fix: $resultFormat=CSV and GeoJSON crash server with 500, parser token missing

[Vishmayraj]

While working on the STA to STAC metadata connector, I needed to fetch observations in CSV format to test the pipeline. Every request with $resultFormat=CSV was returning 500, so I pulled the Docker logs and traced the error.

The root cause was in lexer.py where the RESULT_FORMAT_VALUE token pattern only matched dataArray. So when the parser saw CSV, it tokenized it as an EXPAND_IDENTIFIER and threw an exception. That exception had no dedicated handler in read.py, so it fell through to the generic except Exception block and returned 500.

Changes

lexer.py => extend the pattern to cover all three spec-defined values:

# before
"RESULT_FORMAT_VALUE": r"\bdataArray\b",
# after
"RESULT_FORMAT_VALUE": r"\bdataArray\b|\bCSV\b|\bGeoJSON\b",

read.py => wrap convert_query in its own try/except returning 400. Parser exceptions are always caused by malformed client input, not server faults, so 500 was the wrong code here regardless.

Note

While tracing this I found that CSV and GeoJSON serialization is not actually implemented. The parser scaffolding exists, ResultFormatNode, parse_result_format, and the entity validation in convert_query are all there, but result_format is never passed through to the return dict or the streaming layer in read.py, so the response is always JSON regardless of what $resultFormat is set to. Only dataArray appears to have any downstream handling.

This PR only fixes the 500 and the wrong error code. The actual serialization is a separate feature that needs to be built. I am opening a separate issue for that.

References

OGC SensorThings API Part 1, Section 9.3.4 => valid $resultFormat values are dataArray, CSV, and GeoJSON.