CLN/BUG: Consolidate Index.astype and fix tz aware bugs

[jschendel]

• closes CLN: consolidate Index.astype #18704
• closes BUG: DatetimeIndex.astype doesn't fully handle tz aware dtypes #18951
• tests added / passed
• passes git diff upstream/master -u -- "*.py" | flake8 --diff
• whatsnew entry

Only behavioral changes:

• Allowed .astype(str) on TimedeltaIndex and PeriodIndex, which previously raised.
  • Couldn't see a reason why it shouldn't be supported.
• Fixed issues related to tz-aware conversion in BUG: DatetimeIndex.astype doesn't fully handle tz aware dtypes #18951
• RangeIndex.astype('int64') now remains a RangeIndex
  • Previously returned a Int64Index

[codecov]

Codecov Report

Merging #18937 into master will decrease coverage by <.01%.
The diff coverage is 100%.

@@            Coverage Diff             @@
##           master   #18937      +/-   ##
==========================================
- Coverage   91.59%   91.59%   -0.01%     
==========================================
  Files         150      150              
  Lines       48964    48939      -25     
==========================================
- Hits        44850    44826      -24     
+ Misses       4114     4113       -1

Flag	Coverage Δ
#multiple	89.95% <100%> (-0.01%)	⬇️
#single	41.16% <51.11%> (+0.02%)	⬆️

Impacted Files	Coverage Δ
pandas/core/indexes/datetimes.py	95.57% <100%> (-0.05%)	⬇️
pandas/core/indexes/interval.py	93.81% <100%> (-0.08%)	⬇️
pandas/core/indexes/base.py	96.45% <100%> (+0.01%)	⬆️
pandas/core/indexes/datetimelike.py	97.19% <100%> (+0.06%)	⬆️
pandas/core/indexes/numeric.py	97.28% <100%> (-0.09%)	⬇️
pandas/core/indexes/period.py	93% <100%> (+0.06%)	⬆️
pandas/core/indexes/timedeltas.py	91.08% <100%> (-0.18%)	⬇️

---

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 a18d725...d7c7b9d. Read the comment docs.

[sinhrks]

can it be merged to datetimelike using is_dtype_equal(self.dtype, dtype)?

[jschendel]

Nice catch. I think it can be merged all the way down to base.

[jreback]

lgtm. ex @sinhrks comment. ping when resolved.

[pep8speaks]

Hello @jschendel! Thanks for updating the PR.

Cheers ! There are no PEP8 issues in this Pull Request. 🍻

Comment last updated on December 27, 2017 at 04:17 Hours UTC

[jschendel]

Made the change suggested by @sinhrks, and fixed #18951 at the same time, since it was closely tied to code I was changing without any other dependencies.

Bringing is_dtype_equal(self.dtype, dtype) check down to base.py appears to have slightly changed the behavior of RangeIndex. Previous RangeIndex.astype('int64') returned an Int64Index, but with the change it now remains a RangeIndex. This seems fine to me, but not 100% sure.

Bringing is_dtype_equal(self.dtype, dtype) check down to base.py should also allow for further cleaning of IntervalIndex.astype, but there are some larger issues there that don't allow it to work right now. Specifically, IntervalDtype(*) == 'interval' returns False, unlike CategoricalDtype, which breaks things when relying on is_dtype_equal and trying .astype('interval'). Casting between subtypes also doesn't work, e.g IntervalDtype('int64') to IntervalDtype('float64'). These issues seem independent from the changes being made here, and big enough to require a separate PR.

[jreback]

thanks @jschendel nice patch!

keep em coming!