Columnar merge batcher#693
Merged
frankmcsherry merged 1 commit intoTimelyDataflow:master-nextfrom Mar 22, 2026
Merged
Conversation
Replace the chainless_batcher::BatcherStorage implementations for ValStorage and KeyStorage with InternalMerge implementations that work with DD's standard MergeBatcher pipeline. This demonstrates that custom trie-shaped columnar containers can use DD's existing merge infrastructure without modifications to DD itself. Changes: - Implement InternalMerge for ValStorage and KeyStorage - Implement SizableContainer for both storage types - Add TrieChunker: a pass-through chunker for pre-sorted trie data - Update batcher type aliases to use MergeBatcher + InternalMerger - Remove chainless_batcher dependency from the example Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Work with Claude to bring changes from a DD columnar implementation into the
columnar.rsexample. In particular, this walks back that example's use of the chainless merge batcher, which is sufficiently risky to want to avoid. The recent change to internal iteration in DD has made this easier. Other functionality was brought over, and .. it's worth seeing what might be redundant with what is already here (which was working, to some extent). The goal is to get the example to look like what it would be for a repo-external human to assemble the necessary parts for a new container type.The main additions are a pass-through chunker (the trie-shaped containers are ready to go), a
SizeableContainerimplementation so that the merge batcher can snip batches at the right boundaries. Also pivots of the trait methods of the chainless batchersBatcherStoragetoInternalMerge; they are similar, except that merging yields when low on fuel, and the extraction draws out the frontier as well.The
columnarexample vs thespines newexample suggests that there is some potential here, but be warned that some amount of this is because the harness produces sorted data by accident, and that's helpful for the ordered columnar container.