xvrt

GDAL multidimensional VRT from xarray Datasets and STAC catalogues

xvrt writes GDAL multidimensional VRT files: small XML documents that describe a collection of files as one n-dimensional dataset. The VRT holds no data. It records which file supplies which slab of which array, plus the dimensions, coordinates, CRS, units and packing, so any reader of GDAL’s multidimensional API opens the whole collection as a single cube and reads only the files a request touches.

It is pure Python (no GDAL needed to write) and has two front doors:

xvrt 0.1.0

A recipe, not a payload

Opening a long time series of files is expensive work: coordinates are read, dimensions aligned, packing and time decoded. Most tools do this work each time and then throw the result away. A VRT keeps it as a portable text artifact that GDAL, and anything built on GDAL, can reuse directly. It sits between the file collection and the analysis: no bytes are copied, and the description can be versioned, shared and inspected like any other text file.

Install

pip install xvrt
pip install "xvrt[stac]"   # adds pyarrow and pyproj for from_stac()

Requirements are numpy and xarray. Reading stac-geoparquet needs pyarrow; turning an EPSG code into WKT needs pyproj (or GDAL’s Python bindings). Reading the result needs GDAL 3.x with multidimensional support.

Where to go next

  • Get started: a set of daily NetCDF files to one VRT, read back through GDAL.
  • The Guide covers each feature: compositions, packed data, multiple variables, raster (GeoTIFF/COG) sources, catalogue-driven VRTs and CRS handling.
  • GHRSST MUR on source.coop: 8913 days of five variables, built from a STAC catalogue in seconds.
  • Reference: every public function.