buffered strategy to not use eof for the final chunk#7219
Merged
onursatici merged 2 commits intodevelopfrom Apr 1, 2026
Merged
buffered strategy to not use eof for the final chunk#7219onursatici merged 2 commits intodevelopfrom
onursatici merged 2 commits intodevelopfrom
Conversation
Signed-off-by: Onur Satici <onur@spiraldb.com>
github-actions
bot
removed
the
action/benchmark
Merging this PR will not alter performance
Comparing 
Footnotes
|
Contributor
Polar Signals Profiling ResultsLatest Run
Powered by Polar Signals Cloud |
Contributor
Benchmarks: PolarSignals ProfilingVortex (geomean): 1.186x ❌ datafusion / vortex-file-compressed (1.186x ❌, 0↑ 9↓)
|
Contributor
Benchmarks: TPC-H SF=1 on NVMEVerdict: No clear signal (environment too noisy confidence) datafusion / vortex-file-compressed (1.017x ➖, 1↑ 1↓)
datafusion / vortex-compact (0.994x ➖, 1↑ 0↓)
datafusion / parquet (1.029x ➖, 1↑ 2↓)
datafusion / arrow (1.052x ➖, 1↑ 4↓)
duckdb / vortex-file-compressed (1.016x ➖, 0↑ 0↓)
duckdb / vortex-compact (1.045x ➖, 0↑ 0↓)
duckdb / parquet (0.968x ➖, 6↑ 2↓)
duckdb / duckdb (1.061x ➖, 0↑ 6↓)
Full attributed analysis
|
Contributor
Benchmarks: FineWeb NVMeVerdict: No clear signal (low confidence) datafusion / vortex-file-compressed (0.997x ➖, 0↑ 0↓)
datafusion / vortex-compact (1.016x ➖, 0↑ 1↓)
datafusion / parquet (0.985x ➖, 0↑ 0↓)
duckdb / vortex-file-compressed (1.002x ➖, 0↑ 0↓)
duckdb / vortex-compact (0.986x ➖, 0↑ 0↓)
duckdb / parquet (1.004x ➖, 0↑ 0↓)
Full attributed analysis
|
Contributor
Benchmarks: TPC-DS SF=1 on NVMEVerdict: No clear signal (low confidence) datafusion / vortex-file-compressed (0.936x ➖, 29↑ 0↓)
datafusion / vortex-compact (0.964x ➖, 21↑ 1↓)
datafusion / parquet (0.910x ➖, 41↑ 0↓)
duckdb / vortex-file-compressed (0.922x ➖, 41↑ 1↓)
duckdb / vortex-compact (0.993x ➖, 0↑ 1↓)
duckdb / parquet (0.974x ➖, 6↑ 0↓)
duckdb / duckdb (0.986x ➖, 4↑ 0↓)
Full attributed analysis
|
Contributor
Benchmarks: TPC-H SF=10 on NVMEVerdict: No clear signal (environment too noisy confidence) datafusion / vortex-file-compressed (0.918x ➖, 9↑ 0↓)
datafusion / vortex-compact (0.961x ➖, 3↑ 0↓)
datafusion / parquet (0.904x ➖, 12↑ 0↓)
datafusion / arrow (0.892x ✅, 10↑ 0↓)
duckdb / vortex-file-compressed (0.850x ✅, 20↑ 0↓)
duckdb / vortex-compact (0.879x ✅, 16↑ 0↓)
duckdb / parquet (0.937x ➖, 3↑ 0↓)
duckdb / duckdb (0.991x ➖, 0↑ 0↓)
Full attributed analysis
|
Contributor
Benchmarks: TPC-H SF=1 on S3Verdict: No clear signal (environment too noisy confidence) datafusion / vortex-file-compressed (1.114x ➖, 1↑ 6↓)
datafusion / vortex-compact (1.269x ➖, 0↑ 10↓)
datafusion / parquet (1.027x ➖, 0↑ 2↓)
duckdb / vortex-file-compressed (1.018x ➖, 0↑ 2↓)
duckdb / vortex-compact (1.037x ➖, 0↑ 0↓)
duckdb / parquet (1.065x ➖, 0↑ 0↓)
Full attributed analysis
|
Contributor
Benchmarks: FineWeb S3Verdict: No clear signal (low confidence) datafusion / vortex-file-compressed (1.464x ❌, 0↑ 5↓)
datafusion / vortex-compact (0.986x ➖, 0↑ 0↓)
datafusion / parquet (1.061x ➖, 0↑ 0↓)
duckdb / vortex-file-compressed (0.895x ➖, 2↑ 0↓)
duckdb / vortex-compact (0.933x ➖, 0↑ 0↓)
duckdb / parquet (1.065x ➖, 0↑ 0↓)
Full attributed analysis
|
Contributor
Benchmarks: Statistical and Population GeneticsVerdict: No clear signal (low confidence) duckdb / vortex-file-compressed (0.939x ➖, 2↑ 0↓)
duckdb / vortex-compact (0.925x ➖, 1↑ 0↓)
duckdb / parquet (0.976x ➖, 0↑ 0↓)
Full attributed analysis
|
Contributor
Benchmarks: Clickbench on NVMEVerdict: No clear signal (low confidence) datafusion / vortex-file-compressed (0.948x ➖, 5↑ 0↓)
datafusion / parquet (0.930x ➖, 10↑ 1↓)
duckdb / vortex-file-compressed (1.044x ➖, 0↑ 7↓)
duckdb / parquet (1.015x ➖, 0↑ 1↓)
duckdb / duckdb (0.999x ➖, 1↑ 1↓)
Full attributed analysis
|
Contributor
Benchmarks: Random AccessVortex (geomean): 0.905x ➖ unknown / unknown (0.973x ➖, 5↑ 1↓)
|
Contributor
Benchmarks: TPC-H SF=10 on S3Verdict: No clear signal (environment too noisy confidence) datafusion / vortex-file-compressed (1.126x ➖, 0↑ 4↓)
datafusion / vortex-compact (1.069x ➖, 1↑ 3↓)
datafusion / parquet (0.990x ➖, 2↑ 3↓)
duckdb / vortex-file-compressed (1.033x ➖, 0↑ 1↓)
duckdb / vortex-compact (1.081x ➖, 0↑ 1↓)
duckdb / parquet (1.102x ➖, 0↑ 1↓)
Full attributed analysis
|
Contributor
Benchmarks: CompressionVortex (geomean): 0.975x ➖ unknown / unknown (0.949x ➖, 23↑ 3↓)
|
Signed-off-by: Onur Satici <onur@spiraldb.com>
robert3005
approved these changes
Apr 1, 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
3 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
fixes buffered layout writer so it doesn't write the final chunk on the eof pointer. Eof should only be used for data that the writer wants to place at the end of the file. Buffered writer was writing regular buffered data to there which did mess up ordering of some segments.
Previously struct writer was using a transposed stream without spawning a task per column, on that world buffering was deadlocky. That is changed for a while to spawn now, so we should be deadlock safe.
I did try converting all clickbench files repeatedly, as well as the public bi datasets and randomly generated wide tables but I couldn't deadlock this.
fixes #7234
fixes #7236
Testing
add a vortex file test that asserts the dict layout segments are in the right order, as well as zone maps across columns