API/ENH: Detect trying to set inplace on copies in a nicer way, related (GH5597)

[jreback]

related #5597
pretty common error that I have seen, originally from:

http://stackoverflow.com/questions/20508968/series-fillna-in-a-multiindex-dataframe-does-not-fill-is-this-a-bug

In [1]: a = [12, 23]

In [2]: b = [123, None]

In [3]: c = [1234, 2345]

In [4]: d = [12345, 23456]

In [5]: tuples = [('eyes', 'left'), ('eyes', 'right'), ('ears', 'left'), ('ears', 'right')]

In [6]: events = {('eyes', 'left'): a, ('eyes', 'right'): b, ('ears', 'left'): c, ('ears', 'right'): d}

In [7]: multiind = pandas.MultiIndex.from_tuples(tuples, names=['part', 'side'])

In [8]: zed = pandas.DataFrame(events, index=['a', 'b'], columns=multiind)

Shows a warning (the message is comewhat generic)

In [9]: zed['eyes']['right'].fillna(value=555, inplace=True)
/usr/local/bin/ipython:1: SettingWithCopyWarning: A value is trying to be set on a copy of a slice from a DataFrame
  #!/usr/local/bin/python

In [10]: zed
Out[10]: 
part  eyes         ears       
side  left  right  left  right
a       12    123  1234  12345
b       23    NaN  2345  23456

[2 rows x 4 columns]

Correct method

In [11]: zed.loc[:,('eyes','right')].fillna(value=555, inplace=True)

In [12]: zed
Out[12]: 
part  eyes         ears       
side  left  right  left  right
a       12    123  1234  12345
b       23    555  2345  23456

[2 rows x 4 columns]

[jreback]

cc @michaelaye .... I know you use lots of multi-index stuff...can you give this a quick test run and lmk if any SettingWithCopyError/Warns come up (that are unexpected). ? thxs

[michaelaye]

I tested 222bc9d and got no Error/Warns.
But I got to warn you that I don't have much multi-index stuff anymore. The issues we discussed recently were months old, and because of the pending issues that I partly didn't understand, and because they are harder to grasp in general I refrained from using them and solved my stuff mostly with 1-D indexes.

[jreback]

@michaelaye gr8 thanks. It was also trying to do inplace operations when operating on a copy. So you code seems good.