def test_single_shape_should_have_equivalent_counts(self):
# Read in raster projection (applies to all raster tiles).
rast_proj = self.dataset.proj
# Compute counts for a single shape using single-tile computation for reference.
shp = self.px_shps[0]
mask = rasterize(shp, ext_outline=0, int_outline=1).T
minx, miny, maxx, maxy = shp.bounds
pts = (np.argwhere(mask>0) + np.array([minx, miny])).astype(int)
counts_reference = np.bincount(self.rb[pts[:,1],pts[:, 0]], minlength=256)
# Now, compute counts for a single shape using RasterDataset, and compare against single-tile computation.
values = np.array(self.dataset.get_values_for_pixels(pts)).astype(np.uint8)
counts = np.bincount(values, minlength=256)
assert(np.array_equal(counts, counts_reference))
def test_all_shapes_should_have_equivalent_counts(self):
# Read in raster projection (applies to all raster tiles).
rast_proj = self.dataset.proj
# All shapes in the layer should already be in pixel coords. Check that we have the right number of them.
assert (len(self.px_shps) == 23403)
# Compute counts for a single shape using single-tile computation for reference.
for i, shp in enumerate(self.px_shps):
# Rasterize the shape, and find list of all points. This is the same for either case.
mask = rasterize(shp, ext_outline=0, int_outline=1).T
minx, miny, maxx, maxy = shp.bounds
pts = (np.argwhere(mask>0) + np.array([minx, miny])).astype(int)
# Tiled data (in RasterDataset): Compute counts.
values_td = np.array(self.dataset.get_values_for_pixels(pts)).astype(np.uint8)
counts_td = np.bincount(values_td, minlength=256)
# Single-tile Dataset: Compute counts for reference, and compare for equality.
counts_ref = np.bincount(self.rb[pts[:,1],pts[:, 0]], minlength=256)
assert(np.array_equal(counts_td, counts_ref))
def get_counts_for_shape(self, shp):
mask = rasterize(shp, ext_outline=0, int_outline=1).T
minx, miny, maxx, maxy = shp.bounds
pts = (np.argwhere(mask > 0) + np.array([minx, miny])).astype(int)
return np.bincount(self.rb[pts[:, 1], pts[:, 0]], minlength=256)