[3.13.0b4] f-string with precision specifier has `format_spec` of `Constant`, not `JoinedStr`

[jacobtylerwalls]

Bug report

Bug description:

Python 3.12.0 (v3.12.0:0fb18b02c8, Oct  2 2023, 09:45:56) [Clang 13.0.0 (clang-1300.0.29.30)] on darwin
>>> import ast
>>> ast.parse("f'{e:.3}'").body[0].value.values[0].format_spec
<ast.JoinedStr object at 0x1002aa050>

Python 3.13.0b4 (v3.13.0b4:567c38b4eb, Jul 18 2024, 07:28:29) [Clang 15.0.0 (clang-1500.3.9.4)] on darwin
>>> import ast
>>> ast.parse("f'{e:.3}'").body[0].value.values[0].format_spec
<ast.Constant object at 0x104af45d0>

Docs say FormattedValue.format_spec is a JoinedStr.

Reported downstream in pylint-dev/astroid#2478

Not present in 3.13.0b3

CPython versions tested on:

3.13

Operating systems tested on:

Linux, macOS

Linked PRs

• gh-122300: Preserve AST nodes for format specifiers with single elements #122308
• [3.12] gh-122300: Preserve AST nodes for format specifiers with single elements (GH-122308) #122363
• [3.13] gh-122300: Preserve AST nodes for format specifiers with single elements (GH-122308) #122364

[Eclips4]

Bisected to c46d64e
cc @pablogsal

[jacobtylerwalls]

Thanks for the bisect. That PR updated the docs for ast.JoinedStr to show this as the new expected result, so maybe we just need a docs update for ast.FormattedValue?

[pablogsal]

Thanks for the bisect. That PR updated the docs for ast.JoinedStr to show this as the new expected result, so maybe we just need a docs update for ast.FormattedValue?

Note that the docs are not prescriptive over what the shape of the AST will be, it's just an example. The actual shape of the AST can still change due to bug fixes and grammar extensions so I would not recommend looking at those docs as a prescription with any backwards compatibility guarantees.

Being said that I can try to investigate how difficult it's to preserve this because I understand how annoying this can be downstream if you don't have anything you can match against and was like that for several versions.

[pablogsal]

I think #122308 will do the trick

[penguinpee]

Yep. #122308 fixes the test regressions in rope and astor that I reported in #122307. Thanks!

[pablogsal]

Will backport this soon