BUG: Cleaner exception when .iloc called with non-integer list

[kykosic]

• closes ERR: Raise a better message when mixing positional and label indexing in loc/iloc #25753
• tests added / passed
• passes git diff upstream/master -u -- "*.py" | flake8 --diff
• whatsnew entry

Issue #25753 was raised requesting more verbose error messages when calling loc and iloc with mixed dtype arguments. In this pull request, I've added a new error message when calling iloc with a non-numeric dtype list-like object.

This fixes numpy ultimately raising an ambiguous exception when, for example, calling .iloc with a list of string column names:

>>> print(data.iloc[['Colorado', 'Utah'], [3, 0, 1]])
Old:
TypeError: cannot perform reduce with flexible type
New:
IndexError: .iloc requires integer indexers, got ['Colorado' 'Utah']

However, the current exception when calling .loc with integer column names when the column names are actually strings I feel is already accurate:

>>> print(data.loc[['Colorado', 'Utah'], [3, 0, 1]])
KeyError: "None of [Int64Index([3, 0, 1], dtype='int64')] are in the [columns]"

[mroeschke]

Needs a new test to catch that this error message is raised.

[codecov]

Codecov Report

Merging #25759 into master will decrease coverage by <.01%.
The diff coverage is 50%.

@@            Coverage Diff             @@
##           master   #25759      +/-   ##
==========================================
- Coverage   91.25%   91.24%   -0.01%     
==========================================
  Files         172      172              
  Lines       52977    52979       +2     
==========================================
+ Hits        48342    48343       +1     
- Misses       4635     4636       +1

Flag	Coverage Δ
#multiple	89.82% <50%> (-0.01%)	⬇️
#single	41.74% <50%> (ø)	⬆️

Impacted Files	Coverage Δ
pandas/core/indexing.py	90.8% <50%> (-0.08%)	⬇️

---

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 707c720...4cb4667. Read the comment docs.

[codecov]

Codecov Report

Merging #25759 into master will increase coverage by <.01%.
The diff coverage is 100%.

@@            Coverage Diff             @@
##           master   #25759      +/-   ##
==========================================
+ Coverage   91.26%   91.26%   +<.01%     
==========================================
  Files         172      172              
  Lines       52965    52967       +2     
==========================================
+ Hits        48337    48339       +2     
  Misses       4628     4628

Flag	Coverage Δ
#multiple	89.82% <100%> (ø)	⬆️
#single	41.74% <50%> (ø)	⬆️

Impacted Files	Coverage Δ
pandas/core/indexing.py	90.9% <100%> (+0.01%)	⬆️

---

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 c975fc4...a8caad0. Read the comment docs.

[WillAyd]

lgtm @mroeschke

[mroeschke]

One small whatsnew change otherwise looks good to me.

[jreback]

use is_integer_dtype

[kykosic]

This will cause a new exception when an array of float values is passed to .iloc. Currently float values are allowed and the array is cast to integer dtype. This will cause some regression.

[jreback]

then you need to thread this in the appropriate place

[jreback]

alt you could check is_numeric_dtype

[kykosic]

Alright I tried to reorganize it a bit, let me know if I misunderstood what you meant.

[jreback]

also merge master

[jreback]

lgtm. @mroeschke over to you

[jreback]

thanks @kykosic