pluggable registry for input/export arrow kernels#7824
Open
pluggable registry for input/export arrow kernels#7824
Conversation
Signed-off-by: Andrew Duffy <andrew@a10y.dev>
Signed-off-by: Andrew Duffy <andrew@a10y.dev>
Signed-off-by: Andrew Duffy <andrew@a10y.dev>
Closed
Merging this PR will degrade performance by 12.38%
|
| Mode | Benchmark | BASE |
HEAD |
Efficiency | |
|---|---|---|---|---|---|
| ❌ | Simulation | bench_compare_sliced_dict_primitive[(1000, 10000)] |
79.7 µs | 91 µs | -12.38% |
| ❌ | Simulation | bench_compare_sliced_dict_primitive[(2000, 10000)] |
85.2 µs | 95.2 µs | -10.54% |
| ❌ | Simulation | bench_compare_sliced_dict_primitive[(2500, 10000)] |
88 µs | 97.9 µs | -10.19% |
| ❌ | Simulation | bench_compare_sliced_dict_primitive[(3333, 10000)] |
92.5 µs | 102.9 µs | -10.15% |
| ❌ | Simulation | bench_compare_sliced_dict_varbinview[(1000, 10000)] |
111.8 µs | 124.6 µs | -10.28% |
| ❌ | Simulation | canonicalize_compare[(1000, 4, 4)] |
122.1 µs | 135.7 µs | -10% |
Comparing aduffy/arrow-vtable (71d6452) with develop (f3d5f09)
joseph-isaacs
changed the title
Signed-off-by: Andrew Duffy <andrew@a10y.dev>
Signed-off-by: Andrew Duffy <andrew@a10y.dev>
github-actions
Bot
removed
the
action/benchmark-sql
Contributor
Polar Signals Profiling ResultsLatest Run
Previous Runs (2)
Powered by Polar Signals Cloud |
Contributor
Benchmarks: PolarSignals ProfilingVortex (geomean): 0.981x ➖ datafusion / vortex-file-compressed (0.981x ➖, 0↑ 0↓)
|
Contributor
File Sizes: PolarSignals ProfilingNo file size changes detected. |
Contributor
Benchmarks: FineWeb NVMeVerdict: No clear signal (low confidence) datafusion / vortex-file-compressed (1.027x ➖, 0↑ 1↓)
datafusion / vortex-compact (1.015x ➖, 0↑ 0↓)
datafusion / parquet (0.994x ➖, 0↑ 0↓)
duckdb / vortex-file-compressed (1.022x ➖, 0↑ 0↓)
duckdb / vortex-compact (1.000x ➖, 0↑ 0↓)
duckdb / parquet (1.004x ➖, 0↑ 1↓)
Full attributed analysis
|
Contributor
File Sizes: FineWeb NVMeNo file size changes detected. |
Contributor
Benchmarks: TPC-H SF=1 on NVMEVerdict: No clear signal (low confidence) datafusion / vortex-file-compressed (0.990x ➖, 0↑ 0↓)
datafusion / vortex-compact (0.996x ➖, 0↑ 0↓)
datafusion / parquet (1.000x ➖, 1↑ 1↓)
datafusion / arrow (1.022x ➖, 0↑ 1↓)
duckdb / vortex-file-compressed (1.005x ➖, 0↑ 0↓)
duckdb / vortex-compact (0.988x ➖, 0↑ 0↓)
duckdb / parquet (0.993x ➖, 1↑ 0↓)
duckdb / duckdb (0.991x ➖, 0↑ 1↓)
Full attributed analysis
|
Contributor
File Sizes: TPC-H SF=1 on NVMENo file size changes detected. |
Contributor
Benchmarks: TPC-DS SF=1 on NVMEVerdict: No clear signal (low confidence) datafusion / vortex-file-compressed (1.108x ❌, 0↑ 55↓)
datafusion / vortex-compact (1.094x ➖, 1↑ 42↓)
datafusion / parquet (1.099x ➖, 1↑ 50↓)
duckdb / vortex-file-compressed (1.096x ➖, 1↑ 47↓)
duckdb / vortex-compact (1.084x ➖, 0↑ 39↓)
duckdb / parquet (1.055x ➖, 0↑ 14↓)
duckdb / duckdb (1.106x ❌, 0↑ 56↓)
Full attributed analysis
|
Contributor
File Sizes: TPC-DS SF=1 on NVMENo file size changes detected. |
Signed-off-by: Andrew Duffy <andrew@a10y.dev>
Contributor
Benchmarks: FineWeb S3Verdict: No clear signal (environment too noisy confidence) datafusion / vortex-file-compressed (0.992x ➖, 0↑ 0↓)
datafusion / vortex-compact (0.960x ➖, 0↑ 0↓)
datafusion / parquet (0.922x ➖, 0↑ 0↓)
duckdb / vortex-file-compressed (0.990x ➖, 0↑ 0↓)
duckdb / vortex-compact (1.007x ➖, 0↑ 0↓)
duckdb / parquet (0.989x ➖, 0↑ 0↓)
Full attributed analysis
|
Contributor
Benchmarks: Statistical and Population GeneticsVerdict: No clear signal (low confidence) duckdb / vortex-file-compressed (1.002x ➖, 1↑ 0↓)
duckdb / vortex-compact (1.020x ➖, 0↑ 0↓)
duckdb / parquet (1.033x ➖, 0↑ 0↓)
Full attributed analysis
|
Contributor
File Sizes: Statistical and Population GeneticsNo file size changes detected. |
Contributor
Benchmarks: TPC-H SF=10 on NVMEVerdict: No clear signal (low confidence) datafusion / vortex-file-compressed (1.006x ➖, 0↑ 0↓)
datafusion / vortex-compact (1.014x ➖, 0↑ 0↓)
datafusion / parquet (0.988x ➖, 0↑ 0↓)
datafusion / arrow (1.126x ❌, 0↑ 14↓)
duckdb / vortex-file-compressed (1.056x ➖, 0↑ 3↓)
duckdb / vortex-compact (1.035x ➖, 0↑ 1↓)
duckdb / parquet (1.013x ➖, 0↑ 0↓)
duckdb / duckdb (1.068x ➖, 0↑ 5↓)
Full attributed analysis
|
Contributor
File Sizes: TPC-H SF=10 on NVMENo file size changes detected. |
Contributor
Benchmarks: Clickbench on NVMEVerdict: No clear signal (low confidence) datafusion / vortex-file-compressed (1.073x ➖, 0↑ 4↓)
datafusion / parquet (1.055x ➖, 0↑ 3↓)
duckdb / vortex-file-compressed (1.089x ➖, 0↑ 11↓)
duckdb / parquet (1.022x ➖, 0↑ 3↓)
duckdb / duckdb (1.026x ➖, 0↑ 2↓)
Full attributed analysis
|
Contributor
File Sizes: Clickbench on NVMEFile Size Changes (1 files changed, -0.0% overall, 0↑ 1↓)
Totals:
|
Contributor
Benchmarks: TPC-H SF=1 on S3Verdict: No clear signal (environment too noisy confidence) datafusion / vortex-file-compressed (0.954x ➖, 0↑ 0↓)
datafusion / vortex-compact (0.975x ➖, 1↑ 0↓)
datafusion / parquet (0.994x ➖, 1↑ 0↓)
duckdb / vortex-file-compressed (1.009x ➖, 0↑ 0↓)
duckdb / vortex-compact (1.029x ➖, 0↑ 0↓)
duckdb / parquet (1.010x ➖, 0↑ 0↓)
Full attributed analysis
|
Contributor
Benchmarks: TPC-H SF=10 on S3Verdict: No clear signal (environment too noisy confidence) datafusion / vortex-file-compressed (0.993x ➖, 0↑ 0↓)
datafusion / vortex-compact (0.996x ➖, 1↑ 2↓)
datafusion / parquet (0.973x ➖, 0↑ 0↓)
duckdb / vortex-file-compressed (1.000x ➖, 0↑ 0↓)
duckdb / vortex-compact (1.020x ➖, 0↑ 1↓)
duckdb / parquet (1.038x ➖, 0↑ 0↓)
Full attributed analysis
|
Signed-off-by: Andrew Duffy <andrew@a10y.dev>
Signed-off-by: Andrew Duffy <andrew@a10y.dev>
github-actions
Bot
removed
the
action/benchmark-sql
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.
Summary
Adds a pluggable
ArrowSessionregistry onVortexSessionfor round-tripping Vortex extension types in and out of Arrow extension types. Unblocks Arrow round-trip forarrow.uuidtoday, witharrow.parquet.variant, GeoArrow, and tensor types as the next consumers.Part of #7686.
API changes
The session exposes two trait-driven plugin slots:
ArrowExportVTable— dispatched by target Arrow extension name (ARROW:extension:name). Implementations turn a VortexArrayRefinto an ArrowArrayRefshaped to the requestedField. Also providesto_arrow_fieldfor schema inference when only a VortexDTypeis inhand.
ArrowImportVTable— dispatched by source Arrow extension name carried on the incomingField. Implementations turn an ArrowArrayRefback into a VortexArrayRef, including any storage re-encoding (e.g.FixedSizeBinary[16]→FixedSizeList<u8; 16>for UUID).Both traits return
Unsupported(input)to defer to the next plugin or to the canonical fallback, so multiple plugins can register against the same key and probe in order.New session entry points (
vortex-array/src/arrow/session.rs):ArrowSession::to_arrow_field/to_arrow_schema— VortexDType→ ArrowField/Schema, recursing into containers so nested extension fields go through the registered plugin.ArrowSession::from_arrow_field/from_arrow_schema— inverse direction, plugin-aware.ArrowSession::from_arrow_record_batch/execute_record_batch—RecordBatchround-trip.ArrowSessionExtextension trait so anySessionExtcan callsession.arrow().….The default session pre-registers the builtin UUID plugin (
vortex-array/src/extension/uuid/arrow.rs).What's not in the plugin layer
Date,Time, andTimestampare Vortex builtin extensions that map directly to native Arrow temporal types, so they continue to go through the canonical executor (vortex-array/src/arrow/executor/temporal.rs) rather than the plugin registry. The plugin layer is reserved for Arrow extension types that the canonical path can't express.DataFusion wiring
vortex-datafusionnow goes through the session for schema/array conversion:convert/schema.rs::calculate_physical_schemausesArrowSession::to_arrow_fieldso extension metadata survives projection.persistent/format.rsandpersistent/opener.rsroute schema inference through the session.persistent/sink.rsusesfrom_arrow_record_batch, passing the original schema separately fromRecordBatch::schema()to preserveARROW:extension:namemetadata that DataFusion strips at runtime.Tests
Two new end-to-end tests in
vortex-datafusion/src/persistent/tests.rs:arrow_uuid_extension_roundtrip— write Arrow UUID column to a Vortex file via the session,SELECT *it back, assert the field still carries theUuidextension type and the values match.arrow_uuid_extension_roundtrip_nested_struct— same flow with the UUID nested in a top-levelStruct, exercising recursive session-aware schema inference.