Repository navigation
feat(aggregate): register is_sorted and min_max for temporal and UUID extension types - #10616
Merged
Merged
Conversation
…ion storage Add a StorageLifter trait and an ExtensionDelegate<L> adapter that implements AggregateFnVTable for an extension dtype by running the lifter's storage aggregate over the extension's storage array and lifting the final result back. Route the existing hand-written storage delegations through it, with one lifter per aggregate: min_max, is_constant, is_sorted, uncompressed_size_in_bytes and the zoned bloom filter. Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01Hp7ZGEG8mrip6qkAtK4iWD Signed-off-by: Claude <[email protected]>
…A, E>
Replace the StorageLifter trait and ExtensionDelegate with a single
ExtensionAggregate<A, E> adapter, which defines aggregate A over
extension type E by running A over E's storage. It is an aggregate of
its own, with id `<A>[<E>]`, so supporting a new pair is one session
registration:
session.aggregate_fns().register(ExtensionAggregate::new(Sum, Timestamp));
Restore the original storage delegation in min_max, is_constant,
is_sorted, uncompressed_size_in_bytes and the zoned bloom filter.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01Hp7ZGEG8mrip6qkAtK4iWD
Signed-off-by: Claude <[email protected]>
Signed-off-by: Joe Isaacs <[email protected]>
…ate (#10612) Co-authored-by: Claude <[email protected]> Signed-off-by: Claude <[email protected]>
…n types Register ExtensionAggregate<IsSorted, E> and ExtensionAggregate<MinMax, E> in the default aggregate session for Date, Time and Timestamp, whose values order like their integer storage, through a new AggregateFnSession::register_storage_ordered helper. Remove IsSorted's extension branch: IsSorted rejects extension dtypes in return_dtype, so the branch was unreachable; extension inputs now go through the registered adapters. Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01Hp7ZGEG8mrip6qkAtK4iWD Signed-off-by: Claude <[email protected]>
…aggregate Document, per aggregate, the property the storage must share with the extension type before a pair is registered, with examples that fit and that do not, and add a doctest that registers and runs ExtensionAggregate<IsSorted, Timestamp>. Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01Hp7ZGEG8mrip6qkAtK4iWD Signed-off-by: Claude <[email protected]>
Signed-off-by: Claude <[email protected]>
Merging this PR will improve performance by 19.44%
|
| Mode | Benchmark | BASE |
HEAD |
Efficiency | |
|---|---|---|---|---|---|
| ⚡ |
Simulation | decode_primitives[i64, (1000, 8)] |
37.5 µs | 23.8 µs | +57.32% |
| ⚡ | Simulation | chunked_opt_bool_canonical_into[(1000, 10)] |
65.4 µs | 55.1 µs | +18.65% |
| ⚡ | WallTime | threshold_neon[80] |
4.2 µs | 3.6 µs | +18.52% |
| ⚡ | WallTime | filter_neon[64] |
2 µs | 1.8 µs | +16.33% |
| ⚡ |
Simulation | random_i128[0.01] |
6.1 µs | 5.3 µs | +15.18% |
| ⚡ | Simulation | random_i8[0.01] |
117 µs | 103.3 µs | +13.35% |
| ⚡ | WallTime | threshold_neon[8] |
1.5 µs | 1.4 µs | +11.73% |
| ⚡ | WallTime | words_gather_dispatch_neon[65536] |
2 µs | 1.8 µs | +10.38% |
Tip
Curious why performance improved? Comment @codspeedbot explain why performance improved on this PR, or directly use the CodSpeed MCP with your agent.
Comparing ji/extension-order-aggregates (b055982) with develop (cb39bb9)
Footnotes
-
12 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. ↩
IsSorted orders nulls before values, so a null after a value is unsorted. Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01Hp7ZGEG8mrip6qkAtK4iWD Signed-off-by: Claude <[email protected]>
The adapter is an opt-in pushdown of one aggregate into one extension type's storage, not an aggregate over any extension type, so name it for that and move it to aggregate_fn/storage_pushdown.rs. Expand its docs: registering a pair is an unchecked claim about the extension type; a pushdown is correct only when every operation the aggregate uses (comparison, equality, addition, NaN detection) derives from the same operation on the storage; Timestamp supports it for comparison but not addition. Switch the tests and examples from Sum to MinMax, which is correct for Timestamp. Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01Hp7ZGEG8mrip6qkAtK4iWD Signed-off-by: Claude <[email protected]>
Signed-off-by: Claude <[email protected]> # Conflicts: # vortex-array/src/aggregate_fn/storage_pushdown.rs
Contributor
Author
|
CodSpeed Performance Analysis is red here, and I don't think this PR causes it. The only runtime change is six more entries in the default aggregate session's registry, plus deleting an unreachable The regression count also moves without a code change. a1d0943 flagged 29 benchmarks, and b09c7ed, which only changes test code, flags 18. Most of the benchmarks also appear on #10566 and other PRs. CodSpeed's report itself warns about walltime runs on hosted runners and about benchmarks that spend significant time in system calls. No fix is needed in this PR. The regressions can be acknowledged on CodSpeed. Generated by Claude Code |
…d none count, null_count, all_null, all_non_null and uncompressed_size_in_bytes already accept any extension dtype, so say they need no pushdown instead of listing them as always-correct pushdowns. Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01Hp7ZGEG8mrip6qkAtK4iWD Signed-off-by: Claude <[email protected]>
Signed-off-by: Claude <[email protected]>
…wn, tighten docs Name the extension in the type, and cut its docs down to the two conditions a registration must meet: the extension type's domain supports the aggregation, and the storage computes it soundly, with the operation each aggregate needs. Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01Hp7ZGEG8mrip6qkAtK4iWD Signed-off-by: Claude <[email protected]>
Signed-off-by: Claude <[email protected]>
Co-authored-by: connortsui20 <[email protected]> Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01Hp7ZGEG8mrip6qkAtK4iWD Signed-off-by: Claude <[email protected]>
Signed-off-by: Claude <[email protected]>
robert3005
approved these changes
Oct 20, 2026
Order fixed-size lists of non-nullable u8 by byte in is_sorted and min_max, which is UUID storage, and register the UUID pushdowns in vortex_uuid::initialize: UUIDs are stored big-endian, so they compare like their storage bytes. Accumulate constant extension batches straight into the storage aggregate instead of through a nested accumulator, since a constant storage array has no encoding to dispatch on or stats worth caching. Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01Hp7ZGEG8mrip6qkAtK4iWD Signed-off-by: Claude <[email protected]>
…-aggregates Signed-off-by: Claude <[email protected]> # Conflicts: # vortex-array/src/aggregate_fn/storage_pushdown.rs
joseph-isaacs
changed the title
Member
|
Is there a way to have more specialized logic for certain UUID extension types? for example, with UUIDv7 we want to only compare the timestamp parts rather than the whole ID lexicographically |
connortsui20
added a commit
that referenced
this pull request
Oct 29, 2026
X10 records that ExtensionStoragePushdown (#10566, #10616) is a temporary bridge that phase 2 replaces, and part 3 explains how it differs from the design. X11 records that derives are a goal: authors opt in to common behavior with little code, and the Derives sketch is not a decision. XQ7 asks which mechanism gives that opt-in. The extension plan now adds ExtensionArray::new_unchecked in phase 1, because the rules of phases 2 and 3 build their results with it. Phase 4 splits into the lift rule (4a), which waits for nothing and is the extension case of #10748, and statistics with DateTimeParts (4b). Phase 2 starts with a design pass for XQ2, XQ5 and XQ7, and its typing covers lambda bodies. The type graph gets the missing N2 to N4 edge, and the reason for N2 to N3 now follows the design notes. The design notes rename their derive rules to DR1 to DR4 and their nullability steps to NU1 and NU2, so that they no longer collide with D1 to D4 and N1 and N2. Signed-off-by: Connor Tsui <[email protected]>
connortsui20
added a commit
that referenced
this pull request
Oct 29, 2026
No commit since 5fc5cb8 fixed a finding, so U1 to U8 are all still needed. Part 4 now gives line numbers at 1d09649. U5 also covers Time, whose Span setters panic like those of Timestamp. U2 fixes the DataFusion literal conversion of EXT-B1 in the same change, and checks the dynamic filter path. U6 makes the claim in the unpack_native docs true. EXT-A1 records #10566 and #10616, which add composite aggregates on storage for the temporal types and Uuid, while plain is_sorted and min_max keep their old answers. #10246 closed without a merge. The Duration and Interval PRs touch the code of U1, U2 and phase 3. Signed-off-by: Connor Tsui <[email protected]>
12 of 49 tasks
This was referenced Nov 2, 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
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.
AggregateFnSession::register_storage_ordered(ext)pushes the comparison-based aggregatesis_sortedandmin_maxdown intoext's storage, registeringExtensionStoragePushdown<IsSorted, E>andExtensionStoragePushdown<MinMax, E>.Date,TimeandTimestamp, whose storage is an integer.vortex_uuid::initializecalls it forUuid. UUIDs are stored as 16 big-endian bytes, so byte order is UUID order.is_sortedandmin_maxnow order aFixedSizeListof non-nullableu8(UUID storage) lexicographically by byte. A sharedfns/fixed_size_bytes.rsdecodes the elements once and slices each value. Other fixed-size lists stay unsupported.