Congrats to the Google Cloud Storage team for the contribution !
Crazy speed improvement in `datasets` when streaming a dataset using approximate shuffling
Arrow batches are buffered, concatenated, and then rows are shuffled all using optimized Arrow C++ operations now
