ohlc not available for groupby/etc

[hayd]

Was this available at some stage (it's looking in self._cython_functions for it rather than just saying no dice with a plain ol' AttributeError):

In [1]: rng = pd.date_range('2012', periods=3)

In [2]: df = pd.DataFrame([[1, 2], [1, 3], [5, 6]], rng, columns=['A', 'B'])

In [3]: g = df.groupby('A')

In [4]: g.foo()
AttributeError: 'DataFrameGroupBy' object has no attribute 'foo'

In [5]: g.ohlc()
KeyError: 'ohlc'

In [6]: g.ohlc(rule='D')
TypeError: ohlc() got an unexpected keyword argument 'rule'

Can do atm using g.resample('D', how='ohlc').

[jreback]

hmm

I think this is only valid as a resample method

[hayd]

close or leave open? (Do we want a sep method... someday ? - not sure it's needed tbh)

I added the resample version to sub-selection tests, so we're good on that front.

[jreback]

I would blacklist it to get the correct error message
but might break resample

[jreback]

I think this should have a better error message

[hayd]

We could just create the method and raise saying to use resample(how="ohlc")... or better yet just do the resample?

[jreback]

I don't think you should do the resample, but raising a more helpful message is good.

[jreback]

this totally works now (need slight change to syntax as the above is not valid), needing testing/validation

In [31]: df.groupby(pd.Grouper(freq='D')).ohlc()
Out[31]: 
              A                   B               
           open high low close open high low close
2012-01-01    1    1   1     1    2    2   2     2
2012-01-02    1    1   1     1    3    3   3     3
2012-01-03    5    5   5     5    6    6   6     6

In [33]: df.groupby('A').ohlc()
Out[33]: 
     A                   B               
  open high low close open high low close
A                                        
1    1    1   1     1    2    3   2     3
5    5    5   5     5    6    6   6     6

[jreback]

closing as this is already supported with a timegrouper / resample.