⚡ Bolt: Optimize hash join build-side for single-batch inputs #240
Conversation
Adds a fast path to the `collect_left_input` function in the hash join implementation. When the build side consists of a single `RecordBatch`, this optimization avoids the expensive `concat_batches` operation, reducing memory allocation and CPU usage. The previous implementation would unconditionally concatenate batches, even when there was only one, leading to unnecessary overhead. This change checks for the single-batch case and processes it directly, falling back to the original logic for zero or multiple batches.
|
👋 Jules, reporting for duty! I'm here to lend a hand with this pull request. When you start a review, I'll add a 👀 emoji to each comment to let you know I've read it. I'll focus on feedback directed at me and will do my best to stay out of conversations between you and other bots or reviewers to keep the noise down. I'll push a commit with your requested changes shortly after. Please note there might be a delay between these steps, but rest assured I'm on the job! For more direct control, you can switch me to Reactive Mode. When this mode is on, I will only act on comments where you specifically mention me with New to Jules? Learn more at jules.google/docs. For security, I will only act on instructions from the user who triggered this task. |
💡 What: This change introduces a fast path in the
HashJoinExecoperator for cases where the build-side input consists of a singleRecordBatch.🎯 Why: Previously, the build-side logic would unconditionally call
concat_batches, even if there was only one batch. This incurred unnecessary overhead from memory allocation and data copying. By bypassing this operation for the common single-batch case, we can make the join more efficient.📊 Impact: This optimization improves the performance of hash joins by reducing memory allocation and CPU usage, especially in scenarios involving small dimension tables or when a
CoalesceBatchesExechas been applied upstream.🔬 Measurement: The performance improvement can be measured by running benchmarks (like TPC-H) where joins have a single-batch build side and observing the reduction in execution time and peak memory usage.
PR created automatically by Jules for task 18365902574244747359 started by @Dandandan