issues

When using open_dataset( ), it is translating data variables with units of "seconds" to time coordinates. For example, measurements of wave period. I don't believe xarray should treat variables as time coordinates unless their units are of "seconds since...". I have noticed that changing my units to "second" or "sec" or "s" prevents xarray from translating the variable to datetime64 and keeps it float64, as desired. More details and an OPeNDAP example posted on github here: https://stackoverflow.com/questions/46552078/xarray-wave-period-in-seconds-ingested-as-timedelta64

I get the following error when _FillValue and missing_value are both floating-point decimals, even though they are the same value (-999.99):

File "/usr/local/lib/python2.7/dist-packages/xarray/backends/api.py", line 301, in open_dataset return maybe_decode_store(store, lock) File "/usr/local/lib/python2.7/dist-packages/xarray/backends/api.py", line 225, in maybe_decode_store drop_variables=drop_variables) File "/usr/local/lib/python2.7/dist-packages/xarray/conventions.py", line 955, in decode_cf decode_coords, drop_variables=drop_variables) File "/usr/local/lib/python2.7/dist-packages/xarray/conventions.py", line 888, in decode_cf_variables decode_times=decode_times) File "/usr/local/lib/python2.7/dist-packages/xarray/conventions.py", line 800, in decode_cf_variable "and decoding explicitly using " ValueError: ('Discovered conflicting _FillValue and missing_value. Considering opening the offending dataset using decode_cf=False, corrected the attributes', 'and decoding explicitly using xarray.conventions.decode_cf(ds)')

Here is the OPeNDAP end point that causes the error:

http://oos.soest.hawaii.edu/thredds/dodsC/hioos/tide_mhi

The error is unwarranted since the values match. My workaround thus far has been to first open the dataset with decode_cf=False, then I remove the duplicate missing_value attributes, and then decode explicitly using decode_cf(ds).