Support for wildcard property filtering in load_stac?

[soxofaan]

User tries load_collection with SENTINEL2_L2A and a property filter

"tileId":{
  "process_graph": {
    "eq1": {
      "arguments": {"x": {"from_parameter": "value"}, "y": "32U*B"},
      "process_id": "eq",
      "result": true
    }
  }
}

This wildcard match "32U*B" used to work on OpenSearch catalog based collections, as implemented under Open-EO/openeo-opensearch-client#25

But that results now, with load_stac based collections in

OpenEO batch job failed: NoDataAvailableException(status_code=400, code='NoDataAvailable', message='There is no data available for the given extents.', id='no-request')

[JeroenVerstraelen]

From VLCC:

Thank you for submitting this request. It would be greatly appreciated if a solution for this could be prioritized, as this issue has a significant impact on our production workflow.

[soxofaan]

It's unfortunate that this wildcard feature (which is non-standard and did not get test coverage) got in a production workflow.
We basically created a lock-in here and then unknowingly locked ourselves out when changing from OpenSearch to STAC

A better solution is to stick to standard functionality, e.g. in this case where it is feasible to replace the wildcard with a full listing:

from openeo.processes import array_contains
...
cube = connection.load_collection(
    ...
    properties={
        "grid:code": lambda p: array_contains(["MGRS-32ULB", "MGRS-32UMB"], p)
    }

Note: the next release (0.49.0) of openeo python client allows to express this without the cumbersome lambda:

cube = connection.load_collection(
    ...
    properties=[
        openeo.collection_property("grid:code").is_one_of(
            ["MGRS-32ULB", "MGRS-32UMB"]
        )
    ]

If actual wildcard support is necessary, we should do it properly with defining an explicit regex-aware process (e.g. regex_match), instead of arbitrarily overloading a well-defined process like eq

[JeroenVerstraelen]

Related job-id: j-2603031230544f5dbd7a3ebc73728d4f