rasm.nc is not a valid NetCDF 3 file

[dokeeffe]

Not sure if this is a duplicate of #1106 or something crazy that I'm doing....

I'm attempting to play with the xarray_multidimensional_coords.ipynb but I cant load the sample dataset.
Below are my versions (I just installed xarray using pip today)
('numpy version : ', '1.11.2')
('pandas version : ', u'0.19.0')
('xarray version : ', '0.8.2')

Pasting the full trace below

---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
<ipython-input-6-3af9bb5e86b6> in <module>()
----> 1 ds = xr.tutorial.load_dataset('rasm')
      2 ds

/usr/local/lib/python2.7/dist-packages/xarray/tutorial.pyc in load_dataset(name, cache, cache_dir, github_url, **kws)
     53         _urlretrieve(url, localfile)
     54 
---> 55     ds = _open_dataset(localfile, **kws).load()
     56 
     57     if not cache:

/usr/local/lib/python2.7/dist-packages/xarray/backends/api.pyc in open_dataset(filename_or_obj, group, decode_cf, mask_and_scale, decode_times, concat_characters, decode_coords, engine, chunks, lock, drop_variables)
    210                 store = backends.NetCDF4DataStore(filename_or_obj, group=group)
    211             elif engine == 'scipy':
--> 212                 store = backends.ScipyDataStore(filename_or_obj)
    213             elif engine == 'pydap':
    214                 store = backends.PydapDataStore(filename_or_obj)

/usr/local/lib/python2.7/dist-packages/xarray/backends/scipy_.pyc in __init__(self, filename_or_obj, mode, format, group, writer, mmap)
     94             filename_or_obj = BytesIO(filename_or_obj)
     95         self.ds = scipy.io.netcdf_file(
---> 96             filename_or_obj, mode=mode, mmap=mmap, version=version)
     97         super(ScipyDataStore, self).__init__(writer)
     98 

/home/dokeeffe/.local/lib/python2.7/site-packages/scipy/io/netcdf.pyc in __init__(self, filename, mode, mmap, version, maskandscale)
    262 
    263         if mode in 'ra':
--> 264             self._read()
    265 
    266     def __setattr__(self, attr, value):

/home/dokeeffe/.local/lib/python2.7/site-packages/scipy/io/netcdf.pyc in _read(self)
    577         if not magic == b'CDF':
    578             raise TypeError("Error: %s is not a valid NetCDF 3 file" %
--> 579                             self.filename)
    580         self.__dict__['version_byte'] = fromstring(self.fp.read(1), '>b')[0]
    581 

TypeError: Error: /home/dokeeffe/.xarray_tutorial_data/rasm.nc is not a valid NetCDF 3 file

[fmaussion]

You have to install netCDF4 in order to open the rasm dataset (at least until we convert the file to netCDF3 - this might take a bit).

[dokeeffe]

Thanks for the fast response!! Yes that did it. All ok now. I'll close the issue.

[chunweiyuan]

When I do conda install netcdf4, it ends up downgrading my numpy from 1.12.0 to 1.11.3, which triggers other dependency-related problems in my code. Is there a different solution to this problem?

[fmaussion]

Did you try the conda-forge channel to install netcdf4?

https://conda-forge.github.io/

[chunweiyuan]

Just did. Got it to work. Muchas Gracias.

[rabernat]

My students / postdocs have been reporting lots of netcdf-related xarray issues recently. I think there are problems with the default anaconda distribution which we are not catching because all our tests use conda-forge.

…

On Apr 28, 2017, at 1:09 PM, chunweiyuan ***@***.***> wrote: Just did. Got it to work. Muchas Gracias. — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub, or mute the thread.

[dopplershift]

Anaconda netcdf4 has just been flat out broken: ContinuumIO/anaconda-recipes#89

Not sure how that's really an xarray issue, though.

[fmaussion]

Note that since pydata/xarray-data#7 , the rasm file can be read with scipy only:

ds = xr.open_dataset('.xarray_tutorial_data/rasm.nc', engine='scipy')

I guess that @chunweiyuan downloaded the file before that change.

We recommend to use conda-forge in the docs, I don't think there's much more we can do.