Things to complete before releasing xarray v0.9.0

[shoyer]

• Add drop=True argument to isel, sel and squeeze Add drop=True argument to isel, sel and squeeze #1153
• Add set_index, reset_index and reorder_levels methods Add set_index, reset_index and reorder_levels methods #1028
• Variable.copy() should copy dask.arrays Variable.copy() should copy dask.arrays #1166
• Switch to shared Lock (SerializableLock if possible) for reading/writing Switch to shared Lock (SerializableLock if possible) for reading/writing #1179

@pydata/xarray anything else essential?

[max-sixty]

Are we pushing #964? This would be awesome to start building on, although important to get right given it'll be a platform

[shoyer]

I also really want to merge #964 (the "apply" function) but the current API is a little too confusing. We could merge it without exposing it as public API for now and/or start building simplified versions (e.g., with an optional dim argument instead of signature).

I agree that it should be a top priority but I don't want to let it hold up 0.9 any longer, so I think we should release without it. apply does not change any APIs in a breaking way so we can even both put it in a minor release if desried (though obviously it is a big feature).

[jhamman]

I have two smaller features I'd like to get in:

• Unlimited dimensions / Dataset.encoding Dataset.encoding and unlimited dimensions for to_netcdf #1170
• Dataset.info method add info method to dataset #1176

These are both pretty close and could be wrapped up in a day or so.

[shoyer]

What's our status here? I'd like to get the release (maybe a release candidate) out this week if possible.

It would be great if someone else could take a look at #1166 and #1179.

[shoyer]

There is never an ideal time to make a release cut, but this has been dragging on for too long. I just sent out an email announcing a release candidate for v0.9.0:
https://groups.google.com/forum/#!topic/xarray/Yc6A45zXmwM

Unlimited dimensions will have to wait to the v0.9.1 release, but that release can follow pretty shortly given that the changes should be mostly backwards compatible.

[ghisvail]

@shoyer Do you plan to push a release tarball for 0.9.0rc1 on PyPI? I'd prefer to use this version over 0.8.2 as the basis for the initial Debian packaging. I tested both and found 0.9.0rc1 to lead to less errors with the test suite than 0.8.2. I could also wait for 0.9.0 final if its release is planned soon.

[shoyer]

@ghisvail Oops, I forgot to upload the source distribution to pypi (now fixed).

The final release should be some time this week, up to you if you want to wait.

[ghisvail]

The final release should be some time this week, up to you if you want to wait.

Thanks @shoyer, I used the rc to try out the packaging and give you some initial feedback. See #1208 and #1209.

[fmaussion]

If I wrap up this thread, I see we are missing:

• Dataset.encoding and unlimited dimensions for to_netcdf #1170 (Unlimited dimensions / Dataset.encoding)
• New testing module and tests refactor #1222 (testing module)
• List "Unindexed dimensions" separately in __repr__ #1221 (document the repr)

[shoyer]

With the merge of #1222, I plan to issue the new release tonight or tomorrow morning (pending a successful doc build and a final check-over).

[fmaussion]

It is picky but it might be worth to pin mpl to v2 for nicer plots in the docs ;-) Am I allowed to try it out?

[shoyer]

Certainly, give it a try!

…

On Tue, Jan 24, 2017 at 1:28 PM, Fabien Maussion ***@***.***> wrote: It is picky but it might be worth to pin mpl to v2 for nicer plots in the docs ;-) Am I allowed to try it out? — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#1167 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ABKS1qt-_CC_qfNcXHNEOa5tIfTETEHEks5rVm0WgaJpZM4LOzyA> .

[rabernat]

Not that this matters greatly to xarray, but I am happy to report that the latest master causes no problems with xmitgcm, which, because of the custom DataStore, is a somewhat obscure edge case. Full speed ahead with the release!

[shoyer]

@rabernat the DataStore refactor will happen sometime soon I hope, but I will definitely coordinate with you!