Skip to main content

Browse the archive

swh:1:snp:a72e953ecd624a7df6e6196bbdd05851996c5e40
sort by:
Revision Author Date Message Commit Date
20a6dea Matt Bauman 20 February 2024, 15:58:37 UTC Merge remote-tracking branch 'origin/master' into sgj/mapreduce_pairwise 20 February 2024, 15:58:37 UTC
02699bb Jameson Nash 19 February 2024, 12:52:29 UTC fix sysimage-native-code=no option with pkgimages (#53373) Loading pkgimages would try to access the sysimage native code, which will fail. Ensure that no code tries to load if the sysimage native code is not available, as it may try to link against it. Fixes #53147 19 February 2024, 12:52:29 UTC
9c0f1dc Kristoffer Carlsson 19 February 2024, 12:11:08 UTC add being able to pass a `CacheFlags()` to `isprecompiled` to verify a precompile file against a set of custom flags (#53332) 19 February 2024, 12:11:08 UTC
59102aa DilumAluthgeBot 19 February 2024, 02:07:49 UTC 🤖 [master] Bump the Pkg stdlib from 6dd0e7c9e to 76070d295 (#53357) Stdlib: Pkg URL: https://github.com/JuliaLang/Pkg.jl.git Stdlib branch: master Julia branch: master Old commit: 6dd0e7c9e New commit: 76070d295 Julia version: 1.12.0-DEV Pkg version: 1.11.0(Does not match) Bump invoked by: @IanButterworth Powered by: [BumpStdlibs.jl](https://github.com/JuliaLang/BumpStdlibs.jl) Diff: https://github.com/JuliaLang/Pkg.jl/compare/6dd0e7c9e99d578aa5477e2c78c91a161ce4c357...76070d295fc4a1f27f852e05400bbc956962e084 ``` $ git log --oneline 6dd0e7c9e..76070d295 76070d295 Prevent repl crash on invalid command (#3800) d267986c2 RFC: stop testing non-ARM on mac on CI (#3794) 2571be6d1 CI: Add Apple Silicon (macOS aarch64) to the CI matrix (#3793) 3075fb78c Simplify Pkg.Registry APIs. (#3785) e6f1e0902 Explain about Manifest.toml per julia version (#3791) ``` Co-authored-by: Dilum Aluthge <dilum@aluthge.com> 19 February 2024, 02:07:49 UTC
1c25d93 Keno Fischer 19 February 2024, 01:35:56 UTC Refactor CodeInfo/CodeInstance separation and interfaces (#53219) The `CodeInfo` type is one of the oldest types in the system and has grown a bit of cruft. In particular, the `rettype`, `inferred`, `parent`, `edges`, `min_world`, `max_world` fields are not used for the original purpose of representing code, but for one or more of (in decreasing order of badness): 1. Smuggling extra results from inference into the compiler 2. Sumggling extra arguments into OpaqueClosure constructors 3. Passing extra information from generated functions to inference The first of these points in particular causes a fair bit of mixup between caching concerns and compiler concerns and results in external abstract interpreters maintainging their own dummy CodeInfos, just to comply with the interface. Originally, I just wanted to clean up that bit, but it didn't really make sense as a standalone piece, so this PR is more comprehensive. In particular, this PR: 1. Removes the `parent`, `inferred` and `rettype` fields of `CodeInfo`. They are largely vestigal and code accessing these is probably doing the wrong thing. They should instead be looking at either the CodeInstance or remembering the query that was asked of the cache in the first place. 2. Makes `edges`, `min_world` and `max_world` used for generated functions only. All other uses were replaced by appropriate queries on the CodeInstance. In particular, inference no longer sets these. In the future we may want to consider removing these also and having generated functions return some other object, but that is a topic to revisit once the broader compiler plugins landscape is more clear. 3. Makes the external type inference interface return `CodeInstance` rather than `CodeInfo`. This results in a lot of cleanup, because many functions had multiple code paths, some for CodeInstance and others for fallback to inference/CodeInfo. This is all cleaned up now. If you don't have a CodeInstance, you can ask inference for one. This CodeInstance may or may not be in the cache, but you can look at its types, compile it, etc. 4. Moves the main inference entrypoint out of the codegen library. There is still a little bit of entangelement, but this makes codegen much more of an independent system that you give a CodeInstance and it just fills in the invoke pointer for. With these changes, only the third use of the above mentioned fields remains. The overall theme here is decoupling. Over time, various parties have wanted to use the julia compiler with custom IR datastructure, backend code generators, caches, etc. This doesn't quite get us all the way there, but makes inference and codegen much more independent with a clear IR-format-independent interface (CodeInstance). --------- Co-authored-by: Valentin Churavy <v.churavy@gmail.com> 19 February 2024, 01:35:56 UTC
c0a93f8 Collin Wittenstein 18 February 2024, 16:01:59 UTC fixed typo in performance-tips.md (#53379) I have ESL, but I'm quite certain in should be "repeatedly accessing it" and not "repeated access it" 18 February 2024, 16:01:59 UTC
4505855 Zhuoqian Zhou 18 February 2024, 15:44:13 UTC fix typos in docs (#53378) fix typos in docs 18 February 2024, 15:44:13 UTC
2a9c95a Aravindh Krishnamoorthy 18 February 2024, 13:10:42 UTC Implement `circshift(::Tuple, ::Integer)` (#52438) 18 February 2024, 13:10:42 UTC
16871e7 Jishnu Bhattacharya 18 February 2024, 08:40:51 UTC Orthogonalize re-indexing for FastSubArrays (#53369) By separating out the re-indexing step for `FastSubArray`s and specializing this for `FastContiguousSubArray`s, we don't need to define specialized `getindex`, `setindex!` and `isassigned` for `FastContiguousSubArray`s anymore. The fallback method for `FastSubArray`s will correctly handle the special case. 18 February 2024, 08:40:51 UTC
b5221e1 Florian 17 February 2024, 21:09:32 UTC Silence warnings in `test/file.jl` (#53372) Warnings were introduced in #33593. 17 February 2024, 21:09:32 UTC
2ca3753 Florian 17 February 2024, 20:17:29 UTC Add `Base.wrap` to docs (#53342) I thought that's reasonable in the light that `view` was not implemented: https://github.com/JuliaLang/julia/pull/52049#issuecomment-1848711619 17 February 2024, 20:17:29 UTC
1998518 tecosaur 17 February 2024, 19:01:55 UTC Allow AnnotatedStrings in log messages (#51802) Permitting annotated strings allows for styling information to be preserved through to log printing. 17 February 2024, 19:01:55 UTC
4990429 Florian Atteneder 17 February 2024, 18:01:54 UTC silence warnings in tests introduced in #33593 17 February 2024, 18:01:54 UTC
93876c9 Shuhei Kadowaki 17 February 2024, 17:41:58 UTC allow external absint to hold custom data in `codeinst.inferred` (#53300) It has been possible for external abstract interpreters to keep custom data in `codeinst.inferred` (together /w overloading `inlining_policy`). After JuliaLang/julia#52233, when such external absint uses `InternalCodeCache`, this data is passed to `jl_ir_flag_inferred`, leading to segfaults in assertion builds. This commit resolves the issue by omitting `jl_ir_flag_inferred` checks when the `cache_owner` is external. Nonetheless, a better resolution might be necessary. It suggests that the current design of `code_owner` and `InternalCodeCache` for the external cache system is somewhat flawed. A conceivable approach could involve: - Adding a layer similar to `inlining_policy` in `CC.get(::WorldView{InternalCodeCache})` to enable safe redirection of custom data to the native interpreter's implementation. - Prohibiting custom data in the `inferred` field and directing such data to be kept in `analysis_results`. 17 February 2024, 17:41:58 UTC
88587ab omahs 17 February 2024, 16:10:59 UTC Fix some typos in docstrings and comments (#53368) 17 February 2024, 16:10:59 UTC
ddd7afb DilumAluthgeBot 16 February 2024, 16:36:53 UTC 🤖 [master] Bump the SparseArrays stdlib from c9f7293 to cb602d7 (#53361) Stdlib: SparseArrays URL: https://github.com/JuliaSparse/SparseArrays.jl.git Stdlib branch: main Julia branch: master Old commit: c9f7293 New commit: cb602d7 Julia version: 1.12.0-DEV SparseArrays version: 1.11.0(Does not match) Bump invoked by: @dkarrasch Powered by: [BumpStdlibs.jl](https://github.com/JuliaLang/BumpStdlibs.jl) Diff: https://github.com/JuliaSparse/SparseArrays.jl/compare/c9f7293c10e6bea775feceabd4554ab55c34e3a8...cb602d7b7cf46057ddc87d23cda2bdd168a548ac ``` $ git log --oneline c9f7293..cb602d7 cb602d7 Add generic matmatmul for inplace sparse x sparse (#486) 95575c0 Speedup for `findmin()/findmax()` on sparse arrays (#510) (#511) 4cc31f2 Fix error message in getcommon() in cholmod (#509) 1748989 Move `fkeep!` docstring to the right function (#503) 1aa6431 Update ci.yml timeout (#507) 1f88ae1 Update ci.yml - add codecov token (#504) ``` Co-authored-by: Dilum Aluthge <dilum@aluthge.com> 16 February 2024, 16:36:53 UTC
3ed2b49 Michael Abbott 16 February 2024, 14:00:02 UTC Add a lazy `logrange` function and `LogRange` type (#39071) 16 February 2024, 14:00:02 UTC
b4266e8 Kristoffer Carlsson 16 February 2024, 09:09:10 UTC use a separate depot for `--pkgimages` loading test (#53344) Otherwise this will write compile files to your home directory (and you get confused in the debug output next time you run the tests due to old precompile files being there) 16 February 2024, 09:09:10 UTC
7053e4c Kristoffer Carlsson 16 February 2024, 08:58:28 UTC set VERSION to 1.12.0-DEV (#53346) 16 February 2024, 08:58:28 UTC
aecd8fd Jishnu Bhattacharya 16 February 2024, 02:40:57 UTC Propagate inbounds in isassigned with CartesianIndex indices (#53305) Close #53302 --------- Co-authored-by: Matt Bauman <mbauman@juliahub.com> 16 February 2024, 02:40:57 UTC
4118db8 Kristoffer Carlsson 16 February 2024, 02:40:26 UTC update `--pkgimages=existing` option in CLI section of manual (#53343) 16 February 2024, 02:40:26 UTC
e460d35 Jishnu Bhattacharya 15 February 2024, 15:55:51 UTC Support `signbit(::Dates.Period)` (#51073) 15 February 2024, 15:55:51 UTC
1b25d1e Nathan Boyer 15 February 2024, 15:25:31 UTC Add CartesianIndex to getindex docstring (#51567) Co-authored-by: Matt Bauman <mbauman@gmail.com> Co-authored-by: Denis Barucic <barucic.d@gmail.com> Co-authored-by: Jishnu Bhattacharya <jishnub.github@gmail.com> Co-authored-by: Max Horn <max@quendi.de> 15 February 2024, 15:25:31 UTC
c7eb766 Shuhei Kadowaki 15 February 2024, 14:06:58 UTC minor refactoring on the `CustomAbstractInterpreterCaching` test case (#53329) By using `@newinterp`. Also ensures `cache_owner` for external abstract interpreter defined by `@newinterp` is unique always. 15 February 2024, 14:06:58 UTC
61c3521 tecosaur 15 February 2024, 10:26:34 UTC More cautious legacy_color conversion in Logging (#53280) Can be seen as a follow-on from #51829, also wants https://github.com/JuliaLang/StyledStrings.jl/pull/37 to behave as expected. 15 February 2024, 10:26:34 UTC
2197b5c Florian Atteneder 15 February 2024, 09:04:28 UTC add Base.wrap to docs 15 February 2024, 09:04:28 UTC
7fcf9c8 Rafael Fourquet 15 February 2024, 02:38:27 UTC Random: make statistical uniform tests fail less often (#52165) We have a very basic test to check that a distribution might be uniform: all the sampled elements should be unique, under certain conditions. But these conditions allowed the tests to fail too often. So allow one collision if the probability of such an event is too high. Fix #52156. --------- Co-authored-by: Kristoffer Carlsson <kcarlsson89@gmail.com> 15 February 2024, 02:38:27 UTC
2c0098e Valentin Churavy 15 February 2024, 02:22:21 UTC Cleanups for tagged code instance (#53336) Two small cleanups to #52233 - Don't use foreign code-instances for native code caching decision making - don't return a non-native codeinst for jl_method_compiled --------- Co-authored-by: Keno Fischer <keno@juliahub.com> 15 February 2024, 02:22:21 UTC
4c2633c tecosaur 15 February 2024, 02:16:43 UTC Fix results.json generation after chunk-splitting (#53335) Should fix the problem raised in https://github.com/JuliaLang/julia/pull/53323#issuecomment-1943007134. 15 February 2024, 02:16:43 UTC
254f234 tecosaur 15 February 2024, 02:16:03 UTC Merge duplicate entries in results.json (#53334) Should resolve #53198 🙂 15 February 2024, 02:16:03 UTC
c23be9a Yuto Horikawa 15 February 2024, 01:53:14 UTC Update style-guide with invalid constructor example (#53328) 15 February 2024, 01:53:14 UTC
3e6ff3d Jeff Bezanson 14 February 2024, 19:24:25 UTC add ability to specify heap size hint as a percent (#52979) also fix overflowed subtraction with size hints < 250mb --- we should pick some minimum soft heap limit; I chose 1mb here but it could be anything. 14 February 2024, 19:24:25 UTC
8eaf83c mbeltagy 14 February 2024, 18:14:07 UTC doc: clean up some example regex expressions (#26931) Remove some unnecessary parts of regex expressions used as examples in the docs. 14 February 2024, 18:14:07 UTC
a19fbac Kristoffer Carlsson 14 February 2024, 12:47:41 UTC make dependencies of extensions be found in implicit environments (#53314) Alternative to https://github.com/JuliaLang/julia/pull/53293 This does kind of the same thing but it does not rely on `EXT_PRIMED` to identify extensions which I think is a better idea since we don't really want code lookup to depend on the state of Julia itself imo. Fixes https://github.com/JuliaLang/julia/issues/53264 ----------------------- Co-authored by: Mark Kittisopikul <`markkitti@fosstodon.org`> 14 February 2024, 12:47:41 UTC
52006ae Jishnu Bhattacharya 14 February 2024, 12:13:56 UTC Move isassigned within inbounds in copyto_unaliased (#53311) This matches the other branches where the `isassigned` and ` _unsetindex!` are within `@inbounds` blocks, and this might help with performance after https://github.com/JuliaLang/julia/pull/53305 14 February 2024, 12:13:56 UTC
09a27b3 Shuhei Kadowaki 14 February 2024, 10:17:39 UTC AbsInt: add interfaces to customize cases when cached results are used (#53318) For external abstract interpreters like JET, which propagate custom data interprocedurally, it is essential to inject custom behaviors into where cached regular/constant inference results are used. Previously, this was accomplished by overloading both `abstract_call_method` and `get(::WorldView{CustomView}, ...)` (and `const_prop_call` and `cached_lookup`), that method was admittedly hacky and should probably better to be avoided. Moreover, after #52233, doing so has become infeasible when the external abstract interpreter uses `InternalCodeCache`. To address this issue, this commit provides an interface named `return_cached_result`. This allows external abstract interpreters to inject custom interprocedural data propagation during abstract interpretation even when they use `InternalCodeCache`. 14 February 2024, 10:17:39 UTC
4a13e9e Pierre Haessig 14 February 2024, 10:11:53 UTC Clarify the difference between collect and bracket notation for generators (#50619) As a follow-up to my question on the forum [Unclear documentation of `collect`? (similarity and difference with brackets)](https://discourse.julialang.org/t/unclear-documentation-of-collect-similarity-and-difference-with-brackets/), here is a tentative expansion of the docstring of `Base.collect` to clarify both the similarities and differences with the bracket notation for generators. I'm pretty sure it needs so edits, but here are the motivating factors behind the changes, by decreasing order of importance: 1. avoid the presence of bracket notation in the example without explanation (in the present docstring, I find it highly *implicit*) 2. simplification of examples to focus on the topic (and not on range syntax or filtering in comprehensions) 3. avoid wasting vertical space by using longer than needed iterators/generators Feedback on my formulation is welcome! --------- Co-authored-by: Max Horn <max@quendi.de> Co-authored-by: Jishnu Bhattacharya <jishnub.github@gmail.com> 14 February 2024, 10:11:53 UTC
aaed1cc Kristoffer Carlsson 14 February 2024, 06:43:02 UTC allow passing command line flags to `compilecache`. (#53316) Pkg right now has to start a separate process to run precompilation for the test environment which is annoying for multiple reasons Corresponding Pkg PR: https://github.com/JuliaLang/Pkg.jl/pull/3792 14 February 2024, 06:43:02 UTC
90afb7c Ian Butterworth 14 February 2024, 00:14:44 UTC split results.json into 5000 result partitions (#53323) Re. https://github.com/JuliaCI/julia-buildkite/pull/338#issuecomment-1942542712 14 February 2024, 00:14:44 UTC
288912a David Anthoff 13 February 2024, 23:16:48 UTC Add Juliaup docs (#50477) 13 February 2024, 23:16:48 UTC
ba34d89 Daniel Karrasch 13 February 2024, 15:33:06 UTC Add `conj!` for arrays of arrays (#53303) Apparently, `broadcast!(conj!, A, A)` is not allocation-free, so I spelled it out, but perhaps there's a more elegant solution. 13 February 2024, 15:33:06 UTC
d7b9ac8 Jameson Nash 13 February 2024, 15:32:15 UTC change IOBuffer to use Memory internally (#53192) An Array is often still allocated on output, but this gives the compiler a chance to potentially elide that in certain cases. For measurement, it seems about 10% faster as a string builder: ``` julia> @btime repr("hello\nworld"^10); 1.096 μs (10 allocations: 640 bytes) # master 973.000 ns (9 allocations: 608 bytes) # PR 994.000 ns (8 allocations: 576 bytes) # also PR, after Revise-ing Base.wrap ``` 13 February 2024, 15:32:15 UTC
81c3474 Jameson Nash 13 February 2024, 14:45:02 UTC Revert "More consistent return value for annotations" (#53309) Reverts JuliaLang/julia#53281 as it appears to have broken CI on all subsequent commits: https://github.com/JuliaLang/julia/pull/53281#issuecomment-1938071915 According to the CI logs, it is this test that aborts early on in the rr-trace log: ``` Pkg.test(TEST_PKG.name; coverage = coverage_path) ``` 13 February 2024, 14:45:02 UTC
d211398 Alex Arslan 13 February 2024, 12:43:21 UTC Fix FileWatching tests on FreeBSD (#53307) #49937 optimistically enabled some file watcher tests on FreeBSD due to changes made in libuv that suggested it should work, but unfortunately it does not. 13 February 2024, 12:43:21 UTC
38065a4 Aiden Grossman 13 February 2024, 04:26:46 UTC Set LLVM issue tracker URL to Julia's issue tracker (#53128) This patch sets the default issue tracker URL for LLVM to Julia's issue tracker. LLVM issues coming from Julia should probably go through Julia's issue tracker first and then can be reported upstream if there is an issue with something in upstream LLVM. Having this set to the default is moderately confusing for users as it refers to the LLVM issue tracker, which probably isn't wanted. 13 February 2024, 04:26:46 UTC
69f16a2 AllHailTheSheep 13 February 2024, 02:19:55 UTC Update logging.jl to add docstrings for exported symbols (#53299) This is a part of issue #52725. 13 February 2024, 02:19:55 UTC
54e9728 Sebastien Marie 13 February 2024, 02:18:25 UTC llvmunwind: use LLVM_PATH instead of LLVM_CONFIG_PATH variable (#53304) with llvm-16, the `--src-root` argument of llvm-config has been removed. see https://github.com/llvm/llvm-project/commit/c061892fcdbdfe46884c54a7a7bfe6df54d1df12 but `llvmunwind` was using it for configuration. so prefer to use `LLVM_PATH` and explicitly pass the llvm source directory. 13 February 2024, 02:18:25 UTC
1fd0981 Cody Tapscott 13 February 2024, 02:17:37 UTC Trim hash map size when pruning type caches (#53296) This doesn't make much of a difference for the smaller type caches (many of which have exactly 1 element actually), but it does offer up to ~75% savings (8192 -> 2048, 4096 -> 512) for some of the larger type caches in our sysimage. 13 February 2024, 02:17:37 UTC
75cb2a5 N5N3 13 February 2024, 02:16:16 UTC eliminate the dead `iterate` branch in `_unsafe_(get)setindex!`. (#52809) It's sad that compiler can't do this automatically. Some benchmark with `setindex!`: ```julia julia> a = zeros(Int, 100, 100); julia> @btime $a[:,:] = $(1:10000); 1.340 μs (0 allocations: 0 bytes) #master: 3.350 μs (0 allocations: 0 bytes) julia> @btime $a[:,:] = $(view(LinearIndices(a), 1:100, 1:100)); 10.000 μs (0 allocations: 0 bytes) #master: 11.000 μs (0 allocations: 0 bytes) ``` BTW optimization for `FastSubArray` introduced in #45371 still work after this change as the parent array might have their own `copyto!` optimization. 13 February 2024, 02:16:16 UTC
e8bf9bc Philip Tellis 13 February 2024, 02:12:56 UTC Add more @showarg examples (#52410) Adds a few more examples to show how `@showarg` works with literals and interpolated string expressions. 13 February 2024, 02:12:56 UTC
dba215f Nicholas Bauer 13 February 2024, 02:11:58 UTC Strengthen docs for `reinterpret` (#52157) As suggested in #52135, I've added some to the `reinterpret` docs. 13 February 2024, 02:11:58 UTC
e6992f7 Michael Abbott 13 February 2024, 01:50:05 UTC Add `Iterators.cycle(iter, n)` (#47354) At present `Iterators.repeated(iter, n)` is the lazy cousin of `fill(iter, n)`, but there is no iterator for `repeat(iter, n)`. This PR proposes that `Iterators.cycle(iter, n)` should fill that role. This relates to the one-argument form in the same way as `Iterators.repeated`. That is, `cycle(iter)` means `cycle(iter, Inf)` in the same way that `repeated(iter)` means `repeated(iter, Inf)`... or would be if Inf were an integer. The implementation uses `flatten(repeated(xs, n))`. It could instead use `take(cycle(xs), n * length(xs))` but that only works when the contents has known length. `take(cycle...)` tends to be faster, perhaps it should be used when possible? Some timing below. But perhaps this detail is a secondary question. <details> ```julia julia> takecycle(x, n::Int) = Iterators.take(Iterators.cycle(x), n * length(x)); julia> flatrep(x, n::Int) = Iterators.flatten(Iterators.repeated(x, n)); # as in PR, first commit julia> takecycle(1:10, 100) |> length 1000 julia> flatrep(1:10, 100) |> Base.haslength # and won't be helped by 47353 false julia> @btime collect(takecycle(1:10, 100)); min 1.642 μs, mean 2.554 μs (1 allocation, 7.94 KiB) julia> @btime collect(flatrep(1:10, 100)); min 6.617 μs, mean 9.107 μs (6 allocations, 21.86 KiB) julia> flatrep(Tuple(1:10), 100) |> Base.haslength # behaves better with tuples, but not faster: true julia> @btime collect(takecycle($(Tuple(rand(10))), 100)); min 1.100 μs, mean 1.977 μs (1 allocation, 7.94 KiB) julia> @btime collect(flatrep($(Tuple(rand(10))), 100)); min 10.458 μs, mean 11.220 μs (1 allocation, 7.94 KiB) ``` </details> 13 February 2024, 01:50:05 UTC
56e193e Stefan Karpinski 13 February 2024, 00:12:46 UTC tempdir: error on non-directory result (#33593) Potential more graceful handling of https://github.com/JuliaLang/julia/issues/33382. I thought about creating the directory if it doesn't exist, but that seems aa bit questionable. Probably better to let the user know about the situation in a clear way so they can mitigate it. An even better improvement would be if we could tell them which environment variable to look at. I tried setting `TEMP` in the environment on macOS but it didn't seem to have any effect on the result. --------- Co-authored-by: Jameson Nash <vtjnash@gmail.com> Co-authored-by: Dilum Aluthge <dilum@aluthge.com> 13 February 2024, 00:12:46 UTC
81c6526 Christian Guinard 12 February 2024, 18:44:06 UTC Fix typo in `Sys.total_memory` docstring. (#53301) Fixes #53298. 12 February 2024, 18:44:06 UTC
f8d5947 Aravindh Krishnamoorthy 12 February 2024, 17:21:15 UTC Documentation update for LinearAlgebra functions (#52934) Documentation update for LinearAlgebra functions, see #52725 - [X] Pivoting strategies - [X] `NoPivot` - [X] `RowNonZero` - [X] `ColumnNorm` - [X] `RowMaximum` - [X] Matrix multiplication functions - [X] `copy_transpose!` - [X] `copyto!` Note: `copyto!` is not mentioned in the original issue. However, it is felt that `copy_transpose!` without the complementing `copyto!` will be confusing for a reader. - [X] Exceptions - [X] `LAPACKException` - [X] `RankDeficientException` Co-authored-by: Steven G. Johnson <stevenj@mit.edu> Co-authored-by: Daniel Karrasch <daniel.karrasch@posteo.de> 12 February 2024, 17:21:15 UTC
29a58d5 Michael Abbott 12 February 2024, 14:03:37 UTC Improve docstrings for `Int`, `Float64`, `^`, and friends (#45221) This extends the docs for `Int`, `Float64` etc, to * Note that Int is the default, mention that it can overflow * Explain what `1f0` means, and that Float64 is the default Similarly extends docs for `^`, `+`, `*` aiming to * Point out that `1.2 * 10^3` is a bad habit, warn about overflow * Give an example for what `literal_pow` is doing since the explanation is quite technical * Also point out overflow in `+`, as [suggested here](https://github.com/JuliaLang/julia/pull/45141#discussion_r862808888) * While there, mention that you can add vectors, and that vararg `+(1,2,3,4)` has a default binary behaviour * Similarly mention that vararg `*(1,2,3,4)` has a default order, from the left * While there, mention `1/2pi` and `v'v` as examples of `*`. --------- Co-authored-by: Lilith Orion Hafner <lilithhafner@gmail.com> 12 February 2024, 14:03:37 UTC
61bf208 Shuhei Kadowaki 12 February 2024, 10:56:26 UTC follow up #52868 and fix up doc/src/base/base.md (#53294) 12 February 2024, 10:56:26 UTC
0a89164 Jishnu Bhattacharya 12 February 2024, 02:14:23 UTC Restrict vector indexing of views to 1-based ranges (#53249) This will be an offset array in general, so we need to restrict the method to 1-based ranges. This restores the behavior on `v"1.10"`. The method was added in #45371 12 February 2024, 02:14:23 UTC
d6396f1 Neven Sajko 12 February 2024, 02:11:28 UTC doc: in constructors, note the correspondence with callable objects (#53051) 12 February 2024, 02:11:28 UTC
987d6f3 Zentrik 12 February 2024, 01:43:52 UTC Update probes.md to remove workaround mention (#53160) Given #41616 merged, remove reference to it in doc example 12 February 2024, 01:43:52 UTC
8500356 inky 12 February 2024, 01:41:41 UTC Add more doc examples for `cat` (#47409) - `cat` 2 2D arrays - `cat` 2 3D arrays - `cat` 2 strings 12 February 2024, 01:41:41 UTC
a459926 Frames White 12 February 2024, 01:41:01 UTC Make zero on array of arrays etc apply recursively (#51458) I wonder if this breaks things, in practice. It shouldn't. Since old code behavior errored for the cases I am aware of. As discussed in #38064, this definition is needed to be consistent with our other linear algebra operations, and with us considering a vector of vectors etc as a vector space. Closes #38064 12 February 2024, 01:41:01 UTC
57f02bf Eric Hanson 12 February 2024, 01:33:29 UTC Add `Lockable` to Base, to bundle a lock with its resource (#52898) I am not sure about a `lock(f, ::Lockable)` method: it is nice because it unpacks the value for you, but it is weird because it unpacks the value for you. For a `Lockable`, `f` must accept one argument, whereas for a `Lock`, `f` must be 0-arg. A `Lockable` is not `<:AbstractLock` here, so maybe this is allowed, but if we deleted this `lock` method, we could inherit from `AbstractLock` and just use the generic one (requiring folks to unpack the value within the locked region themselves, as is the case for `@lock`). I think it is preferred these days to use `@lock` anyway, so having the `lock(f, ::Lockable)` method may be of limited value anyway. I searched Base and came up with two places that could currently use this internally, `TEMP_CLEANUP` and `env_dict`. We didn't add them both as usages yet, to avoid external breakage from delaying this PR. Similarly, this is not exported yet, to avoid breakage with older releases of ConcurrentUtilities.jl. redo of https://github.com/JuliaLang/julia/pull/34400 First commit copied from https://github.com/JuliaServices/ConcurrentUtilities.jl/blob/main/src/lockable.jl, Closes #52897 Co-authored-by: Jacob Quinn <quinn.jacobd@gmail.com> Co-authored-by: Dilum Aluthge <dilum@aluthge.com> 12 February 2024, 01:33:29 UTC
a86edf1 Michael Abbott 12 February 2024, 01:24:59 UTC Add existing 3-arg `*` docstring to the manual (#53288) PR #37898 added methods to `*` for chained matrix multiplication. They have a descriptive docstring but I don't think this is mentioned in the manual. 12 February 2024, 01:24:59 UTC
3412c86 tecosaur 12 February 2024, 01:08:47 UTC Fix seekend method of AnnotatedIOBuffer (#53282) A minor oversight had it returning the underlying IOBuffer originally. 12 February 2024, 01:08:47 UTC
e1a76bb tecosaur 12 February 2024, 00:56:30 UTC More consistent return value for `annotations` (#53281) This is a generally nice simplification, that tweaks the `annotations` API. It also requires the adjustment in https://github.com/JuliaLang/StyledStrings.jl/pull/38 to be made, hence the bump. 12 February 2024, 00:56:30 UTC
86c0191 Florian 11 February 2024, 23:25:48 UTC Tests: rename Foo test pkg (#53289) Encountered another instance of a flaky `Foo.jl` test pkg which caused me trouble locally. Xref: https://github.com/JuliaLang/julia/issues/53103#issuecomment-1915458694 11 February 2024, 23:25:48 UTC
002f07a tecosaur 11 February 2024, 22:42:31 UTC Improve the annotated join method and dispatch (#51914) With the initial implementation, join could work for AnnotatedStrings, however only when the eltype of the iterator or delim was itself a Annotated{String,Char}. This was better than nothing, but seems inconsistent in the face of mixed iterators. Furthermore, the implementation of an annotated join was far from optimised, relying on zipping and then calling annotatedstring(xs...). By contrast, the non-annotated implementation relies on printing to IO and even has manually defined alternative methods for optional arguments to minimise code generation. With the advent of AnnotatedIOBuffer and _isannotated, we can improve on both those aspects. The new AnnotatedIOBuffer type allows us to re-use the optimised join(::IO, ...) methods, and we can more reliably dispatch to them with _isannotated. Since this is a type-based decision, the Julia compiler is kind enough to work out which branch is taken at compile-time, making this zero-overhead in the unannotated case. 11 February 2024, 22:42:31 UTC
efa77cc Michael Abbott 11 February 2024, 22:33:43 UTC Predicate versions of `allequal` & `allunique` (#47679) First, use `Iterators.peel` to improve `allequal(itr)` as suggested by jlapeyre in here: https://github.com/JuliaLang/julia/issues/47005#issuecomment-1276477572 After that, `allequal(f, xs) = allequal(f(x) for x in xs)` should never call `f` twice on the same `x`, nor more times than necessary. For `allunique`, at present this has a path which collects short iterators (<32) as this is faster than making a `Set`. Thus calling `allunique(f(x) for x in xs)` will sometimes call `f` more times than it has to. This PR doesn't change that, just adds a method `allunique(f, xs)`. Since generators tend to have eltype `Any`, it pays to use `Set{@default_eltype(C)}()` now. I think this ought to be safe. Finally, methods `allequal(f, xs::Tuple)` and `allunique(f, xs::Tuple)` seem to be worthwhile. [This gist](https://gist.github.com/mcabbott/69c0b85cdf0a37d97eecba3edb18bdc0) times a few variants. Closes #47005, and does the same for `allunique` 11 February 2024, 22:33:43 UTC
c4bec9a Katharine Hyatt 11 February 2024, 18:44:25 UTC Missing tests for TwicePrecision (#34938) Co-authored-by: Tim Holy <tim.holy@gmail.com> 11 February 2024, 18:44:25 UTC
3560036 Dilum Aluthge 11 February 2024, 18:43:47 UTC Docs: add `Logging.BelowMinLevel` and `Logging.AboveMaxLevel` to the manual page for the Logging stdlib (#52232) According to the source code file, `Logging.BelowMinLevel` and `Logging.AboveMaxLevel` are already part of the public API of the Logging stdlib: https://github.com/JuliaLang/julia/blob/c8ca350832030992fca113ed56d979b6a8ff7fd3/stdlib/Logging/src/Logging.jl#L85-L91 Additionally, Logging currently exports `BelowMinLevel` and `AboveMaxLevel`: https://github.com/JuliaLang/julia/blob/c8ca350832030992fca113ed56d979b6a8ff7fd3/stdlib/Logging/src/Logging.jl#L61-L81 However, `Logging.BelowMinLevel` and `Logging.AboveMaxLevel` are not currently in the manual. This PR adds them to the manual. --------- Co-authored-by: Jameson Nash <vtjnash@gmail.com> 11 February 2024, 18:43:47 UTC
1f1436d Max Horn 11 February 2024, 18:16:40 UTC Prefer `# Examples` in docstrings (#53276) ... as suggested by the Julia manual and by `CONTRIBUTING.md` In one case `# Example` was dropped (instead of appending an `s`) because it was redundant. 11 February 2024, 18:16:40 UTC
790a4eb Neven Sajko 11 February 2024, 18:16:25 UTC document exactness for `typeintersect(T, S)` when `T <: S` (#53283) As per the discussion in #53265. 11 February 2024, 18:16:25 UTC
a37d626 Maarten Pronk 11 February 2024, 15:52:54 UTC Clarify that command-line switches need to be manually set in `addprocs` (#50843) Co-authored-by: Max Horn <max@quendi.de> 11 February 2024, 15:52:54 UTC
37ed82b Valentin Churavy 11 February 2024, 14:08:13 UTC Mark codeinstance.specsigflags as atomic (#53275) A correctness fix from #53255 This shouldn't matter much as that field is only accessed from the runtime, but this synchronizes the C-header definition, with the Julia native definition. 11 February 2024, 14:08:13 UTC
af90dac Éric Thiébaut 11 February 2024, 03:56:08 UTC Add unary negation in table of operator precedence and associativity (#49438) Co-authored-by: Jameson Nash <vtjnash@gmail.com> 11 February 2024, 03:56:08 UTC
b8540d1 Keno Fischer 11 February 2024, 03:53:23 UTC [REPL] Fix typo in using/import completion (#53273) Co-authored-by: Keno Fischer <keno@juliahub.com> 11 February 2024, 03:53:23 UTC
4be67e4 Kristoffer Carlsson 11 February 2024, 01:34:59 UTC export `IdSet` and document it (#53262) It feels kind of odd to have `IdDict` exported and documented and not `IdSet`. 11 February 2024, 01:34:59 UTC
a3e0b62 Valentin Churavy 10 February 2024, 22:35:36 UTC Use tagged CodeInstances for AbstractInterpreter caching (#52233) Currently external AbstractInterpreters all follow the pattern that `CACHE = IdDict{MethodInstance,CodeInstance}`. The NativeInterpreter has the benefit that each `MethodInstance` carries an inline cache of `CodeInstances`. `CodeInstances` pull triple duty, they contain validity information (`min_world`, `max_world`), the inference state cache and the compilation state cache. When we currently create a CodeInstance we don't record the owner and we thus construct detached CodeInstances. In this PR I introduce the notion of tagging the code instances with the owner (nothing meaning native), thus allowing foreign code instances to be stored in the inline cache. GPUCompiler/CUDA would change it's caching strategy from `IdDict{MethodInstance, CodeInstance}` to `IdDict{CodeInstance, LLVM.Module}` or similar. Ideally we would allow polymorphism for the compilation state part of the code instance, but that seems to invasive for now. This has the benefit that CodeInstances are handled by caching and it would allow us to cache inference results from foreign abstract interpreters across sessions. The downside is that the cache walk now is a bit slower since we need to filter on owner. If this turns out to be a large bottleneck we could bifurcate the cache on owner. Co-authored-by: Shuhei Kadowaki <aviatesk@gmail.com> 10 February 2024, 22:35:36 UTC
e5496e0 Jameson Nash 10 February 2024, 20:26:47 UTC update libuv to v2-1.48.0 (#49937) Notable additions: - Linux now uses io_uring for many fs calls - Windows now supports WTF-8 - new cross-platform uv_clock_gettime API - Windows now emits a coredump when sending SIGQUIT, if configured - Windows now can execute file given by an exact path name that do not contain a `.` (unlike cmd.exe, which cannot) Closes #47611 Fixes #33486 10 February 2024, 20:26:47 UTC
604609a Lars Göttgens 10 February 2024, 19:26:18 UTC Add docstring to constants in `Base.Filesystem` (#53247) A first step towards https://github.com/JuliaLang/julia/issues/52725 for `Base.Filesystem`. 10 February 2024, 19:26:18 UTC
a78b8e3 Antoine Levitt 10 February 2024, 19:17:53 UTC bypass assignments in edit, code_warntype and friends (#52046) It was suggested in the original issue that the binding should still occur, but for consistency with `@edit f(2)` which doesn't actually call `f`, this does not. Fixes #20270 10 February 2024, 19:17:53 UTC
5114441 Mosè Giordano 10 February 2024, 19:10:09 UTC [GMP_jll] Upgrade to v6.3.0 (#53270) 10 February 2024, 19:10:09 UTC
a59f219 inky 10 February 2024, 18:26:28 UTC doc: float: add note for NaN compare with egal (#53194) Current documentation seems to imply that one should use `===` for NaN comparisons. But non-normalized NaNs will give unexpected results: `NaN32p1 !== NaN32`. 10 February 2024, 18:26:28 UTC
8e70bf1 Neal Gafter 10 February 2024, 18:23:09 UTC Add documentation for some documentation forms. (#50679) There are a couple of documentation forms that are used in the wild but not covered by the documentation. Here I describe them. 10 February 2024, 18:23:09 UTC
65ffa79 Steven G. Johnson 10 February 2024, 18:11:40 UTC `names` docstring: qualify non-exported cross-refs (#52745) It seems like cross-references to non-exported symbols, like `ispublic`, should be qualified — otherwise it is confusing since looking them up fails. Co-authored-by: Max Horn <max@quendi.de> 10 February 2024, 18:11:40 UTC
35768c2 Keno Fischer 10 February 2024, 17:41:57 UTC Fix interpreter_exec.jl test (#53218) This test was supposed to check that we correctly handled PhiNodes in uninferred code in both the interpreter and the compiler. However, the compiler path wasn't actually exercised, because the `inferred=true` part of this forced it to be skipped. Drop that and fix the exposed issues in the compiler where we didn't handle PhiNodes properly. --------- Co-authored-by: Shuhei Kadowaki <aviatesk@gmail.com> Co-authored-by: Jameson Nash <vtjnash@gmail.com> Co-authored-by: Shuhei Kadowaki <40514306+aviatesk@users.noreply.github.com> 10 February 2024, 17:41:57 UTC
3a283d9 Steven G. Johnson 10 February 2024, 14:55:04 UTC document Base.Chars arguments of startswith/endswith (#52741) Since Julia 0.3 (#6779), the `startswith` and `endswith` functions have accepted not only a string as the prefix/suffix, but also a character or a tuple/vector/set of characters. However, this wasn't fully documented. (The docstring actually mentioned vectors and sets of characters, but this was inconsistent with the given signature.) Also updates `Base.Chars` to allow `AbstractSet` types, not just `Set`, which seems to have been an oversight. 10 February 2024, 14:55:04 UTC
baa36e4 Simon Byrne 10 February 2024, 14:47:20 UTC add --compiled-modules=strict option (#52174) This is useful for ensuring that all modules were precompiled successfully. 10 February 2024, 14:47:20 UTC
a333f1c Yuto Horikawa 10 February 2024, 09:44:38 UTC Update `unique` and `unique!` docstrings to mention `hash` function (#52159) 10 February 2024, 09:44:38 UTC
5d9a472 Steven G. Johnson 10 February 2024, 09:06:15 UTC additional clarification on cfunction embedding (#52315) Additional clarification to #45208 — in addition to some punctuation tweaks, I added a sentence noting that using `@cfunction` to construct and C function pointers, unlike `jl_call`, eliminates the overhead of dynamic dispatch. See also [this discourse thread](https://discourse.julialang.org/t/jl-call-function-call-latency/106791). 10 February 2024, 09:06:15 UTC
e177fc3 Lilith Orion Hafner 10 February 2024, 06:37:32 UTC Fix sortperm(::OffsetRange) (#53208) 10 February 2024, 06:37:32 UTC
8922eaa Elliot Saba 10 February 2024, 03:48:27 UTC Allow overriding default value of `JL_STACK_SIZE` (#53254) Occasionally, users may want to be able to customize the default per-task stack size. While this is a somewhat unusual thing to do, this plumbs through support for this to be set at the `Make.user` level, rather than having to modify `CFLAGS` or `src/options.h` directly. 10 February 2024, 03:48:27 UTC
bdf9830 Jishnu Bhattacharya 10 February 2024, 03:46:32 UTC Tests for isassigned for a SubArray (#53267) 10 February 2024, 03:46:32 UTC
7baaafe Jaakko Ruohio 10 February 2024, 03:45:25 UTC Truecolor terminal capability detection for Windows (#53257) ``` for i in 0:255 print("\x1b[38;2;$i;0;0m█") end ``` shows on Windows Terminal ![image](https://github.com/JuliaLang/julia/assets/7318249/d742b739-9645-4daa-8572-d4c28803013e) so it does have true-color support. With this PR, `Base.ttyhastruecolor()` returns `true` on Windows Terminal and Cygwin64 Terminal, those are the ones I tried out. Notes: https://devblogs.microsoft.com/commandline/24-bit-color-in-the-windows-console/ says > in Windows 10 Insiders Build # 14931, we’ve updated the Windows Console to support full, glorious 24-bit RGB true color! 10 February 2024, 03:45:25 UTC
3952e78 Jeremie Knuesel 10 February 2024, 03:12:10 UTC LU factorization: add allowsingular keyword argument (#52957) Currently the `check` argument of `lu` conflates failed factorization (invalid factors) and valid factorizations that have a rank-deficient factor, and the `show` value for the factorization is misleading in the latter case. This PR adds an `allowsingular` keyword argument to `lu` and the corresponding `issuccess` method to treat valid but rank-deficient factors as success. It also changes `show` to show the factors as usual in such cases, but with a "rank-deficient" note. The invalid vs rank-defficient factors are distinguished using negative `info` values in the rank-deficient case, as proposed in #27657. Fixes #27657 Co-authored-by: Daniel Karrasch <daniel.karrasch@posteo.de> Co-authored-by: Jishnu Bhattacharya <jishnub.github@gmail.com> 10 February 2024, 03:12:10 UTC
250916f tecosaur 10 February 2024, 03:09:31 UTC Use StyledStrings for Logging (#51829) Transition from printstyled to the new approach to styling provided by StyledStrings. This both makes it possible for the styling used to be customised, and allows for other styled content to inherit/re-use the logging styles. 10 February 2024, 03:09:31 UTC
9284e57 Nicholas Bauer 10 February 2024, 03:08:04 UTC Add `Vector{<:AbstractString}` method for `Cmd()` (#49531) Was annoyed that `Cmd` wouldn't take a vector containing SubStrings. So I here add a constructor that converts an `AbstractString` vector to `String`s. 10 February 2024, 03:08:04 UTC
5547305 Jameson Nash 10 February 2024, 03:03:26 UTC Fix duplicate error when using generator in Dict (#53151) Fixes: #33147 Replaces/Closes: #40445 The difference here, compared to past implementations, is that we use the zero-cost `isiterable` check on every intermediate step, instead of wrapping the call in a try/catch and then trying to re-approximate the `isiterable` afterwards. Some samples: ```julia julia> Dict(i for i in 1:3) ERROR: ArgumentError: AbstractDict(kv): kv needs to be an iterator of 2-tuples or pairs Stacktrace: [1] _throw_dict_kv_error() @ Base ./dict.jl:118 [2] grow_to! @ ./dict.jl:132 [inlined] [3] dict_with_eltype @ ./abstractdict.jl:592 [inlined] [4] Dict(kv::Base.Generator{UnitRange{Int64}, typeof(identity)}) @ Base ./dict.jl:120 [5] top-level scope @ REPL[1]:1 julia> Dict(i => error("$i") for i in 1:3) ERROR: 1 Stacktrace: [1] error(s::String) @ Base ./error.jl:35 [2] (::var"#3#4")(i::Int64) @ Main ./none:0 [3] iterate @ ./generator.jl:48 [inlined] [4] grow_to! @ ./dict.jl:124 [inlined] [5] dict_with_eltype @ ./abstractdict.jl:592 [inlined] [6] Dict(kv::Base.Generator{UnitRange{Int64}, var"#3#4"}) @ Base ./dict.jl:120 [7] top-level scope @ REPL[2]:1 ``` The other unrelated change here is that `dest = empty(dest, typeof(k), typeof(v))` is made conditional, so we do not unconditionally construct an empty Dict in order to discard it and allocate an exact duplicate of it, but only do so if inference wasn't precise originally. Co-authored-by: Curtis Vogt <curtis.vogt@gmail.com> 10 February 2024, 03:03:26 UTC
b43edb7 Jameson Nash 10 February 2024, 03:02:29 UTC Updated MethodError to show closest candidates more reliably (#53165) Updated version of #33793. Always show up to 3 methods, even if no arguments types match on some of them, but rank ones with fewer arguments before those with more arguments. Closes #33793 Fixes #33793 Fixes #46236 Co-authored-by: Eric Wright <efwright@udel.edu> (this diff best viewed with whitespace ignored) 10 February 2024, 03:02:29 UTC
back to top