[DISCUSSION] How to deal with limited annotation support in `dask-expr` about dask HOT 3 OPEN

My guess is that query optimization will make more users happy than losing annotations will make users sad. Because of this, I'm in favor of merging in query optimization even if it disrupts annotations. I think that it's ok to make some existing users sad.

Of course I'd be happy to see a better solution here, but I wouldn't want this to slow things down much.

from dask.

I agree that query optimization will be of much bigger benefit than annotations for most users. However, I don't think breaking things silently will be a good UX for those users who are affected. To move forward quickly, I think that adding a warning to dask.annotate, which contains more information, might be a nice compromise. Similar to #10925, users should be able to silence this warning.

from dask.

Yup, I don't mind seeing improvements to annotations UX. However, I don't think that the release of query planning should be slowed down by this. I agree with scope, I wouldn't agree with a blocking change (not that you're proposing that) and I'm not sure how highly I would prioritize this (I don't see many users with annotations in dataframes). However, if you have different knowledge (for example if there has been evidence that people use annotations with dask.dataframe) then I could agree that it's worth prioritizing. If there has not been evidence that this is in use then I probably wouldn't bother (we have plenty of other high priority work to do) unless it's very easy.

from dask.