Cache the scope name prefix to prevent scope traversal in a tight loop #708
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
facebook-github-bot
added
the
CLA Signed
Codecov Report
@@ Coverage Diff @@
## main #708 +/- ##
=======================================
Coverage 94.80% 94.80%
=======================================
Files 247 247
Lines 25701 25711 +10
=======================================
+ Hits 24365 24375 +10
Misses 1336 1336
Continue to review full report at Codecov.
|
lpetre
commented
Jun 20, 2022
| @@ -667,6 +657,7 @@ def __init__( | |||
| self.name = name | |||
| self.node = node | |||
| self._scope_overwrites = {} | |||
| self._name_prefix = self._make_name_prefix() | |||
There was a problem hiding this comment.
Pyre is unhappy with this line. Any suggestions @zsol
There was a problem hiding this comment.
Seems like a bug to me, might be worth raising with Pyre folks internally
zsol
approved these changes
Jun 20, 2022
libcst/metadata/scope_provider.py
Outdated
| return {QualifiedName(".".join(parts), QualifiedNameSource.LOCAL)} | ||
| return { | ||
| QualifiedName( | ||
| ".".join(filter(None, [self.scope._name_prefix, full_name])), |
There was a problem hiding this comment.
Neither of these can be None, and this is a pretty hot function. This should be much faster
Suggested change
| ".".join(filter(None, [self.scope._name_prefix, full_name])), | |
| f"{self.scope._name_prefix}.{full_name}", |
There was a problem hiding this comment.
The filter(None,...) also removes empty strings, which is what I need here. Happy to consider other ways to do it.
There was a problem hiding this comment.
f"{a}.{b}" if a else b is about 4x faster still
| @@ -667,6 +657,7 @@ def __init__( | |||
| self.name = name | |||
| self.node = node | |||
| self._scope_overwrites = {} | |||
| self._name_prefix = self._make_name_prefix() | |||
There was a problem hiding this comment.
Seems like a bug to me, might be worth raising with Pyre folks internally
|
Maybe worth putting a |
lpetre
force-pushed
the
fix_assignment_name_prefix
branch
from
e81678c to
8860e95
Compare
netbsd-srcmastr
pushed a commit
to NetBSD/pkgsrc
that referenced
this pull request
Sep 16, 2022
0.4.7 - 2022-07-12 Fixed * Fix get_qualified_names_for matching on prefixes of the given name by @lpetre in Instagram/LibCST#719 Added * Implement lazy loading mechanism for expensive metadata providers by @Chenguang-Zhu in Instagram/LibCST#720 0.4.6 - 2022-07-04 New Contributors - @superbobry made their first contribution in Instagram/LibCST#702 Fixed - convert_type_comments now preserves comments following type comments by @superbobry in Instagram/LibCST#702 - QualifiedNameProvider optimizations - Cache the scope name prefix to prevent scope traversal in a tight loop by @lpetre in Instagram/LibCST#708 - Faster qualified name formatting by @lpetre in Instagram/LibCST#710 - Prevent unnecessary work in Scope.get_qualified_names_for_ by @lpetre in Instagram/LibCST#709 - Fix parsing of parenthesized empty tuples by @zsol in Instagram/LibCST#712 - Support whitespace after ParamSlash by @zsol in Instagram/LibCST#713 - [parser] bail on deeply nested expressions by @zsol in Instagram/LibCST#718 0.4.5 - 2022-06-17 New Contributors - @zzl0 made their first contribution in Instagram/LibCST#704 Fixed - Only skip supported escaped characters in f-strings by @zsol in Instagram/LibCST#700 - Escaping quote characters in raw string literals causes a tokenizer error by @zsol in Instagram/LibCST#668 - Corrected a code example in the documentation by @zzl0 in Instagram/LibCST#703 - Handle multiline strings that start with quotes by @zzl0 in Instagram/LibCST#704 - Fixed a performance regression in libcst.metadata.ScopeProvider by @lpetre in Instagram/LibCST#698 0.4.4 - 2022-06-13 New Contributors - @adamchainz made their first contribution in Instagram/LibCST#688 Added - Add package links to PyPI by @adamchainz in Instagram/LibCST#688 - native: add overall benchmark by @zsol in Instagram/LibCST#692 - Add support for PEP-646 by @zsol in Instagram/LibCST#696 Updated - parser: use references instead of smart pointers for Tokens by @zsol in Instagram/LibCST#691
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.
Summary
We don't need to recompute this name prefix every time we try to lookup a qualified name for a node. This value is immutable when the scope is constructed.
Context
There is a certain type of file that causes the scope provider to take an extraordinary amount of time.
These files tend to fit a specific pattern:
Here is an example: https://github.com/FreeOpcUa/python-opcua/blob/0.98.11/opcua/server/standard_address_space/standard_address_space_part9.py
I've been running these w/ yappi to figure out where the scope provider might be improved.
Test Plan
Previously this file would take ~70 seconds on my laptop (Apple M1 Pro, 32GB RAM). Now it takes ~30 seconds.
Here is the top hits from yappi (sorted by tsub) before and after:
Before
After