VarBinBuilder: eagerly set IsSorted stat on offsets #6452
Conversation
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
VarBinBuilder guarantees monotonically increasing offsets, but finish() wasn't setting the IsSorted stat. This caused an O(n) recomputation every time a deserialized VarBinArray was validated, which showed up as a hot path in production profiles. Signed-off-by: Dimitar Dimitrov <[email protected]>
dimitarvdimitrov
changed the title
Contributor
|
Can also do this for runend varbinview if not |
Signed-off-by: Dimitar Dimitrov <[email protected]>
Author
i couldn't find where in the VarBinViewArrayBuilder we create runend offsets for VarBin(View) arrays. Can you point me pls? |
Contributor
|
varbinview: also maybe sequence array: |
RunEndArray::validate didn't check that ends are strictly increasing, which is required for binary search to work correctly. Also caches IsSorted and IsStrictSorted on the ends after construction so downstream consumers don't recompute it. Signed-off-by: Dimitar Dimitrov <[email protected]>
Merging this PR will degrade performance by 10.64%
Performance Changes
Comparing 
Footnotes
|
SequenceArray (A[i] = base + i * multiplier) can determine sortedness from the sign of the multiplier at construction time. Set both stats eagerly so we skip the kernel entirely. Signed-off-by: Dimitar Dimitrov <[email protected]>
Author
Contributor
joseph-isaacs
left a comment
joseph-isaacs
left a comment
There was a problem hiding this comment.
we need to be careful on validation cost on the release path
| let stats_set = ArrayStats::default(); | ||
| stats_set.set(Stat::IsSorted, StatPrecision::Exact(is_sorted.into())); | ||
| stats_set.set( | ||
| Stat::IsStrictSorted, |
Contributor
There was a problem hiding this comment.
think we can get away with a statically alloced CoW here so we don't need to alloc a stats set here?
Not sure.
Maybe also include min and max and other stats?
| // and strictly sorted iff multiplier > 0. | ||
| let (is_sorted, is_strict_sorted) = match_each_native_ptype!(ptype, |P| { | ||
| let m = multiplier.cast::<P>(); | ||
| (m >= zero::<P>(), m > zero::<P>()) |
Contributor
There was a problem hiding this comment.
not sure we allow a zero multiplier. if no we should not allow?
| // Run ends are always strictly sorted (and therefore sorted) by invariant. | ||
| // Cache this so downstream consumers don't need to recompute it. | ||
| ends.statistics() | ||
| .set(Stat::IsStrictSorted, StatPrecision::Exact(true.into())); |
Contributor
There was a problem hiding this comment.
does this run in release? it's very slow.
Can we set this on compression?
| } | ||
|
|
||
| // Run ends must be strictly sorted for binary search to work correctly. | ||
| if let Some(is_strict_sorted) = ends.statistics().compute_is_strict_sorted() { |
Contributor
There was a problem hiding this comment.
this is too expensive
joseph-isaacs
added
the
action/benchmark
github-actions
bot
removed
the
action/benchmark
Contributor
Polar Signals Profiling ResultsLatest Run
Powered by Polar Signals Cloud |
Contributor
Benchmarks: PolarSignals ProfilingSummary
Detailed Results Table
|
Contributor
Benchmarks: TPC-H SF=1 on NVMESummary
Detailed Results Table
|
Contributor
Benchmarks: FineWeb NVMeSummary
Detailed Results Table
|
Contributor
Benchmarks: Random AccessSummary
Detailed Results Table
|
Contributor
Benchmarks: TPC-H SF=1 on S3Summary
Detailed Results Table
|
Contributor
Benchmarks: TPC-DS SF=1 on NVMESummary
Detailed Results Table
|
Contributor
Benchmarks: TPC-H SF=10 on NVMESummary
Detailed Results Table
|
Contributor
Benchmarks: FineWeb S3Summary
Detailed Results Table
|
Contributor
Benchmarks: Statistical and Population GeneticsSummary
Detailed Results Table
|
Contributor
Benchmarks: TPC-H SF=10 on S3Summary
Detailed Results Table
|
Contributor
Benchmarks: CompressionSummary
Detailed Results Table
|
Contributor
Benchmarks: Clickbench on NVMESummary
Detailed Results Table
|
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.
Does this PR closes an open issue or discussion?
no
What this PR does?
VarBinBuilder guarantees monotonically increasing offsets, but finish() wasn't setting the IsSorted stat. This caused an O(n) recomputation every time a deserialized VarBinArray was validated.
How is this change tested?
unit tests
Are there any user-facing changes?
VarBin arrays now persist stats for their offsets child array. Existing persisted arrays will still be verified as sorted upon reading from disk.