodes n backslashes followed by a quote character (#22926).
reprmime(mime, x) has been renamed to repr(mime, x), and along with repr(x) and sprint it now accepts an optional context keyword for IOContext attributes. stringmime has been moved to the Base64 stdlib package (#25990).
The syntax (x...) for constructing a tuple is deprecated; use (x...,) instead (#24452).
Non-parenthesized interpolated variables in strings, e.g. "$x", must be followed by a character that will never be an allowed identifier character (currently operators, space/control characters, or common punctuation characters) (#25231).
The syntax using A.B can now only be used when A.B is a module, and the syntax using A: B can only be used for adding single bindings (#8000).
=> now has its own precedence level, giving it strictly higher precedence than = and , (#25391).
The conditions under which unary operators followed by ( are parsed as prefix function calls have changed (#26154).
begin is disallowed inside indexing expressions, in order to enable the syntax a[begin] (for selecting the first element) in the future (#23354).
Underscores for _italics_ and __bold__ are now supported by the Base Markdown parser. (#25564)
… (dots) and ⁝ (tricolon) are now parsed as binary operators (#26262).
Assignment syntax (a=b) inside square bracket expressions (e.g. A[...], [x, y]) is deprecated. It will likely be reclaimed in a later version for passing keyword arguments. Note this does not affect updating operators like += (#25631).
try blocks without catch or finally are no longer allowed. An explicit empty catch block should be written instead (#27554).
AbstractArray types that use unconventional (not 1-based) indexing can now support size, length, and @inbounds. To optionally enforce conventional indices, you can @assert !has_offset_axes(A).
Module pre-compilation is now the default for code loading. Adding a __precompile__() declaration is no longer necessary, although __precompile__(false) can still be used to opt-out ([#26991]).
Breaking changes:
This section lists changes that do not have deprecation warnings.
The package manager Pkg has been replaced with a new one. See the manual entries on "Code Loading" and "Pkg" for documentation.
replace(s::AbstractString, pat=>repl) for function repl arguments formerly passed a substring to repl in all cases. It now passes substrings for string patterns pat, but a Char for character patterns (when pat is a Char, collection of Char, or a character predicate) (#25815).
readuntil now does not include the delimiter in its result, matching the behavior of readline. Pass keep=true to get the old behavior (#25633).
lu methods now return decomposition objects such as LU rather than tuples of arrays or tuples of numbers (#26997, #27159, #27212).
schur methods now return decomposition objects such as Schur and GeneralizedSchur rather than tuples of arrays (#26997, #27159, #27212).
lq methods now return decomposition objects such as LQ rather than tuples of arrays (#26997, #27159, #27212).
qr methods now return decomposition objects such as QR, QRPivoted, and QRCompactWY rather than tuples of arrays (#26997, #27159, #27212).
svd methods now return decomposition objects such as SVD and GeneralizedSVD rather than tuples of arrays or tuples of numbers (#26997, #27159, #27212).
countlines now always counts the last non-empty line even if it does not end with EOL, matching the behavior of eachline and readlines (#25845).
getindex(s::String, r::UnitRange{Int}) now throws StringIndexError if last(r) is not a valid index into s (#22572).
ntuple(f, n::Integer) throws ArgumentError if n is negative. Previously an empty tuple was returned (#21697).
⋮, ⋱, ⋰, and ⋯ are now parsed as binary operators, not ordinary identifiers. ≔, ≕, and ⩴ now parse with assignment rather than comparison precedence (#26262).
Juxtaposing string literals (e.g. "x"y) is now a syntax error (#20575).
finalizer(function, object) now returns object rather than nothing (#24679).
The constructor of SubString now checks if the requested view range is defined by valid indices in the parent AbstractString (#22511).
Macro calls with for expressions are now parsed as generators inside function argument lists (#18650).
@__DIR__ returns the current working directory rather than nothing when not run from a file (#21759).
@__FILE__ and @__DIR__ return information relative to the file that it was parsed from, rather than from the task-local SOURCE_PATH global when it was expanded.
All macros receive an extra argument __source__::LineNumberNode which describes the parser location in the source file for the @ of the macro call. It can be accessed as a normal argument variable in the body of the macro. This is implemented by inserting an extra leading argument into the Expr(:macrocall, :@name, LineNumberNode(...), args...) surface syntax. (#21746)
Passing the same keyword argument multiple times is now a syntax error (#16937).
getsockname on a TCPSocket now returns the locally bound address and port of the socket. Previously the address of the remote endpoint was being returned (#21825).
The ~/.juliarc.jl file has been moved to ~/.julia/config/startup.jl and /etc/julia/juliarc.jl file has been renamed to /etc/julia/startup.jl (#26161).
Using ARGS within startup.jl files or within a .jl file loaded with --load will no longer contain the script name as the first argument. Instead, the script name will be assigned to PROGRAM_FILE. (#22092)
The format for a ClusterManager specifying the cookie on the command line is now --worker=<cookie>. --worker <cookie> will not work as it is now an optional argument.
The representation of CartesianRange has changed to a tuple-of-AbstractUnitRanges; the start and stop fields are no longer present. Use first(R) and last(R) to obtain start/stop. (#20974)
The Diagonal, Bidiagonal, Tridiagonal and SymTridiagonal type definitions have changed from Diagonal{T}, Bidiagonal{T}, Tridiagonal{T} and SymTridiagonal{T} to Diagonal{T,V<:AbstractVector{T}}, Bidiagonal{T,V<:AbstractVector{T}}, Tridiagonal{T,V<:AbstractVector{T}} and SymTridiagonal{T,V<:AbstractVector{T}} respectively (#22718, #22925, #23035, #23154).
The immediate supertype of BitArray is now simply AbstractArray. BitArray is no longer considered a subtype of DenseArray and StridedArray (#25858).
When called with an argument that contains NaN elements, findmin and findmax now return the first NaN found and its corresponding index. Previously, NaN elements were ignored. The new behavior matches that of min, max, minimum, and maximum.
isapprox(x,y) now tests norm(x-y) <= max(atol, rtol*max(norm(x), norm(y))) rather than norm(x-y) <= atol + ..., and rtol defaults to zero if an atol > 0 is specified (#22742).
Spaces are no longer allowed between @ and the name of a macro in a macro call (#22868).
Juxtaposition of a non-literal with a macro call (x@macro) is no longer valid syntax (#22868).
On a cluster, all files are now loaded from the local file system rather than node 1 (#22588). To load the same file everywhere from node 1, one possible alternative is to broadcast a call to include_string: @everywhere include_string(Main, $(read("filename", String)), "filename"). Improving upon this API is left as an opportunity for packages.
randperm(n) and randcycle(n) now always return a Vector{Int} (independent of the type of n). Use the corresponding mutating functions randperm! and randcycle! to control the array type (#22723).
Hermitian now ignores any imaginary components in the diagonal instead of checking the diagonal. (#17367)
Worker-worker connections are setup lazily for an :all_to_all topology. Use keyword arg lazy=false to force all connections to be setup during a addprocs call. (#22814)
In joinpath(a, b) on Windows, if the drive specifications of a and b do not match, joinpath now returns b instead of throwing an ArgumentError. joinpath(path...) is defined to be left associative, so if any argument has a drive path which does not match the drive of the join of the preceding paths, the prior ones are dropped. (#20912)
^(A::AbstractMatrix{<:Integer}, p::Integer) now throws a DomainError if p < 0, unless A == one(A) or A == -one(A) (same as for ^(A::Integer, p::Integer)) (#23366).
^(A::AbstractMatrix{<:Integer}, p::Integer) now promotes the element type in the same way as ^(A::Integer, p::Integer). This means, for instance, that [1 1; 0 1]^big(1) will return a Matrix{BigInt} instead of a Matrix{Int} (#23366).
The element type of the input is now preserved in unique. Previously the element type of the output was shrunk to fit the union of the type of each element in the input. (#22696)
The promote function now raises an error if its arguments are of different types and if attempting to convert them to a common type fails to change any of their types. This avoids stack overflows in the common case of definitions like f(x, y) = f(promote(x, y)...) (#22801).
indmin and indmax have been renamed to argmin and argmax, respectively (#25654).
findmin, findmax, argmin, and argmax used to always return linear indices. They now return CartesianIndexes for all but 1-d arrays, and in general return the keys of indexed collections (e.g. dictionaries) (#22907).
The openspecfun library is no longer built and shipped with Julia, as it is no longer used internally (#22390).
All loaded packages used to have bindings in Main (e.g. Main.Package). This is no longer the case; now bindings will only exist for packages brought into scope by typing using Package or import Package (#17997).
The rules for mixed-signedness integer arithmetic (e.g. Int32(1) + UInt64(1)) have been simplified: if the arguments have different sizes (in bits), then the type of the larger argument is used. If the arguments have the same size, the unsigned type is used (#9292).
All command line arguments passed via -e, -E, and -L will be executed in the order given on the command line (#23665).
I now yields UniformScaling{Bool}(true) rather than UniformScaling{Int64}(1) to better preserve types in operations involving I (#24396).
The return type of reinterpret has changed to ReinterpretArray. reinterpret on sparse arrays has been discontinued.
Base.find_in_path is now Base.find_package or Base.find_source_file (#24320).
finalizer now takes functions or pointers as its first argument, and the object being finalized as its second (rather than the reverse). For the majority of use cases deprecation warnings will be triggered. However, deprecation warnings will not trigger where (1) the callable argument is not a subtype of Function; or (2) both arguments are Functions or Ptr{Cvoid}s (#24605).
The kill function now throws errors on user error (e.g. on permission errors), but returns successfully if the process had previously exited. Its return value has been removed. Use the process_running function to determine if a process has already exited.
The logging system has been redesigned - info and warn are deprecated and replaced with the logging macros @info, @warn, @debug and @error. The logging function is also deprecated and replaced with AbstractLogger and the functions from the new standard Logging library. (#24490)
The RevString type has been removed from the language; reverse(::String) returns a String with code points (or fragments thereof) in reverse order. In general, reverse(s) should return a string of the same type and encoding as s with code points in reverse order; any string type overrides reverse to return a different type of string must also override reverseind to compute reversed indices correctly.
eachindex(A, B...) now requires that all inputs have the same number of elements. When the chosen indexing is Cartesian, they must have the same axes.
AbstractRange objects are now considered as equal to other AbstractArray objects by == and isequal if all of their elements are equal (#16401). This has required changing the hashing algorithm: ranges now use an O(N) fallback instead of a O(1) specialized method unless they define the Base.RangeStepStyle trait; see its documentation for details. Types which support subtraction (operator -) must now implement widen for hashing to work inside heterogeneous arrays.
findn(x::AbstractArray) has been deprecated in favor of findall(!iszero, x), which now returns cartesian indices for multidimensional arrays (see below, #25532).
Broadcasting operations are no longer fused into a single operation by Julia's parser. Instead, a lazy Broadcasted object is created to represent the fused expression and then realized with copy(bc::Broadcasted) or copyto!(dest, bc::Broadcasted) to evaluate the wrapper. Consequently, package authors generally need to specialize copy and copyto! methods rather than broadcast and broadcast!. This also allows for more customization and control of fused broadcasts. See the Interfaces chapter for more information.
find has been renamed to findall. findall, findfirst, findlast, findnext now take and/or return the same type of indices as keys/pairs for AbstractArray, AbstractDict, AbstractString, Tuple and NamedTuple objects (#24774, #25545). In particular, this means that they use CartesianIndex objects for matrices and higher-dimensional arrays instead of linear indices as was previously the case. Use LinearIndices(a)[findall(f, a)] and similar constructs to compute linear indices.
The find* functions, i.e. findnext, findprev, findfirst, and findlast, as well as indexin, now return nothing when no match is found rather than 0 or 0:-1 (#25472, #25662, #26149)
The Base.HasShape iterator trait has gained a type parameter N indicating the number of dimensions, which must correspond to the length of the tuple returned by size (#25655).
AbstractSet objects are now considered equal by == and isequal if all of their elements are equal (#25368). This has required changing the hashing algorithm for BitSet.
the default behavior of titlecase is changed in two ways (#23393)
characters not starting a word are converted to lowercase; a new keyword argument strict is added which allows to get the old behavior when it's false.
any non-letter character is considered as a word separator; to get the old behavior (only "space" characters are considered as word separators), use the keyword wordsep=isspace.
writedlm in the standard library module DelimitedFiles now writes numeric values using print rather than print_shortest (#25745).
The tempname function used to create a file on Windows but not on other platforms. It now never creates a file (#9053).
The fieldnames and propertynames functions now return a tuple rather than an array (#25725).
indexin now returns the first rather than the last matching index (#25998).
parse(::Type, ::Char) now uses a default base of 10, like other number parsing methods, instead of 36 (#26576).
isequal for Ptrs now compares element types; == still compares only addresses (#26858).
widen on 8- and 16-bit integer types now widens to 16- and 32-bit types, respectively. (#28045).
mv,cp, touch, mkdir, mkpath, chmod and chown now return the path that was created/modified rather than nothing (#27071).
Regular expressions now default to UCP mode. Escape sequences such as w will now match based on unicode character properties, e.g. r"w+" will match café (not just caf). Add the a modifier (e.g. r"w+"a) to restore the previous behavior (#27189).
@sync now waits only for lexically enclosed (i.e. visible directly in the source text of its argument) @async expressions. If you need to wait for a task created by a called function f, have f return the task and put @async wait(f(...)) within the @sync block. This change makes @schedule redundant with @async, so @schedule has been deprecated (#27164).
norm(A::AbstractMatrix, p=2) computes no longer the operator/matrix norm but the norm of A as for other iterables, i.e. as if it were a vector. Especially, norm(A::AbstractMatrix) is the Frobenius norm. To compute the operator/matrix norm, use the new function opnorm (#27401).
dot(u, v) now acts recursively. Instead of sum(u[i]' * v[i] for i in ...), it computes sum(dot(u[i], v[i]) for i in ...), similarly to vecdot before (#27401).
Sys.CPU_CORES has been renamed to Sys.CPU_THREADS; it still gives the number of "logical cores" (including hyperthreading) rather than the number of physical cores present on the CPU. Similarly, the environment variable JULIA_CPU_CORES is deprecated in favor of JULIA_CPU_THREADS (#27856).
WeakKeyDict does not convert keys on insertion anymore (#24941).
Library improvements:
The function thisind(s::AbstractString, i::Integer) returns the largest valid index less or equal than i in the string s or 0 if no such index exists (#24414).
Support for Unicode 11 (#28266).
Char is now a subtype of AbstractChar, and most of the functions that take character arguments now accept any AbstractChar (#26286).
pathof(module) returns the path a module was imported from ([#28310]).
bytes2hex now accepts an optional io argument to output to a hexadecimal stream without allocating a String first (#27121).
String(array) now accepts an arbitrary AbstractVector{UInt8}. For Vector inputs, it "steals" the memory buffer, leaving them with an empty buffer which is guaranteed not to be shared with the String object. For other types of vectors (in particular immutable vectors), a copy is made and the input is not truncated (#26093).
Irrational is now a subtype of AbstractIrrational (#24245).
Introduced the empty function, the functional pair to empty! which returns a new, empty container (#24390).
Jump to first/last history entries in the REPL via "Alt-<" and "Alt->" (#22829).
REPL LaTeX-like tab completions have been simplified for several Unicode characters, e.g. 𝔸 is now bbA rather than BbbA (#25980).
The function chop now accepts two arguments head and tail allowing to specify number of characters to remove from the head and tail of the string (#24126).
get(io, :color, false) can now be used to query whether a stream io supports ANSI color codes (#25067), rather than using the undocumented Base.have_color global flag.
print_with_color has been deprecated in favor of printstyled([io], xs...; bold=false, color=:normal) for printing styled text (#25522).
Functions first and last now accept nchar argument for AbstractString. If this argument is used they return a string consisting of first/last nchar characters from the original string (#23960).
Expressions x^-n where n is an integer literal now correspond to inv(x)^n. For example, x^-1 is now essentially a synonym for inv(x), and works in a type-stable way even if typeof(x) != typeof(inv(x)) (#24240).
New Iterators.reverse(itr) for reverse-order iteration (#24187). Iterator types T can implement start etc. for Iterators.Reverse{T} to support this.
The functions nextind and prevind now accept nchar argument that indicates the number of characters to move (#23805).
The functions strip, lstrip and rstrip now return SubString (#22496).
The functions strwidth and charwidth have been merged into textwidth(#20816).
The functions base and digits digits now accept a negative base (like ndigits did) (#21692).
The function randn now accepts complex arguments (Complex{T <: AbstractFloat}) (#21973).
parse(Complex{T}, string) can parse complex numbers in some common formats (#24713).
The function rand can now pick up random elements from strings, associatives and sets (#22228, #21960, #18155, #22224).
It's now possible to specify the characters to pick from in the randstring function (#22222).
Allow multidimensional arrays in shuffle and shuffle! functions (#22226).
Method lists are now printed as a numbered list. In addition, the source code of a method can be opened in an editor by entering the corresponding number in the REPL and pressing ^Q (#22007).
getpeername on a TCPSocket returns the address and port of the remote endpoint of the TCP connection (#21825).
resize! and sizehint! methods no longer over-reserve memory when the requested array size is more than double of its current size (#22038).
The crc32c function for CRC-32c checksums is now exported (#22274).
eye(::Type{Diagonal{T}}, m::Integer) has been deprecated in favor of Diagonal{T}(I, m) (#24413).
The output of versioninfo is now controlled with keyword arguments (#21974).
The function LibGit2.set_remote_url now always sets both the fetch and push URLs for a git repo. Additionally, the argument order was changed to be consistent with the git command line tool (#22062).
Added unique! which is an inplace version of unique (#20549).
@test isequal(x, y) and @test isapprox(x, y) now prints an evaluated expression when the test fails (#22296).
Uses of Val{c} in Base has been replaced with Val{c}(), which is now easily accessible via the efficient constructor Val(c). Functions are defined as f(::Val{c}) = ... and called by f(Val(c)). Notable affected functions include: ntuple, Base.literal_pow, sqrtm, lufact, lufact!, qrfact, qrfact!, cholfact, cholfact!, _broadcast!, reshape, cat and cat_t.
A new @macroexpand1 macro for non recursive macro expansion (#21662).
Chars can now be concatenated with Strings and/or other Chars using * (#22532).
Diagonal, Bidiagonal, Tridiagonal and SymTridiagonal are now parameterized on the type of the wrapped vectors, allowing Diagonal, Bidiagonal, Tridiagonal and SymTridiagonal matrices with arbitrary AbstractVectors (#22718, #22925, #23035, #23154).
Mutating versions of randperm and randcycle have been added: randperm! and randcycle! (#22723).
BigFloat random numbers can now be generated (#22720).
The efficiency of random generation for MersenneTwister RNGs has been improved for integers, Float64 and ranges; as a result, given a seed, the produced stream of numbers has changed (#27560, #25277, #25197, #25058, #25047).
REPL Undo via Ctrl-/ and Ctrl-_
diagm now accepts several diagonal index/vector Pairs (#24047).
isequal, ==, and in have one argument "curried" forms. For example isequal(x) returns a function that compares its argument to x using isequal (#26436).
reinterpret now works on any AbstractArray using the new ReinterpretArray type. This supersedes the old behavior of reinterpret on Arrays. As a result, reinterpreting arrays with different alignment requirements (removed in 0.6) is once again allowed (#23750).
The keys of an Associative are now an AbstractSet. Base.KeyIterator{<:Associative} has been changed to KeySet{K, <:Associative{K}} <: AbstractSet{K} (#24580).
New function ncodeunits(s::AbstractString) gives the number of code units in a string. The generic definition is constant time but calls lastindex(s) which may be inefficient. Therefore custom string types may want to define direct ncodeunits methods.
reverseind(s::AbstractString, i::Integer) now has an efficient generic fallback, so custom string types do not need to provide their own efficient definitions. The generic definition relies on ncodeunits however, so for optimal performance you may need to define a custom method for that function.
The global RNG is being re-seeded with its own seed at the beginning of each @testset, and have its original state restored at the end (#24445). This is breaking for testsets relying implicitly on the global RNG being in a specific state.
permutedims(m::AbstractMatrix) is now short for permutedims(m, (2,1)), and is now a more convenient way of making a "shallow transpose" of a 2D array. This is the recommended approach for manipulating arrays of data, rather than the recursively defined, linear-algebra function transpose. Similarly, permutedims(v::AbstractVector) will create a row matrix (#24839).
A new replace(A, old=>new) function is introduced to replace old by new in collection A. There is also another method with a different API, and a mutating variant, replace! (#22324, #25697, #26206, #27944).
Adding integers to CartesianIndex objects is now deprecated. Instead of i::Int + x::CartesianIndex, use i*one(x) + x (#26284).
The linearindices function has been deprecated in favor of the new LinearIndices type, which additionally provides conversion from cartesian indices to linear indices using the normal indexing operation. (#24715, #26775).
IdDict{K,V} replaces ObjectIdDict. It has type parameters like other AbstractDict subtypes and its constructors mirror the ones of Dict. (#25210)
IOBuffer can take the sizehint keyword argument to suggest a capacity of the buffer (#25944).
lstrip and rstrip now accept a predicate function that defaults to isspace (#27309).
trunc, floor, ceil, and round specify digits, sigdigits and base using keyword arguments. (#26156, #26670)
Sys.which() provides a cross-platform method to find executable files, similar to the Unix which command. (#26559)
Added an optimized method of vecdot for taking the Frobenius inner product of sparse matrices. (#27470)
Added an optimized method of kron for taking the tensor product of two Diagonal matrices. ([27581])
An official API for extending rand is now defined (#23964, #25002).
The constructor MersenneTwister() is re-enabled, producing a randomly initialized RNG (similar to Random.seed!(MersenneTwister(0))) (#21909).
BitSet can now store any Int (instead of only positive ones) (#25029).
The initial element v0 in reduce(op, v0, itr) has been replaced with an init optional keyword argument, as in reduce(op, itr; init=v0). Similarly for foldl, foldr, mapreduce, mapfoldl, mapfoldr, accumulate and accumulate!. (#27711, #27859)
CartesianRange changes (#24715):
Inherits from AbstractArray, and linear indexing can be used to provide linear-to-cartesian conversion (#24715)
It has a new constructor taking an array
several missing set-like operations have been added (#23528): union, intersect, symdiff, setdiff are now implemented for all collections with arbitrary many arguments, as well as the mutating counterparts (union! etc.). The performance is also much better in many cases. Note that this change is slightly breaking: all the non-mutating functions always return a new object even if only one argument is passed. Moreover the semantics of intersect and symdiff is changed for vectors:
intersect doesn't preserve the multiplicity anymore (use filter for the old behavior)
symdiff has been made consistent with the corresponding methods for other containers, by taking the multiplicity of the arguments into account. Use unique to get the old behavior.
Compiler/Runtime improvements:
The inlining heuristic now models the approximate runtime cost of a method (using some strongly-simplifying assumptions). Functions are inlined unless their estimated runtime cost substantially exceeds the cost of setting up and issuing a subroutine call. (#22210, #22732)
Inference recursion-detection heuristics are now more precise, allowing them to be triggered less often, but being more aggressive when they are triggered to drive the inference computation to a solution (#23912).
Inference now propagates constants inter-procedurally, and can compute various constants expressions at compile-time (#24362).
The LLVM SLP Vectorizer optimization pass is now enabled at the default optimization level.
Deprecated or removed:
The JULIA_HOME environment variable has been renamed to JULIA_BINDIR and Base.JULIA_HOME has been moved to Sys.BINDIR (#20899).
The keyword immutable is fully deprecated to struct, and type is fully deprecated to mutable struct (#19157, #20418).
lufact, schurfact, lqfact, qrfact, ldltfact, svdfact, bkfact, hessfact, eigfact, and cholfact have respectively been deprecated to lu, schur, lq, qr, ldlt, svd, bunchkaufman, hessenberg, eigen, and cholesky (#26997, #27159, #27212).
lufact!, schurfact!, lqfact!, qrfact!, ldltfact!, svdfact!, bkfact!, hessfact!, and eigfact! have respectively been deprecated to lu!, schur!, lq!, qr!, ldlt!, svd!, bunchkaufman!, hessenberg!, and eigen! (#26997, #27159, #27212).
eig(A[, args...]) has been deprecated in favor of eigen(A[, args...]). Whereas the former returns a tuple of arrays, the latter returns an Eigen object. So for a direct replacement, use (eigen(A[, args...])...,). But going forward, consider using the direct result of eigen(A[, args...]) instead, either destructured into its components (vals, vecs = eigen(A[, args...])) or as an Eigen object (X = eigen(A[, args...])) (#26997, #27159, #27212).
eig(A::AbstractMatrix, B::AbstractMatrix) and eig(A::Number, B::Number) have been deprecated in favor of eigen(A, B). Whereas the former each return a tuple of arrays, the latter returns a GeneralizedEigen object. So for a direct replacement, use (eigen(A, B)...,). But going forward, consider using the direct result of eigen(A, B) instead, either destructured into its components (vals, vecs = eigen(A, B)), or as a GeneralizedEigen object (X = eigen(A, B)) (#26997, #27159, #27212).
ordschur(T::StridedMatrix{Ty}, Z::StridedMatrix{Ty}, select::Union{Vector{Bool},BitVector}) and ordschur(S::StridedMatrix{Ty}, T::StridedMatrix{Ty}, Q::StridedMatrix{Ty}, Z::StridedMatrix{Ty}, select::Union{Vector{Bool},BitVector}) and their respective inplace versions have been deprecated. Use ordschur(schur::Schur, select::Union{Vector{Bool},BitVector}) and ordschur(gschur::GeneralizedSchur, select::Union{Vector{Bool},BitVector}) instead (#28155).
Indexing into multidimensional arrays with more than one index but fewer indices than there are dimensions is no longer permitted when those trailing dimensions have lengths greater than 1. Instead, reshape the array or add trailing indices so the dimensionality and number of indices match (#14770, #23628).
The use of a positional dimension argument has largely been deprecated in favor of a dims keyword argument. This includes the functions sum, prod, maximum, minimum, all, any, findmax, findmin, mean, varm, std, var, cov, cor, median, mapreducedim, reducedim, sort, accumulate, accumulate!, cumsum, cumsum!, cumprod, cumprod!, flipdim, dropdims, and cat (#25501, #26660, #27100).
indices(a) and indices(a,d) have been deprecated in favor of axes(a) and axes(a, d) (#25057).
EnvHash has been renamed to EnvDict (#24167).
Uninitialized Array constructors of the form Array[{T,N}](shape...) have been deprecated in favor of equivalents accepting undef (an alias for UndefInitializer()) as their first argument, as in Array[{T,N}](undef, shape...). For example, Vector(3) is now Vector(undef, 3), Matrix{Int}((2, 4)) is now, Matrix{Int}(undef, (2, 4)), and Array{Float32,3}(11, 13, 17) is now Array{Float32,3}(undef, 11, 13, 17) (#24781).
Previously setindex!(A, x, I...) (and the syntax A[I...] = x) supported two different modes of operation when supplied with a set of non-scalar indices I (e.g., at least one index is an AbstractArray) depending upon the value of x on the right hand side. If x is an AbstractArray, its contents are copied elementwise into the locations in A selected by I and it must have the same number of elements as I selects locations. Otherwise, if x is not an AbstractArray, then its value is implicitly broadcast to all locations to all locations in A selected by I. This latter behavior—implicitly broadcasting "scalar"-like values across many locations—is now deprecated in favor of explicitly using the broadcasted assignment syntax A[I...] .= x or fill!(view(A, I...), x) (#26347).
broadcast_getindex(A, I...) and broadcast_setindex!(A, v, I...) are deprecated in favor of getindex.((A,), I...) and setindex!.((A,), v, I...), respectively (#27075).
LinAlg.fillslots! has been renamed LinAlg.fillstored! (#25030).
fill!(A::Diagonal, x) and fill!(A::AbstractTriangular, x) have been deprecated in favor of Base.LinAlg.fillstored!(A, x) (#24413).
eye has been deprecated in favor of I and Matrix constructors. Please see the deprecation warnings for replacement details (#24438).
zeros(D::Diagonal[, opts...]) has been deprecated (#24654).
Using Bool values directly as indices is now deprecated and will be an error in the future. Convert them to Int before indexing if you intend to access index 1 for true and 0 for false.
slicedim(A, d, i) has been deprecated in favor of copy(selectdim(A, d, i)). The new selectdim function now always returns a view into A; in many cases the copy is not necessary. Previously, slicedim on a vector V over dimension d=1 and scalar index i would return the just selected element (unless V was a BitVector). This has now been made consistent: selectdim now always returns a view into the original array, with a zero-dimensional view in this specific case (#26009).
whos has been renamed varinfo, and now returns a markdown table instead of printing output (#12131).
Uninitialized RowVector constructors of the form RowVector{T}(shape...) have been deprecated in favor of equivalents accepting undef (an alias for UndefInitializer()) as their first argument, as in RowVector{T}(undef, shape...). For example, RowVector{Int}(3) is now RowVector{Int}(undef, 3), and RowVector{Float32}((1, 4)) is now RowVector{Float32}(undef, (1, 4)) (#24786).
writecsv(io, a; opts...) has been deprecated in favor of writedlm(io, a, ','; opts...) (#23529).
The method srand(rng, filename, n=4) has been deprecated (#21359).
readcsv(io[, T::Type]; opts...) has been deprecated in favor of readdlm(io, ','[, T]; opts...) (#23530).
sparse(s::UniformScaling, m::Integer) has been deprecated in favor of the three-argument equivalent sparse(s::UniformScaling, m, n) (#24472).
The cholfact/cholfact! methods that accepted an uplo symbol have been deprecated in favor of using Hermitian (or Symmetric) views (#22187, #22188).
The thin keyword argument for orthogonal decomposition methods has been deprecated in favor of full, which has the opposite meaning: thin == true if and only if full == false (#24279).
isposdef(A::AbstractMatrix, UL::Symbol) and isposdef!(A::AbstractMatrix, UL::Symbol) have been deprecated in favor of isposdef(Hermitian(A, UL)) and isposdef!(Hermitian(A, UL)) respectively (#22245).
The bkfact/bkfact! methods that accepted uplo and issymmetric symbols have been deprecated in favor of using Hermitian (or Symmetric) views (#22605).
The function current_module is deprecated and replaced with @__MODULE__. This caused the deprecation of some reflection methods (such as macroexpand and isconst), which now require a module argument. And it caused the bugfix of other default arguments to use the Main module (including whos, which) (#22064).
expand(ex) and expand(module, ex) have been deprecated in favor of Meta.lower(module, ex) (#22064, #24278).
ones(A::AbstractArray[, opts...]) and zeros(A::AbstractArray[, opts...]) methods have been deprecated. For zeros(A), consider zero(A). For ones(A) or zeros(A), consider ones(size(A)), zeros(size(A)), fill(v, size(A)) for v an appropriate one or zero, fill!(copy(A), {1|0}), fill!(similar(A), {1|0}), or any of the preceding with different element type and/or shape depending on opts.... Where strictly necessary, consider fill!(similar(A[, opts...]), {one(eltype(A)) | zero(eltype(A))}). For an algebraic multiplicative identity, consider one(A) (#24656).
The similar(dims->f(..., dims...), [T], axes...) method to add offset array support to a function f that would otherwise create a non-offset array has been deprecated. Instead, call f(..., axes...) directly and, if needed, the offset array implementation should add offset axis support to the function f directly (#26733).
The functions ones and zeros used to accept any objects as dimensional arguments, implicitly converting them to Ints. This is now deprecated; only Integers or AbstractUnitRanges are accepted as arguments. Instead, convert the arguments before calling ones or zeros (#26733).
The variadic size(A, dim1, dim2, dims...) method to return a tuple of multiple dimension lengths of A has been deprecated (#26862).
The Operators module is deprecated. Instead, import required operators explicitly from Base, e.g. import Base: +, -, *, / (#22251).
Bindings to the FFTW library have been removed from Base. The DFT framework for building FFT implementations is now in AbstractFFTs.jl, the bindings to the FFTW library are in FFTW.jl, and the Base signal processing functions which used FFTs are now in DSP.jl (#21956).
The corrected positional argument to cov has been deprecated in favor of a keyword argument with the same name (#21709).
Omitting spaces around the ? and the : tokens in a ternary expression has been deprecated. Ternaries must now include some amount of whitespace, e.g. x ? a : b rather than x?a:b (#22523 and #22712).
? can no longer be used as an identifier name (#22712)
The method replace(s::AbstractString, pat, r, [count]) is deprecated in favor of replace(s::AbstractString, pat => r; [count]) (#25165). Moreover, count cannot be negative anymore (use typemax(Int) instead (#22325).
read(io, type, dims) is deprecated to read!(io, Array{type}(undef, dims)) (#21450).
read(::IO, ::Ref) is now a method of read!, since it mutates its Ref argument (#21592).
nb_available is now bytesavailable (#25634).
skipchars(io::IO, predicate; linecomment=nothing) is deprecated in favor of skipchars(predicate, io::IO; linecomment=nothing) (#25667).
Bidiagonal constructors now use a Symbol (:U or :L) for the upper/lower argument, instead of a Bool or a Char (#22703).
Bidiagonal, Tridiagonal and SymTridiagonal constructors that automatically converted the input vectors to the same type are deprecated in favor of explicit conversion (#22925, #23035, #23154.
Calling nfields on a type to find out how many fields its instances have is deprecated. Use fieldcount instead. Use nfields only to get the number of fields in a specific object (#22350).
fieldnames now operates only on types. To get the names of fields in an object, use fieldnames(typeof(x)) (#22350).
InexactError, DomainError, and OverflowError now take arguments. InexactError(func::Symbol, type, -3) now prints as "ERROR: InexactError: func(type, -3)", DomainError(val, [msg]) prints as "ERROR: DomainError with val:nmsg", and OverflowError(msg) prints as "ERROR: OverflowError: msg". (#20005, #22751, #22761)
The operating system identification functions: is_linux, is_bsd, is_apple, is_unix, and is_windows, have been deprecated in favor of Sys.islinux, Sys.isbsd, Sys.isapple, Sys.isunix, and Sys.iswindows, respectively (#22182).
The forms of read, readstring, and eachline that accepted both a Cmd object and an input stream are deprecated. Use e.g. read(pipeline(stdin, cmd)) instead (#22762).
The unexported type AbstractIOBuffer has been renamed to GenericIOBuffer (#17360 #22796).
IOBuffer(data::AbstractVector{UInt8}, read::Bool, write::Bool, maxsize::Integer), IOBuffer(read::Bool, write::Bool), and IOBuffer(maxsize::Integer) are deprecated in favor of constructors taking keyword arguments (#25872).
Display has been renamed to AbstractDisplay (#24831).
Remaining vectorized methods over SparseVectors, particularly floor, ceil, trunc, round, and most common transcendental functions such as exp, log, and sin variants, have been deprecated in favor of dot-syntax (#22961).
The method String(io::IOBuffer) is deprecated to String(take!(copy(io))) (#21438).
The function readstring is deprecated in favor of read(io, String) (#22793)
The function showall is deprecated. Showing entire values is the default, unless an IOContext specifying :limit=>true is in use (#22847).
issubtype has been deprecated in favor of <: (which used to be an alias for issubtype).
Calling write on non-isbits arrays is deprecated in favor of explicit loops or serialize (#6466).
The default startup.jl file on Windows has been removed. Now must explicitly include the full path if you need access to executables or libraries in the Sys.BINDIR directory, e.g. joinpath(Sys.BINDIR, "7z.exe") for 7z.exe (#21540).
sqrtm has been deprecated in favor of sqrt (#23504).
expm has been deprecated in favor of exp (#23233).
logm has been deprecated in favor of log (#23505).
full has been deprecated in favor of more specific, better defined alternatives. On structured matrices A, consider instead Matrix(A), Array(A), SparseMatrixCSC(A), or sparse(A). On sparse arrays S, consider instead Vector(S), Matrix(S), or Array(S) as appropriate. On factorizations F, consider instead Matrix(F), Array(F), AbstractMatrix(F), or AbstractArray(F). On implicit orthogonal factors Q, consider instead Matrix(Q) or Array(Q); for implicit orthogonal factors that can be recovered in square or truncated form, see the deprecation message for square recovery instructions. On Symmetric, Hermitian, or AbstractTriangular matrices A, consider instead Matrix(S), Array(S), SparseMatrixCSC(S), or sparse(S). On Symmetric matrices A particularly, consider instead LinAlg.copytri!(copy(parent(A)), A.uplo). On Hermitian matrices A particularly, consider instead LinAlg.copytri!(copy(parent(A)), A.uplo, true). On UpperTriangular matrices A particularly, consider instead triu!(copy(parent(A))). On LowerTriangular matrices A particularly, consider instead tril!(copy(parent(A))) (#24250).
speye has been deprecated in favor of I, sparse, and SparseMatrixCSC constructor methods (#24356).
Calling union with no arguments is deprecated; construct an empty set with an appropriate element type using Set{T}() instead (#23144).
Vectorized DateTime, Date, and format methods have been deprecated in favor of dot-syntax (#23207).
Base.cpad has been removed; use an appropriate combination of rpad and lpad instead (#23187).
ctranspose and ctranspose! have been deprecated in favor of adjoint and adjoint!, respectively (#23235).
filter and filter! on dictionaries now pass a single key=>value pair to the argument function, instead of two arguments (#17886).
rol, rol!, ror, and ror! have been deprecated in favor of specialized methods for circshift/circshift! (#23404).
Base.SparseArrays.SpDiagIterator has been removed (#23261).
The function cfunction, has been deprecated in favor of a macro form @cfunction. Most existing uses can be upgraded simply by adding a @. The new syntax now additionally supports allocating closures at runtime, for dealing with C APIs that don't provide a separate void* env-type callback argument. (#26486)
diagm(v::AbstractVector, k::Integer=0) has been deprecated in favor of diagm(k => v) (#24047).
diagm(x::Number) has been deprecated in favor of fill(x, 1, 1) (#24047).
diagm(A::SparseMatrixCSC) has been deprecated in favor of spdiagm(sparsevec(A)) (#23341).
diagm(A::BitMatrix) has been deprecated, use diagm(0 => vec(A)) or BitMatrix(Diagonal(vec(A))) instead (#23373, #24047).
ℯ (written as mscre<TAB> or euler<TAB>) is now the only (by default) exported name for Euler's number, and the type has changed from Irrational{:e} to Irrational{:ℯ} (#23427).
The mathematical constants π, pi, ℯ, e, γ, eulergamma, catalan, φ and golden have been moved from Base to a new module; Base.MathConstants. Only π, pi and ℯ are now exported by default from Base (#23427).
eu (previously an alias for ℯ) has been deprecated in favor of ℯ (or MathConstants.e) (#23427).
GMP.gmp_version(), GMP.GMP_VERSION, GMP.gmp_bits_per_limb(), and GMP.GMP_BITS_PER_LIMB have been renamed to GMP.version(), GMP.VERSION, GMP.bits_per_limb(), and GMP.BITS_PER_LIMB, respectively. Similarly, MPFR.get_version(), has been renamed to MPFR.version() (#23323). Also, LinAlg.LAPACK.laver() has been renamed to LinAlg.LAPACK.version() and now returns a VersionNumber.
select, select!, selectperm and selectperm! have been renamed respectively to partialsort, partialsort!, partialsortperm and partialsortperm! (#23051).
The Range abstract type has been renamed to AbstractRange (#23570).
map on dictionaries previously operated on key=>value pairs. This behavior is deprecated, and in the future map will operate only on values (#5794).
map on sets previously returned a Set, possibly changing the order or number of elements. This behavior is deprecated and in the future map will preserve order and number of elements (#26980).
Previously, broadcast defaulted to treating its arguments as scalars if they were not arrays. This behavior is deprecated, and in the future broadcast will default to iterating over all its arguments. Wrap arguments you wish to be treated as scalars with Ref() or a 1-tuple. Package developers can choose to allow a non-iterable type T to always behave as a scalar by implementing broadcastable(x::T) = Ref(x) (#26212).
Automatically broadcasted + and - for array + scalar, scalar - array, and so-on have been deprecated due to inconsistency with linear algebra. Use .+ and .- for these operations instead (#22880, #22932).
flipbits!(B) is deprecated in favor of using in-place broadcast to negate each element: B .= .!B (#27067).
isleaftype is deprecated in favor of the simpler predicates isconcretetype and isdispatchtuple. Concrete types are those that might equal typeof(x) for some x; isleaftype included some types for which this is not true. Those are now categorized more precisely as "dispatch tuple types" and "!has_free_typevars" (not exported). (#17086, #25496)
contains(eq, itr, item) is deprecated in favor of any with a predicate (#23716).
spdiagm(x::AbstractVector) has been deprecated in favor of sparse(Diagonal(x)) alternatively spdiagm(0 => x) (#23757).
spdiagm(x::AbstractVector, d::Integer) and spdiagm(x::Tuple{<:AbstractVector}, d::Tuple{<:Integer}) have been deprecated in favor of spdiagm(d => x) and spdiagm(d[1] => x[1], d[2] => x[2], ...) respectively. The new spdiagm implementation now always returns a square matrix (#23757).
spones(A::AbstractSparseArray) has been deprecated in favor of LinAlg.fillstored!(copy(A), 1) (#25037).
Constructors for LibGit2.UserPasswordCredentials and LibGit2.SSHCredentials which take a prompt_if_incorrect argument are deprecated. Instead, prompting behavior is controlled using the allow_prompt keyword in the LibGit2.CredentialPayload constructor (#23690).
gradient is deprecated and will be removed in the next release (#23816).
The timing functions tic, toc, and toq are deprecated in favor of @time and @elapsed (#17046).
Methods of findfirst, findnext, findlast, and findprev that accept a value to search for are deprecated in favor of passing a predicate (#19186, #10593).
find functions now operate only on booleans by default. To look for non-zeros, use x->x!=0 or !iszero (#23120).
The ability of reinterpret to yield Arrays of different type than the underlying storage has been removed. The reinterpret function is still available, but now returns a ReinterpretArray. The three argument form of reinterpret that implicitly reshapes has been deprecated (#23750).
bits has been deprecated in favor of bitstring (#24281, #24263).
num2hex and hex2num have been deprecated in favor of reinterpret combined with parse/hex (#22088).
copy! is deprecated for AbstractSet and AbstractDict, with the intention to re-enable it with a cleaner meaning in a future version (#24844).
copy! (resp. unsafe_copy!) is deprecated for AbstractArray and is renamed copyto! (resp. unsafe_copyto!); it will be re-introduced with a different meaning in a future version (#24808).
a:b is deprecated for constructing a StepRange when a and b have physical units (Dates and Times). Use a:s:b, where s = Dates.Day(1) or s = Dates.Second(1).
trues(A::AbstractArray) and falses(A::AbstractArray) are deprecated in favor of trues(size(A)) and falses(size(A)) respectively (#24595).
workspace is discontinued, check out Revise.jl for an alternative workflow (#25046).
cumsum, cumprod, accumulate, their mutating versions, and diff all now require a dim argument instead of defaulting to using the first dimension unless there is only one dimension (#24684, #25457).
The sum_kbn and cumsum_kbn functions have been moved to the KahanSummation package (#24869).
isnumber has been renamed to isnumeric (#25021).
isalpha has been renamed to isletter (#26932).
is_assigned_char and normalize_string have been renamed to isassigned and normalize, and moved to the new Unicode standard library module. graphemes has also been moved to that module (#25021).
Sparse array functionality has moved to the SparseArrays standard library module (#25249).
Linear algebra functionality, and specifically the LinAlg module has moved to the LinearAlgebra standard library module (#25571).
@printf and @sprintf have been moved to the Printf standard library (#23929,#25056).
The Libdl module has moved to the Libdl standard library module (#25459).
The aliases Complex32, Complex64 and Complex128 have been deprecated in favor of ComplexF16, ComplexF32 and ComplexF64 respectively (#24647).
Base.parentindexes and SharedArrays.localindexes have been renamed to parentindices and localindices, respectively. Similarly, the indexes field in the SubArray type has been renamed to indices without deprecation (#25088).
Associative has been deprecated in favor of AbstractDict (#25012).
Void has been renamed back to Nothing with an alias Cvoid for use when calling C with a return type of Cvoid or a return or argument type of Ptr{Cvoid} (#25162).
Nullable{T} has been deprecated and moved to the Nullables package (#23642). Use Union{T, Nothing} instead, or Union{Some{T}, Nothing} if nothing is a possible value (i.e. Nothing <: T). isnull(x) can be replaced with x === nothing and unsafe_get/get can be dropped or replaced with coalesce. NullException has been removed.
unshift! and shift! have been renamed to pushfirst! and popfirst! (#23902)
ipermute! has been deprecated in favor of invpermute! (#25168).
CartesianRange has been renamed CartesianIndices (#24715).
sub2ind and ind2sub are deprecated in favor of using CartesianIndices and LinearIndices (#24715).
getindex(F::Factorization, s::Symbol) (usually seen as e.g. F[:Q]) is deprecated in favor of dot overloading (getproperty) so factors should now be accessed as e.g. F.Q instead of F[:Q] (#25184).
search and rsearch have been deprecated in favor of findfirst/findnext and findlast/findprev respectively, in combination with curried isequal and in predicates for some methods (#24673).
search(buf::IOBuffer, delim::UInt8) has been deprecated in favor of either occursin(delim, buf) (to test containment) or readuntil(buf, delim) (to read data up to delim) (#26600).
ismatch(regex, str) has been deprecated in favor of contains(str, regex) (#24673).
matchall has been deprecated in favor of collect(m.match for m in eachmatch(r, s)) (#26071).
similar(::Associative) has been deprecated in favor of empty(::Associative), and similar(::Associative, ::Pair{K, V}) has been deprecated in favour of empty(::Associative, K, V) (#24390).
findin(a, b) has been deprecated in favor of findall(in(b), a) (#24673).
module_name has been deprecated in favor of a new, general nameof function. Similarly, the unexported Base.function_name and Base.datatype_name have been deprecated in favor of nameof methods (#25622).
The module Random.dSFMT is renamed Random.DSFMT (#25567).
Random.RandomDevice(unlimited::Bool) (on non-Windows systems) is deprecated in favor of Random.RandomDevice(; unlimited=unlimited) (#25668).
The generic implementations of strides(::AbstractArray) and stride(::AbstractArray, ::Int) have been deprecated. Subtypes of AbstractArray that implement the newly introduced strided array interface should define their own strides method (#25321).
module_parent, Base.datatype_module, and Base.function_module have been deprecated in favor of parentmodule ([#TODO]).
rand(t::Tuple{Vararg{Int}}) is deprecated in favor of rand(Float64, t) or rand(t...); rand(::Tuple) will have another meaning in the future (#25429, #25278).
randjump, which produced an array, is deprecated in favor of the scalar version Future.randjump used with accumulate (#27746).
The assert function (and @assert macro) have been documented that they are not guaranteed to run under various optimization levels and should therefore not be used to e.g. verify passwords.
ObjectIdDict has been deprecated in favor of IdDict{Any,Any} (#25210).
gc and gc_enable have been deprecated in favor of GC.gc and GC.enable (#25616).
Base.@gc_preserve has been deprecated in favor of GC.@preserve (#25616).
print_shortest has been discontinued, but is still available in the Base.Grisu submodule (#25745).
scale! has been deprecated in favor of mul!, lmul!, and rmul! (#25701, #25812).
The remove_destination keyword argument to cp, mv, and the unexported cptree has been renamed to force (#25979).
contains has been deprecated in favor of a more general occursin function, which takes its arguments in reverse order from contains (#26283).
Regex objects are no longer callable. Use occursin instead (#26283).
The methods of range based on positional arguments have been deprecated in favor of keyword arguments (#25896).
linspace has been deprecated in favor of range with stop and length keyword arguments (#25896).
LinSpace has been renamed to LinRange (#25896).
logspace has been deprecated to its definition (#25896).
endof(a) has been renamed to lastindex(a), and the end keyword in indexing expressions now lowers to either lastindex(a) (in the case with only one index) or lastindex(a, d) (in cases where there is more than one index and end appears at dimension d) (#23554, #25763).
DateTime(), Date(), and Time() have been deprecated, instead use DateTime(1), Date(1) and Time(0) respectively (#23724).
The fallback method ^(x, p::Integer) is deprecated. If your type relied on this definition, add a method such as ^(x::MyType, p::Integer) = Base.power_by_squaring(x, p) (#23332).
DevNull, STDIN, STDOUT, and STDERR have been renamed to devnull, stdin, stdout, and stderr, respectively (#25786).
wait and fetch on Task now resemble the interface of Future.
showcompact(io, x...) has been deprecated in favor of show(IOContext(io, :compact => true), x...) (#26080). Use sprint(show, x..., context=:compact => true) instead of sprint(showcompact, x...).
isupper, islower, ucfirst and lcfirst have been deprecated in favor of isuppercase, islowercase, uppercasefirst and lowercasefirst, respectively (#26442).
signif has been deprecated in favor of the sigdigits keyword argument to round.
Base.IntSet has been deprecated in favor of Base.BitSet (#24282).
setrounding has been deprecated for Float32 and Float64, as the behaviour was too unreliable (#26935).
gamma, lgamma, beta, lbeta and lfact have been moved to SpecialFunctions.jl (#27459, #27473).
atan2 is now a 2-argument method of atan (#27248).
The functions eigs and svds have been moved to the Arpack.jl package (#27616).
vecdot and vecnorm are deprecated in favor of dot and norm, respectively (#27401).
clipboard has been moved to the InteractiveUtils standard library package (along with other utilities mostly used at the interactive prompt, such as edit and less) (#27635).
ndigits(n, b, [pad]) is deprecated in favor of ndigits(n, base=b, pad=pad) (#27908).
squ

New in Julia Language 0.6.0 (Aug 8, 2018)

New language features:
New type system capabilities (#8974, #18457):
Type parameter constraints can refer to previous parameters, e.g. type Foo{R<:Real, A<:AbstractArray{R}}. Can also be used in method definitions.
New syntax Array{T} where T<:Integer, indicating a union of types over all specified values of T (represented by a UnionAll type). This provides behavior similar to parametric methods or typealias, but can be used anywhere a type is accepted. This syntax can also be used in method definitions, e.g. function inv(M::Matrix{T}) where T<:AbstractFloat. Anonymous functions can have type parameters via the syntax ((x::Array{T}) where T<:Real) -> 2x.
Implicit type parameters, e.g. Vector{<:Real} is equivalent to Vector{T} where T<:Real, and similarly for Vector{>:Int} (#20414).
Much more accurate subtype and type intersection algorithms. Method sorting and identification of equivalent and ambiguous methods are improved as a result.
Language changes:
"Inner constructor" syntax for parametric types is deprecated.
The keywords used to define types have changed (#19157, #20418).
immutable changes to struct
type changes to mutable struct
abstract changes to abstract type ... end
bitstype 32 Char changes to primitive type Char 32 end
In 0.6, immutable and type are still allowed as synonyms without a deprecation warning.
Multi-line and single-line nonstandard command literals have been added. A nonstandard command literal is like a nonstandard string literal, but the syntax uses backquotes (`) instead of double quotes, and the resulting macro called is suffixed with _cmd. For instance, the syntax q`xyz` is equivalent to @q_cmd "xyz" (#18644).
Nonstandard string and command literals can now be qualified with their module. For instance, Base.r"x" is now parsed as Base.@r_str "x". Previously, this syntax parsed as an implicit multiplication (#18690).
For every binary operator ⨳, a .⨳ b is now automatically equivalent to the broadcast call (⨳).(a, b). Hence, one no longer defines methods for .* etcetera. This also means that "dot operations" automatically fuse into a single loop, along with other dot calls f.(x) (#17623). Similarly for unary operators (#20249).
Newly defined methods are no longer callable from the same dynamic runtime scope they were defined in (#17057).
isa is now parsed as an infix operator with the same precedence as in (#19677).
@. is now parsed as @__dot__, and can be used to add dots to every function call, operator, and assignment in an expression (#20321).
The identifier _ can be assigned, but accessing its value is deprecated, allowing this syntax to be used in the future for discarding values (#9343, #18251, #20328).
The typealias keyword is deprecated, and should be replaced with Vector{T} = Array{T,1} or a const assignment (#20500).
Experimental feature: x^n for integer literals n (e.g. x^3 or x^-3) is now lowered to Base.literal_pow(^, x, Val{n}), to enable compile-time specialization for literal integer exponents (#20530, #20889).
Breaking changes
This section lists changes that do not have deprecation warnings.
readline, readlines and eachline return lines without line endings by default. You must use readline(s, chomp=false), etc. to get the old behavior where returned lines include trailing end-of-line character(s) (#19944).
Strings no longer have a .data field (as part of a significant performance improvement). Use Vector{UInt8}(str) to access a string as a byte array. However, allocating the Vector object has overhead. You can also use codeunit(str, i) to access the ith byte of a String. Use sizeof(str) instead of length(str.data), and pointer(str) instead of pointer(str.data) (#19449).
Operations between Float16 and Integers now return Float16 instead of Float32 (#17261).
Keyword arguments are processed left-to-right: if the same keyword is specified more than once, the rightmost occurrence takes precedence (#17785).
The lgamma(z) function now uses a different (more standard) branch cut for real(z) < 0, which differs from log(gamma(z)) by multiples of 2π in the imaginary part (#18330).
broadcast now handles tuples, and treats any argument that is not a tuple or an array as a "scalar" (#16986).
broadcast now produces a BitArray instead of Array{Bool} for functions yielding a boolean result. If you want Array{Bool}, use broadcast! or .= (#17623).
Broadcast A[I...] .= X with entirely scalar indices I is deprecated as its behavior will change in the future. Use A[I...] = X instead.
Operations like .+ and .* on Range objects are now generic broadcast calls (see above) and produce an Array. If you want a Range result, use + and *, etcetera (#17623).
broadcast now treats Ref (except for Ptr) arguments as 0-dimensional arrays (#18965).
broadcast now handles missing data (Nullables) allowing operations to be lifted over mixtures of Nullables and scalars, as if the Nullable were like an array with zero or one element (#16961, #19787).
The runtime now enforces when new method definitions can take effect (#17057). The flip-side of this is that new method definitions should now reliably actually take effect, and be called when evaluating new code (#265).
The array-scalar methods of /, , *, +, and - now follow broadcast promotion rules. (Likewise for the now-deprecated array-scalar methods of div, mod, rem, &, |, and xor; see "Deprecated or removed" below.) (#19692).
broadcast!(f, A) now calls f() for each element of A, rather than doing fill!(A, f()) (#19722).
rmprocs now throws an exception if requested workers have not been completely removed before waitfor seconds. With a waitfor=0, rmprocs returns immediately without waiting for worker exits.
quadgk has been moved from Base into a separate package (#19741).
The Collections module has been removed, and all functions defined therein have been moved to the DataStructures package (#19800).
The RepString type has been moved to the LegacyStrings.jl package.
In macro calls with parentheses, e.g. @m(a=1), assignments are now parsed as = expressions, instead of as kw expressions (#7669).
When used as an infix operator, ~ is now parsed as a call to an ordinary operator with assignment precedence, instead of as a macro call (#20406).
(µ "micro" and ɛ "latin epsilon") are considered equivalent to the corresponding Greek characters in identifiers. varepsilon now tab-completes to U+03B5 (greek small letter epsilon) (#19464).
retry now inputs the keyword arguments delays and check instead of n and max_delay. The previous functionality can be achieved setting delays to ExponentialBackOff (#19331).
transpose(::AbstractVector) now always returns a RowVector view of the input (which is a special 1×n-sized AbstractMatrix), not a Matrix, etc. In particular, for v::AbstractVector we now have (v.').' === v and v.' * v is a scalar (#19670).
Parametric types with "unspecified" parameters, such as Array, are now represented as UnionAll types instead of DataTypes (#18457).
Union types have two fields, a and b, instead of a single types field. The empty type Union{} is represented by a singleton of type TypeofBottom (#18457).
The type NTuple{N} now refers to tuples where every element has the same type (since it is shorthand for NTuple{N,T} where T). To get the old behavior of matching any tuple, use NTuple{N,Any} (#18457).
FloatRange has been replaced by StepRangeLen, and the internal representation of LinSpace has changed. Aside from changes in the internal field names, this leads to several differences in behavior (#18777):
Both StepRangeLen and LinSpace can represent ranges of arbitrary object types---they are no longer limited to floating-point numbers.
For ranges that produce Float64, Float32, or Float16 numbers, StepRangeLen can be used to produce values with little or no roundoff error due to internal arithmetic that is typically twice the precision of the output result.
To take advantage of this precision, linspace(start, stop, len) now returns a range of type StepRangeLen rather than LinSpace when start and stop are FloatNN. LinSpace(start, stop, len) always returns a LinSpace.
StepRangeLen(a, step, len) constructs an ordinary-precision range using the values and types of a and step as given, whereas range(a, step, len) will attempt to match inputs a::FloatNN and step::FloatNN to rationals and construct a StepRangeLen that internally uses twice-precision arithmetic. These two outcomes exhibit differences in both precision and speed.
A=>B expressions are now parsed as calls instead of using => as the expression head (#20327).
The count function no longer sums non-boolean values (#20404)
The generic getindex(::AbstractString, ::AbstractVector) method's signature has been tightened to getindex(::AbstractString, ::AbstractVector{<:Integer}). Consequently, indexing into AbstractStrings with non-AbstractVector{<:Integer} AbstractVectors now throws a MethodError in the absence of an appropriate specialization. (Previously such cases failed less explicitly with the exception of AbstractVector{Bool}, which now throws an ArgumentError noting that logical indexing into strings is not supported.) (#20248)
Bessel, Hankel, Airy, error, Dawson, eta, zeta, digamma, inverse digamma, trigamma, and polygamma special functions have been moved from Base to the SpecialFunctions.jl package (#20427). Note that airy, airyx and airyprime have been deprecated in favor of more specific functions (airyai, airybi, airyaiprime, airybiprimex, airyaix, airybix, airyaiprimex, airybiprimex) (#18050).
When a macro is called in the module in which that macro is defined, global variables in the macro are now correctly resolved in the macro definition environment. Breakage from this change commonly manifests as undefined variable errors that do not occur under 0.5. Fixing such breakage typically requires sprinkling additional escs in the offending macro (#15850).
write on an IOBuffer now returns a signed integer in order to be consistent with other buffers (#20609).
The <:Integer division fallback /(::Integer, ::Integer), which formerly inappropriately took precedence over other division methods for some mixed-integer-type division calls, has been removed (#19779).
@async, @spawn, @spawnat, @fetch and @fetchfrom no longer implicitly localize variables. Previously, the expression would be wrapped in an implicit let block (#19594).
parse no longer accepts IPv4 addresses including leading zeros, octal, or hexadecimal. Convert IPv4 addresses including octal or hexadecimal to decimal, and remove leading zeros in decimal addresses (#19811).
Closures shipped for remote execution via @spawn or remotecall now automatically serialize globals defined under Main. For details, please refer to the paragraph on "Global variables" under the "Parallel computing" chapter in the manual (#19594).
homedir now determines the user's home directory via libuv's uv_os_homedir, rather than from environment variables (#19636).
Workers now listen on an ephemeral port assigned by the OS. Previously workers would listen on the first free port available from 9009 (#21818).
Library improvements:
A new @views macro was added to convert a whole expression or block of code to use views for all slices (#20164).
max, min, and related functions (minmax, maximum, minimum, extrema) now return NaN for NaN arguments (#12563).
oneunit(x) function to return a dimensionful version of one(x) (which is clarified to mean a dimensionless quantity if x is dimensionful) (#20268).
The chop and chomp functions now return a SubString (#18339).
Numbered stackframes printed in stacktraces can now be opened in an editor by entering the corresponding number in the REPL and pressing ^Q (#19680).
The REPL now supports something called prompt pasting (#17599). This activates when pasting text that starts with julia> into the REPL. In that case, only expressions starting with julia> are parsed, the rest are removed. This makes it possible to paste a chunk of code that has been copied from a REPL session without having to scrub away prompts and outputs. This can be disabled or enabled at will with Base.REPL.enable_promptpaste(::Bool).
The function print_with_color can now take a color represented by an integer between 0 and 255 inclusive as its first argument (#18473). For a number-to-color mapping, please refer to this chart. It is also possible to use numbers as colors in environment variables that customizes colors in the REPL. For example, to get orange warning messages, simply set ENV["JULIA_WARN_COLOR"] = 208. Please note that not all terminals support 256 colors.
The function print_with_color no longer prints text in bold by default (#18628). Instead, the function now take a keyword argument bold::Bool which determines whether to print in bold or not. On some terminals, printing a color in non bold results in slightly darker colors being printed than when printing in bold. Therefore, light versions of the colors are now supported. For the available colors see the help entry on print_with_color.
The default text style for REPL input and answers has been changed from bold to normal (#11250). They can be changed back to bold by setting the environment variables JULIA_INPUT_COLOR and JULIA_ANSWER_COLOR to "bold". For example, one way of doing this is adding ENV["JULIA_INPUT_COLOR"] = :bold and ENV["JULIA_ANSWER_COLOR"] = :bold to the .juliarc.jl file. See the manual section on customizing colors for more information.
The default color for info messages has been changed from blue to cyan (#18442), and for warning messages from red to yellow (#18453). This can be changed back to the original colors by setting the environment variables JULIA_INFO_COLOR to "blue" and JULIA_WARN_COLOR to "red".
Iteration utilities that wrap iterators and return other iterators (enumerate, zip, rest, countfrom, take, drop, cycle, repeated, product, flatten, partition) have been moved to the module Base.Iterators (#18839).
BitArrays can now be constructed from arbitrary iterables, in particular from generator expressions, e.g. BitArray(isodd(x) for x = 1:100) (#19018).
hcat, vcat, and hvcat now work with UniformScaling objects, so you can now do e.g. [A I] and it will concatenate an appropriately sized identity matrix (#19305).
New accumulate and accumulate! functions were added, which generalize cumsum and cumprod. Also known as a scan operation (#18931).
reshape now allows specifying one dimension with a Colon() (:) for the new shape, in which case that dimension's length will be computed such that its product with all the other dimensions is equal to the length of the original array (#19919).
The new to_indices function provides a uniform interface for index conversions, taking an array and a tuple of indices as arguments and returning a tuple of integers and/or arrays of supported scalar indices. It will throw an ArgumentError for any unsupported indices, and the returned arrays should be iterated over (and not indexed into) to support more efficient logical indexing (#19730).
Using colons (:) to represent a collection of indices is deprecated. They now must be explicitly converted to a specialized array of integers with the to_indices function. As a result, the type of SubArrays that represent views over colon indices has changed.
Logical indexing is now more efficient. Logical arrays are converted by to_indices to a lazy, iterable collection of indices that doesn't support indexing. A deprecation provides indexing support with O(n) lookup.
The performance of indexing with CartesianIndexes is also improved in many situations.
A new titlecase function was added, to capitalize the first character of each word within a string (#19469).
any and all now always short-circuit, and mapreduce never short-circuits (#19543). That is, not every member of the input iterable will be visited if a true (in the case of any) or false (in the case of all) value is found, and mapreduce will visit all members of the iterable.
Additional methods for ones and zeros functions were added to support the same signature as the similar function (#19635).
count now has a count(itr) method equivalent to count(identity, itr) (#20403).
Methods for map and filter with Nullable arguments have been implemented; the semantics are as if the Nullable were a container with zero or one elements (#16961).
New @test_warn and @test_nowarn macros were added in the Base.Test module to test for the presence or absence of warning messages (#19903).
logging can now be used to redirect info, warn, and error messages either universally or on a per-module/function basis (#16213).
New function Base.invokelatest(f, args...) to call the latest version of a function in circumstances where an older version may be called instead (e.g. in a function calling eval) (#19784).
A new iszero(x) function was added, to quickly check whether x is zero (or is all zeros, for an array) (#19950).
notify now returns a count of tasks woken up (#19841).
A new nonstandard string literal raw"..." was added, for creating strings with no interpolation or unescaping (#19900).
A new Dates.Time type was added that supports representing the time of day with up to nanosecond resolution (#12274).
Raising one or negative one to a negative integer power formerly threw a DomainError. One raised to any negative integer power now yields one, negative one raised to any negative even integer power now yields one, and negative one raised to any negative odd integer power now yields negative one. Similarly, raising true to any negative integer power now yields true rather than throwing a DomainError (#18342).
A new @macroexpand macro was added as a convenient alternative to the macroexpand function (#18660).
invoke now supports keyword arguments (#20345).
A new ConjArray type was added, as a wrapper type for lazy complex conjugation of arrays. Currently, it is used by default for the new RowVector type only, and enforces that both transpose(vec) and ctranspose(vec) are views not copies (#20047).
rem now accepts a RoundingMode argument via rem(x, y, r::RoundingMode), yielding x - y*round(x/y, r) without intermediate rounding. In particular, rem(x, y, RoundNearest) yields a value in the interval [-abs(y)/2, abs(y)/2]), which corresponds to the IEE754 remainder function. Similarly, rem2pi(x, r::RoundingMode) now exists as well, yielding rem(x, 2pi, r::RoundingMode) but with greater accuracy (#10946).
map[!] and broadcast[!] now have dedicated methods for sparse/structured vectors/matrices. Specifically, map[!] and broadcast[!] over combinations including one or more SparseVector, SparseMatrixCSC, Diagonal, Bidiagonal, Tridiagonal, or SymTridiagonal, and any number of broadcast scalars, Vectors, or Matrixs, now efficiently yield SparseVectors or SparseMatrixs as appropriate (#19239, #19371, #19518, #19438, #19690, #19724, #19926, #19934, #20009).
The operators ! and ∘ (circ<tab> at the REPL and in most code editors) now respectively perform predicate function negation and function composition. For example, map(!iszero, (0, 1)) is now equivalent to map(x -> !iszero(x), (0, 1)) and map(uppercase ∘ hex, 250:255) is now equivalent to map(x -> uppercase(hex(x)), 250:255) (#17155).
enumerate now supports the two-argument form enumerate(::IndexStyle, iterable). This form allows specification of the returned indices' style. For example, enumerate(IndexLinear, iterable) yields linear indices and enumerate(IndexCartesian, iterable) yields cartesian indices (#16378).
Compiler/Runtime improvements:
ccall is now implemented as a macro, removing the need for special code-generator support for Intrinsics (#18754).
ccall gained limited support for a llvmcall calling-convention. This can replace many uses of llvmcall with a simpler, shorter declaration (#18754).
All Intrinsics are now Builtin functions instead and have proper error checking and fall-back static compilation support (#18754).
Deprecated or removed:
ipermutedims(A::AbstractArray, p) has been deprecated in favor of permutedims(A, invperm(p)) (#18891).
Linear indexing is now only supported when there is exactly one non-cartesian index provided. Allowing a trailing index at dimension d to linearly access the higher dimensions from array A (beyond size(A, d)) has been deprecated as a stricter constraint during bounds checking. Instead, reshape the array such that its dimensionality matches the number of indices (#20079).
Multimedia.@textmime "mime" has been deprecated. Instead define Multimedia.istextmime(::MIME"mime") = true (#18441).
isdefined(a::Array, i::Int) has been deprecated in favor of isassigned (#18346).
The three-argument SubArray constructor (which accepts dims::Tuple as its third argument) has been deprecated in favor of the two-argument equivalent (the dims::Tuple argument being superfluous) (#19259).
is has been deprecated in favor of === (which used to be an alias for is) (#17758).
Ambiguous methods for addition and subtraction between UniformScalings and Numbers, for example (+)(J::UniformScaling, x::Number), have been deprecated in favor of unambiguous, explicit equivalents, for example J.λ + x (#17607).
num and den have been deprecated in favor of numerator and denominator respectively (#19233,#19246).
delete!(ENV::EnvDict, k::AbstractString, def) has been deprecated in favor of pop!(ENV, k, def). Be aware that pop! returns k or def, whereas delete! returns ENV or def (#18012).
infix operator $ has been deprecated in favor of infix ⊻ or function xor (#18977).
The single-argument form of write (write(x), with implicit STDOUT output stream), has been deprecated in favor of the explicit equivalent write(STDOUT, x) (#17654).
Dates.recur has been deprecated in favor of filter (#19288)
A number of ambiguous convert operations between Numbers (especially Reals) and Date, DateTime, and Period types have been deprecated in favor of unambiguous convert and explicit constructor calls. Additionally, ambiguous colon construction of <:Period ranges without step specification, for example Dates.Hour(1):Dates.Hour(2), has been deprecated in favor of such construction including step specification, for example Dates.Hour(1):Dates.Hour(1):Dates.Hour(2) (#19920).
cummin and cummax have been deprecated in favor of accumulate (#18931).
The Array constructor syntax Array(T, dims...) has been deprecated in favor of the forms Array{T,N}(dims...) (where N is known, or particularly Vector{T}(dims...) for N = 1 and Matrix{T}(dims...) for N = 2), and Array{T}(dims...) (where N is not known). Likewise for SharedArrays (#19989).
sumabs and sumabs2 have been deprecated in favor of sum(abs, x) and sum(abs2, x), respectively. maxabs and minabs have similarly been deprecated in favor of maximum(abs, x) and minimum(abs, x). Likewise for the in-place counterparts of these functions (#19598).
The array-reducing form of isinteger (isinteger(x::AbstractArray)) has been deprecated in favor of all(isinteger, x) (#19925).
produce, consume and iteration over a Task object have been deprecated in favor of using Channels for inter-task communication (#19841).
The negate keyword has been deprecated from all functions in the Dates adjuster API (adjust, tonext, toprev, Date, Time, and DateTime). Instead use predicate function negation via the ! operator (see Library Improvements) (#20213).
@test_approx_eq x y has been deprecated in favor of @test isapprox(x,y) or @test x ≈ y (#4615).
Matrix() and Matrix{T}() have been deprecated in favor of the explicit forms Matrix(0, 0) and Matrix{T}(0, 0) (#20330).
Vectorized functions have been deprecated in favor of dot syntax (#17302, #17265, #18558, #19711, #19712, #19791, #19802, #19931, #20543, #20228).
All methods of character predicates (isalnum, isalpha, iscntrl, isdigit, isnumber, isgraph, islower, isprint, ispunct, isspace, isupper, isxdigit) that accept AbstractStrings have been deprecated in favor of all. For example, isnumber("123") should now be expressed all(isnumber, "123") (#20342).
A few names related to indexing traits have been changed: LinearIndexing and linearindexing have been deprecated in favor of IndexStyle. LinearFast has been deprecated in favor of IndexLinear, and LinearSlow has been deprecated in favor of IndexCartesian (#16378).
The two-argument forms of map (map!(f, A)) and asyncmap! (asyncmap!(f, A)) have been deprecated in anticipation of future semantic changes (#19721).
unsafe_wrap(String, ...) has been deprecated in favor of unsafe_string (#19449).
zeros and ones methods accepting an element type as the first argument and an array as the second argument, for example zeros(Float64, [1, 2, 3]), have been deprecated in favor of equivalent methods with the second argument instead the size of the array, for example zeros(Float64, size([1, 2, 3])) (#21183).
Base.promote_eltype_op has been deprecated (#19669, #19814, #19937).
isimag has been deprecated (#19949).
The tuple-of-types form of invoke, invoke(f, (types...), ...), has been deprecated in favor of the tuple-type form invoke(f, Tuple{types...}, ...) (#18444).
Base._promote_array_type has been deprecated (#19766).
broadcast_zpreserving has been deprecated (#19533, #19720).
Methods allowing indexing of tuples by AbstractArrays with more than one dimension have been deprecated. (Indexing a tuple by such a higher-dimensional AbstractArray should yield a tuple with more than one dimension, but tuples are one-dimensional.) (#19737).
@test_approx_eq a b has been deprecated in favor of @test a ≈ b (or, equivalently, @test ≈(a, b) or @test isapprox(a, b)). @test_approx_eq_eps has been deprecated in favor of new @test syntax: @test now supports the syntax @test f(args...) key=val ... for @test f(args..., key=val...). This syntax allows, for example, writing @test a ≈ b atol=c in place of @test ≈(a, b, atol=c) (and hence @test_approx_eq_eps a b c) (#19901).
takebuf_array has been deprecated in favor of take!, and takebuf_string(x) has been deprecated in favor of String(take!(x)) (#19088).
convert methods from Diagonal and Bidiagonal to subtypes of AbstractTriangular have been deprecated (#17723).
Base.LinAlg.arithtype has been deprecated. If you were using arithtype within a promote_op call, instead use promote_op(Base.LinAlg.matprod, Ts...). Otherwise, consider defining equivalent functionality locally (#18218).
Special characters (#{}()[]<>|&*?~;) should now be quoted in commands. For example, `export FOO=1;` should replace `export FOO=1;` and `cd $dir '&&' $thingie` should replace `cd $dir && $thingie` (#19786).
Zero-argument Channel constructors (Channel(), Channel{T}()) have been deprecated in favor of equivalents accepting an explicit Channel size (Channel(2), Channel{T}(2)) (#18832).
The zero-argument constructor MersenneTwister() has been deprecated in favor of the explicit MersenneTwister(0) (#16984).
Base.promote_type(op::Type, Ts::Type...) has been removed as part of an overhaul of broadcast's promotion mechanism. If you need the functionality of that Base.promote_type method, consider defining it locally via Core.Compiler.return_type(op, Tuple{Ts...}) (#18642).
bitbroadcast has been deprecated in favor of broadcast, which now produces a BitArray instead of Array{Bool} for functions yielding a boolean result (#19771).
To complete the deprecation of histogram-related functions, midpoints has been deprecated. Instead use the StatsBase.jl package's midpoints function (#20058).
Passing a type argument to LibGit2.cat has been deprecated in favor of a simpler, two-argument method for LibGit2.cat (#20435).
The LibGit2.owner function for finding the repository which owns a given Git object has been deprecated in favor of LibGit2.repository (#20135).
The LibGit2.GitAnyObject type has been renamed to LibGit2.GitUnknownObject to clarify its intent (#19935).
The LibGit2.GitOid type has been renamed to LibGit2.GitHash for clarity (#19878).
Finalizing LibGit2 objects with finalize has been deprecated in favor of using close (#19660).
Parsing string dates from a Dates.DateFormat object has been deprecated as part of a larger effort toward faster, more extensible date parsing (#20952).
Command-line option changes:
In polly builds (USE_POLLY := 1), the new flag --polly={yes|no} controls whether @polly declarations are respected. (With --polly=no, @polly declarations are ignored.) This flag is also available in non-polly builds (USE_POLLY := 0), but has no effect (#18159).

Julia Language Changelog

What's new in Julia Language 1.11.0 Alpha 1

New in Julia Language 1.10.0 (Dec 28, 2023)

New in Julia Language 1.10.0 Alpha 1 (Jul 7, 2023)

New in Julia Language 1.9.0 (May 11, 2023)

New in Julia Language 1.8.0 (Aug 18, 2022)

New in Julia Language 1.8.0 Beta 1 (Feb 27, 2022)

New in Julia Language 1.7.0 (Dec 1, 2021)

New in Julia Language 1.6.0 (Mar 26, 2021)

New in Julia Language 1.6.0 RC 1 (Feb 8, 2021)

New in Julia Language 1.5.1 (Aug 27, 2020)

New in Julia Language 1.5.0 (Aug 3, 2020)

New in Julia Language 1.4.0 (Mar 22, 2020)

New in Julia Language 1.4.0 RC 2 (Feb 27, 2020)

New in Julia Language 1.3.0 (Nov 27, 2019)

New in Julia Language 1.3.0 RC2 (Sep 17, 2019)

New in Julia Language 1.3.0 Alpha (Jul 24, 2019)

New in Julia Language 1.1.0 (Jan 22, 2019)

New in Julia Language 1.1.0 RC2 (Jan 15, 2019)

New in Julia Language 1.1.0 RC1 (Jan 3, 2019)

New in Julia Language 1.0.1 (Oct 22, 2018)

New in Julia Language 0.7.0 (Aug 8, 2018)

New in Julia Language 0.6.0 (Aug 8, 2018)

New in Julia Language 0.5.0 (Jun 20, 2017)

New in Julia Language 0.4.0 (Oct 9, 2015)

New in Julia Language 0.3.0 (Feb 19, 2015)