yigitcan-ozturk commented on issue #10392: URL: https://github.com/apache/arrow-rs/issues/10392#issuecomment-5599721170
I’d like to work on this if the scope is still available. Based on the discussion, I’d keep the first PR narrowly focused on an opt-in IPC reader configuration that copies uncompressed body buffers into independent allocations, preserving the existing zero-copy behavior by default. I’ll use the downstream DataFusion work in apache/datafusion#24592 as a reference, but keep the Arrow implementation at the IPC-reader layer rather than introducing DataFusion-specific behavior. I’d add regression coverage that verifies (1) the default path still shares the backing body allocation, (2) the opt-in path produces independently owned buffers for projected/retained columns, and (3) compressed IPC behavior and round-trip correctness remain unchanged. If that direction matches what you had in mind, I’m happy to take it through implementation and review. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
