Add LRU cache eviction to CombinePerKeyPrecombineOperator#37466
Add LRU cache eviction to CombinePerKeyPrecombineOperator#37466damccorm merged 2 commits intoapache:masterfrom
Conversation
- Added touchKey() method to move accessed keys to end (most recently used) - Added flushLRU() method to evict entries from front (least recently used) - Updated process() to track key usage for proper LRU ordering JavaScript Maps preserve insertion order, so we implement LRU by: - On access: delete and re-insert the key (moves to end) - On eviction: iterate from front (oldest entries first) Fixes apache#37465
Summary of ChangesHello @junaiddshaukat, I'm Gemini Code Assist1! I'm currently reviewing this pull request and will post my feedback shortly. In the meantime, here's a summary to help you and other reviewers quickly get up to speed! This pull request significantly enhances the Highlights
🧠 New Feature in Public Preview: You can now enable Memory to help Gemini Code Assist learn from your team's feedback. This makes future code reviews more consistent and personalized to your project's style. Click here to enable Memory in your admin console. Using Gemini Code AssistThe full guide for Gemini Code Assist can be found on our documentation page, here are some quick tips. Invoking Gemini You can request assistance from Gemini at any point by creating a comment using either
Customization To customize Gemini Code Assist for GitHub experience, repository maintainers can create a configuration file and/or provide a custom code review style guide (such as PEP-8 for Python) by creating and adding files to a Limitations & Feedback Gemini Code Assist may make mistakes. Please leave feedback on any instances where its feedback is incorrect or counter productive. You can react with 👍 and 👎 on @gemini-code-assist comments. If you're interested in giving your feedback about your experience with Gemini Code Assist for Github and other Google products, sign up here. You can also get AI-powered code generation, chat, as well as code reviews directly in the IDE at no cost with the Gemini Code Assist IDE Extension. Footnotes
|
|
Hi @damccorm, added LRU to PrecombineOperator. Please have a look when you got time, Thanks. |
|
Assigning reviewers: R: @Abacn added as fallback since no labels match configuration Note: If you would like to opt out of this review, comment Available commands:
The PR bot will only process comments in the main thread (not review comments). |
|
/gemini review |
![gemini-code-assist[bot]](https://avatars.githubusercontent.com/in/956858?s=60&v=4){kind=small}
There was a problem hiding this comment.
Code Review
This pull request implements LRU cache eviction for CombinePerKeyPrecombineOperator, which is a valuable enhancement. The implementation is generally sound, but I've identified a couple of areas for improvement. Firstly, the logic in the process method for updating the cache can be simplified and made more efficient, which would also eliminate the need for the new touchKey helper method. Secondly, there is considerable code duplication between the new flushLRU method and the existing flush method, which should be refactored to enhance maintainability. My specific comments provide detailed suggestions for addressing these points.
| const existingAccumulator = this.groups.get(wkey); | ||
| const newAccumulator = | ||
| existingAccumulator !== undefined | ||
| ? this.combineFn.addInput(existingAccumulator, wvalue.value.value) | ||
| : this.combineFn.addInput( | ||
| this.combineFn.createAccumulator(), | ||
| wvalue.value.value, | ||
| ); | ||
| // Move to end (most recently used) by delete + set | ||
| this.touchKey(wkey, newAccumulator); |
There was a problem hiding this comment.
The logic for updating the accumulator and managing the LRU behavior can be simplified and made slightly more performant. The current approach uses a touchKey helper method which is called for every element. This results in a delete operation even for new keys, which is unnecessary. By inlining this logic into the process method, we can avoid the helper and the extra operation for new keys, making the code more concise and efficient. This change would also make the touchKey method (lines 477-480) obsolete, allowing for its removal.
const accumulator = this.groups.get(wkey);
if (accumulator !== undefined) {
this.groups.delete(wkey);
}
this.groups.set(
wkey,
this.combineFn.addInput(
accumulator ?? this.combineFn.createAccumulator(),
wvalue.value.value
)
);| flushLRU(target: number): ProcessResult { | ||
| const result = new ProcessResultBuilder(); | ||
| const toDelete: string[] = []; | ||
| // Iterate from the front (oldest/least recently used entries) | ||
| for (const [wkey, values] of this.groups) { | ||
| if (this.groups.size - toDelete.length <= target) { | ||
| break; | ||
| } | ||
| const parts = wkey.split(" "); | ||
| const encodedWindow = parts[0]; | ||
| const encodedKey = parts[1]; | ||
| const window = decodeFromBase64(encodedWindow, this.windowCoder); | ||
| result.add( | ||
| this.receiver.receive({ | ||
| value: { | ||
| key: decodeFromBase64(encodedKey, this.keyCoder), | ||
| value: values, | ||
| }, | ||
| windows: [window], | ||
| timestamp: window.maxTimestamp(), | ||
| pane: PaneInfoCoder.ONE_AND_ONLY_FIRING, | ||
| }), | ||
| ); | ||
| toDelete.push(wkey); | ||
| } | ||
| for (const wkey of toDelete) { | ||
| this.groups.delete(wkey); | ||
| } | ||
| return result.build(); | ||
| } |
There was a problem hiding this comment.
There is significant code duplication between flushLRU and the flush method. The logic for iterating, decoding keys, sending results, and collecting keys for deletion is nearly identical. While the loop termination condition is checked at a different point, the core processing logic is the same. To improve maintainability and reduce redundancy, this common logic should be extracted into a private helper method that both flushLRU and flush can call.
There was a problem hiding this comment.
Do we even need to keep the existing flush implementation? We could probably replace it with a clear operation that is a simpler equivalent of flush(0)
There was a problem hiding this comment.
Great suggestion! You're absolutely right, since flush is only called with target=0 (in finishBundle), and flushLRU(0) would achieve the same result of flushing all entries, we can simply replace flush(target) with flushLRU(0). Let me fix that!
|
@damccorm, I have addressed all review feedback:
Let me know if further changes needed! |
2 participants
Summary
This PR implements LRU (Least Recently Used) cache eviction for
CombinePerKeyPrecombineOperator, addressing the TODO at line 487 ofoperators.ts.Changes
touchKey()method - Moves accessed keys to the end of the Map (most recently used position)flushLRU()method - Evicts entries from the front (least recently used) until cache is under limitprocess()method - Each key access now updates its LRU positionTechnical Details
JavaScript Maps preserve insertion order, so we implement LRU by:
Related Issues
operators.ts:487Testing
Fixes: #37465
Please add a meaningful description for your change here
Thank you for your contribution! Follow this checklist to help us incorporate your contribution quickly and easily:
addresses #123), if applicable. This will automatically add a link to the pull request in the issue. If you would like the issue to automatically close on merging the pull request, commentfixes #<ISSUE NUMBER>instead.CHANGES.mdwith noteworthy changes.See the Contributor Guide for more tips on how to make review process smoother.
To check the build health, please visit https://github.com/apache/beam/blob/master/.test-infra/BUILD_STATUS.md
GitHub Actions Tests Status (on master branch)
See CI.md for more information about GitHub Actions CI or the workflows README to see a list of phrases to trigger workflows.