Fix complex int16 dn dtype slc #5
Conversation
| @@ -433,7 +432,11 @@ def _load_digital_number(self, resolution=None, chunks=None, resampling=rasterio | |||
| # sort chunks keys like map_dims | |||
| chunks = dict(sorted(chunks.items(), key=lambda pair: list(map_dims.keys()).index(pair[0]))) | |||
| chunks_rio = {map_dims[d]: chunks[d] for d in map_dims.keys()} | |||
|
|
|||
| riot_dtyp = rio.dtypes[0] | |||
There was a problem hiding this comment.
@Grouny , can you replace with:
try:
rio_dtyp = self._rio_dtypes[rio.dtypes[0]]
except KeyError:
rio_dtyp = rio.dtypes[0]and define:
self._rio_dtypes = {'complex_int16' : 'complex'}somewhere in init ?
| @@ -479,11 +482,13 @@ def _load_digital_number(self, resolution=None, chunks=None, resampling=rasterio | |||
| rio.height // resolution['atrack'] * resolution['atrack']) | |||
|
|
|||
| if self.s1meta.driver == 'GTiff': | |||
| logging.info('rio dtypes zero: %s',rio.dtypes[0]) | |||
There was a problem hiding this comment.
To be removed if it's works as expected.
|
Hi, xr.open_dataset(f,chunks=chunks_rio, parse_coordinates=False,engine='rasterio')this syntax benefits from xr.open_rasterio(f, chunks=chunks_rio, parse_coordinates=False)is not working for SLC data. |
|
Yes, it would be nice to drop xr.open_rasterio in favor of rioxarray in xsar. So for now, I think that this PR is the best solution. |
|
Going back on this PR, I cannot find the internal types to decode the .tiff files for SLC products since it is now integrated in the |
|
It's seems that this problem has been fixed upstream in rasterio. @agrouaze , are you ok to close this PR without merge ? |
|
Ok to close this PR without merge. |
small fix to allow
rasteriocomplex_int16to be interpreted asnp.complex_in order to fix the dtype casting error in case of coarse resolution options on SLC (complex Digital Number).For instance
xsar.open_dataset(wv_slc_meta.subdatasets[good_indice],resolution=resolutionZ,resampling=rasterio.enums.Resampling.rms)
is now possible.
Note that rasterio will be probably replaced by rioxarray in xarray backends: pydata/xarray#5491