[request] pl.describe() improvements about polars HOT 3 OPEN

See also:

• #10270

We're generally in favour of adding n_unique, but there are performance considerations for large data frames; I think the only thing holding it back from being added is that nobody has actually quantified the impact on large frames yet 😉

If the impact is relatively minimal then we could certainly add this. Alternatively we could perhaps provide approx_n_unique results instead (labeled appropriately) if the impact is more significant.

from polars.

I am in favor of a unique count! However, I would call it "distinct count" since unique and distinct are not the same 😀

For the others I am not sure. I can see them being usefull for specific use cases. Maybe add some parameters to describe to optionally add more statistics.

I really like the idea of describe being an one-stop-shop for all commonly used statistics instead of sprinkling around 100x cells in the notebook all the time 🤣

from polars.

Yes, I have found describe to not be enough for real life application (you know those hardcoded "999999"). I'd like to get other opinions, then i'll try to handle it.

from polars.