What is the status of array expressions?

Hi Dask community!

From the docs I am not quite sure what to expect from dask arrays.

There has been the GitHub - dask/dask-expr project which included array expressions (I don’t know to which extend), that now advertises that it is merged upstream and deprecated, but also that it is the default backend for dask dataframes.

It doesn’t say anything about arrays, and I can’t seem to find any info on array expressions in dask’s docs, though it seems that the array_expr functionality has been really merged into dask.

Are array expressions actually used? Are they maybe hiding under the name of HighLevelGraph?

Hi @vladidobro, welcome to Dask community!

dask-expr has been merged upstream and is the default backend for DataFrame. The same work of adding Query optimization or Expr into Array backend is on-going, directly in dask repo, but I think it isn’t done yet (see @fjetter comment here).

So no, array expressions are not used by default, and Expr backend is a replacement for HighLevelGraph. HLG are the internal currently used with Array.

Thank you for the explanation, looking forward to array Expr very much!

1 Like