Untangle reshape imports --> Define pivot_table in DataFrame #17215
Conversation
so that core.frame can import these safely at the top of the module. This allows us to define DataFrame.pivot_table inside DataFrame instead of pinning it on from reshape.pivot This also lets us put docstrings in _shared_docs adjacent to the implementations without having to contort ourselves to make sure imports are ordered correctly.
…master/pandas into reshape_imports
Codecov Report
@@ Coverage Diff @@
## master #17215 +/- ##
==========================================
- Coverage 91.02% 90.98% -0.05%
==========================================
Files 162 162
Lines 49517 49539 +22
==========================================
Hits 45075 45075
- Misses 4442 4464 +22
Continue to review full report at Codecov.
|
Codecov Report
@@ Coverage Diff @@
## master #17215 +/- ##
==========================================
- Coverage 91.02% 90.98% -0.05%
==========================================
Files 162 162
Lines 49517 49539 +22
==========================================
Hits 45075 45075
- Misses 4442 4464 +22
Continue to review full report at Codecov.
|
Codecov Report
@@ Coverage Diff @@
## master #17215 +/- ##
==========================================
- Coverage 91.02% 90.98% -0.05%
==========================================
Files 162 162
Lines 49517 49539 +22
==========================================
Hits 45075 45075
- Misses 4442 4464 +22
Continue to review full report at Codecov.
|
|
I like the other soln better. This is going backwards. |
This is an alternative to #17174. This is ultimately a better solution to the issue there, but involves a more invasive refactoring.
To recap:
Right now
DataFrame.pivot_tableis defined incore.reshape.pivotwithDataFrame.pivot_table = pivot_table. The original and main goal of this (and move pivot_table doc-string to DataFrame #17174) is to avoid that, and instead defineDataFrame.pivot_tablewithin theDataFrameclass definition.Doing this makes the docstrings tricky, since we want
pivot_table.__doc__to be shared byreshape.pivot.pivot_tableandDataFrame.pivot_table. To resolve this cleanly, we need to importreshape.pivotat the top ofcore.frame. Doing this requires getting the import ofDataFrameout ofreshape.pivot(and anything else it imports, namelyreshape.concat).Bonus: If/when this is approved, all of the mid-module imports of
reshape.concatcan be moved to the top ofcore.frame, and a handful of other docstrings can be put into_shared_docsin more convenient locations.Bonus 2: If/when this is approved, doing the same thing in
reshape.mergeandreshape.tileis very easy.reshape.reshapeis doable but take a bit more work.git diff upstream/master -u -- "*.py" | flake8 --diff