rio-stac-io

rio-stac-io is a rasterio extension to open STAC Items, ItemCollections, client search results, and STAC-compliant GeoDataFrame tables (e.g. from STAC GeoParquet) using native GDAL drivers including STACIT, STACTA and GTI. The library is build on top of rasterio and pystac. Full notes and a GeoParquet example are in the documentation.

Documentation

https://planetlabs.github.io/rio-stac-io

Installation

pip install rio-stac-io

When using the GTI driver you will need to install gti extras. Your GDAL binaries need to be compiled with geoparquet support.

pip install rio-stac-io[gti]

Usage

from pystac_client import Client

import rio_stac_io as stacio

client = Client.open(...)
search = client.search(...)

with stacio.open(search, asset_key="data") as src:
    data = src.read()

GeoDataFrame input requires pip install rio-stac-io[gti]; rows must follow a STAC item layout (as in stac-geoparquet). The library tries GTI first, then STACIT if GTI is not usable; the use_gti argument is ignored for this input type. Example: read a GeoParquet file, filter rows, then open the asset key:

import geopandas as gpd
import rio_stac_io as stacio

gdf = gpd.read_parquet("items.parquet")
gdf = gdf[gdf["collection"] == "s2"]

with stacio.open(gdf, asset_key="cog") as src:
    data = src.read()

Development

This repository requires Pixi v0.67.2 or later.

git clone git@github.com:planetlabs/rio-stac-io.git
cd rio-stac-io
pixi shell -e dev

Multiple Pixi environments pin different libgdal versions (e.g. dev, dev-gdal310, dev-gdal312, dev-gdal313, dev-gdal311-noparquet; see pyproject.toml for the full list). Rasterio is installed from source ([tool.pixi.pypi-options] no-binary = ["rasterio"]) so it links against the conda GDAL in that environment.

Troubleshooting: rasterio linked to the wrong GDAL

Switching between any two GDAL environments locally (for example moving between any of the dev-gdal* variants) can leave Pixi's PyPI/build cache holding a rasterio build compiled against a different libgdal. Symptoms include odd GTI or CRS failures, or rasterio.__gdal_version__ not matching gdalinfo --version in the same shell.

Check:

pixi run -e <env> verify-gdal

If that fails, clear the cached build and reinstall the environment (not usually needed on CI—runners are isolated per job):

pixi clean cache --pypi --build -y
rm -rf .pixi/envs/<env>
pixi install -e <env>
pixi run -e <env> verify-gdal

Replace <env> with the environment you use (dev, dev-gdal310, dev-gdal312, dev-gdal313, dev-gdal311-noparquet, …).