REF: EA quantile logic to EA._quantile

[jbrockmendel]

• closes #xxxx
• tests added / passed
• Ensure all linting tests pass, see here for how to run them
• whatsnew entry

[jorisvandenbossche]

I am personally not super fond of moving this code. The ExtensionArray class also serves as a base class and example implementation for external EA authors, and for that purpose I am not sure this is useful (it allows to override it, of course (I don't know if we had feature requests for that up to now?), but for that purpose I would also simplify the base implementation).

(the dual use case as base class for internal usage vs external usage is sometimes a bit conflicting ..)

Is it really annoying to have the code in array_algos/quantile.py ?

[jorisvandenbossche]

This docstring would need to be updated to be written for EA implementors (eg for those it doesn't matter this is a compatility layer for quantile_with_mask)

[jbrockmendel]

Good point, will update

[jbrockmendel]

(the dual use case as base class for internal usage vs external usage is sometimes a bit conflicting ..)

i sympathize with this

Is it really annoying to have the code in array_algos/quantile.py ?

A bit, yes. The idea of array_algos is to not depend on EA implementation details (the docstring in array_algos.__init__ hints at this). More importantly, the special-treatment of pandas-internal (i.e. MaskedArray and a bit more subtly Sparse) is a pattern we've been trying to avoid. I don't see any other way around this, am open to ideas.

[jreback]

am fine with this, though anyway to combine some of the added code that is in base & masked?

[jbrockmendel]

though anyway to combine some of the added code that is in base & masked?

there might be, but i haven't found it. trouble is we call the constructors differently in the masked code