| 73b5249 | Keno Fischer | 21 September 2018, 19:39:56 UTC | Allow eliminating nothrow intrinsics during optimization | 21 September 2018, 19:50:44 UTC |
| bd5bb17 | Keno Fischer | 21 September 2018, 17:39:17 UTC | Inlining thing | 21 September 2018, 17:39:17 UTC |
| 81436ee | Keno Fischer | 18 September 2018, 21:50:33 UTC | WIP: Extend code_typed to be able to debug constant prop I'm currently debugging code that uses a lot of arrays with dimensions in their type parameters. This type of code heavily relies on constant propagation to lift dimensions from the value domain into the type domain. Unfortunately, it's a bit hard to discover what exactly causes inference to drop information from constants, because there's no way to feed in constants for a particular invocation using code_typed. This is a quick hack to remidy that, by making `$`-interpolated expressions available as constants to type inference, e.g. ```julia julia> @code_typed 1+1 CodeInfo( 53 1 ─ %1 = (Base.add_int)(x, y)::Int64 └── return %1 ) => Int64 julia> @code_typed $(1)+$(1) CodeInfo( 53 1 ─ return 2 ) => Int64 ``` | 21 September 2018, 16:10:50 UTC |
| f07eb96 | Keno Fischer | 21 September 2018, 16:09:46 UTC | Recursion fixup | 21 September 2018, 16:10:08 UTC |
| be397e5 | Keno Fischer | 21 September 2018, 03:49:26 UTC | Make inference more aggressive | 21 September 2018, 16:10:07 UTC |
| 0bf816a | Keno Fischer | 21 September 2018, 03:45:06 UTC | RFC: Allow structural recursion without triggering edge cycle limiting This attempts to fix inference for the case in #29293 (the one returning `Any`). It does not fix the cache poisoning part of that issue, which is a separate concern. The idea here is that we avoid applying limiting if the argtypes of the frame become strictly simpler (thus guaranteeing eventual termination). It is important that the complexity relation be transitive and anti-reflexive. | 21 September 2018, 16:10:07 UTC |
| 647d809 | Keno Fischer | 18 September 2018, 23:12:45 UTC | Improve inference of `tail`-like functions Previously these sorts of function would block constant propagation. Hopfully #28955 will just fix this, but until then, add a surgical fix and a test. | 21 September 2018, 16:10:07 UTC |
| 6f81da6 | Keno Fischer | 18 September 2018, 22:00:52 UTC | WIP | 18 September 2018, 23:17:40 UTC |
| f9f9cf3 | Keno Fischer | 15 September 2018, 19:13:09 UTC | Add a method of run_passes that skips SSA conversion | 18 September 2018, 23:17:40 UTC |
| 8c4d61e | Keno Fischer | 15 September 2018, 19:12:47 UTC | Align StmtRange field names with UnitRange | 18 September 2018, 23:17:40 UTC |
| 8dd3326 | Thomas Christensen | 18 September 2018, 17:35:07 UTC | ComplexF64 division: combine four if-statements into two if-elseif-statements (#29042) * ComplexF64 division: combine four if-statements into two if-elseif-statements * add @fastmath to magnitude check; NaN handling not needed * don't use @fastmath; use explicit route instead * remove unnecessary two & half variables | 18 September 2018, 17:35:07 UTC |
| abe38f1 | Carl Gay | 18 September 2018, 15:00:13 UTC | Base.TCPServer --> Sockets.TCPServer (#29230) | 18 September 2018, 15:00:13 UTC |
| 3a15e6b | Valentin Churavy | 18 September 2018, 01:19:03 UTC | Merge pull request #29233 from aghitza/patch-2 fix awkward phrasing in parallel-computing.md | 18 September 2018, 01:19:03 UTC |
| 157a6d1 | Alex Ghitza | 17 September 2018, 21:55:35 UTC | fix awkward phrasing in parallel-computing.md | 17 September 2018, 21:55:35 UTC |
| 88f74b7 | Stefan Karpinski | 17 September 2018, 20:28:25 UTC | fix reinterpret(Char, ::UInt32) for "unnatural" values (fix #29181) (#29192) This code was assuming that character values only have bit-patterns that decoding a string can produce, but of course `reinterpret` can produce any bit pattern in a `Char` whatsoever. The fix doesn't use that assumption and only uses the cache for actual ASCII characters. | 17 September 2018, 20:28:25 UTC |
| 1324ceb | Peter Ahrens | 17 September 2018, 17:12:45 UTC | Propagate constant calls to new! (#28284) * this is the thing that infers new. * more stuff! * Incorporating Jeff's suggestions. * Added tests and used Jeff's implementation. | 17 September 2018, 17:12:45 UTC |
| 2c15430 | Alex Ames | 17 September 2018, 16:19:10 UTC | Make `using LinearAlgebra` explicit (#29188) Ref. https://discourse.julialang.org/t/error-undefvarerror-eigvecs-not-defined/14923/2?u=stillyslalom | 17 September 2018, 16:19:10 UTC |
| 3255f28 | Jeff Bezanson | 17 September 2018, 15:05:47 UTC | fix #29175, invalid lowered IR from repeating code for declared types (#29194) | 17 September 2018, 15:05:47 UTC |
| c0afddf | Jeff Bezanson | 17 September 2018, 15:03:43 UTC | fix #28481, faster reading of primitive types from IOStream and IOBuffer (#29186) | 17 September 2018, 15:03:43 UTC |
| 632ca40 | Jeff Bezanson | 17 September 2018, 15:01:28 UTC | rm an unused function and fix jl_eof_error (#29185) | 17 September 2018, 15:01:28 UTC |
| 16516b5 | Stefan Karpinski | 17 September 2018, 12:51:29 UTC | fix digits(n::Unsigned) with neg base for `n > typemax(n)÷2` (#29205) Alternative based on https://github.com/JuliaLang/julia/pull/29187 Tests from rforquet's PR linked above. | 17 September 2018, 12:51:29 UTC |
| 7d6bc6b | Keno Fischer | 13 September 2018, 02:16:01 UTC | Fix patch | 17 September 2018, 01:57:55 UTC |
| ef8bbd3 | Valentin Churavy | 12 September 2018, 19:30:21 UTC | bump to LLVM_BB rel 6.0.1-3 | 17 September 2018, 01:57:55 UTC |
| 34284f6 | Peter Ahrens | 29 July 2018, 00:09:10 UTC | added a test for codegen on quoted constant structs. | 17 September 2018, 01:57:55 UTC |
| 366e2d6 | Keno Fischer | 09 September 2018, 21:17:55 UTC | Carry LLVM patch implementing byval attribute on Win64 | 17 September 2018, 01:57:55 UTC |
| 7507511 | Jon Malmaud | 16 September 2018, 14:51:31 UTC | Have 'cat' only accept positive dimensions. (#29189) | 16 September 2018, 14:51:31 UTC |
| ec44cf4 | Rafael Fourquet | 16 September 2018, 10:17:00 UTC | optimize slightly BigFloat(-NaN) (#29128) This operation was allocating internally one unnecessary temporary BigFloat object, doubling the construction time. | 16 September 2018, 10:17:00 UTC |
| 5dda994 | tomtom | 16 September 2018, 08:41:07 UTC | correct the docs of `Nothing` (#29210) | 16 September 2018, 08:41:07 UTC |
| e2c9878 | Hamza Yusuf Çakır | 15 September 2018, 12:35:30 UTC | Add `undef` to Vector allocations (#29184) Without the `undef` initializer, the corresponding samples give error for julia version >= 1.0. | 15 September 2018, 12:35:30 UTC |
| 02aa9bb | Katharine Hyatt | 14 September 2018, 21:07:50 UTC | Xrefs and a few examples for math stuff (#29150) | 14 September 2018, 21:07:50 UTC |
| cb269aa | Jeff Bezanson | 14 September 2018, 19:54:01 UTC | fix #25474, support more forms in edit, which, etc. macros (#29159) | 14 September 2018, 19:54:01 UTC |
| c3660f7 | Jeff Bezanson | 14 September 2018, 19:53:06 UTC | fix #29145, error for `new{}` with incomplete type (#29154) | 14 September 2018, 19:53:06 UTC |
| 3b02991 | Stefan Karpinski | 12 September 2018, 18:25:02 UTC | define ncodeunits(c::Char) as fast equivalent of ncodeunits(string(c)) There was a non-public `codelen(c::Char)` method which previously did this. This also replaces internal uses of this with `ncodeunits(c)`. | 14 September 2018, 15:11:29 UTC |
| fc04d73 | Stefan Karpinski | 12 September 2018, 18:23:30 UTC | base/char.jl: tweak doc strings | 14 September 2018, 15:11:29 UTC |
| b9c7a72 | Rafael Fourquet | 14 September 2018, 13:27:26 UTC | BigFloat tests: consolidate two files into one (#29144) merge test/bigfloat.jl into test/mpfr.jl | 14 September 2018, 13:27:26 UTC |
| cda41ae | cormullion | 14 September 2018, 13:22:39 UTC | fixed some typos (#29164) | 14 September 2018, 13:22:39 UTC |
| 26b6a58 | Matt Bauman | 13 September 2018, 21:44:51 UTC | Improve support for constructing zero-step float ranges (#29056) * Improve support for constructing zero-step float ranges Fixes `0.0 * (1:4)`, for example. From https://github.com/JuliaLang/julia/issues/29052#issuecomment-418825887 * simplify | 13 September 2018, 21:44:51 UTC |
| 77ec1ec | Rafael Fourquet | 13 September 2018, 19:31:58 UTC | fix unsigned ndigits with neg base for `n > typemax(n)÷2` (#29148) It relies on the fact that `cld(x, b) == -fld(x, -b)` but does the conversion from unsigned to signed before negating the unsigned quotient; since `-b ≥ 2` the quotient always fits in the signed type. | 13 September 2018, 19:31:58 UTC |
| fd0ce5d | Jeff Bezanson | 13 September 2018, 15:24:52 UTC | Merge pull request #29156 from JuliaLang/jb/10xtests add tests for a couple fixed issues | 13 September 2018, 15:24:52 UTC |
| 3143d89 | Jeff Bezanson | 13 September 2018, 15:23:53 UTC | fix #22592, specificity of union compared to a subtype of it (#29139) We considered Union{A,B} more specific than B if A was more specific than B (but not a subtype of it). Clearly, it should not be. | 13 September 2018, 15:23:53 UTC |
| d974f93 | Keno Fischer | 13 September 2018, 15:04:18 UTC | See if building our own PCRE fixes the travis issues (#29162) | 13 September 2018, 15:04:18 UTC |
| e724864 | Bogumił Kamiński | 13 September 2018, 12:46:39 UTC | fix order of operations in ranges to avoid overflow (#28893) | 13 September 2018, 12:46:39 UTC |
| b4c370d | Klaus Crusius | 12 September 2018, 21:35:11 UTC | Allow `Iterators.flatten` for empty tuple (#29112) | 12 September 2018, 21:35:11 UTC |
| 62de472 | Bogumił Kamiński | 12 September 2018, 21:27:56 UTC | move codelen and first_utf8_byte to Char.jl (#28894) | 12 September 2018, 21:27:56 UTC |
| 5e4ca86 | Stephan Hilb | 12 September 2018, 20:41:56 UTC | doc: fix usage of NTuple{T} (#29104) | 12 September 2018, 20:41:56 UTC |
| c2b61df | Katharine Hyatt | 12 September 2018, 19:35:03 UTC | Some xrefs and examples for Distiributed (#29137) | 12 September 2018, 19:35:03 UTC |
| dc4798f | Jeff Bezanson | 12 September 2018, 19:11:53 UTC | add test case from #26037 closes #26037 | 12 September 2018, 19:11:53 UTC |
| 2be7e09 | Jeff Bezanson | 12 September 2018, 19:09:33 UTC | add test case from #26729 Fixed by the new optimizer. This closes #26729. | 12 September 2018, 19:09:33 UTC |
| e918a92 | Kristoffer Carlsson | 12 September 2018, 15:53:48 UTC | allow building julia without extra precompile statements (#29132) | 12 September 2018, 15:53:48 UTC |
| 592a474 | Mark | 12 September 2018, 14:43:51 UTC | Tried to fix the English of the first few paras (#29050) I don't have time to do this whole document, but it could do with an edit by a native English speaker. | 12 September 2018, 14:43:51 UTC |
| 6acaa10 | Kristoffer Carlsson | 12 September 2018, 14:43:32 UTC | fix perf in exp(::Matrix) (#29116) | 12 September 2018, 14:43:32 UTC |
| 2cd1bf8 | Kristoffer Carlsson | 12 September 2018, 14:38:11 UTC | fix perf regression from not specializing on iterate on tuples (#29133) | 12 September 2018, 14:38:11 UTC |
| 3608e50 | Katharine Hyatt | 12 September 2018, 14:34:02 UTC | Fix some signatures and add examples for Random docs (#29119) | 12 September 2018, 14:34:02 UTC |
| 773540d | laborg | 11 September 2018, 21:21:09 UTC | RFC: Iterate over smaller set for setdiff[!](a,b) (#29048) * Iterate over smaller set for setdiff[!] * Add comment and change to integer multiplication | 11 September 2018, 21:21:09 UTC |
| d8809f1 | Stefan Karpinski | 11 September 2018, 19:36:25 UTC | LibGit2: only call shutdown once [fix #28306] (#29121) | 11 September 2018, 19:36:25 UTC |
| 7142712 | Stephan Hilb | 11 September 2018, 14:40:27 UTC | base/sort: add sort! for multidimensional arrays (#28902) | 11 September 2018, 14:40:27 UTC |
| 70c7298 | Morten Piibeleht | 11 September 2018, 14:05:27 UTC | current_project(): don't bail at Git repos (#29108) | 11 September 2018, 14:05:27 UTC |
| e85af88 | André Lage | 11 September 2018, 07:08:04 UTC | Solving Issue #29101 (#29122) See Issue #29101 | 11 September 2018, 07:08:04 UTC |
| 448bc7d | Katharine Hyatt | 11 September 2018, 06:22:06 UTC | Some more examples/formatting/xrefs for Distributed docs (#29011) | 11 September 2018, 06:22:06 UTC |
| 6caabc9 | Stefan Karpinski | 10 September 2018, 22:08:27 UTC | Revert "RFC: prettier IR-show for line number and inlining information" (#29118) | 10 September 2018, 22:08:27 UTC |
| 3b428c9 | Milan Bouchet-Valat | 10 September 2018, 19:06:57 UTC | Fix libunwind version check with patch releases (#29082) libunwind patch releases like 1.2.1 have UNW_VERSION_MINOR = 2.1, which makes the comparison fail. Use a more robust comparison which excludes all versions between 1.0 and 1.2 instead. https://savannah.nongnu.org/bugs/index.php?52529 | 10 September 2018, 19:06:57 UTC |
| f1ce96e | Iblis Lin | 10 September 2018, 19:03:23 UTC | cleanup testing leftovers (#29069) | 10 September 2018, 19:03:23 UTC |
| 513db98 | Tim Holy | 10 September 2018, 18:37:58 UTC | Check world age bounds on candidate ambiguities. Fixes #28899 (#28916) | 10 September 2018, 18:37:58 UTC |
| 3c8119f | Spencer Russell | 10 September 2018, 17:02:09 UTC | Adds some clarification to `download` (#28989) | 10 September 2018, 17:02:09 UTC |
| 999bcc5 | Jameson Nash | 10 September 2018, 13:44:05 UTC | Merge pull request #28390 from JuliaLang/jn/irshow3 prettier IR-show for line number and inlining information | 10 September 2018, 13:44:05 UTC |
| 97864d3 | Niklas Gustafsson | 09 September 2018, 19:05:38 UTC | Adding text on how to embed Julia on Windows. (#28974) * Adding text on how to embed Julia on Windows. * Fixed trailing whitespace. | 09 September 2018, 19:05:38 UTC |
| ecff881 | Justin Willmert | 08 September 2018, 17:46:21 UTC | Avoid duplicated =value in Enum errors. | 09 September 2018, 02:49:26 UTC |
| 1ab5319 | Keno Fischer | 08 September 2018, 17:38:59 UTC | Allow eliminating ifelse if proven nothrow | 09 September 2018, 00:11:40 UTC |
| 68db9b4 | Malcolm Parsons | 08 September 2018, 21:10:15 UTC | This comma confused me. Remove it | 09 September 2018, 00:11:29 UTC |
| 0c73ad1 | Iblis Lin | 08 September 2018, 23:01:18 UTC | doc/interface: correct signature of `size` (#29094) | 08 September 2018, 23:01:18 UTC |
| fda953b | Keno Fischer | 01 September 2018, 22:41:25 UTC | Kill useless phi nodes | 08 September 2018, 20:11:08 UTC |
| 3ed9599 | Keno Fischer | 01 September 2018, 21:28:15 UTC | Eliminate useless PiNodes during compaction | 08 September 2018, 20:11:08 UTC |
| 92cef07 | Keno Fischer | 30 August 2018, 18:58:10 UTC | Allow CFG transforms during compaction The design taken here is that CFG transformations are allowed during compacting, but BBs are only removed (i.e. BB numbers are only changed at the beginning of compaction). | 08 September 2018, 20:11:08 UTC |
| 3057ebc | Jeff Bezanson | 07 September 2018, 03:37:03 UTC | slightly more efficient lowering of destructuring assignment There is no need to fetch the next state after the last element. | 08 September 2018, 19:47:23 UTC |
| f922a63 | Jeff Bezanson | 07 September 2018, 03:32:38 UTC | fix #29036, poor inference of `val,i = iterate(x,i)` In this case, the result of `iterate` has not been checked for `nothing`, so we try to call `indexed_iterate` (for destructuring assignment) on a Union of Nothing and the tuple returned by `iterate`. That has two method matches, and so was excluded from constant propagation. This commit fixes that by generalizing the constant prop heuristic from requiring one method match to requiring one non-Bottom method match. This issue caused a large slowdown in DelimitedFiles, where the inner loop consists of ``` while idx <= slen val,idx = iterate(dbuff, idx) ``` | 08 September 2018, 19:47:23 UTC |
| 65e4c13 | Jeff Bezanson | 07 September 2018, 17:16:55 UTC | fix #29083, not eliminating kwfunc due to logic error in builtin_nothrow | 08 September 2018, 15:25:12 UTC |
| 42e8920 | Jeff Bezanson | 07 September 2018, 16:45:00 UTC | fix logic in `getfield_nothrow` to allow removing more getfields | 08 September 2018, 15:24:24 UTC |
| a572504 | dehann | 08 September 2018, 03:48:41 UTC | Documentation `mynorm` should not divide by n (#29091) Test with: ```julia a = randn(10) abs(norm(a) - mynorm(a)) < 1e10 ``` Tested here: ```julia-repl _ _ _(_)_ | A fresh approach to technical computing (_) | (_) (_) | Documentation: https://docs.julialang.org _ _ _| |_ __ _ | Type "?help" for help. | | | | | | |/ _` | | | | |_| | | | (_| | | Version 0.6.4 (2018-07-09 19:09 UTC) _/ |\__'_|_|_|\__'_| | Official http://julialang.org/ release |__/ | x86_64-pc-linux-gnu julia> function mynorm(u::Vector) n = length(u) T = eltype(u) s = zero(T) @fastmath @inbounds @simd for i in 1:n s += u[i]^2 end @fastmath @inbounds return sqrt(s) end mynorm (generic function with 1 method) julia> function mynorm_old(u::Vector) n = length(u) T = eltype(u) s = zero(T) @fastmath @inbounds @simd for i in 1:n s += u[i]^2 end @fastmath @inbounds return sqrt(s/n) end mynorm_old (generic function with 1 method) julia> a = randn(10) 10-element Array{Float64,1}: -0.564163 -3.45236 -0.50901 -0.030296 0.114156 0.661961 -1.19972 2.20883 1.74529 0.82515 julia> abs(norm(a)-mynorm(a)) 0.0 julia> abs(norm(a)-mynorm_old(a)) 3.2787896367314344 ``` | 08 September 2018, 03:48:41 UTC |
| bb7d043 | Jeff Bezanson | 07 September 2018, 21:18:46 UTC | fix #25955, hygiene of arg name of function defined by type (`(f::T)(...)`) (#29072) | 07 September 2018, 21:18:46 UTC |
| cfca833 | Jeff Bezanson | 07 September 2018, 18:25:26 UTC | fix #29064, bug in field name same as struct type name (#29070) | 07 September 2018, 18:25:26 UTC |
| 4b103ef | Jeff Bezanson | 07 September 2018, 17:39:14 UTC | stronger warnings about changing constants in help and docs (#28711) [ci skip] | 07 September 2018, 17:39:14 UTC |
| 5c1d4a3 | Kristoffer Carlsson | 07 September 2018, 13:55:09 UTC | comment out assertion | 07 September 2018, 13:55:09 UTC |
| bda5dd8 | Bogumił Kamiński | 06 September 2018, 19:26:41 UTC | add local to outer description (#29010) | 06 September 2018, 19:26:41 UTC |
| 88d536a | Kristoffer Carlsson | 06 September 2018, 18:03:05 UTC | work around a splatting penalty in twiceprecision (#29060) * work around a splatting penalty in twiceprecision * add allocation test | 06 September 2018, 18:03:05 UTC |
| e5cc750 | Jeff Bezanson | 06 September 2018, 16:04:54 UTC | fix #28833, crash in macros returning certain top-level forms (#28882) | 06 September 2018, 16:04:54 UTC |
| a42f4d5 | Kenta Sato | 06 September 2018, 15:33:00 UTC | fix docstring of open(command) (close #24649) (#25563) | 06 September 2018, 15:33:00 UTC |
| be14985 | Harry | 06 September 2018, 14:39:53 UTC | Fix typo in mapreduce docstring (#29049) | 06 September 2018, 14:39:53 UTC |
| 53f8e18 | zy | 06 September 2018, 14:23:18 UTC | Fix typo on file.jl (#29063) cp() doc : dest->dst | 06 September 2018, 14:23:18 UTC |
| 4c02077 | Kristoffer Carlsson | 06 September 2018, 01:19:52 UTC | make sure that the read task is done when precompiling (#29024) | 06 September 2018, 01:19:52 UTC |
| 11e32a4 | Jeff Bezanson | 05 September 2018, 21:50:34 UTC | fix #28991, error when adding default defs to constructed module Exprs (#29044) The code for this expected the first element of a block to be a line number; check to make sure it is before using it. | 05 September 2018, 21:50:34 UTC |
| 1fb86d2 | Jeff Bezanson | 05 September 2018, 21:49:57 UTC | fix #28900, macro hygiene bug inside keyword arg values (#29043) | 05 September 2018, 21:49:57 UTC |
| 8d99356 | Patrick Belliveau | 05 September 2018, 10:09:00 UTC | Fix dispatch of SparseMatrixCSC*Diagonal multiplication (#29045) * Fix type signature of mul! methods for multiplying SparseMatrixCSCs with Diagonal matrices. Type signature for diagonal matrices was wrong, causing fallback to generic Matmul. * Add SparseMatrixCSC*Diagonal dispatch test * Fix trailing whitespace * Don't copy with deepcopy | 05 September 2018, 10:09:00 UTC |
| 807ae71 | Miguel Madrid Mencía | 05 September 2018, 09:25:19 UTC | Fix: Enum values doesn't display UInt8 correctly (#29035) | 05 September 2018, 09:25:19 UTC |
| 3fbdc9b | Alex Arslan | 04 September 2018, 20:51:05 UTC | Merge pull request #28822 from JuliaLang/kc/pkg_bomb Update Pkg | 04 September 2018, 20:51:05 UTC |
| c43793c | ExpandingMan | 04 September 2018, 20:09:39 UTC | fixed regression in sparse of adjoint of sparse (#28948) (#28954) * fixed regression in sparse of adjoint | 04 September 2018, 20:09:39 UTC |
| d4cdde1 | Jeff Bezanson | 04 September 2018, 18:43:43 UTC | fix #28915, identity conversion to union of tuple types (#28951) | 04 September 2018, 18:43:43 UTC |
| 14fb104 | wfrgra | 04 September 2018, 18:15:02 UTC | Fix download agent search relying on throwing of Sys.which(). Update t #28157 (#28682) * Fix download agent search relying on throwing of Sys.which() | 04 September 2018, 18:15:02 UTC |
| 714b2c3 | Kristoffer Carlsson | 04 September 2018, 13:20:34 UTC | use default proxy credentials when downloading via powershell (cherry picked from commit d495fed46606ff29685135bc63bffb5f7cc6d463) | 04 September 2018, 17:29:17 UTC |
| e7d7259 | Keno Fischer | 03 September 2018, 00:13:48 UTC | Fix out of bounds write in array deserialization The array was allocated based on the serialized `elsize` of the array, however, unions get an extra selector array after the regular storage which was not allocated (because we didn't know it was gonna be a union array at the time when we allocated it). According to a48eeefd6207090898f5b1877684aa10f217800a we cannot look at the element type to allocate the array, so we need to serialize a bit to indicate that we will have a union array. Fixes #28998 | 04 September 2018, 16:55:32 UTC |
| c8450d8 | Pablo San-Jose | 04 September 2018, 10:24:29 UTC | CartesianIndices and LinearIndices docstring (#29023) | 04 September 2018, 10:24:29 UTC |
