issues

There might be cleaner ways to fix this than aligning twice, but this will do for now.

• [x] closes #1350
• [x] tests added / passed
• [x] passes git diff upstream/master | flake8 --diff
• [x] whatsnew entry

These results appear to be incorrect unless I'm missing something: ```python In [1]: import xarray as xr

In [2]: import numpy as np

In [3]: array = xr.DataArray(np.zeros((1,2,3)), dims=['time','x','y'], coords={'x':np.arange(2)})

In [4]: array[0,1,1] = 1

In [5]: array.where(array !=0, drop=True) Out[5]: <xarray.DataArray (time: 1, x: 1, y: 1)> array([[[ 0.]]]) Coordinates: * x (x) int64 1 Dimensions without coordinates: time, y

In [5]: array.where(array !=0, drop=True).values Out[5]: array([[[ 0.]]])

In [7]: array.values[array.values !=0] Out[7]: array([ 1.]) ```

This worked in xarray v0.8.2, but no longer works in v0.9.1: ds.to_netcdf(f, engine='scipy') when f is a file-like object.

The traceback looks like: File "...../xarray/core/dataset.py", line 953, in to_netcdf unlimited_dims=unlimited_dims) File "...../xarray/backends/api.py", line 546, in to_netcdf path = _normalize_path(path) File "...../xarray/backends/api.py", line 51, in _normalize_path if is_remote_uri(path): File "...../xarray/core/utils.py", line 439, in is_remote_uri return bool(re.search('^https?\://', path)) File "...../re.py", line 146, in search return _compile(pattern, flags).search(string) TypeError: expected string or buffer

The problem is that to_netcdf does not check to make sure the path is a string before calling _normalize_path: https://github.com/pydata/xarray/blob/b3fc6c4e4fafdf4f075b791594633970a787ad79/xarray/backends/api.py#L545-L548

For now, it's easy enough to work around this by creating a byte-string with to_netcdf and then writing the file separately, but it would be nice to support writing to the file-like object directly again.

With xarray v0.9.1, I get the confusing error message: ValueError: invalid engine for creating bytes with to_netcdf: 'scipy'. Only the default engine or engine='scipy' is supported

This check should be elif engine != 'scipy' instead: https://github.com/pydata/xarray/blob/b3fc6c4e4fafdf4f075b791594633970a787ad79/xarray/backends/api.py#L541

The tests cannot be run from the release tarball without this file.

Hi,

There's now a autoclose parameter to the open_dataset function. But open_dataarray uses positional arguments only, so now the arguments are no longer at their place, and engine, for instance, can't be specified.

Thanks !