Skip to content

[python] Pipeline split reads in Arrow batch reader#8807

Draft
XiaoHongbo-Hope wants to merge 3 commits into
apache:masterfrom
XiaoHongbo-Hope:split_streaming
Draft

[python] Pipeline split reads in Arrow batch reader#8807
XiaoHongbo-Hope wants to merge 3 commits into
apache:masterfrom
XiaoHongbo-Hope:split_streaming

Conversation

@XiaoHongbo-Hope

Copy link
Copy Markdown
Contributor

Purpose

Pipeline split reads in to_arrow_batch_reader without materializing the full result.

The reader uses bounded per-split queues, preserves input split order, and follows parallelism / read.parallelism. This benefits multi-split streaming consumers such as Ray, BLOB streaming, and batch-oriented training readers.

Tests

  • pypaimon/tests/reader_parallel_test.py
  • pypaimon/tests/ray_data_test.py
  • DedicatedFormatWriterTest#test_to_arrow_batch_reader
  • BlobParallelismTest#test_to_arrow_batch_reader_blob_parallelism
  • MultimodalTableTest#test_scan_stream_blobs

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant