rasterizer 0.3

A Python package to rasterize GeoDataFrames

Rasterizer

rasterizer is a lightweight Python package for rasterizing geopandas GeoDataFrames.

Features

• Rasterize lines into a binary (presence/absence) or length-based grid.
• Rasterize polygons into a binary (presence/absence) or area-based grid.
• Hybrid polygon rasterization for large polygon bounding boxes: exact clipping on boundary cells, faster scanline filling for interior cells.
• Weighted rasterization: Rasterize geometries while weighting the output by a numerical column in the GeoDataFrame.
• Works with geopandas GeoDataFrames.
• Outputs an xarray.DataArray for easy integration with other scientific Python libraries.
• No GDAL dependency for the rasterization algorithm itself.

For detailed usage and API documentation, please see the full documentation.

Usage

Here are some examples of what you can do with rasterizer.

import geopandas as gpd
from rasterizer import rasterize_polygons

polys = gpd.read_file("polygons.gpkg")
area_raster = rasterize_polygons(polys, your_x_grid, your_y_grid, polys.crs, mode="area")

Rasterizing Lines

You can rasterize lines in either binary or length mode.

Binary Mode	Length Mode

Rasterizing Polygons

You can rasterize polygons in either binary or area mode.

For polygon workloads, rasterizer now uses two internal strategies. Small polygon bounding boxes are handled with exact per-cell clipping. Larger ones switch to a hybrid path that still clips boundary cells exactly, but fills interior spans with a scanline pass to reduce the amount of geometric clipping required. The resulting area and binary outputs stay exact at cell boundaries while scaling better on large polygons.

Binary Mode	Area Mode

Installation

You can install the package directly from PyPI:

pip install rasterizer

Why rasterizer

This package provides functionalities that are not present in rasterio.features, such as area and length-based rasterization. It is also lighter and faster than using GDAL-based solutions. GDAL's rasterization only burns values per pixel; it cannot return exact fractional area or length contributions without an expensive workaround. The common workaround is to rasterize at a much finer resolution and then downsample with averaging, which approximates the true area/length but is not exact and can be slow, e.g.:

gdal_rasterize -burn 1 -tr 1 1 -ot Float32 -of GTiff input.gpkg tmp_fine.tif
gdalwarp -tr 10 10 -r average tmp_fine.tif out_area_approx.tif

Doing this purely in geopandas by generating one polygon per grid cell and overlaying it with the input geometry is also slow because it creates a huge number of tiny geometries, triggers expensive overlay operations, and scales poorly with grid size.