Reduce single file format overhead by lazily importing modules and loading .gitignore / --exclude #3211
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
For what it is worth, I'd prefer a rebase merge for this PR given the commits are mostly independent (other than the changelog entry). |
|
Coolio, diff-shades is failing and I don't know why at all ... |
`black.reformat_many` depends on a lot of slow-to-import modules. When formatting simply a single file, the time paid to import those modules is totally wasted. So I moved `black.reformat_many` and its helpers to `black.concurrency` which is now *only* imported if there's more than one file to reformat. This way, running Black over a single file is snappier Here are the numbers before and after this patch running `python -m black --version`: - interpreted: 411 ms +- 9 ms -> 342 ms +- 7 ms: 1.20x faster - compiled: 365 ms +- 15 ms -> 304 ms +- 7 ms: 1.20x faster Co-authored-by: Fabio Zadrozny <fabiofz@gmail.com>
Loading .gitignore and compiling the exclude regex can take more than 15ms. We shouldn't and don't need to pay this cost if we're simply formatting files given on the command line directly. I would've loved to lazily import pathspec, but the patch won't be clean until the file collection and discovery logic is refactored first. Co-authored-by: Fabio Zadrozny <fabiofz@gmail.com>
os.cpu_count() can return None (sounds like a super arcane edge case though) so the type annotation for the `workers` parameter of `black.main` is wrong. This *could* technically cause a runtime TypeError since it'd trip one of mypyc's runtime type checks so we might as well fix it. Reading the documentation (and cross-checking with the source code), you are actually allowed to pass None as `max_workers` to `concurrent.futures.ProcessPoolExecutor`. If it is None, the pool initializer will simply call os.cpu_count() [^1] (defaulting to 1 if it returns None [^2]). It'll even round down the worker count to a level that's safe for Windows. ... so theoretically we don't even need to call os.cpu_count() ourselves, but our Windows limit is 60 (unlike the stdlib's 61) and I'd prefer not accidentally reintroducing a crash on machines with many, many CPU cores. [^1]: https://docs.python.org/3/library/concurrent.futures.html#concurrent.futures.ProcessPoolExecutor [^2]: https://github.com/python/cpython/blob/a372a7d65320396d44e8beb976e3a6c382963d4e/Lib/concurrent/futures/process.py#L600
ichard26
force-pushed
the
reduce-single-file-format-overhead
branch
from
aafd5b9 to
2c0d427
Compare
|
PTAL. Please note I'd prefer a rebase merge here if you approve and chose to merge right after. |
ichard26
commented
Aug 13, 2022
| workers: Optional[int], | ||
| ) -> None: | ||
| """Reformat multiple files using a ProcessPoolExecutor.""" | ||
| maybe_install_uvloop() |
There was a problem hiding this comment.
I chose to keep the helper function instead of always attempting an import of uvloop in the module scope just as insurance if we ever accidentally break the invariant black.concurrency should only be imported via black.main if there's more than one file to format.
|
This works on Windows with PyInstaller (just tested locally) so I'm merging. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Description
Fixes #2564 and closes #2656.
Checklist - did you ...