AttributeError: 'Dataset' object has no attribute 'rest'

[fabricebrito]

Hello,

xpublish looks very promising and I want to use it for serving a few datasets in an experiment. I've installed xpublish in a conda environment

I do run into the exception

AttributeError: 'Dataset' object has no attribute 'rest' when running the simple script:

#!/opt/anaconda/envs/env_xpublish/bin/python

import click
import sys
import pandas as pd
import numpy as np
import xarray as xr
import xpublish

ds = xr.Dataset({'foo': (('x', 'y'), np.random.rand(4, 5))},
                 coords={'x': [10, 20, 30, 40],
                        'y': pd.date_range('2000-01-01', periods=5),
                        'z': ('x', list('abcd'))})

ds.rest.serve(host='0.0.0.0', port=9000)

Any help/tips is really appreciated

[jhamman]

Hi @fabricebrito - thanks for trying out xpublish. Can I ask you to include a bit more information in your issue description? When I run your example, I get a server as expected:

In [1]: import click
   ...: import sys
   ...: import pandas as pd
   ...: import numpy as np
   ...: import xarray as xr
   ...: import xpublish
   ...:
   ...: ds = xr.Dataset({'foo': (('x', 'y'), np.random.rand(4, 5))},
   ...:                  coords={'x': [10, 20, 30, 40],
   ...:                         'y': pd.date_range('2000-01-01', periods=5),
   ...:                         'z': ('x', list('abcd'))})
   ...:
   ...:
   ...: ds.rest.serve(host='0.0.0.0', port=9000)

INFO:     Started server process [93713]
INFO:     Waiting for application startup.
INFO:     Application startup complete.
INFO:     Uvicorn running on http://0.0.0.0:9000 (Press CTRL+C to quit)

If you can, please add:

• the full traceback
• the output of xr.show_versions()

You may also try setting up the server this way to see if things work better for you:

In [2]: rest = xpublish.RestAccessor(ds)

In [3]: rest.serve()
INFO:     Started server process [93815]
INFO:     Waiting for application startup.
INFO:     Application startup complete.
INFO:     Uvicorn running on http://0.0.0.0:9000 (Press CTRL+C to quit)

[fabricebrito]

Hi @jhamman

I installed xpublish with conda and that's the behaviour I got. I'll reproduce it in a clean environment.

On the other hand, using pip -e with the source (as you do in .binder) everything is ok:

conda create -y -q -n env_xpublish
conda install -y -n env_xpublish -c conda-forge netcdf4 cachey dask fastapi numcodecs "numpy>=1.17" toolz uvicorn "xarray>=0.15" zarr
conda init bash
source ${HOME}/.bashrc{{conda }}
activate env_xpublish
git clone https://github.com/jhamman/xpublish.git
cd xpublish
pip install -e .

The xpublish approach looks quite promising and we're going to use it in the OGC Testbed 16 for Data Access and Processing API (DAPA) for Geospatial Data (https://portal.ogc.org/files/?artifact_id=91644#PartDAPA)

I'll keep you posted. Thanks!

[jhamman]

@fabricebrito - thanks for updating this issue. After looking into this a bit more, I'm fairly sure there is something odd going on with your install script. I'm not able to reproduce the problem with this install path:

conda create -n xpublish_test xpublish netcdf4
conda activate xpublish_test
...
# run your script above

[fabricebrito]

Hello @jhamman - The installation was ok in other environments. I can't really tell why it didn't before. Closing the issue.

PS having fun with intake-stac

[fabricebrito]

@jhamman I believe the conda package should be updated with the latest code. While trying to serve landsat data via xpublish and intake-stac, I got an exception line 62 with the conda package v0.2 at the URL .zmetadata:

  File "/home/fbrito/.conda/envs/env_intake/lib/python3.8/site-packages/xpublish/rest.py", line 155, in get_zmetadata
    return self.zmetadata_json()
  File "/home/fbrito/.conda/envs/env_intake/lib/python3.8/site-packages/xpublish/rest.py", line 83, in zmetadata_json
    zjson = copy.deepcopy(self.zmetadata)
  File "/home/fbrito/.conda/envs/env_intake/lib/python3.8/site-packages/xpublish/rest.py", line 78, in zmetadata
    self._zmetadata = self._get_zmetadata()
  File "/home/fbrito/.conda/envs/env_intake/lib/python3.8/site-packages/xpublish/rest.py", line 62, in _get_zmetadata
    encoded_da, self._encoding.get(key, {}), da.encoding["dtype"]
KeyError: 'dtype'

I removed the conda package and installed via git clone + pip install -e . and the .zmetadata works as expected

Hope this helps

[benbovy]

@fabricebrito xpublish 0.1.0 has been released and conda and pip packages are now available.

I guess it should solve your issue?