BUG: to_timedelta of a scalar returns a scalar, closes #5410.

[danielballan]

closes #5410

[jreback]

@danielballan I think this is just a trivial test fix...can you have a look...want to do this for 0.13

[danielballan]

Yep, the key was if not _np_version_under1p7. If the build passes, we should be good to go.

Note to selves: Testing on to_timedelta is pretty light.

[danielballan]

Hrm. This passes on my box with numpy 1.7.1. Weird. Will take another looks later.

[jreback]

you are comparnig that result with assert_series_equal, but you should be using assert_almost_equal (as they are arrays and not series)

[jreback]

this looks fine....pls give 1 more rebase and can merge it

[danielballan]

Done.

[jreback]

feel free to add more tests to to_timedelta (its not so light...but more tests are better!)

[danielballan]

Yeah, that was my mistake. I was looking at the smattering of tests under pandas/tests, completely missing pandas/tseries/tests!

[jreback]

np...in theory they should be consolidated..but no big deal really

[jreback]

thanks@!