Issue with xarray dimensions

[juanfcocontreras]

With the following xarray file, I get the following exception when trying to access the xarray dimensions:

ValueError: If using all scalar values, you must pass an index

I think the problem is on line 2718 of the views.py file:

dim = pd.DataFrame({"value": ds.coords[dim].data})

And I think it would be enough to change it for the next one to fix it:

dim = pd.Series({"value": ds.coords[dim].data}).to_frame()

[aschonfeld]

@juanfcocontreras Thanks for catching this and I plan on adding this to the release this weekend. Here's how I'll do it:

@dtale.route("/xarray-dimension-values/<data_id>/<dim>")
@exception_decorator
def get_xarray_dimension_values(data_id, dim):
    ds = global_state.get_dataset(data_id)
    try:
        dim = pd.DataFrame({"value": ds.coords[dim].data})
    except ValueError:
        dim = pd.Series({"value": ds.coords[dim].data}).to_frame()
    dim_f, _ = build_formatters(dim)
    return jsonify(data=dim_f.format_dicts(dim.itertuples()))

Let me know if you're alright with that and I'll move forward 🙏

[juanfcocontreras]

After that fix, dtale gives another exception in:

Input:

ds = xr.open_dataset("./roi.nc")
show(ds)

Output:

Traceback (most recent call last):
  File "/usr/local/Caskroom/miniconda/base/envs/conda-forge/lib/python3.7/site-packages/dtale/views.py", line 79, in _handle_exceptions
    return func(*args, **kwargs)
  File "/usr/local/Caskroom/miniconda/base/envs/conda-forge/lib/python3.7/site-packages/dtale/views.py", line 2723, in get_xarray_dimension_values
    return jsonify(data=dim_f.format_dicts(dim.itertuples()))
  File "/usr/local/Caskroom/miniconda/base/envs/conda-forge/lib/python3.7/site-packages/dtale/utils.py", line 373, in format_dicts
    return list(map(self.format_dict, lsts))
  File "/usr/local/Caskroom/miniconda/base/envs/conda-forge/lib/python3.7/site-packages/dtale/utils.py", line 369, in format_dict
    for idx, name, f in self.fmts
  File "/usr/local/Caskroom/miniconda/base/envs/conda-forge/lib/python3.7/site-packages/dtale/utils.py", line 369, in <dictcomp>
    for idx, name, f in self.fmts
  File "/usr/local/Caskroom/miniconda/base/envs/conda-forge/lib/python3.7/site-packages/dtale/utils.py", line 323, in f
    return json_string(x)
  File "/usr/local/Caskroom/miniconda/base/envs/conda-forge/lib/python3.7/site-packages/dtale/utils.py", line 196, in json_string
    if pd.isnull(x):
ValueError: The truth value of an array with more than one element is ambiguous. Use a.any() or a.all()

[juanfcocontreras]

It's been a while since I've used dtale for xarray, but I think there have been several changes, probably in pandas that have affected its functionality.

If a single-point dataset is selected, it gives another exception. I don't know if both exceptions are related.

Input:

poi = ds.isel(rlat=0, rlon=0)
show(poi)

Output:

Traceback (most recent call last):
  File "/usr/local/Caskroom/miniconda/base/envs/conda-forge/lib/python3.7/site-packages/xarray/core/utils.py", line 542, in __len__
    return self.shape[0]
IndexError: tuple index out of range

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "/usr/local/Caskroom/miniconda/base/envs/conda-forge/lib/python3.7/site-packages/dtale/views.py", line 79, in _handle_exceptions
    return func(*args, **kwargs)
  File "/usr/local/Caskroom/miniconda/base/envs/conda-forge/lib/python3.7/site-packages/dtale/views.py", line 2709, in get_xarray_coords
    for coord, info in ds.coords.items()
  File "/usr/local/Caskroom/miniconda/base/envs/conda-forge/lib/python3.7/site-packages/dtale/views.py", line 2709, in <listcomp>
    for coord, info in ds.coords.items()
  File "/usr/local/Caskroom/miniconda/base/envs/conda-forge/lib/python3.7/site-packages/xarray/core/dataarray.py", line 555, in __len__
    return len(self.variable)
  File "/usr/local/Caskroom/miniconda/base/envs/conda-forge/lib/python3.7/site-packages/xarray/core/utils.py", line 544, in __len__
    raise TypeError("len() of unsized object")
TypeError: len() of unsized object

[aschonfeld]

Ok, thanks. Maybe i’ll hold off on this release and dig back into xarray for the next one. Thanks for your help

[aschonfeld]

@juanfcocontreras good news, I did some digging and the issue was with how I was obtaining the dimensions. I was assuming that everything in ds.coords were dimensions, but what I really needed to be iterating was ds.coords.dims. Now lat & lon won't show up as dimensions, which they shouldn't. I'll have this ready for the release 👍

[juanfcocontreras]

That's great, thank you very much!

By the way, I don't know if to put it right here or open a new issue, when open_browser=True, it shouldn't appear dtale embedded in Jupyter (or nteract in my case), right? The browser is open, but still appears in the output of the cell.

[aschonfeld]

Hmm, I never actually tried doing open_browser=True in jupyter before. It worked for me but left this cell behind it:

I can see if I can update it to no longer create that extra cell in that case

[aschonfeld]

Ok I think I have a fix for that too

[juanfcocontreras]

Thas is awesome!! Thank you so much!!

PD: I've played a little bit with the xarray fix and it seems to work fine!

[aschonfeld]

Terrific, I'll let you know when I publish 1.23.0

[aschonfeld]

Ok v1.23.0 should be on conda in the next 15 minutes. Just merged the version update

[juanfcocontreras]

Both fixes seem to be working great. Thank you very much!