You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
But WIT.process_data has simultaneously been supplied an xarray of raster data with the wrong spatial extent, at least according to data.geobox: lon/lat truncate to 168,-28
Consequently, after invoking a (via wrapper) rasterio.features.geometry_mask (to overlay the vector geometry feature onto the raster defined by the xarray geobox), WIT.process_data finds no pixels (and logs the polygon area as zero), and ultimately returns an empty pandas.DataFrame (with column headings but no index rows).
This appears to be resolved by 23c085e (alas at the expense of #111). Ideally would still narrow this down (and feed into a regression test, to ensure compatibility with the next stable release of datacube-core).
WIT appears to be outputting many parquet files that contain only column headers, and no data rows.
Appears to be related to geopolygon transformations. Possibly a reprojection (CRS84 vs EPSG:4326 vs EPSG:3577) bug in ODC core 1.8.3 virtual products?
One example:
HTTP request (click to expand): lon/lat truncate to 153,-27
Logs of
WIT.process_data(data, parameters)
seem to show the method call is invoked with the expected vector input, supplied inparameters['feature']
:But
WIT.process_data
has simultaneously been supplied anxarray
of raster data with the wrong spatial extent, at least according todata.geobox
: lon/lat truncate to 168,-28Consequently, after invoking a (via wrapper)
rasterio.features.geometry_mask
(to overlay the vector geometry feature onto the raster defined by the xarray geobox),WIT.process_data
finds no pixels (and logs the polygon area as zero), and ultimately returns an emptypandas.DataFrame
(with column headings but no index rows).h/t @emmaai
The text was updated successfully, but these errors were encountered: