fixtures: fix catastrophic performance problem in reorder_items #12409
Conversation
This file contains hidden or 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
This makes some minor clarity and performance improvements to the code.
nicoddemus approved these changes Jun 3, 2024
changelog/12355.bugfix.rst Outdated
| @@ -0,0 +1 @@ | |||
| Fix possible catasrophic performance slowdown on a certain parametrization pattern involving many higher-scoped parameters. | |||
There was a problem hiding this comment.
Suggested change
| Fix possible catasrophic performance slowdown on a certain parametrization pattern involving many higher-scoped parameters. | |
| Fix possible catastrophic performance slowdown on a certain parametrization pattern involving many higher-scoped parameters. |
Fix pytest-dev#12355. In the issue, it was reported that the `reorder_items` has quadratic (or worse...) behavior with certain simple parametrizations. After some debugging I found that the problem happens because the "Fix items_by_argkey order" loop keeps adding the same item to the deque, and it reaches epic sizes which causes the slowdown. I don't claim to understand how the `reorder_items` algorithm works, but if as far as I understand, if an item already exists in the deque, the correct thing to do is to move it to the front. Since a deque doesn't have such an (efficient) operation, this switches to `OrderedDict` which can efficiently append from both sides, deduplicate and move to front.
bluetech force-pushed the reorder-items-perf branch from 7a230b0 to e89d23b Compare 
webknjaz reviewed Mar 19, 2025
| other_scoped_items_by_argkey = items_by_argkey[other_scope] | ||
| for argkey in argkeys_by_item[other_scope].get(i, ()): | ||
| other_scoped_items_by_argkey[argkey][i] = None | ||
| other_scoped_items_by_argkey[argkey].move_to_end( |
There was a problem hiding this comment.
@bluetech I've hit some corner case on PyPy where move_to_end() seems to be raising a KeyError and crashing the item collection entirely...
There was a problem hiding this comment.
Oh, nevermind. This is a regression of v8.2.2 and this PR only hit release in v8.3.0.
Nope, it was backported into 8.2.2: #13312
webknjaz added a commit to webknjaz/multidict that referenced this pull request Mar 19, 2025
This patch temporarily restricts pytest version below 8.2.2 under PyPy due to a discovered regression that it introduced [[1]]. The regression has been observed on at least `pypy3.9-7.3.16`, `pypy3.10-7.3.19` and `pypy3.11-7.3.19`. It can be triggered by running the following in affected runtimes: pytest --collect-only --no-cov tests/test_abc.py tests/test_copy.py tests/test_incorrect_args.py tests/test_multidict.py tests/test_mypy.py tests/test_pickle.py tests/test_types.py tests/test_update.py tests/test_version.py [1]: pytest-dev/pytest#13312 [2]: pytest-dev/pytest#12414 [3]: pytest-dev/pytest#12409
2 tasks
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
3 participants
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.
First commit is a cleanup commit (no functional changes intended).
The second commit fixes #12355.
In the issue, it was reported that the
reorder_itemshas quadratic (or worse...) behavior with certain simple parametrizations. After some debugging I found that the problem happens because the "Fix items_by_argkey order" loop keeps adding the same item to the deque, and it reaches epic sizes which causes the slowdown.I don't claim to understand how the
reorder_itemsalgorithm works, but if as far as I understand, if an item already exists in the deque, the correct thing to do is to move it to the front. Since a deque doesn't have such an (efficient) operation, this switches toOrderedDictwhich can efficiently append from both sides, deduplicate and move to front.