ENH: Dask backend

[zbrookle]

Are there any plan to create a dask backend? I imagine the implementation would be near identical to the pandas one that already exists

[jreback]

would surely take this, and agree its pretty darn straightforward to do.

[goodwanghan]

Hey, can you also add the support for Ray (modin?) and Rapids AI? I think they are all similar

[acivitillo]

One difference between pandas and dask is that in dask you can build a computational graph programmatically. We use this pattern to create an extract graph to run ETL jobs from various backends (Redshift, Salesforce, etc.). We define the columns and other attributes in a .json and then pass that to an Extract() to generate and submit the graph to a cluster. Since we are planning to support various enterprise backends, we use Arrow to store data in memory before sending it to Parquet over S3.

[cpcloud]

This exists.