Sort broken after unique -> Series with object array

[jseabold]

I have no idea why this fails.

pandas.Series(pandas.Series(["a","c","b"]).unique()).sort()

but this doesn't

pandas.Series(np.array(["a", "c", "b"], dtype=object)).sort()

[changhiskhan]

the unique method in hashtable sets the base to be something other than an ndarray.
@wesm is setting uniques.base necessary? I have it setup locally right now so sort checks whether base is an instance of ndarray, but if uniques.base isn't necessary I'll just get rid of those in hashtable.pyx and we should be good to go.

[wesm]

Well, the issue is ownership of the data buffer. I think there is a way to transfer ownership of the buffer that doesn't require setting uniques.base. See http://gael-varoquaux.info/blog/?p=157 (this is where I got the method from that I'm using as I recall)(

[wesm]

Having a look at this. If you resize the databuffer and manually set the array flags to own the data, it should work

[wesm]

alrighty all set now, was fairly simple actually