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/316#issuecomment-74811395,https://api.github.com/repos/pydata/xarray/issues/316,74811395,MDEyOklzc3VlQ29tbWVudDc0ODExMzk1,1217238,2015-02-18T04:45:25Z,2015-02-18T04:45:25Z,MEMBER,"Made a new issue for datetime decoding error handling: #323
","{""total_count"": 0, ""+1"": 0, ""-1"": 0, ""laugh"": 0, ""hooray"": 0, ""confused"": 0, ""heart"": 0, ""rocket"": 0, ""eyes"": 0}",,56817968
https://github.com/pydata/xarray/issues/316#issuecomment-73756512,https://api.github.com/repos/pydata/xarray/issues/316,73756512,MDEyOklzc3VlQ29tbWVudDczNzU2NTEy,1217238,2015-02-10T18:40:12Z,2015-02-10T18:40:12Z,MEMBER,"I agree, this is not ideal. We really should try to decode time units at the time a dataset is opened, not just when the values are read (which open leads to a failed `repr()`). Adding some sort of more helpful error message to catch exception that arise would also help (e.g., ""cannot read non-ISO time units: netCDF4 not installed"").
","{""total_count"": 0, ""+1"": 0, ""-1"": 0, ""laugh"": 0, ""hooray"": 0, ""confused"": 0, ""heart"": 0, ""rocket"": 0, ""eyes"": 0}",,56817968
https://github.com/pydata/xarray/issues/316#issuecomment-73732815,https://api.github.com/repos/pydata/xarray/issues/316,73732815,MDEyOklzc3VlQ29tbWVudDczNzMyODE1,141709,2015-02-10T16:37:16Z,2015-02-10T16:37:16Z,CONTRIBUTOR,"I noticed that you already have a fallback option and I simply made use of that, which solves this problem in a way that looks reasonable to me.

However, I'm still not quite happy with the way errors were/are handled: before this change, reading a file with the non-standard datetime format would work without errors, but trying to do anything with the dataset after opening it would produce a ValueError, which from an end-user perspective doesn't help much to figure out where the error is coming from.

I suppose with the change in my PR, for cases where python-netCDF4 is not installed, one would instead get an ImportError without explanation (one wouldn't know why we tried to import the optional dependency). Maybe something like a wrapper to import netCDF4 which raises a context-specific ImportError message would be useful?
","{""total_count"": 0, ""+1"": 0, ""-1"": 0, ""laugh"": 0, ""hooray"": 0, ""confused"": 0, ""heart"": 0, ""rocket"": 0, ""eyes"": 0}",,56817968
https://github.com/pydata/xarray/issues/316#issuecomment-73619315,https://api.github.com/repos/pydata/xarray/issues/316,73619315,MDEyOklzc3VlQ29tbWVudDczNjE5MzE1,1217238,2015-02-10T00:12:57Z,2015-02-10T00:12:57Z,MEMBER,"Yes, if you want to give it a shot, that would be great!

The only thing to watch out for is that `netCDF4`/`netcdftime` is an optional dependency, so you need to ensure that test methods are properly decorated with `requires_netCDF4`; otherwise the CI tests will fail. I'm not too concerned about being able to decode datetime units without netCDF4 installed, though I suppose that would be nice to maintain if possible (just import netcdftime locally when the fallback is used).
","{""total_count"": 0, ""+1"": 0, ""-1"": 0, ""laugh"": 0, ""hooray"": 0, ""confused"": 0, ""heart"": 0, ""rocket"": 0, ""eyes"": 0}",,56817968
https://github.com/pydata/xarray/issues/316#issuecomment-73519971,https://api.github.com/repos/pydata/xarray/issues/316,73519971,MDEyOklzc3VlQ29tbWVudDczNTE5OTcx,141709,2015-02-09T14:47:39Z,2015-02-09T14:47:39Z,CONTRIBUTOR,"I think falling back to `netcdftime` would be preferrable to a `datetime_format` argument since it would just automagically work (although a `datetime_format` could be useful for other, trickier netCDF files). I could have a go at doing this later in the week and submit a PR, if it's helpful.
","{""total_count"": 0, ""+1"": 0, ""-1"": 0, ""laugh"": 0, ""hooray"": 0, ""confused"": 0, ""heart"": 0, ""rocket"": 0, ""eyes"": 0}",,56817968
https://github.com/pydata/xarray/issues/316#issuecomment-73289825,https://api.github.com/repos/pydata/xarray/issues/316,73289825,MDEyOklzc3VlQ29tbWVudDczMjg5ODI1,1217238,2015-02-06T18:42:32Z,2015-02-06T18:42:32Z,MEMBER,"It looks like `netcdftime` (part of the `netCDF4` package) can parse this string correctly, so we should probably fall back to using that:

```
In [26]: netcdftime.num2date(0, 'hours since 1981-1-10 0')
Out[26]: datetime.datetime(1981, 1, 10, 0, 0)
```

The other possibility is adding a `datetime_format` argument to `open_dataset`.
","{""total_count"": 0, ""+1"": 0, ""-1"": 0, ""laugh"": 0, ""hooray"": 0, ""confused"": 0, ""heart"": 0, ""rocket"": 0, ""eyes"": 0}",,56817968
https://github.com/pydata/xarray/issues/316#issuecomment-73276393,https://api.github.com/repos/pydata/xarray/issues/316,73276393,MDEyOklzc3VlQ29tbWVudDczMjc2Mzkz,1217238,2015-02-06T17:21:18Z,2015-02-06T17:21:18Z,MEMBER,"What a hassle! Unfortunately my experience has been that non-ISO conformant time units are not uncommon. I'll take a look into this.

We did add one fix to improve the time unit parsing since the last release. You might want to try installing the dev version off github to see if it fixes your issue. Otherwise you can also do the sort of fixup you describe by loading the dataset with decode_cf=False, fixing the metadata and then calling xray.decode_cf.
","{""total_count"": 0, ""+1"": 0, ""-1"": 0, ""laugh"": 0, ""hooray"": 0, ""confused"": 0, ""heart"": 0, ""rocket"": 0, ""eyes"": 0}",,56817968