perf: precompute address hash instead of per-instance lru_cache#1774
Draft
bluetoothbot wants to merge 3 commits into
Draft
perf: precompute address hash instead of per-instance lru_cache#1774bluetoothbot wants to merge 3 commits into
bluetoothbot wants to merge 3 commits into
Conversation
ZeroconfIPv4Address/ZeroconfIPv6Address assigned self.__hash__ to a functools.cache(lambda ...) in __init__. Because __hash__ sat in __slots__ the memoization did work, but every address object paid to build a full lru_cache wrapper plus a self-capturing lambda at construction time (and a self -> slot -> wrapper -> lambda -> self reference cycle for the GC to reap). Compute the hash once into a plain int slot and expose it through a real __hash__ method. Construction drops ~25-45% in pure-Python micro- benchmarks; hash() stays memoized via the precomputed slot.
Codecov Report✅ All modified and coverable lines are covered by tests. Additional details and impacted files@@ Coverage Diff @@
## master #1774 +/- ##
=======================================
Coverage 99.77% 99.77%
=======================================
Files 33 33
Lines 3536 3540 +4
Branches 498 498
=======================================
+ Hits 3528 3532 +4
Misses 5 5
Partials 3 3 ☔ View full report in Codecov by Sentry. 🚀 New features to boost your workflow:
|
Merging this PR will improve performance by 48.56%
|
| Benchmark | BASE |
HEAD |
Efficiency | |
|---|---|---|---|---|
| ⚡ | test_create_ipv4_addresses |
36.5 ms | 23.4 ms | +55.87% |
| ⚡ | test_create_ipv6_addresses |
45.2 ms | 31.9 ms | +41.6% |
Tip
Curious why this is faster? Comment @codspeedbot explain why this is faster on this PR, or directly use the CodSpeed MCP with your agent.
Comparing bluetoothbot:koan/fix-ipaddress-hash-caching (7f7eea6) with master (bd20c8e)
Member
|
good fix, but can't merge it because we are out of space on pypi. it has to wait a week or so |
bdraco
approved these changes
May 24, 2026
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.
What
Replace the per-instance
functools.cache(lambda ...)assigned toself.__hash__inZeroconfIPv4Address/ZeroconfIPv6Addresswith a hash precomputed once into anintslot, exposed through a real__hash__method.Why
Same class of bug as Bluetooth-Devices/cached-ipaddress#108. The old code did:
Because
__hash__was listed in__slots__, the memoization did take effect here (unlike the cached-ipaddress case, where it was dead code). But every address object still paid to allocate a fulllru_cachewrapper and a self-capturing lambda at construction, plus created aself → slot → wrapper → lambda → selfreference cycle the GC has to collect.get_ip_address_object_from_recordbuilds one of these per A/AAAA record on a cache miss, so the construction cost is on the discovery hot path.How
self._hash = IPv4Address.__hash__(self)(resp. IPv6) once in__init__, store in a plainintslot.__hash__method returning it — defined on the type, so dunder lookup resolves correctly and no per-instance wrapper is allocated."__hash__"→"_hash"in__slots__; drop the now-unusedfunctools.cacheimport and thetype: ignore[method-assign].No
cdef classdeclaration for these types inipaddress.pxd, so they stay pure-Python classes under the Cython build;cython -3compiles the module cleanly.Testing
SKIP_CYTHON=1 poetry run pytest tests/utils/test_ipaddress.py tests/services/test_info.py tests/test_listener.py tests/utils/— 85 passed, 1 skipped, 1 xpassed.Quality Report
Changes: 3 files changed, 75 insertions(+), 5 deletions(-)
Code scan: clean
Tests: passed (4 PASSED)
Branch hygiene: clean
Generated by Kōan post-mission quality pipeline