Interesting my tests pass on pandas 0.16.1 but not 0.16.2

We could also raise an exception when people call ufuncs on dask.arrays/dataframes. Would that be preferable to silently forcing computation?

Where is the failure in pandas 0.16.1.

I would be for raising a warning.

Note that the underlying problem of #580 is that we don't have a way to say

log(myseries)

for any function log. Arguably the dask.array "ufuncs" should operate smoothly on dask.dataframe Series objects. That's probably the full solution in the near term.

@mrocklin the in 0.16.1 the way I was creating the output series here worked without the name kwarg because the name was not being checked in assert_series_equal. In 0.16.2 assert_series_equal started to check names.

OK. Merging this. If you feel like some adventure it'd still be good to get the dask.array ufuncs to work on dask.dataframe objects. I think that the best way to do this is likely some sort of protocol so that dask.dataframe (or dask.bag) logic doesn't find its way into dask.array