STY: proposed isort settings [ci skip] [skip ci] [ciskip] [skipci]

[jbrockmendel]

(not sure which tag(s) skip the CI)

I propose we change the isort settings to arrange within-pandas imports by dependency structure. i.e.

• pandas._libs, pandas.compat, pandas.util._*, pandas.errors are for the most part not dependent on pandas.core, so get their own section(s)
• pandas.core.dtypes for the most part does not depend on the rest of pandas.core, so it goes next
• then the rest of pandas.core
• then pandas non-core: pandas.io, pandas.plotting, pandas.tseries

Within blocks I propose we stick to alphabetical ordering. e.g. right now in core.series we have import pandas.core.indexes.base as ibase 35 lines away from all the other pandas.core.indexes imports. These should be adjacent.

The main thing I'd still like to fix/improve is that in this PR isort is failing to put from pandas import compat with the other pandas.compat imports.

Putting the trailing parentheses on the same line as the imports instead of a separate line is less important, but I prefer it because it makes it cleaner when I wrap lines in sublimetext. Happy to revert that if it is a sticking point.

[pep8speaks]

Hello @jbrockmendel! Thanks for submitting the PR.

• There are no PEP8 issues in the file pandas/core/series.py !

[jreback]

this is going to massively change the in-progress isort cleanups no?

[jbrockmendel]

this is going to massively change the in-progress isort cleanups no?

Yes. If there is an agreement that this format is better (it is), then better to implement it before more isort work is done.

Also a bunch of modules already use approx this pattern, e.g. core.internals, datetimelike arrays

[jreback]

ok does this break anything i the current build?

[jbrockmendel]

ok does this break anything i the current build?

Yes, a bunch of files. If this format is agreed upon, I can either add those back to the ignore list in setup.cfg or go through and fix them all in one swoop now.

[jreback]

either ok

[jbrockmendel]

updated, now passes isort -rc pandas

[jreback]

is the -rc needed? if so can u update auto check and instructions

[codecov]

Codecov Report

Merging #23366 into master will not change coverage.
The diff coverage is 100%.

@@           Coverage Diff           @@
##           master   #23366   +/-   ##
=======================================
  Coverage   92.18%   92.18%           
=======================================
  Files         161      161           
  Lines       51160    51160           
=======================================
  Hits        47161    47161           
  Misses       3999     3999

Flag	Coverage Δ
#multiple	90.61% <100%> (ø)	⬆️
#single	42.24% <97.93%> (ø)	⬆️

Impacted Files	Coverage Δ
pandas/io/sas/sasreader.py	86.2% <ø> (ø)	⬆️
pandas/core/groupby/categorical.py	95.45% <ø> (ø)	⬆️
pandas/io/packers.py	88.08% <100%> (+0.03%)	⬆️
pandas/io/parquet.py	73.72% <100%> (ø)	⬆️
pandas/tseries/api.py	100% <100%> (ø)	⬆️
pandas/core/groupby/ops.py	96.79% <100%> (ø)	⬆️
pandas/core/groupby/groupby.py	96.48% <100%> (ø)	⬆️
pandas/core/groupby/base.py	91.83% <100%> (ø)	⬆️
pandas/core/groupby/grouper.py	98.16% <100%> (ø)	⬆️
pandas/io/clipboards.py	100% <100%> (ø)	⬆️
... and 10 more

---

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 96d321c...51834f0. Read the comment docs.

[jreback]

thanks @jbrockmendel let's hope this works :>

[alimcmaster1]

I like this alot - thanks @jbrockmendel ! I did actually try this myself since you requested it back in #23048 . But isort seemed to produce non-deterministic sort for me. But I think i wasn't sorting within sections like you added force_sort_within_sections=True

[jorisvandenbossche]

One disadvantage is that it will be harder to know how to correctly do the imports manually, so I think we need to explain the logic in the guidelines (@alimcmaster1 if you can do that in #23364, that would be great)

[alimcmaster1]

Sure happy to @jorisvandenbossche , but I hope most people will use isort to sort imports opposed to doing so manually

[jorisvandenbossche]

Yes, but still, if I just need to add an import to an already correctly sorted file, I think it is nice to know how to do this correctly instead of each time to have to run isort after making a small change (although you could maybe configure your editor to do so, or with a git commit hook)