split netcdf4time into standalone package

[jhamman]

I'd like to propose splitting netcdf4time into a standalone package. The motivation for this request is coming from ongoing xarray development (pydata/xarray#1084) where we would like to use netcdftime without the complete netCDF dependency. This would not require any change to user code as netcdf-python would still be able to provide users the netcdftime module. I'd be happy to perform this task, including setting up TravisCI, Pypi, etc...

cc @shoyer @spencerahill @spencerkclark @darothen

xref: pydata/xarray#1084, pydata/xarray#789

[shoyer]

To give a clearer motivation: netCDF-C can be hard to install. But netcdftime (a name I would prefer to netcdf4time) is still useful even if you only have scipy, pydap or h5netcdf for reading netCDF-like data. It's also useful even if you never read or write a netCDF files, because it contains all the necessary logic for working with custom datetime types (now including arithmetic).

[jswhit]

This is fine with me, as long as installing netcdf4-python installs netcdftime as well. I can ask the Unidata folks to create a netcdf4time project.

[shoyer]

Let's call it netcdftime, which matches the existing name of the Python
module.
On Mon, Nov 7, 2016 at 7:35 AM Jeff Whitaker notifications@github.com
wrote:

This is fine with me, as long as installing netcdf4-python installs
netcdftime as well. I can ask the Unidata folks to create a netcdf4time
project.

—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
#601 (comment),
or mute the thread
https://github.com/notifications/unsubscribe-auth/ABKS1nLZOpkRaGveFd6Gb4XTBOVoEpcWks5q70UmgaJpZM4KqseB
.

[dopplershift]

So do we have consensus on netcdftime ? I can create the repo as soon as I know what to call it.

[jhamman]

netcdftime is fine. I would actually prefer cftime since that's what it is
but I understand the namespace confusion is an issue.

On Mon, Nov 7, 2016 at 12:09 PM, Ryan May notifications@github.com wrote:

So do we have consensus on netcdftime ? I can create the repo as soon as
I know what to call it.

—
You are receiving this because you authored the thread.
Reply to this email directly, view it on GitHub
#601 (comment),
or mute the thread
https://github.com/notifications/unsubscribe-auth/ACVILWNfMfkej8N5A1FmLFvcEzyf4uf_ks5q71tBgaJpZM4KqseB
.

[dopplershift]

Repository is up at: https://github.com/Unidata/netcdftime

Jeff, you should have everything you need to give people permission as necessary--though I'm happy to do more if you want.

[jswhit]

Thanks Ryan - I don't have a "Settings" button on the main project page (as I do for netcdf4-python) so I can't add @jhamman as a contributor.

[dopplershift]

Accept my invite to the repo. 😁

[jswhit]

OK, done. I invited @jhamman to be a collaborator.

[jhamman]

A first swing at the stand-alone netcdftime package is up: https://github.com/Unidata/netcdftime

For that package, there are a bunch of issues I need to work through including documentation, testing, and pypi/conda distributions.

Once I have that all done, I'll issue a PR here removing the netcdftime module.

[jswhit]

Sounds like I should go ahead with a 1.2.6 release with netcdfime, and target removal for the 1.2.7 release?

[pelson]

netcdftime was created. It is now called cftime. https://github.com/Unidata/cftime

I think this can now be closed! 🎉