Removed unnecessary params in cum_func

[gfyoung]

Picks up from #13167 by properly removing the parameters and ensuring that numpy compatibility has been maintained. The current test suite does a good job of checking that already, so no tests were added.

Closes #13541.

[codecov-io]

Current coverage is 84.34%

Merging #13550 into master will increase coverage by <.01%

@@             master     #13550   diff @@
==========================================
  Files           138        138          
  Lines         51113      51119     +6   
  Methods           0          0          
  Messages          0          0          
  Branches          0          0          
==========================================
+ Hits          43109      43115     +6   
  Misses         8004       8004          
  Partials          0          0          

Powered by Codecov. Last updated by a01644c...6ba7a77

[jorisvandenbossche]

Looks good to me!

[jreback]

do need a test for this?
as prior treats didn't catch this

[gfyoung]

Catch what? We already have tests in fact to ensure that the numpy call won't break as well as one for non-default arguments in test_generic.py.

[jreback]

what I mean is I suspect we have more functions that have named numpy args - they work but we are specifically naming them

[gfyoung]

AFAIU cum_func is a template for all of the accumulation functions for DataFrame and Series at the very least. What other functions are you referring to?

[jreback]

no my point is that there might be others that accept the dtype argument (or out) and yet are validated, but we really should remove the actual named arguments. This one was not caught, so there might be others. You might have to inspect them manually. Not sure you can tell the difference between a named arg and a kwarg once your function sees it.

[gfyoung]

@jreback : GitHub search indicates that out=None is essentially nonexistent in the code-base except for cum_func, algorithms.py and function.py (my numpy compat module).

However, the same cannot be said with dtype=None. Manual inspection of every dtype=None? That's just being ridiculous.

Fortunately, a GitHub search indicates that all other instances of dtype=None are called from functions that I know definitely do not exist in numpy, so I think we are good to go on that.

[jreback]

ok, then lgtm. yeah the out in algorithms.py is ok.