ENH: NaTType can be unpickled properly

[sinhrks]

Closes #4606.

I'm not sure this also be a solution for #5388, but seems to be a workaround.

[jreback]

@sinhrks ok...this is ok for #4606 (though do you really need that separate function _nat_upickle?

This 'solves' #5388 but not for the right reason. This is of course object type, but can represented if transposed into 'correct' blocks.

I am not sure what to do about this. Multi-dtypes have to be inserted into a Panel on only certain axes otherwise it forces object.

So pls pull out the tests for #5388. Let's keep that as a separate issue.

In [36]: wp.to_frame() Out[36]: i1 i2 major minor 0 A 1 1 B 2010-01-01 00:00:00 2010-01-01 00:00:00 1 A 2 2 [3 rows x 2 columns] In [37]: wp.to_frame().dtypes Out[37]: i1 object i2 object dtype: object In [38]: wp.transpose(2,0,1).to_frame().dtypes Out[38]: A object B datetime64[ns] dtype: object In [39]: wp.transpose(2,0,1).to_frame() Out[39]: A B major minor i1 0 1 2010-01-01 i2 0 1 2010-01-01 [2 rows x 2 columns] 

[jreback]

you can take this out of here.

[jreback]

merged via 2ada054

thanks!

[sinhrks]

Thanks to merge.

though do you really need that separate function _nat_upickle?

I believe so, because pickle module search the function under the module namespace and raise PicklingError if it cannot find it.

[jreback]

got it...thanks