datafusion-contrib/arrow-zarr

Implement proper async reader

Opened this issue · 0 comments

Right now, the async reader "blocks" the thread when decompressing chunks of data (and there might also be other things blocking the thread, like generating the record batches, I'm not sure but something is taking a significant amount of time).

I think the right way to do this is to spawn a tokio blocking task for the part where the chunks is decompressed and everything that follows, so that the async reader can truly be async.