function and provided the polygon coords in a" GeoJSON-like dict or an object that implements the Python geo interface protocol".
"AttributeError: 'numpy.ndarray' object has no attribute 'transform'"
when trying to acces to the dataset attribute 'transform'. Datset is a numpy.ndarray and, since it does not have such attribute, the error appears.
I have created a quick ndarray with numpy to see is normally this object has this attribute, and the same error appears when accessing such attribute, like follows:
>>> "AttributeError: 'numpy.ndarray' object has no attribute 'transform'"
I guess if there is a deprecation error from rasterio then, and maybe numpy has updated the name of its attributes or something similar.
I would be very appreciate if someone could shed some light on this :)
Here I share my code:
--------------------------------------------------
>>>> import rasterio
>>>> import json
>>> import geopandas as gpd
>>>> from rasterio.plot import show
>>>> from rasterio.merge import merge
>>>> from rasterio.mask import mask
>>> corrientes_gdf = gpd.read_file('corrientes.shp')
>>> alos_1 = rasterio.open('data\Layers\ALOS\ALOS-0000000000-0000000000.tif')
>>> alos_2 = rasterio.open('data\Layers\ALOS\ALOS-0000000000-0000046592.tif')
>>> alos = merge([alos_1, alos_2])[0]
>>> coords = [json.loads(corrientes_gdf .to_json())['features'][0]['geometry']]
>>> alos_clip, alos_transform = mask(alos, shapes=corriente_coords, crop=True, nodata=0)
Thank you ver much.
Kind regards,
Miguel GonzálezJ
Well, rasterio.merge returns a numpy array and a transform, not a rasterio.dataset. It's important that mask receives a rasterio.dataset because it needs a transform. And, yeah, numpy arrays aren't geolocated and don't have a geotransform.
There is, however a "dst_path" argument to merge, so you can just:
```
merge([alos_1, alos_2], dst_path='/tmp/merged.tif')
merged = rasterio.open('/tmp/merged.tif')
alos_clip, alos_transform = mask(merged, shapes=coords, crop=True, nodata=0)
```
Or, you could create an in-memory file, fill it with the raster data and use that for mask. Maybe something like this:
```
import io
# other code ...
alos_np, alos_transform = merge([alos_1, alos_2])
profile = {
**alos_1.profile,
'width': alos_np.shape[1],
'height': alos_np.shape[0]
}
fp = io.BytesIO()
with rasterio.open(fp, 'w', **profile) as tif:
tif.write(alos_np)
alos_clip, alos_transform = mask(fp, shapes=coords, crop=True, nodata=0)
```
Well, rasterio.merge returns a numpy array and a transform, not a rasterio.dataset. It's important that mask receives a rasterio.dataset because it needs a transform. And, yeah, numpy arrays aren't geolocated and don't have a geotransform.
There is, however a "dst_path" argument to merge, so you can just:
```
merge([alos_1, alos_2], dst_path='/tmp/merged.tif')
merged = rasterio.open('/tmp/merged.tif')
alos_clip, alos_transform = mask(merged, shapes=coords, crop=True, nodata=0)
```
Or, you could create an in-memory file, fill it with the raster data and use that for mask. Maybe something like this:
```
import io
# other code ...
alos_np, alos_transform = merge([alos_1, alos_2])
profile = {
**alos_1.profile,
'width': alos_np.shape[1],
'height': alos_np.shape[0]
}
fp = io.BytesIO()
with rasterio.open(fp, 'w', **profile) as tif:
tif.write(alos_np)
alos_clip, alos_transform = mask(fp, shapes=coords, crop=True, nodata=0)
```
