XArrayInterface improvements: dimension units and labels #2431
Conversation
…nsion labels for all key and value dimensions, added test to make sure xarray.Datasets and xarray.DataArrays are treated equivalently
|
Looks good so far and the other suggestions sound good too although I suspect it might get tricky. |
|
I'm tempted to agree; the first two points should probably be implemented together, and the second one could be a real head scratcher. I'll try to put some thought into it, but if it's too ambitious, I'll just add the documentation I had in mind and leave the issue open for another time. Anyway, I would like to get this in the next release, if possible. Any ideas when that will be? |
drs251
changed the title
drs251
changed the title
|
We were hoping to have a feature freeze at the end of this week, with a release following a few days later. |
|
OK, then I'll probably concentrate on the docs for now. |
|
@philippjfr OK, I added a simple example of how visualize an xarray Dataset. I also found that something similar already exists for GeoViews, so I added some links. I don't think I'll have time right now to work on the other ideas I have for XArrayInterface, so I would consider this PR to be finished, but I'll leave issue #2319 open and hopefully I'll manage to come back to it sometime. |
|
At a quick glance this looks great! Would you mind clearing the notebook and committing that. We generally don't store notebooks with output and it will make it easier to review. |
| "cell_type": "markdown", | ||
| "metadata": {}, | ||
| "source": [ | ||
| "Note, that this approach immediately converts all available data to images, which will take up a lot of RAM for large datasets. For these situations, consider using a [DynamicMap](./06-Live_Data.ipynb) in conjunction with [xarray's dask support](http://xarray.pydata.org/en/stable/dask.html) instead.\n", |
There was a problem hiding this comment.
Maybe mention the dynamic argument to .to here, it will generate a DynamicMap for you.
|
One minor comment, otherwise looks good to merge. |
|
BTW: Does anyone know how to change the vdim label in this case? I've tried |
|
You mean changing a label after it has already been set? It was a deliberate decision not to allow that although @jlstevens was more insistent on that than I was. You can override the existing label like this though: xr_ds = xr.tutorial.load_dataset("air_temperature")
hv.Dataset(xr_ds, vdims=('air', 'air_temp')).redim(air=('air', 'air_temperature')).vdims[0].label |
|
Sorry, but I still don't get it: I'm absolutely unable to replace the rather long-winded "4xDaily Air temperature at sigma level 995" axis label in the plot within Holoviews. On the other hand, it's not essential for this PR though. |
|
Right, that's a separate issue in any case. I'll play around once merged and if I can reproduce I'll raise an issue. |
|
Thanks for the contribution! |
|
This pull request has been automatically locked since there has not been any recent activity after it was closed. Please open a new issue for related bugs. |
This addresses the first two points in #2319. As @philippjfr suggested, it's relatively straightforward to read units and dimension labels from xarray and attach them to the appropriate key and value dimensions (as far as I can tell, the attributes
unitsandlong_nameare a standard for netcdf files). I could not reproduce the the problem with the swapped kdims for xarray.DataArray vs. xarrayDataset, but I added a test to guard against it.I would like to try to add some stuff to this PR with your input (see #2319):
Support for dimensions without coordinatesPerhaps a better way to deal with non-coordinate dimensions