REGR: fix DataFrame sum and prod with min_count and numeric_only

[simonjayhawkins]

• closes BUG: Reduction operations fail #41074
• tests added / passed
• Ensure all linting tests pass, see here for how to run them
• whatsnew entry

it appears that the clean-up in #35899 to simplify _reduce may have been a bit premature, have reinstated a few lines of the removed code for backport purposes.

The longer term fix would probably be to support min_count in _mgr.reduce, #40143 (comment)

[jbrockmendel]

this should still hold?

[jbrockmendel]

oh i guess it should be assert numeric_only is None or min_count != 0?

[simonjayhawkins]

correct we are only using the old code path when min_count is given until _mgr.reduce can handle min_count.

why do we need to add an assert?

[jbrockmendel]

not needed, just took me a minute to understand what had changed

[jreback]

this overlaps with #41706 ?

[jreback]

ok so this is an alt to #41711 ?

[simonjayhawkins]

this overlaps with #41706 ?

#41711? yes and no. we could just do #41711 and not backport. do this, backport this and then merge #41711. or merge #41711 and I reopen this directly against 1.2.x

those options don't include do #41711, backport it and close this one.

[simonjayhawkins]

ok so this is an alt to #41711 ?

no really, this intended as a backport? #41711 was opened after is the alt.

[jreback]

this overlaps with #41706 ?

#41711? yes and no. we could just do #41711 and not backport. do this, backport this and then merge #41711. or merge #41711 and I reopen this directly against 1.2.x

those options don't include do #41711, backport it and close this one.

#41711 and I reopen this directly against 1.2.x

why is this not the preferred option?

[simonjayhawkins]

#41711 and I reopen this directly against 1.2.x

why is this not the preferred option?

because we normally merge to master and backport. If #41711 gets merged first. then our options reduce.

[simonjayhawkins]

closing in favor of #41711 (comment)

The change here was effectively only

        else:
            if numeric_only:
                data = _get_data()
                labels = data._get_agg_axis(axis)
                values = data.values
            result = func(values)

when min_count and numeric_only is given, without touching any other library code