Plan for supporting data destinations

[wangxiaoying]

Other than Pandas.DataFrame, we plan to support the following destinations:

• Dask.DataFrame
• Modin.DataFrame
• PyArrow

Reply here if you have other ideas!

[ritchie46]

I think that using arrow as data destination gives you a lot of future destinations for free. I believe arrow has the fastest algorithm to ingest data into pandas without triggering block manager re-allocations. This would give you zero copy transfer to products build around apache arrow like Polars and Dremio. And probably fast ingestion to other products.

I could help on the arrow interop.