8000 mypy_primer is very slow · Issue #6481 · python/typeshed · GitHub
[go: up one dir, main page]
Skip to content

mypy_primer is very slow #6481

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its main 8000 tainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Closed
srittau opened this issue Dec 3, 2021 · 5 comments
Closed

mypy_primer is very slow #6481

srittau opened this issue Dec 3, 2021 · 5 comments
Assignees
@hauntsaninja
Labels
project: infrastructure typeshed build, test, documentation, or distribution related

Comments

@srittau
Copy link
Collaborator
srittau commented Dec 3, 2021

Within the last week, mypy_primer has become very slow. Where previously, it ran in about 10 minutes, it now needs over 45. What could be the cause for that?
@srittau srittau added the project: infrastructure typeshed build, test, documentation, or distribution related label Dec 3, 2021
@AlexWaygood
Copy link
Member

#6451 (comment)

@srittau
Copy link
Collaborator Author
srittau commented Dec 3, 2021

So what can we do to bring it back to run under 10 minutes again? Everything above that, and it just becomes a hindrance more than a help. Can we segment it? Or only run for projects that are potentially affected by the changes?

@hauntsaninja
Copy link
Collaborator

Options:

  • 0.920 should speed it up a lot (could use a prerelease wheel)
  • we could shard it more
  • we could try adding some GitHub Actions caching
  • we could reduce the number of projects we run against (note mypy_primer has an option that lets you measure project runtimes)

Not sure how you would determine projects potentially affected by changes (note we don't run primer on changes to third party stubs anyway, since modular typeshed)

@srittau
Copy link
Collaborator Author
srittau commented Dec 3, 2021

One idea to determine potentially affected projects: Cache the modules imported by a project (basically do a grep -r ^import | cut -f 2 | cut -d . -f 1 once a day) and compare to the changes files.

Do we have any performance data what the bottlenecks are? What would make sense to cache?

@Akuli Akuli mentioned this issue Dec 7, 2021
Merged
@hauntsaninja
Copy link
Collaborator

Given that mypy 0.920 is like 2x faster for mypy_primer like usage, I was surprised to not see the gain.

It looks like I broke the logic that made typeshed use wheels from PyPI, so we have been running pure Python mypy for the last couple weeks :-) Anyway, all fixed, and I expect that should resolve this.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees

@hauntsaninja hauntsaninja

Labels
project: infrastructure typeshed build, test, documentation, or distribution related
Projects
None yet
Milestone
No milestone
Development

No branches or pull requests
3 participants
@srittau @hauntsaninja @AlexWaygood
0