OBOFormat: Configuration option to skip validation before writing OBO file

[hdietze]

By default the OBOFormatWriter validates the converted OBO document before writing the file. For certain use cases (e.g. debugging), the writer has already an option to skip that validation step.

What would be the best way to expose this configuration option though the OWL-API?

[cmungall]

Just for background, there are somethings that are "syntactically valid" (ie they conform to the grammar) but "structurally invalid" (e.g. multiple rdfs:labels per class)

Cardinality constraints on annotation assertion axioms:
http://oboformat.googlecode.com/svn/trunk/doc/obo-syntax.html#4.4

Prohibiting rendering of obo-structurally-invalid ontologies may be too harsh. E.g. a naive user may have ended up with two labels on a class and may not know how to resolve. They should be allowed to save their file ontologies so that others can look at it and help resolve.

I think these would be better separated out from the business of rendering and treated as a separate lint like check (the checks are currently done on the obo document model, in theory they could be migrated to run without any obo renderer dependency).

The exception is for the logical axioms (intersection and union of)

Whilst as Horrocks originally noted,

id: X
intersection_of: Y

is just another way of writing

EquivalentClasses(X ObjectIntersectionOf(Y))

Which is equivalent to

EquivalentClasses(X Y)

However, we really want to prohibit this, as 99.9999% of the time when we have found an obo file with a class with a single intersection-of tag it has been an error. There's no reason to allow this, so I favor keeping this in the rendering.

[ignazio1977]

Currently OWLDocumentFormat has a parameter map that is supposed to be used for format specific "parameters", i.e., generic stuff that a parser/renderer for a format might use for whatever purpose. I think this use case can make use of these contextual parameters, so that the OBO storer would know to check these parameters. The default can be decided by what's the expected common case, i.e., whether validation should be on or off.

[ignazio1977]

Regarding the type of parameter, it might be a good idea to have an enum to describe the validations to perform. Or, thinking of OWL profiles, a set of validations to run on the OBO document?

[ignazio1977]

This commit allows to set a boolean on OBODocumentFormat to enable/disable checks. Default is true, which is the current behaviour.
In the current implementation this needs to be set manually on each format. We can decide on a default value to be set and add providers to simplify use.

[hdietze]

Thank you for creating the patch. This might seem like nitpicking, but why are you modifying the OBOFormatWriter? In theory it would be enough to set the flag on the writer instance in the OBOFormatRenderer.render function. We already create an OBOFormatWriter instance there anyway. No need to push it down into the writer (yet).

[ignazio1977]

Good point, I'll rework it.

[ignazio1977]

ffd0fcb

[hdietze]

Perfect. Thank you.

[ignazio1977]

Forgot to use the format actually associated with the ontology, fixed in 1043f2c