With tabular data query formulations (including CSV), it makes sense to throw an error when a column reference is not found.
However, with hierarchical formulations like JSON and XML, I'm doubting whether throwing an error makes sense, since it is not guaranteed that the scheme of the source data is known upfront.
Also, IMO a reference can also be a more complex expression when using XPath or JSONPath, which, even if the engine is schema-aware, is hard, and possibly infeasible, to check.
Proposed:
Change 'Error expected' to 'No' for these test cases.
Change 'Expected knowledge graph' to 'None' for these test cases.
Hi all, I'm looking in to getting CARML to comply with the test cases. I've got a couple of RFC proposals for which I will create an issue
RFC for:
With tabular data query formulations (including CSV), it makes sense to throw an error when a column reference is not found. However, with hierarchical formulations like JSON and XML, I'm doubting whether throwing an error makes sense, since it is not guaranteed that the scheme of the source data is known upfront.
Also, IMO a reference can also be a more complex expression when using XPath or JSONPath, which, even if the engine is schema-aware, is hard, and possibly infeasible, to check.
Proposed: