TypeError: Type is not JSON serializable: numpy.float64

[cordmaur]

Hi Guys,
Any reason for the methods .bounds() and .transform() return the values in numpy.float64 instead of standard python float?
This error was being raised when creating a STAC catalog and using rioxarray to get footprint, bounds, etc. and the code was working fine when using plain rasterio. Debugging was tricky as the representation of these values is the same in the dictionary and a manual type casting was necessary for each single value. :(

[snowman2]

I don't see anything in the code explicitly casting to float64. It may be something to do with being an xarray extension and the way the response gets handled under the hood.

This might be helpful: https://github.com/pyproj4/pyproj/blob/8f0e1f7ddc5ba5a1992f939f91bb9ea023316219/pyproj/utils.py#L34-L52

[cordmaur]

Thanks @snowman2. This class is helpful, but the JSON encoding is called by pystac. Tracking down how things work in rioxarray, the _internal_bounds() functions is casting the top, left, right bottom values to float (as expected).

rioxarray/rioxarray/rioxarray.py

Line 874 in a0322e9

def _internal_bounds(self):

The "issue" is in the _chached_transform() function:

rioxarray/rioxarray/rioxarray.py

Line 503 in a0322e9

def _cached_transform(self):

When calling Affine.from_gdal() and passing the np.fromstring as argument. From this point, all computations will be in numpy.float64 dtype...

[snowman2]

Good catch 👍. You are welcome to submit a PR ensuring it is a Python float.

[cordmaur]

@snowman2, PR is ready!
#475 (comment)