- New option
--trim
creates smaller binaries by removing code that was not proven to be reachable from the entry points. Entry points can be marked usingBase.Experimental.entrypoint
([#55047]). - A new keyword argument
usings::Bool
has been added tonames
. By using this, we can now find all the names available in moduleA
bynames(A; all=true, imported=true, usings=true)
. ([#54609]) - the
@atomic(...)
macro family supports now the reference assignment syntax, e.g.@atomic :monotonic v[3] += 4
modifiesv[3]
atomically with monotonic ordering semantics. ([#54707]) The supported syntax allows- atomic fetch (
x = @atomic v[3]
), - atomic set (
@atomic v[3] = 4
), - atomic modify (
@atomic v[3] += 2
), - atomic set once (
@atomiconce v[3] = 2
), - atomic swap (
x = @atomicswap v[3] = 2
), and - atomic replace (
x = @atomicreplace v[3] 2=>5
).
- atomic fetch (
-
When methods are replaced with exactly equivalent ones, the old method is no longer deleted implicitly simultaneously, although the new method does take priority and become more specific than the old method. Thus if the new method is deleted later, the old method will resume operating. This can be useful to mocking frameworks (such as in SparseArrays, Pluto, and Mocking, among others), as they do not need to explicitly restore the old method. While inference and compilation still must be repeated with this, it also may pave the way for inference to be able to intelligently re-use the old results, once the new method is deleted. ([#53415])
-
Macro expansion will no longer eagerly recurse into into
Expr(:toplevel)
expressions returned from macros. Instead, macro expansion of:toplevel
expressions will be delayed until evaluation time. This allows a later expression within a given:toplevel
expression to make use of macros defined earlier in the same:toplevel
expression. ([#53515]) -
Trivial infinite loops (like
while true; end
) are no longer undefined behavior. Infinite loops that actually do things (e.g. have side effects or sleep) were never and are still not undefined behavior. ([#52999])
-
Generated LLVM IR now uses actual pointer types instead of passing pointers as integers. This affects
llvmcall
: Inline LLVM IR should be updated to usei8*
orptr
instead ofi32
ori64
, and remove unneededptrtoint
/inttoptr
conversions. For compatibility, IR with integer pointers is still supported, but generates a deprecation warning. ([#53687]) -
A new exception
FieldError
is now introduced to raise/handlegetfield
exceptions. Previouslygetfield
exception was captured by fallback generic exceptionErrorException
. Now thatFieldError
is more specificgetfield
related exceptions that can occur should useFieldError
exception instead. ([#54504])
- The
-m/--module
flag can be passed to run themain
function inside a package with a set of arguments. Thismain
function should be declared using@main
to indicate that it is an entry point. - Enabling or disabling color text in Julia can now be controlled with the
NO_COLOR
orFORCE_COLOR
environment variables. These variables are also honored by Julia's build system ([#53742], [#56346]). --project=@temp
starts Julia with a temporary environment.- New
--trace-compile-timing
option to report how long each method reported by--trace-compile
took to compile, in ms. ([#54662]) --trace-compile
now prints recompiled methods in yellow or with a trailing comment if color is not supported ([#55763])- New
--trace-dispatch
option to report methods that are dynamically dispatched ([#55848]).
- New types are defined to handle the pattern of code that must run once per process, called
a
OncePerProcess{T}
type, which allows defining a function that should be run exactly once the first time it is called, and then always return the same result value of typeT
every subsequent time afterwards. There are alsoOncePerThread{T}
andOncePerTask{T}
types for similar usage with threads or tasks. ([#TBD])
- There are new
Makefile
s to build Julia and LLVM using the Binary Optimization and Layout Tool (BOLT), seecontrib/bolt
andcontrib/pgo-lto-bolt
([#54107]).
logrange(start, stop; length)
makes a range of constant ratio, instead of constant step ([#39071])- The new
isfull(c::Channel)
function can be used to check ifput!(c, some_value)
will block. ([#53159]) waitany(tasks; throw=false)
andwaitall(tasks; failfast=false, throw=false)
which wait multiple tasks at once ([#53341]).uuid7()
creates an RFC 9652 compliant UUID with version 7 ([#54834]).insertdims(array; dims)
allows to insert singleton dimensions into an array which is the inverse operation todropdims
- The new
Fix
type is a generalization ofFix1/Fix2
for fixing a single argument ([#54653]).
invmod(n, T)
whereT
is a native integer type now computes the modular inverse ofn
in the modular integer ring thatT
defines ([#52180]).invmod(n)
is an abbreviation forinvmod(n, typeof(n))
for native integer types ([#52180]).replace(string, pattern...)
now supports an optionalIO
argument to write the output to a stream rather than returning a string ([#48625]).sizehint!(s, n)
now supports an optionalshrink
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 forProcess
spawn now works as expected, synchronized withwait
orsuccess
, so aBase.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. Callwait
on the process instead to ensure the content is fully written, then callclosewrite
manually to avoid data-races. Or use the callback form ofopen
to have all that handled automatically. @timed
now additionally returns the elapsed compilation and recompilation time ([#52889])escape_string
takes additional keyword argumentsascii=true
(to escape all non-ASCII characters) andfullhex=true
(to require full 4/8-digit hex numbers for u/U escapes, e.g. for C compatibility) [#55099]).filter
can now act on aNamedTuple
([#50795]).tempname
can now take a suffix string to allow the file name to include a suffix and include that suffix in the uniquing checking ([#53474])RegexMatch
objects can now be used to constructNamedTuple
s andDict
s ([#50988])Lockable
is now exported ([#54595])Base.require_one_based_indexing
andBase.has_offset_axes
are now public ([#56196])- New
ltruncate
,rtruncate
andctruncate
functions for truncating strings to text width, accounting for char widths ([#55351]) isless
(and thuscmp
, sorting, etc.) is now supported for zero-dimensionalAbstractArray
s ([#55772])
gcdx(0, 0)
now returns(0, 0, 0)
instead of(0, 1, 0)
([#40989]).fd
returns aRawFD
instead of anInt
([#55080]).
- A new standard library for applying syntax highlighting to Julia code, this
uses
JuliaSyntax
andStyledStrings
to implement ahighlight
function that creates anAnnotatedString
with syntax highlighting applied.
rank
can now take aQRPivoted
matrix to allow rank estimation via QR factorization ([#54283]).- Added keyword argument
alg
toeigen
,eigen!
,eigvals
andeigvals!
for self-adjoint matrix types (i.e., the type unionRealHermSymComplexHerm
) that allows one to switch between different eigendecomposition algorithms ([#49355]). - Added a generic version of the (unblocked) pivoted Cholesky decomposition
(callable via
cholesky[!](A, RowMaximum())
) ([#54619]). - The number of default BLAS threads now respects process affinity, instead of using total number of logical threads available on the system ([#55574]).
- A new function
zeroslike
is added that is used to generate the zero elements for matrix-valued banded matrices. Custom array types may specialize this function to return an appropriate result ([#55252]). - The matrix multiplication
A * B
callsmatprod_dest(A, B, T::Type)
to generate the destination. This function is now public ([#55537]). - The function
haszero(T::Type)
is used to check if a typeT
has a unique zero element defined aszero(T)
. This is now public.
Profile.take_heap_snapshot
takes a new keyword argument,redact_data::Bool
, that istrue
by default. When set, the contents of Julia objects are not emitted in the heap snapshot. This currently only applies to strings. ([#55326])Profile.print()
now colors Base/Core/Package modules similarly to how they are in stacktraces. Also paths, even if truncated, are now clickable in terminals that support URI links to take you to the specifiedJULIA_EDITOR
for the given file & line number. ([#55335])
- Using the new
usings=true
feature of thenames()
function, REPL completions can now complete names that have been explicitlyusing
-ed. ([#54610]) - REPL completions can now complete input lines like
[import|using] Mod: xxx|
e.g. completeusing Base.Experimental: @op
tousing Base.Experimental: @opaque
. ([#54719]) - the REPL will now warn if it detects a name is being accessed from a module which does not define it (nor has a submodule which defines it),
and for which the name is not public in that module. For example,
map
is defined in Base, and executingLinearAlgebra.map
in the REPL will now issue a warning the first time occurs. ([#54872]) - When an object is printed automatically (by being returned in the REPL), its display is now truncated after printing 20 KiB.
This does not affect manual calls to
show
,print
, and so forth. ([#53959])
- New macros
@trace_compile
and@trace_dispatch
for running an expression with--trace-compile=stderr --trace-compile-timing
and--trace-dispatch=stderr
respectively enabled. ([#55915])
- The terminal info database,
terminfo
, is now vendored by default, providing a better REPL user experience whenterminfo
is not available on the system. Julia can be built without vendoring the database using the Makefile optionWITH_TERMINFO=0
. ([#55411])
- A wall-time profiler is now available for users who need a sampling profiler that captures tasks regardless of their scheduling or running state. This type of profiler enables profiling of I/O-heavy tasks and helps detect areas of heavy contention in the system ([#55889]).