API: arithmetic ops could accept an index as appropriate

[jreback]

Not sure if this will cause weirdness but haveing the arithmetic operations able to accept an index (when appropriate type), e.g. a datecolumn should be able to interact with a DatetimeIndex, but prob should prohibit an int column from interacting with a IntIndex?

e.g. in below df['date']-df.index should work

This actually DOES work now! (and I take it back, does make some sense)
should df[1]-df.reset_index().index work????

http://stackoverflow.com/questions/18368442/how-to-directly-use-pandas-date-time-index-in-calculations/18368793?noredirect=1#comment26971689_18368793

In [1]: df = DataFrame(randn(5,2),index=date_range('20130101',periods=5))

In [3]: df['date'] = Timestamp('20130102')

In [4]: df
Out[4]: 
                   0         1                date
2013-01-01  2.406932 -0.313473 2013-01-02 00:00:00
2013-01-02  0.034162 -0.708450 2013-01-02 00:00:00
2013-01-03 -1.585031  0.587243 2013-01-02 00:00:00
2013-01-04  1.454818  1.089208 2013-01-02 00:00:00
2013-01-05 -0.778016 -0.994242 2013-01-02 00:00:00

In [5]: df['td'] = df['date']-df.index.to_series()

In [6]: df
Out[6]: 
                   0         1                date                td
2013-01-01  2.406932 -0.313473 2013-01-02 00:00:00  1 days, 00:00:00
2013-01-02  0.034162 -0.708450 2013-01-02 00:00:00          00:00:00
2013-01-03 -1.585031  0.587243 2013-01-02 00:00:00 -1 days, 00:00:00
2013-01-04  1.454818  1.089208 2013-01-02 00:00:00 -2 days, 00:00:00
2013-01-05 -0.778016 -0.994242 2013-01-02 00:00:00 -3 days, 00:00:00

[jtratner]

@jreback why would you expect this not to work? Index is just an ndarray with some ops disabled. It won't work with a MultiIndex and I don't think there's a clear reason to make that work.

[jreback]

@jtratner no i expect it TO work

I guess there is some setting going on, so need to just fix it (e.g. just do index.to_series) is prob enough
I think only matters for DatetimeIndex (as it hits that special loop)

In [16]: df['date']-df.index
---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
<ipython-input-16-bdf30f864601> in <module>()
----> 1 df['date']-x

/usr/local/lib/python2.7/site-packages/pandas-0.12.0-py2.7-linux-x86_64.egg/pandas/core/series.pyc in wrapper(self, other, name)
    154 
    155             lvalues = lvalues.view('i8')
--> 156             rvalues = rvalues.view('i8')
    157 
    158         if isinstance(rvalues, Series):

AttributeError: can't set attribute

[jtratner]

@jreback fun, no traceback either?

[jreback]

yep..just bad error message....think need to add a check if the lvalues/rvalues are an Index, :( sigh

[jtratner]

@jreback are you sure this isn't something wrong with the way Index is currently defined (e.g., something overriding view(), like an aliasing to copy that then thinks that it's the first positional argument tot the function and tries to set it)?

[jreback]

@jtratner #4641 fixes it...no biggie