Support data URIs in load_url to support process graphs with embedded geoparquet, ...

[soxofaan]

This came up recently a couple of times: the need to specify geometry in a certain CRS (e.g. UTM), without having to worry about reprojections of lon-lat GeoJSON.

At the moment we support that with load_url and an externally hosted (for example) GeoParquet file. It's however cumbersome to have that file hosted publicly over HTTP(s), compared to GeoJSON which can be embedded directly in the process graph.

I created Open-EO/openeo-processes#498 to extend to scope of load_url to also support data URI, which would be a simple standardized way to embed binary data in a process graph. For example

  "lu": {
    "process_id": "load_url",
    "arguments": {
      "url": "data:application/vnd.apache.parquet;base64,UEFSMRUEFRAVFEwVAhUAEgAACBwqAAAAAAAAABUAFRIVFiwVAhUQFQYVBhwYCCoAAAAAAAAAGAgqAAAAAAAAABYAKAgqAAAAAAAAABgIKgAAAAAAAAAAAAAJIAIA...",  

It should be fairly straightforward to add support for this in our vector cube loading logic