Data error handling (e.g., lenient mode)

[bjdmeest]

There are a couple of MUST statements in the spec that depend on the actual data that could be revisited to support some kind of 'lenient' mode of processing the mapping document.

For example: I have a mapping file with 10 triples maps that map 10 tables, which I use for mapping databases A and B (and many more others). 3 of those tables are optional in database B, crashing the mapping engine due to the fact that a logical source description MUST resolve to a data source (see eg R2RML spec)

• lenient mode: process only the 7 resolvable triples maps
• similar for The referenced columns of all term maps of a triples map (subject map, predicate maps, object maps, graph maps) MUST be column names that exist in the term map's logical table.: don't crash if you don't have all references, but generate all others

It's an open question on whether this is functionality of the mapping engine or a feature of the language, however I think it would be good to clarify which errors should result in exiting the engine vs handling gracefully

[dachafra]

I think this issue is very related to the recent issue closed #32. IMHO, at this moment, it is out of the scope of the CG but we can transfer this issue to best-practices and/or mark it as working-group. In a more general way, but these issues are clarified in Data Errors section.