separate _libs/src/reduce.pyx to _libs.reduction

[jbrockmendel]

• closes #xxxx
• tests added / passed
• passes git diff upstream/master -u -- "*.py" | flake8 --diff
• whatsnew entry

[jreback]

this is fine, I guess we can't use reduce as a name?

[jreback]

you don't like reduce?

[jbrockmendel]

I try to avoid overlap with built-in names.

[jreback]

needs a rebase

[codecov]

Codecov Report

Merging #19306 into master will decrease coverage by 0.02%.
The diff coverage is 100%.

@@            Coverage Diff             @@
##           master   #19306      +/-   ##
==========================================
- Coverage   91.53%    91.5%   -0.03%     
==========================================
  Files         150      150              
  Lines       48738    48738              
==========================================
- Hits        44611    44599      -12     
- Misses       4127     4139      +12

Flag	Coverage Δ
#multiple	89.87% <100%> (-0.03%)	⬇️
#single	41.72% <37.5%> (ø)	⬆️

Impacted Files	Coverage Δ
pandas/core/groupby.py	92.14% <100%> (ø)	⬆️
pandas/core/apply.py	99.42% <100%> (ø)	⬆️
pandas/plotting/_converter.py	65.22% <0%> (-1.74%)	⬇️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update f0cd23c...a63da0f. Read the comment docs.

[jreback]

remove this, and instead where .reduce() is called you can do this on the python side.

[jbrockmendel]

We'd also need to do that everywhere SeriesGrouper.get_result or SeriesBinGrouper.get_result is called. What's the upside that makes the duplicated code worthwhile?

[jreback]

this completely removes this module from the import time dependency chain.

you could do this in the top-level function reduce() I suppose (inside the function)

[jreback]

or could expose maybe_convert_objects in a pxd and then call it directly in cython (but that actually puts this back in the dep chain).

[jbrockmendel]

you could do this in the top-level function reduce() I suppose (inside the function)

That would be better than trying to do it after every place in the code-base where the function is called. But note that reduce is not the only part of this module that is used externally, so we'd need to do this run-time import in three places in this module.

Since the dependency is one-way and not a c-dep, I don't see a huge downside to the import-time import of maybe_convert_objects.

[jreback]

ok, this at least de-couples some lib.pyx, which is good.