kylebarron commented on PR #10431: URL: https://github.com/apache/arrow-rs/pull/10431#issuecomment-5699931452
I understand that it was important to close the unsoundness hole, but this change is frustrating as it forces me to vendor all of the `FFI_ArrowArrayStream` FFI code into my project, since it's now currently impossible to create an `FFI_ArrowArrayStream` except through a `RecordBatchReader`. In particular, it's impossible in arrow-rs to define a stream of _Arrays_; arrow-rs only supports defining a stream of _RecordBatches_. (This is required to describe a concept like a "Chunked Array"). This is documented in https://github.com/apache/arrow-rs/issues/6586, which also has more context. As it stands, I have my own [`ArrowArrayStreamReader`](https://github.com/kylebarron/arro3/blob/d65a1a5eeee0af3a0e292111ed077c0c181fb7fc/pyo3-arrow/src/ffi/from_python/ffi_stream.rs) (a counterpart to the `arrow-rs` [`ArrowArrayStreamReader`](https://docs.rs/arrow/latest/arrow/array/ffi_stream/struct.ArrowArrayStreamReader.html) but which manages a more generic `FieldRef` instead of the stricter `SchemaRef`). But I had been able to use the public fields on `FFI_ArrowArrayStream` without vendoring that code. I guess the best way to solve this is to push for inclusion of an `ArrayReader` concept in arrow-rs. -- 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]
