skmap.io.base.read_rasters_cpp#

read_rasters_cpp(raster_files=[], band=1, window=None, n_jobs=8, out_data=None, out_idx=None, dtype=<class 'numpy.float32'>, gdal_opts={}, verbose=False)[source]#

Read rasters in parallel using the C++ backend, aggregating them into a single array.

Parameters:

  • raster_files (Union[List[Union[str, Path]], str, Path]) – A list with the raster paths.

  • band (Union[List[int], int]) – The band to be read from each raster file

  • window (Optional[Window]) – The window (if any) to read from the raster

  • n_jobs (int) – The number threads to read in parallel

  • out_data (Optional[ndarray[tuple[Any, ...], dtype[float32]]]) – a pre-allocated array to write into

  • out_idx (Optional[List]) – permutation array

  • dtype (type) – Datatype (currently only np.float32 is supported)

  • gdal_opts (dict) – additional options to be passed to GDAL

  • verbose – Whether to print extra output

Returns:

A 2D array of n_bands by n_pixels

Return type:

NDArray[np.float32]

Examples

>>> import rasterio
>>> import tempfile
>>> import numpy as np
>>> from skmap.io.base import read_rasters_cpp
>>> from pathlib import Path
>>> # Create a dummy raster
>>> with tempfile.TemporaryDirectory() as tempdir:
...     band_1 = np.random.rand(100,100).astype(np.float32)
...     band_2 = np.random.rand(100,100).astype(np.float32)
...     transform = rasterio.transform.from_origin(0,0,1,1)
...     raster_name = Path(tempdir)/"example.tif"
...     with rasterio.open(
...         raster_name,
...         "w",
...         height=100,
...         width=100,
...         count=2,
...         dtype=np.float32,
...         crs="EPSG:4326",
...         transform=transform
...     ) as raster:
...         raster.write(band_1, 1)
...         raster.write(band_2, 2)
...     # read the raster bands in parallel
...     bands = read_rasters_cpp(raster_files=[raster_name, raster_name], band=[1,2])
...     # the shape is bands x n_pix
...     assert bands.shape == (2,100*100)
...     # bands are selected as above
...     np.testing.assert_equal(band_1.reshape(100*100), bands[0,:])
...     np.testing.assert_equal(band_2.reshape(100*100), bands[1,:])