perf(specifiers): keep range caches across canonicalization - #1253
Merged
Merged
Conversation
Contributor
Author
|
@notatallshaw I'll leave this one up to you, if it overlaps with your work, feel free to close. |
henryiii
commented
Jun 13, 2026
henryiii
marked this pull request as ready for review
June 13, 2026 04:58
notatallshaw
approved these changes
Jun 15, 2026
SpecifierSet._canonical_specs() deduplicated and sorted self._specs and then reset both self._is_unsatisfiable and self._ranges. Canonicalization is semantics-preserving: it only changes the order and removes exact duplicates of the spec tuple, never the set of constraints. Both caches depend only on the multiset-as-set of specifiers: - _ranges is the intersection of every spec's individual ranges. Each spec's _to_ranges() depends only on that single spec, and intersect_ranges implements set intersection, which is commutative, associative, and idempotent. Reordering the fold or dropping a duplicate spec (A intersect A == A) cannot change the result. - _is_unsatisfiable derives from _get_ranges() plus the arbitrary-equality and prerelease-only checks, all of which are order- and duplicate-insensitive. Resetting the caches therefore discarded valid, expensive work: calling contains() (which builds ranges) followed by str()/hash()/== forced the next contains() to recompute the intersection. Leaving the caches intact keeps them warm. The prereleases.setter reset is unrelated and left alone. Assisted-by: ClaudeCode:claude-opus-4-8
henryiii
force-pushed
the
perf/specifierset-keep-caches
branch
from
June 15, 2026 23:51
5975a98 to
4ab044e
Compare
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 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.
🤖 AI text below 🤖
SpecifierSet._canonical_specs()reset_is_unsatisfiableand_rangeswhenever it canonicalized, so callingfilter()/is_unsatisfiable()(which compute the range intersection) followed bystr()/hash()/==(which canonicalize) threw the expensive intersection away, and the next range-based call recomputed it.Canonicalization is just
dict.fromkeys(sorted(specs, key=str))— a reorder plus exact-duplicate dedup. Both caches depend only on the set of constraints: eachSpecifier._to_ranges()is a pure function of that one spec, andintersect_rangesis mathematical set intersection (commutative, associative, idempotent), so reordering or deduplicating specs cannot change the result._is_unsatisfiablederives from the ranges plus symmetric checks over the===specs. The reset in theprereleasessetter stays, since prereleases genuinely affects satisfiability.The existing asv suite never interleaves the two operation families, so it shows no change (and no regressions). A targeted microbenchmark of
is_unsatisfiable()→str()→is_unsatisfiable()on an 8-spec set (Python 3.14, Apple M5 Pro): 67.4μs → 42.0μs per round (1.6×). The cost on main is bounded — one extra range computation per object lifetime — so this mostly matters for many short-lived sets, e.g. resolvers hashing or printing specifier sets they also evaluate.Regression tests pin that the cache object survives canonicalization and that
contains()results are identical before/after for duplicate, unsorted, and post-release-!=sets.Part of #1239.
🤖 Generated with Claude Code