WIP: convert to/from cdms2 variables

[shoyer]

Fixes #133

@DamienIrving am I missing anything obvious here?

[DamienIrving]

Your approach looks good to me. All people need to be able to do is move to and from cdms2 transient variables and then they'll have access to all the convenience functions defined in cdat-lite libraries like cdutil and genutil.

[shoyer]

So I understand cdat uses masked arrays for missing values. How does it handle dates?

On Mon, Sep 22, 2014 at 3:13 PM, Damien Irving [email protected]
wrote:

Your approach looks good to me. All people need to be able to do is move to and from cdms2 transient variables and then they'll have access to all the convenience functions defined in cdat-lite libraries like cdutil and genutil.

Reply to this email directly or view it on GitHub:
#236 (comment)

[DamienIrving]

I'm not sure exactly how it handles dates under the hood, however as a user before using a convenience function from the cdutil library to calculate a seasonal climatology or something, you have to use cdutil.setTimeBounds() to specify the time bounds of your data (i.e. whether it is daily, monthly, etc).

To view the time axis of a cdms2 transient variable in a convenient YYYY-MM-DD HH:MM:SS format (e.g. let's say that variable's name is data), you can use data.getTime().asComponentTime().

Does that help?

[shoyer]

@DamienIrving Yes, that does help.

The main thing is I'm pretty sure CDAT's climatological related functions can't handle dates in the form of a np.datetime64 array (which xray uses), so we'll probably need to convert back into CF conventions style "days since YYYY-MM-DD".

[shoyer]

@DamienIrving I'm going to merge this now. I did add support for encoding/decoding time information, but you should probably still consider this experimental for now. Please do let me know how it works for you when you get the chance to try it out.

I'll be releasing v0.3.2 with this within the next few days.