html_url,issue_url,id,node_id,user,created_at,updated_at,author_association,body,reactions,performed_via_github_app,issue https://github.com/pydata/xarray/issues/2251#issuecomment-417008887,https://api.github.com/repos/pydata/xarray/issues/2251,417008887,MDEyOklzc3VlQ29tbWVudDQxNzAwODg4Nw==,1217238,2018-08-29T16:06:28Z,2018-08-29T16:06:28Z,MEMBER,I filed a bug report for netCDF4-Python: https://github.com/Unidata/netcdf4-python/issues/841,"{""total_count"": 0, ""+1"": 0, ""-1"": 0, ""laugh"": 0, ""hooray"": 0, ""confused"": 0, ""heart"": 0, ""rocket"": 0, ""eyes"": 0}",,335608017 https://github.com/pydata/xarray/issues/2251#issuecomment-400428831,https://api.github.com/repos/pydata/xarray/issues/2251,400428831,MDEyOklzc3VlQ29tbWVudDQwMDQyODgzMQ==,1217238,2018-06-26T19:08:57Z,2018-06-26T19:08:57Z,MEMBER,"If I remember correctly, netCDF intentionally does not support multi-dimensional array attributes. I'm not sure why it silently changes shape, though. An error message would seem much more user-friendly to me. Possibly this is worth raising on the netCDF4 issue tracker: https://github.com/unidata/netcdf4-python","{""total_count"": 0, ""+1"": 0, ""-1"": 0, ""laugh"": 0, ""hooray"": 0, ""confused"": 0, ""heart"": 0, ""rocket"": 0, ""eyes"": 0}",,335608017 https://github.com/pydata/xarray/issues/2251#issuecomment-400411869,https://api.github.com/repos/pydata/xarray/issues/2251,400411869,MDEyOklzc3VlQ29tbWVudDQwMDQxMTg2OQ==,10050469,2018-06-26T18:12:45Z,2018-06-26T18:16:53Z,MEMBER,"> I have some metadata that is in the form of numpy arrays. The short answer is: storing arrays (or other objects) as attributes is not easy. When you do so, xarray has to convert them at serialization, and the roundtrip conserving the shape cannot be achieved because it isn't available in the datamodel of NetCDF. What is it you are storing as 2D arrays? Would a coordinate variable be an option? ","{""total_count"": 1, ""+1"": 1, ""-1"": 0, ""laugh"": 0, ""hooray"": 0, ""confused"": 0, ""heart"": 0, ""rocket"": 0, ""eyes"": 0}",,335608017