BUG: TDI / TDI should work

[jreback]

In [1]: pd.__version__
Out[1]: '0.20.3'

In [2]: i = pd.timedelta_range('1 day', periods=3)

In [3]: i / i
TypeError: cannot evaluate a numeric op with a non-numeric dtype

master

In [1]: pd.__version__
Out[1]: '0.24.0.dev0+550.g70c9003c9'

In [2]: i = pd.timedelta_range('1 day', periods=3)

In [3]: i / i
TypeError: unsupported operand type(s) for /: 'TimedeltaIndex' and 'TimedeltaIndex'

In [4]: i / 2
Out[4]: TimedeltaIndex(['0 days 12:00:00', '1 days 00:00:00', '1 days 12:00:00'], dtype='timedelta64[ns]', freq='12H')

In [5]: i / pd.Timedelta('1 day')
Out[5]: Float64Index([1.0, 2.0, 3.0], dtype='float64')

# conversion then op works
In [8]: i.astype('i8') / i.astype('i8')
Out[8]: Float64Index([1.0, 1.0, 1.0], dtype='float64')

[jreback]

cc @jbrockmendel

I really thought this DID work in 0.20.3, though my example is not reproducible ATM.

of course this works for a scalar TD or numeric on the rhs.

Is there a reason we didn't allow TDI / TDI though?

[jreback]

cc @joshuastorck

[jbrockmendel]

Is there a reason we didn't allow TDI / TDI though?

Not that I can think of. In trying this myself found #22632.