Feature #3036: Window Spooling #14181
Merged
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
Contributor
hawkfish
commented
Oct 1, 2024
- Add a Seek method for built on a new PrevScanIndex method to support random access to collections. The scans are linear for now, but that should be good enough for the current task.
- Plumb through everything needed to accumulate WindowDataChunk tuples in a ColumnDataCollection.
- Convert the value storage for the "value" window functions to use collections. Note that the validity masks for IGNORE NULLS and EXCLUDE are still memory resident, but they are a lot smaller.
- Create an explicit class for WindowAggregatorLocalState that can handle building and reading the argument collection in parallel instead of using a DataChunk and locks.
- Convert the naïve aggregator to use collections.
- Convert the segment tree aggregator to use collections.
- Move all of the WindowDataChunk collection scanning functionality into a single class.
- Convert WindowDistinctAggregator to use paging collections.
- Create a wrapper for multi-threaded appending to WindowDataChunk objects.
- Move RANGE values to a collection.
- Convert the custom window functions to use collections instead of in-memory DataChunks.
- Track the insert data validity and pass it down to avoid checks.
- Pass down context so we can get real memory flushing sizes.
added 18 commits
September 11, 2024 11:26
Add a Seek method for built on a new PrevScanIndex method to support random access to collections. The scans are linear for now, but that should be good enough for the current task.
Plumb through everything needed to accumulate WindowDataChunk tuples in a ColumnDataCollection. Not used yet.
Convert the value storage for the "value" window functions to use collections. Note that the validity masks for IGNORE NULLS and EXCLUDE are still memory resident, but they are a lot smaller.
Create an explicit class for this that can handle building and reading the argument collection in parallel instead of using a DataChunk and locks.
Convert the naïve aggregator to use collections. Also fix loop invariant mistake in PrevScanIndex.
Convert the segment tree aggregator to use collections.
Move all of the WindowDataChunk collection scanning functionality into a single class.
Convert WindowDistinctAggregator to use paging collections.
Create a wrapper for multi-threaded appending to WindowDataChunk objects.
Move RANGE values to a collection.
Remove this helper as it relies on DataChunks.
Convert the custom window functions to use collections instead of in-memory DataChunks.
Track the insert data validity and pass it down to avoid checks. Also pass down context so we can get real memory flushing sizes.
lnkuiper
suggested changes
Oct 2, 2024
There was a problem hiding this comment.
Code changes look good, this is awesome! So mad, mode and quantile can now be computed on partitions if that single partition is larger than memory? Could you add a test for this? What if we have an evil query with multiple like this:
SELECT mad(c0) OVER (<window with one large partition that doesn't fit in memory>),
mode(c0) OVER (<same window>),
quantile(c0, 0.5) OVER (<same window>)
FROM tbl;Could you add a test for that?
Add tests to validate that the custom aggregators work with spooling. Also fix PR feedback and add missing PrepareMergeStage call.
lnkuiper
approved these changes
Oct 7, 2024
|
Thanks! |
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.