Nodata value type causes error

[jackieleng]

In [1]: import rioxarray

In [2]: import xarray as xr

In [3]: d = xr.open_rasterio('./data/ASTGTMV003_N01E103_dem.tif')

In [4]: g = {"type": "Polygon", "coordinates": [[[103.85548, 1.27909], [103.86977, 1.27909], [103.85548, 1.29498], [103.85548, 1.27909]]]}

In [5]: d
Out[5]:
<xarray.DataArray (band: 1, y: 3601, x: 3601)>
[12967201 values with dtype=int16]
Coordinates:
  * band     (band) int64 1
  * y        (y) float64 2.0 2.0 1.999 1.999 1.999 ... 1.001 1.001 1.001 1.0 1.0
  * x        (x) float64 103.0 103.0 103.0 103.0 ... 104.0 104.0 104.0 104.0
Attributes:
    transform:      (0.000277777777777778, 0.0, 102.999861111111, 0.0, -0.000...
    crs:            +init=epsg:4326
    res:            (0.000277777777777778, 0.000277777777777778)
    is_tiled:       1
    nodatavals:     (nan,)
    scales:         (1.0,)
    offsets:        (0.0,)
    descriptions:   ('Band 1',)
    AREA_OR_POINT:  Area
    Band_1:         Band 1

In [6]: d.sel(band=1).rio.clip([g])
/##/.venv/lib/python3.8/site-packages/pyproj/crs/crs.py:280: FutureWarning: '+init=<authority>:<code>' syntax is deprecated. '<authority>:<code>' is the preferred initialization method. When making the change, be mindful of axis order changes: https://pyproj4.github.io/pyproj/stable/gotchas.html#axis-order-changes-in-proj-6
  projstring = _prepare_from_string(projparams)
---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
<ipython-input-6-49e09f098c16> in <module>
----> 1 d.sel(band=1).rio.clip([g])

~/##/.venv/lib/python3.8/site-packages/rioxarray/rioxarray.py in clip(self, geometries, crs, all_touched, drop, invert)
   1531                 )
   1532             )
-> 1533         if self.nodata is not None and not np.isnan(self.nodata):
   1534             cropped_ds = cropped_ds.fillna(self.nodata)
   1535

~/##/.venv/lib/python3.8/site-packages/rioxarray/rioxarray.py in nodata(self)
   1107             return None
   1108
-> 1109         self._nodata = _ensure_nodata_dtype(self._nodata, self._obj.dtype)
   1110         return self._nodata
   1111

~/##/.venv/lib/python3.8/site-packages/rioxarray/rioxarray.py in _ensure_nodata_dtype(original_nodata, new_dtype)
    265     """
    266     original_nodata = float(original_nodata)
--> 267     nodata = np.dtype(new_dtype).type(original_nodata)
    268     if not np.isnan(nodata) and original_nodata != nodata:
    269         warnings.warn(

ValueError: cannot convert float NaN to integer

Problem description

I'm not sure if this is the correct place to post this, or if this issue belongs to the xarray repo. But I'm getting an error if I do a clipping after a selection (band=1 in this case). If I do not do a .sel, there seem to be no problem.

The problem seems to be because the data has a dtype of int16, but it has a 'nodatavals' attribute that is nan, which has a dtype of float. The nodatavals seems to be set by xarray. The _ensure_nodata_dtype tries to convert the nan into an integer, which causes the error.

Expected Output

No error.

Environment Information

• python -c "import rioxarray; rioxarray.show_versions()"

(.venv)~/### (master *%)$ python -c "import rioxarray; rioxarray.show_versions()"
rioxarray (0.1.1) deps:
  rasterio: 1.1.8
    xarray: 0.16.2
      GDAL: 2.4.4

Other python deps:
     scipy: 1.5.4
    pyproj: 3.0.0.post1

System:
    python: 3.8.5 (default, Jul 28 2020, 12:59:40)  [GCC 9.3.0]
executable: /###/.venv/bin/python
   machine: Linux-4.19.128-microsoft-standard-x86_64-with-glibc2.29

[snowman2]

I would recommend using rioxarray.open_rasterio.

[snowman2]

A few years ago, I added the nodatavals: pydata/xarray#1740

nan was used because None was not serializable with to_netcdf (not because nan was the nodata value). This is no longer the case when you use rioxarray.open_rasterio as it just doesn't store the no data value when there isn't one. Other benefits of using rioxarray.open_rasterio here.

The current functionality for the data type checking works as expected and so I don't plan on changing anything there.