BUG: in _nsorted for frame with duplicated values index

[Tux1]

• closes BUG: in _nsorted for frame with duplicated values index #13412
• tests added / passed
• passes git diff upstream/master | flake8 --diff
• whatsnew entry

[Tux1]

what doesn't pass tests ?

[sinhrks]

You must fix test to pass flake8 check.

[sinhrks]

Describe the problem from user's point of view. Users don't care _nsorted.

[codecov-io]

Current coverage is 84.23%

Merging #13428 into master will increase coverage by <.01%

@@             master     #13428   diff @@
==========================================
  Files           138        138          
  Lines         50805      50810     +5   
  Methods           0          0          
  Messages          0          0          
  Branches          0          0          
==========================================
+ Hits          42796      42801     +5   
  Misses         8009       8009          
  Partials          0          0          

Powered by Codecov. Last updated by 62b4327...9e47bbe

[TomAugspurger]

git diff upstream/master | flake8 --diff will help you track down the changes you have to make.

[jreback]

I think a simple soln will work here

(Pdb) p self.loc[ser.index].head(n).sort_values(columns, ascending=ascending,kind='mergesort')
   a  b
1  3  2
(Pdb) p self.loc[ser.index]
   a  b
1  3  2
1  4  1

[Tux1]

@jreback what do you mean ? in _nsorted ? I don't think so

Testing. I think you're right

[jreback]

maybe the .head() should go at the end. I don't recall the exact guarantees of this.

[Tux1]

I tested and your soln doesn't work and doesn't pass the test with this case :
df = pd.DataFrame({'a': [1, 2, 3, 4], 'b': [4, 3, 2, 1]}, index=[0, 0, 1, 1])

Any other suggestion ?

[jreback]

@Tux1 well then play around with it. What you are doing is WAY too complicated for a simple take.

[jreback]

can you rebase / update?