BUG: CategoricalIndex.format

[topper-123]

• closes BUG: output of df.to_string depends on whether columns is a CategoricalIndex or not #35439
• tests added / passed
• passes black pandas
• passes git diff upstream/master -u -- "*.py" | flake8 --diff
• whatsnew entry

I've temporarily put the whatsnewentry in the v.1.1.0 release note, because there isn't a v.1.1.1 version yet. I'll move it, before this is merged.

[simonjayhawkins]

Thanks @topper-123 for the PR.

The regression was caused by #35118. Categorical types other than object were also affected. maybe need to parameterise test with other values for cols

>>> pd.__version__
'1.2.0.dev0+10.g3b1d4f1ee'
>>>
>>> data = [[4, 2], [3, 2], [4, 3]]
>>> cols = [1, None]
>>> res = pd.DataFrame(data, columns=cols)
>>> print(res)
   1  NaN
0  4    2
1  3    2
2  4    3
>>>
>>> res = pd.DataFrame(data, columns=pd.CategoricalIndex(cols))
>>> print(res)
   1    NaN
0    4    2
1    3    2
2    4    3
>>>

>>> pd.__version__
'1.0.5'
>>>
>>> data = [[4, 2], [3, 2], [4, 3]]
>>> cols = [1, None]
>>> res = pd.DataFrame(data, columns=cols)
>>> print(res)
   1  NaN
0  4    2
1  3    2
2  4    3
>>>
>>> res = pd.DataFrame(data, columns=pd.CategoricalIndex(cols))
>>> print(res)
   1.0  NaN
0    4    2
1    3    2
2    4    3
>>>

I've temporarily put the whatsnewentry in the v.1.1.0 release note, because there isn't a v.1.1.1 version yet.

doc\source\whatsnew\v1.1.1.rst now merged to master

[topper-123]

The added tests revealed that this method in master made the output from RangeIndex.format different than for Int64Index.format:

>>> pd.RangeIndex(0, 18, 2).format()
['0', '2', '4', '6', '8', '10', '12', '14', '16']
>>> pd.Int64Index(range(0, 18, 2)).format()
 ['0 ', '2 ', '4 ', '6 ', '8 ', '10', '12', '14', '16']

Notice the extra space for one-digit scalars in the Int64Index case. The outputs from the two methods are identical after merging this PR.

[topper-123]

Updated.

[jreback]

thanks @topper-123 very nice!

[simonjayhawkins]

@meeseeksdev backport to 1.1.x