docs(vortex-layout): mark the bloom filter aggregate as unstable - #9753
Conversation
Merging this PR will regress 2 benchmarks
|
| Mode | Benchmark | BASE |
HEAD |
Efficiency | |
|---|---|---|---|---|---|
| ❌ | Simulation | random_i8[0.5] |
67 µs | 90.6 µs | -26.02% |
| ❌ | Simulation | allocate_drop_bytes[0] |
466 ns | 520.2 ns | -10.41% |
| ⚡ | WallTime | arrow_checked_add_u32_neon[16384] |
20.4 µs | 12.7 µs | +60.18% |
| ⚡ | Simulation | random_i16[0.8] |
92.7 µs | 74.3 µs | +24.76% |
Tip
Investigate this regression by commenting @codspeedbot fix this regression on this PR, or directly use the CodSpeed MCP with your agent.
Comparing claude/aggregate-fn-unstable-comment-j95028 (6dd8679) with develop (3170657)
Footnotes
-
206 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩
|
Things not in editions cannot be persisted anyway. The editions are the reference of what’s persistable. Happy with the docs fwiw just letting you know there’s layers to safety now |
The SBBF hash function, block layout, and salt order can still change, and no edition declares `vortex.bloom_filter.sbbf`. Say so on the type so nobody persists bloom partials into a file. Signed-off-by: Connor Tsui <connor.tsui20@gmail.com> Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_019yqkmRmsfqAzRj1KXh8HZU
5a1b2c0 to
6dd8679
Compare
Summary
ZonedLayout#8901The bloom filter aggregate from #9398 does not have a settled wire format. The hash function, the block layout, and the salt order can all still change, and no edition declares
vortex.bloom_filter.sbbf, so nothing must write bloom partials into a file yet.Changes
Adds a
Stabilitysection to theBloomFilterdoc comment saying that the aggregate is unstable and thatvortex.bloom_filter.sbbfpartials must not be persisted. Comment only, no behavior change.