stubgen: Preserve simple defaults in function signatures

[hamdanal]

trafficstars

Fixes #13238

See also https://github.com/python/typeshed/issues/8988

cc @AlexWaygood @JelleZijlstra

[hamdanal]

Some questions for Alex and Jelle:

1. Do you want full parity with stubdefaulter?
2. ~Should I implement built in container types?~ (added)
3. What about name expression default like def f(x=dict)?
4. I left a todo in the code about what should be the length limit for a "simple default", currently it is 200 characters. What do you think.

[hamdanal]

I installed the following packages

matplotlib==3.7.1
numpy==1.24.3
pandas==2.0.2
scipy==1.10.1
seaborn==0.12.2

and generated their stubs with mypy master then added defaults with stubdefaulter then compared the result with defaults added by this PR. The only differences I saw were related to the runtime vs static nature of the two tools:

1. stubdefaulter not being able to add defaults to a file due to runtime errors, stubgen defaults were added
2. if a default is set to a module variable, stubdefaulter uses its value as the default while stubgen ignores it

Please let me know if you have other ideas to test this change.

[coditamar]

@CodiumAI-Agent please review

[CodiumAI-Agent]

PR Analysis

• 🎯 Main theme: Preserving simple defaults in function signatures
• 🔍 Description and title: Yes
• 📌 Type of PR: Bug fix
• 🧪 Relevant tests added: Yes
• ✨ Minimal and focused: Yes, the PR is focused on fixing the issue of preserving simple defaults in function signatures.
• 🔒 Security concerns: No, the changes made in this PR do not introduce any security concerns.

PR Feedback

• 💡 General PR suggestions: The PR is well-structured and the changes are focused on the issue at hand. The added function get_str_default_of_node is a good addition and helps in preserving the simple defaults in function signatures. The changes are reflected in the tests as well, which is a good practice.

• 🤖 Code suggestions:

  • relevant file: mypy/stubgen.py suggestion content: Consider adding a docstring to the get_str_default_of_node function to explain its purpose and the parameters it takes. This would improve the readability and maintainability of the code. [medium]

  • relevant file: mypy/stubgen.py suggestion content: In the get_str_default_of_node function, you have a check for the length of the default value string. It would be good to make this limit a constant at the beginning of the file or function, to make it easily adjustable. [medium]

  • relevant file: mypy/stubgen.py suggestion content: In the get_str_default_of_node function, there are several if conditions checking the instance of rvalue. Consider using a dictionary mapping to simplify this structure and make it more maintainable. [medium]

  • relevant file: mypy/stubgen.py suggestion content: In the get_str_default_of_node function, when handling TupleExpr, ListExpr, SetExpr, and DictExpr, there is repeated code that could be extracted into a helper function to reduce redundancy. [important]

How to use

Tag me in a comment '@CodiumAI-Agent' to ask for a new review after you update the PR. You can also tag me and ask any question, for example '@CodiumAI-Agent is the PR ready for merge?'

[hauntsaninja]

@coditamar Please stop, or I'll have to report your account

[coditamar]

@coditamar Please stop, or I'll have to report your account

OK. I apologize. thank you

[hoechenberger]

Hello! May I ask what's the status of this? 🙂 thank you!

[hamdanal]

Hello! May I ask what's the status of this? 🙂 thank you!

I don’t have an answer for you but if you need default values in stub files now I invite you to try https://github.com/JelleZijlstra/stubdefaulter. I use it regularly and it works very well.

[github-actions[bot]]

According to mypy_primer, this change doesn't affect type check results on a corpus of open source code. ✅

[github-actions[bot]]

Diff from mypy_primer, showing the effect of this PR on open source code:

discord.py (https://github.com/Rapptz/discord.py): typechecking got 1.07x faster (189.3s -> 177.3s)
(Performance measurements are based on a single noisy sample)