API/DOC: unclear scope of index.asof

[jorisvandenbossche]

From http://stackoverflow.com/questions/26918045/python-pandas-typeerror-cannot-compare-type-timestamp-with-type-float

The docstring of Index.asof says:

In [73]: pd.Index.asof?
Type:        instancemethod
String form: <unbound method Index.asof>
File:        c:\anaconda\lib\site-packages\pandas\core\index.py
Definition:  pd.Index.asof(self, label)
Docstring:
For a sorted index, return the most recent label up to and including
the passed label. Return NaN if not found

while asof only works for a time series.

• Clarify the docs that asof only works for DatetimeIndex (and PeriodIndex?)

• Why is this actually restricted to timestamps? This functions seems also possibly useful to me for other index types? Something like

  In [74]: idx = pd.Index([1,2,3,8,9])
  
  In [75]: idx.asof(5)    # gives now TypeError
  3
  

  Instead of having to do:

  In [78]: idx[idx.searchsorted(5) - 1]
  Out[78]: 3
  

[jreback]

actually a much bigger issue is that this should raise on a non-monotonic index (or at least show a PerformanceWarning). and prob should provide a side argument as well (and thus its very close to searchsorted, except when the label is actually in the index).

[jreback]

@shoyer now that you fixed .asof close?

[shoyer]

Indeed, it's fixed!