raster - Python equivalent of gdalbuildvrt

Is there a way to perform the same task as the gdalbuildvrt utility using GDAL Python bindings? So far I have not found any way to do this other than creating a vrt of a single dataset and manually editing the xml. I would like to create a vrt from multiple rasters (essentially performing a mosaic). Is this possible using pure Python? My other option is to use subprocess to simply call gdalbuildvrt.

Answer

Honestly it's easier to do this by using gdalbuildvrt in a subprocess or os.system.

Should you wish to do this through Python it can be done. Using the standard dataset creation methods within GDAL Python we can easily create the base dataset VRT.

from osgeo import gdal

drv = gdal.GetDriverByName("VRT")
vrt = drv.Create("test.vrt", x_size, y_size, 0)

Note that we are creating the dataset with no bands initially. From the documentation on VRTs that VRT datasets are one of the few dataset types that can accept AddBand arguments.

vrt.AddBand(gdal.GDT_Float32)
band = vrt.GetRasterBand(1)

Now for each band we have to set the metadata items manually:

simple_source = '%s' % source_path + \
    '%i' % source_band + \
    '' % (x_size, y_size, x_block, y_block) + \
    '' % (x_offset, y_offset, x_source_size, y_source_size) + \
    '' % (dest_x_offset, dest_y_offset, x_dest_size, y_dest_size)
band.SetMetadataItem("SimpleSource", simple_source)
band.SetMetadataItem("NoDataValue", -9999)

SetMetadatItem takes two arguments, the first a string of the metadata item, the second the item itself. This means that you can't subset a metadata item, so for data sources you have to set the entire contents as a string.

Note that we can use this method to create complex sources (ComplexSource) that contain look-up-tables of values, Kernel filter sources (KernelFilteredSource) of arbitrary sizes and shapes, and Mask Bands (MaskBand).