model.GridType differences from 3.1.1 about pydap

[MBARIMike]

It wasn't too hard for me to make a work-around in my code, but it'd be nice to have the same behavior that we have 3.1.1.

[MBARIMike]

Upon further inspection I see that the GridType returned is different. In Version 3.1.1 I get a dictionary:

>>> from pydap.client import open_url
>>> ds = open_url('http://dods.mbari.org/opendap/data/auvctd/surveys/2010/netcdf/Dorado389_2010_300_00_300_00_decim.nc')
>>> ds['temperature']
{'temperature': <pydap.model.BaseType object at 0x7f18464a7650>, 'time': <pydap.model.BaseType object at 0x7f18464a7750>}

With Version 3.2.0 I get something different:

>>> from pydap.client import open_url
>>> ds = open_url('http://dods.mbari.org/opendap/data/auvctd/surveys/2010/netcdf/Dorado389_2010_300_00_300_00_decim.nc')
>>> ds['temperature']
<GridType with array 'temperature' and maps 'time'>

Perhaps this is an improvement and I should change my code?

[taeold]

This explains some of the compatibility issue I saw in pydata/xarray#1174 - namely, datatype returned by pydap.client in version 3.2 is dictionary-like but not quite so, missing couple of functions like __len__ and iteritems/items.

Does it make sense to implement these functions? From my understanding of the code, GridType/StructureType appeared well equipped to implement those functions (am I naive in thinking so?).

[jenfly]

I'd love to have this compatibility as well! I'm new to Pydap - I've been using Xarray and now I'm trying to get it to work with engine=pydap in order to handle NASA's new authentication procedures. But the engine=pydap option doesn't work in xarray.open_dataset() anymore because of these changes to GridType. I can try to make my own versions with workarounds, but going forward it would be great to restore compatibility between Xarray and Pydap again. Is there any particular reason these attributes were removed?

[laliberte]

I'm going to close this one following PR #48. If it turns out to be insufficient, I'll reopen.