Make more types jl_static_show readably #58512

xal-0 · 2025-05-23T19:39:47Z

Makes more types survive jl_static_show:

Symbols
- Symbols printed in the :var"foo" form use raw string escaping, fixing :var"a\b", :var"a\\", :var"$a", etc.
- Symbols that require parens use parens (:(=), ...)
Signed integers: Except for Int, signed integers print like Int8(1).
Floats: floats are printed in a naive but reversible (TODO: double check) way. Inf(16|32|) and NaN(16|32|) are printed, and Float16/Float32 print the type (Float32(1.5)). Float64s are printed with a trailing .0 if it is necessary to disambiguate from Int.

Fixes #52677, #58484 (comment), #58484 (comment), and the specific case mentioned in #58484. Improves the situation for #38902 but does not close it, because a few cases still do not round-trip (inexhaustive list):

Non-canonical NaNs
BFloat16
User-defined primitive types. This one is tricky, because they can have a size different from any type we have literals for.
Without don't print field names in static_show of structs #40652 or something similar, precompile statements will still fail with field names.

KristofferC · 2025-05-23T19:58:15Z

For #40652, can we thread something through to not show the field name in precompile printing contexts? Seems unnecessary to have the correctness of the printing be bike shredded on how people want it to print in the debugger.

nickrobinson251 · 2025-05-23T20:11:12Z

Fixes ... #58484 (comment),

Please can you add tests for Val(Inf) (and maybe Val(-Inf) and Inf and -Inf)?

src/rtutils.c

topolarity · 2025-05-27T18:37:38Z

It might be worth printing these floating-point numbers as hexadecimal literals (specifically in a type context):

julia> 0x1p0
1.0

julia> 0x1.8p3
12.0

julia> x = 0x.4p-1
0.125

These can be easy to parse / print bit-accurately, if we have trouble finding floating point <-> decimal conversion routines that we trust to be digit-/bit-accurate.

Of course, if we trust the decimal conversion on all platforms, that's just as good and more human-readable.

n backslashes followed by " => 2n+1 backslashes and " n backslahes at the end => 2n backslashes

xal-0 · 2025-05-28T18:00:57Z

Unfortunately we have no hex Float32 literals, and I would like to preserve readability. That's also why I prefer %.17g (with trailing .0 when required) over %.17e: 123.0 is nicer than 1.23000000000000000e+02, 0.5 vs 5.00000000000000000e-01, etc. It's not as nice as Ryu (we still print 0.10000000000000001 instead of 0.1), but it's the best we can do with libc.

src/rtutils.c

Co-authored-by: Jameson Nash <vtjnash@gmail.com>

topolarity · 2025-05-28T18:32:12Z

Unfortunately we have no hex Float32 literals

No, but it losslessly casts to Float32 in the same way that Int64 -> Int32 does

That said, I'm all in favor of the readability

topolarity · 2025-05-29T14:40:33Z

src/rtutils.c

 #ifdef _P64
        n += jl_printf(out, "0x%016" PRIx64, *(uint64_t*)v);
 #else
        n += jl_printf(out, "0x%08" PRIx32, *(uint32_t*)v);


Can you fix-up Pointer types as well?

julia> ccall(:jl_, Cvoid, (Any,), Val(Ptr{Cvoid}(1))) Base.Val{0x0000000000000001}()

It's a little awkward, because their repr doesn't round-trip either:

julia> Val(Ptr{Cvoid}(1)) Val{Ptr{Nothing} @0x0000000000000001}() julia> Val{Ptr{Nothing} @0x0000000000000001}() ERROR: ParseError: # Error @ REPL[10]:1:18 Val{Ptr{Nothing} @0x0000000000000001}() # └─────────────────┘ ── Expected `}`

but regardless of whether it parses properly, I don't think we should be throwing out the type information

We should also consider fixing the jl_tvar_type case also. Right now it only prints the name after checking that the parent UnionAll is present, but we should also check that that name is unique, and if not, to find some way to make it unique in printing.

Small follow-up to JuliaLang#58512

Makes more types survive `jl_static_show` unambiguously: - Symbols - Symbols printed in the `:var"foo"` form use raw string escaping, fixing `:var"a\b"`, `:var"a\\"`, `:var"$a"`, etc. - Symbols that require parens use parens (`:(=)`, ...) - Signed integers: Except for `Int`, signed integers print like `Int8(1)`. - Floats: floats are printed in a naive but reversible (TODO: double check) way. `Inf(16|32|)` and `NaN(16|32|)` are printed, and `Float16`/`Float32` print the type (`Float32(1.5)`). `Float64`s are printed with a trailing `.0` if it is necessary to disambiguate from `Int`. Fixes #52677, #58484 (comment), #58484 (comment), and the specific case mentioned in #58484. Improves the situation for round-trip (inexhaustive list): - Non-canonical NaNs - BFloat16 - User-defined primitive types. This one is tricky, because they can have a size different from any type we have literals for. (cherry picked from commit b03ef6b)

Small follow-up to #58512 (cherry picked from commit 36bd3ad)

Makes more types survive `jl_static_show` unambiguously: - Symbols - Symbols printed in the `:var"foo"` form use raw string escaping, fixing `:var"a\b"`, `:var"a\\"`, `:var"$a"`, etc. - Symbols that require parens use parens (`:(=)`, ...) - Signed integers: Except for `Int`, signed integers print like `Int8(1)`. - Floats: floats are printed in a naive but reversible (TODO: double check) way. `Inf(16|32|)` and `NaN(16|32|)` are printed, and `Float16`/`Float32` print the type (`Float32(1.5)`). `Float64`s are printed with a trailing `.0` if it is necessary to disambiguate from `Int`. Fixes #52677, #58484 (comment), #58484 (comment), and the specific case mentioned in #58484. Improves the situation for round-trip (inexhaustive list): - Non-canonical NaNs - BFloat16 - User-defined primitive types. This one is tricky, because they can have a size different from any type we have literals for. (cherry picked from commit b03ef6b)

Small follow-up to #58512 (cherry picked from commit 36bd3ad)

Makes more types survive `jl_static_show` unambiguously: - Symbols - Symbols printed in the `:var"foo"` form use raw string escaping, fixing `:var"a\b"`, `:var"a\\"`, `:var"$a"`, etc. - Symbols that require parens use parens (`:(=)`, ...) - Signed integers: Except for `Int`, signed integers print like `Int8(1)`. - Floats: floats are printed in a naive but reversible (TODO: double check) way. `Inf(16|32|)` and `NaN(16|32|)` are printed, and `Float16`/`Float32` print the type (`Float32(1.5)`). `Float64`s are printed with a trailing `.0` if it is necessary to disambiguate from `Int`. Fixes #52677, #58484 (comment), #58484 (comment), and the specific case mentioned in #58484. Improves the situation for round-trip (inexhaustive list): - Non-canonical NaNs - BFloat16 - User-defined primitive types. This one is tricky, because they can have a size different from any type we have literals for. (cherry picked from commit b03ef6b)

Small follow-up to #58512 (cherry picked from commit 36bd3ad)

Makes more types survive `jl_static_show` unambiguously: - Symbols - Symbols printed in the `:var"foo"` form use raw string escaping, fixing `:var"a\b"`, `:var"a\\"`, `:var"$a"`, etc. - Symbols that require parens use parens (`:(=)`, ...) - Signed integers: Except for `Int`, signed integers print like `Int8(1)`. - Floats: floats are printed in a naive but reversible (TODO: double check) way. `Inf(16|32|)` and `NaN(16|32|)` are printed, and `Float16`/`Float32` print the type (`Float32(1.5)`). `Float64`s are printed with a trailing `.0` if it is necessary to disambiguate from `Int`. Fixes #52677, #58484 (comment), #58484 (comment), and the specific case mentioned in #58484. Improves the situation for round-trip (inexhaustive list): - Non-canonical NaNs - BFloat16 - User-defined primitive types. This one is tricky, because they can have a size different from any type we have literals for. (cherry picked from commit b03ef6b)

Small follow-up to #58512 (cherry picked from commit 36bd3ad)

* static-show: improve accuracy of some printings (JuliaLang#52799) - Show strings with escaping, rather than trying to output the text unmodified. - Show symbols with the same formatting as Strings - Avoid accidentally defining a broken Core.show method for NamedTuple * Make more types jl_static_show unambiguously (JuliaLang#58512) Makes more types survive `jl_static_show` unambiguously: - Symbols - Symbols printed in the `:var"foo"` form use raw string escaping, fixing `:var"a\b"`, `:var"a\\"`, `:var"$a"`, etc. - Symbols that require parens use parens (`:(=)`, ...) - Signed integers: Except for `Int`, signed integers print like `Int8(1)`. - Floats: floats are printed in a naive but reversible (TODO: double check) way. `Inf(16|32|)` and `NaN(16|32|)` are printed, and `Float16`/`Float32` print the type (`Float32(1.5)`). `Float64`s are printed with a trailing `.0` if it is necessary to disambiguate from `Int`. Fixes JuliaLang#52677, JuliaLang#58484 (comment), JuliaLang#58484 (comment), and the specific case mentioned in JuliaLang#58484. Improves the situation for round-trip (inexhaustive list): - Non-canonical NaNs - BFloat16 - User-defined primitive types. This one is tricky, because they can have a size different from any type we have literals for. * Use `julia__gnu_h2f_ieee` instead of `julia_half_to_float` `julia_half_to_float` came in with an LLVM version upgrade after v1.10. --------- Co-authored-by: Jameson Nash <vtjnash@gmail.com> Co-authored-by: Sam Schweigel <33556084+xal-0@users.noreply.github.com>

Makes more types survive `jl_static_show` unambiguously: - Symbols - Symbols printed in the `:var"foo"` form use raw string escaping, fixing `:var"a\b"`, `:var"a\\"`, `:var"$a"`, etc. - Symbols that require parens use parens (`:(=)`, ...) - Signed integers: Except for `Int`, signed integers print like `Int8(1)`. - Floats: floats are printed in a naive but reversible (TODO: double check) way. `Inf(16|32|)` and `NaN(16|32|)` are printed, and `Float16`/`Float32` print the type (`Float32(1.5)`). `Float64`s are printed with a trailing `.0` if it is necessary to disambiguate from `Int`. Fixes #52677, #58484 (comment), #58484 (comment), and the specific case mentioned in #58484. Improves the situation for round-trip (inexhaustive list): - Non-canonical NaNs - BFloat16 - User-defined primitive types. This one is tricky, because they can have a size different from any type we have literals for. (cherry picked from commit b03ef6b)

Small follow-up to #58512 (cherry picked from commit 36bd3ad)

Small follow-up to JuliaLang#58512

Makes more types survive `jl_static_show` unambiguously: - Symbols - Symbols printed in the `:var"foo"` form use raw string escaping, fixing `:var"a\b"`, `:var"a\\"`, `:var"$a"`, etc. - Symbols that require parens use parens (`:(=)`, ...) - Signed integers: Except for `Int`, signed integers print like `Int8(1)`. - Floats: floats are printed in a naive but reversible (TODO: double check) way. `Inf(16|32|)` and `NaN(16|32|)` are printed, and `Float16`/`Float32` print the type (`Float32(1.5)`). `Float64`s are printed with a trailing `.0` if it is necessary to disambiguate from `Int`. Fixes #52677, #58484 (comment), #58484 (comment), and the specific case mentioned in #58484. Improves the situation for round-trip (inexhaustive list): - Non-canonical NaNs - BFloat16 - User-defined primitive types. This one is tricky, because they can have a size different from any type we have literals for. (cherry picked from commit b03ef6b)

Small follow-up to #58512 (cherry picked from commit 36bd3ad)

KristofferC · 2025-06-25T12:49:14Z

Backporting this to 1.11 gives:

Test Failed at /home/kc/julia1.11/test/show.jl:1562
  Expression: v === eval(Meta.parse(static_shown(v)))
   Evaluated: Float16(1.0) === Inf16

Test Failed at /home/kc/julia1.11/test/show.jl:1562
  Expression: v === eval(Meta.parse(static_shown(v)))
   Evaluated: Float16(1.5) === Inf16

Test Failed at /home/kc/julia1.11/test/show.jl:1562
  Expression: v === eval(Meta.parse(static_shown(v)))
   Evaluated: Float16(0.4893) === Inf16

Test Failed at /home/kc/julia1.11/test/show.jl:1562
  Expression: v === eval(Meta.parse(static_shown(v)))
   Evaluated: Float16(0.00010014) === Inf16
....

Am I missing some other PR that is needed?

kpamnany · 2025-06-25T13:14:11Z

We needed #52799 also, to backport this to our 1.10 build.

KristofferC · 2025-06-25T13:26:42Z

That should already be on 1.11 though?

kpamnany · 2025-06-25T13:45:17Z

Ah, okay. Then I don't know. Sorry.

topolarity · 2025-06-27T15:12:36Z

Probably related to the Float16 intrinsic fixes that @xal-0 made in 1.12?

KristofferC · 2025-06-27T15:35:00Z

I think it was due to a missing header declaration of a function which defined it implicitly and maybe caused undefined behavior? On CI this errored and after fixing that it seems like things are OK. Could that be a possible scenario?

Makes more types survive `jl_static_show` unambiguously: - Symbols - Symbols printed in the `:var"foo"` form use raw string escaping, fixing `:var"a\b"`, `:var"a\\"`, `:var"$a"`, etc. - Symbols that require parens use parens (`:(=)`, ...) - Signed integers: Except for `Int`, signed integers print like `Int8(1)`. - Floats: floats are printed in a naive but reversible (TODO: double check) way. `Inf(16|32|)` and `NaN(16|32|)` are printed, and `Float16`/`Float32` print the type (`Float32(1.5)`). `Float64`s are printed with a trailing `.0` if it is necessary to disambiguate from `Int`. Fixes #52677, #58484 (comment), #58484 (comment), and the specific case mentioned in #58484. Improves the situation for round-trip (inexhaustive list): - Non-canonical NaNs - BFloat16 - User-defined primitive types. This one is tricky, because they can have a size different from any type we have literals for. (cherry picked from commit b03ef6b)

Small follow-up to #58512 (cherry picked from commit 36bd3ad)

Backported PRs: - [x] #54840  - [x] #42080  - [x] #58127  - [x] #58202  - [x] #58584  - [x] #58637  - [x] #58837  - [x] #57826  - [x] #58293  - [x] #58887  - [x] #58937  - [x] #58978  - [x] #58998  - [x] #59120  - [x] #59170  Need manual backport: - [ ] #56329  - [ ] #56956  - [ ] #57035  - [ ] #57089  - [ ] #57249  - [ ] #58011  - [ ] #58062  - [ ] #58157  - [ ] #58209  - [ ] #58108  - [ ] #58356  - [ ] #58415  - [ ] #58510  - [ ] #58110  - [ ] #58965  - [ ] #58971  - [ ] #59066  Contains multiple commits, manual intervention needed: - [ ] #55877  - [ ] #56755  - [ ] #57809  - [ ] #57398  - [ ] #56351  - [ ] #57129  - [ ] #58134  - [ ] #58401  - [ ] #58435  - [ ] #58483  - [ ] #58512  - [ ] #58012  - [ ] #58683  - [ ] #59112  Non-merged PRs with backport label: - [ ] #59329  - [ ] #58848  - [ ] #58535  - [ ] #58038  - [ ] #57604  - [ ] #57366  - [ ] #56890  - [ ] #56823  - [ ] #55958  - [ ] #55956  - [ ] #55534  - [ ] #55499  - [ ] #55458  - [ ] #55457  - [ ] #55220  - [ ] #53957  - [ ] #51479  - [ ] #50813  - [ ] #50157  --------- Co-authored-by: Kiran Pamnany <kpamnany@users.noreply.github.com> Co-authored-by: adienes <51664769+adienes@users.noreply.github.com> Co-authored-by: Gabriel Baraldi <baraldigabriel@gmail.com> Co-authored-by: Keno Fischer <keno@juliacomputing.com> Co-authored-by: Simeon David Schaub <simeon@schaub.rocks> Co-authored-by: Jameson Nash <vtjnash@gmail.com> Co-authored-by: Alex Arslan <ararslan@comcast.net> Co-authored-by: Fons van der Plas <fonsvdplas@gmail.com> Co-authored-by: Ian Butterworth <i.r.butterworth@gmail.com> Co-authored-by: JonasIsensee <jonas.isensee@web.de> Co-authored-by: Curtis Vogt <curtis.vogt@gmail.com> Co-authored-by: Dilum Aluthge <dilum@aluthge.com> Co-authored-by: DilumAluthgeBot <43731525+DilumAluthgeBot@users.noreply.github.com> Co-authored-by: DilumAluthge <5619885+DilumAluthge@users.noreply.github.com>

Make more types jl_static_show readably

b6996d2

xal-0 requested a review from topolarity May 23, 2025 19:39

xal-0 added the bugfix This change fixes an existing bug label May 23, 2025

xal-0 added 2 commits May 23, 2025 12:40

Typo fix

e73b687

Typo fix

02f54e4

Fix jl_static_show for negative floats with no fractional part

8a9e425

topolarity added backport 1.10 Change should be backported to the 1.10 release backport 1.11 Change should be backported to release-1.11 backport 1.12 Change should be backported to release-1.12 labels May 23, 2025

kpamnany mentioned this pull request May 23, 2025

Improvements to static-show RelationalAI/julia#239

Merged

3 tasks

vtjnash reviewed May 27, 2025

View reviewed changes

src/rtutils.c Outdated Show resolved Hide resolved

vtjnash reviewed May 27, 2025

View reviewed changes

src/rtutils.c Outdated Show resolved Hide resolved

vtjnash reviewed May 27, 2025

View reviewed changes

src/rtutils.c Outdated Show resolved Hide resolved

vtjnash reviewed May 27, 2025

View reviewed changes

src/rtutils.c Outdated Show resolved Hide resolved

vtjnash reviewed May 27, 2025

View reviewed changes

src/rtutils.c Outdated Show resolved Hide resolved

Correct jl_static_show_string raw escaping

c4d2117

n backslashes followed by " => 2n+1 backslashes and " n backslahes at the end => 2n backslashes

KristofferC mentioned this pull request May 28, 2025

Backports for 1.12.0-beta4 #58369

Merged

58 tasks

Print floats with the correct precision, don't rely on fabs/modf

208a56c

vtjnash reviewed May 28, 2025

View reviewed changes

src/rtutils.c Outdated Show resolved Hide resolved

vtjnash reviewed May 28, 2025

View reviewed changes

src/rtutils.c Outdated Show resolved Hide resolved

xal-0 and others added 3 commits May 28, 2025 11:09

Update src/rtutils.c

054ce44

Co-authored-by: Jameson Nash <vtjnash@gmail.com>

Update src/rtutils.c

77c913f

Co-authored-by: Jameson Nash <vtjnash@gmail.com>

Fixes, additional tests, use trailing zeros for floats

77fcca3

Fix -Werror=c++-compat

0d2a833

topolarity reviewed May 29, 2025

View reviewed changes

kpamnany pushed a commit to RelationalAI/julia that referenced this pull request Jun 4, 2025

Make Ptr values static-show w/ type-information (JuliaLang#58584)

79e652e

Small follow-up to JuliaLang#58512

kpamnany pushed a commit to RelationalAI/julia that referenced this pull request Jun 4, 2025

Make Ptr values static-show w/ type-information (JuliaLang#58584)

e322558

Small follow-up to JuliaLang#58512

KristofferC pushed a commit that referenced this pull request Jun 5, 2025

Make Ptr values static-show w/ type-information (#58584)

4719830

Small follow-up to #58512 (cherry picked from commit 36bd3ad)

KristofferC pushed a commit that referenced this pull request Jun 5, 2025

Make Ptr values static-show w/ type-information (#58584)

319b1a7

Small follow-up to #58512 (cherry picked from commit 36bd3ad)

KristofferC pushed a commit that referenced this pull request Jun 5, 2025

Make Ptr values static-show w/ type-information (#58584)

7287cd7

Small follow-up to #58512 (cherry picked from commit 36bd3ad)

KristofferC pushed a commit that referenced this pull request Jun 5, 2025

Make Ptr values static-show w/ type-information (#58584)

296b1bc

Small follow-up to #58512 (cherry picked from commit 36bd3ad)

KristofferC pushed a commit that referenced this pull request Jun 5, 2025

Make Ptr values static-show w/ type-information (#58584)

c3bd899

Small follow-up to #58512 (cherry picked from commit 36bd3ad)

nsajko mentioned this pull request Jun 8, 2025

misleading printing of value type parameters #57189

Open

kpamnany pushed a commit to RelationalAI/julia that referenced this pull request Jun 9, 2025

Make Ptr values static-show w/ type-information (JuliaLang#58584)

9838fac

Small follow-up to JuliaLang#58512

topolarity mentioned this pull request Jun 17, 2025

--trace-compile / static_show fails to round-trip for several types (☔ umbrella issue) #58756

Open

KristofferC pushed a commit that referenced this pull request Jun 25, 2025

Make Ptr values static-show w/ type-information (#58584)

600efd6

Small follow-up to #58512 (cherry picked from commit 36bd3ad)

KristofferC pushed a commit that referenced this pull request Jul 3, 2025

Make Ptr values static-show w/ type-information (#58584)

5d22e4e

Small follow-up to #58512 (cherry picked from commit 36bd3ad)

KristofferC mentioned this pull request Aug 19, 2025

Backports for julia 1.11.7 #59336

Merged

65 tasks

This was referenced Sep 9, 2025

Backports for Julia 1.11.8 #59521

Draft

Backports for Julia 1.10.11 #58889

Draft

Uh oh!

Make more types jl_static_show readably #58512

Make more types jl_static_show readably #58512

Conversation

xal-0 commented May 23, 2025

Uh oh!

KristofferC commented May 23, 2025

Uh oh!

nickrobinson251 commented May 23, 2025

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

topolarity commented May 27, 2025

Uh oh!

xal-0 commented May 28, 2025

Uh oh!

Uh oh!

Uh oh!

topolarity commented May 28, 2025

Uh oh!

topolarity May 29, 2025

Choose a reason for hiding this comment

Uh oh!

vtjnash May 29, 2025

Choose a reason for hiding this comment

Uh oh!

KristofferC commented Jun 25, 2025

Uh oh!

kpamnany commented Jun 25, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

KristofferC commented Jun 25, 2025

Uh oh!

kpamnany commented Jun 25, 2025

Uh oh!

topolarity commented Jun 27, 2025

Uh oh!

KristofferC commented Jun 27, 2025

Uh oh!

Uh oh!

kpamnany commented Jun 25, 2025 •

edited

Loading