Fix parse test case on Python 3.12 #15577
Conversation
This fixes the test case testFStringWithFormatSpecifierExpression on Python 3.12. Strip redundant empty string literal from AST generated from f-string. It's not generated on earlier Python versions and it seems fine to just drop it.
| elif len(strs_to_join.items) > 1: | ||
| last = strs_to_join.items[-1] | ||
| if isinstance(last, StrExpr) and last.value == "": | ||
| # 3.12 can add an empty literal at the end. Delete it for consistency |
There was a problem hiding this comment.
Interesting, why is that needed in CPython 🤔
There was a problem hiding this comment.
Interesting, why is that needed in CPython 🤔
It caused a crash over at flake8-bugbear too: PyCQA/flake8-bugbear#393
There was a problem hiding this comment.
If this has caused issues in two projects, should we report upstream?
|
According to mypy_primer, this change doesn't affect type check results on a corpus of open source code. ✅ |
|
Looks like this might be changed in CPython, so maybe we should just wait for the next 3.12 beta rather than merging this: |
We can wait until the next beta, unless perhaps we want to enable 3.12 in CI before the next beta is available. I'll keep this open for now. |
For now could we maybe just skip the test? I can't remember the pytest equivalent off the top of my head, but for unittest I'd just decorate the test with: @skipIf(
sys.version_info[:4] == (3, 12, 0, "beta") and sys.version_info[4] < 4,
"Early 3.12 betas had a bug"
)Though mypy's test suite is very data-driven I guess, so maybe it's more difficult to be that granular for a single test case |
|
This is still failing for me on Python 3.12.0 beta 4. I'm leaning towards merging this. We can always revert this afterwards if it's no longer needed. |
Makes sense! |
This fixes the test case testFStringWithFormatSpecifierExpression on Python 3.12.
Strip redundant empty string literal from AST generated from f-string. It's not generated on earlier Python versions and it seems fine to just drop it.