Perhaps change handling of failures in sh:or, sh:and etc

[HolgerKnublauch]

@HolgerKnublauch the informative part matches sh:or, but there is a mismatch in the textual definition. That should be aligned. Your definition for sh:someValue matches the informative parts, and probably also the reality of current implementations. I can confirm that shacl-engine skips any validations for sh:or after the first successful validation, unless debug mode is enabled.

Originally posted by @bergos in #524 (comment)

The same question may apply to sh:and and maybe even general validation order of constraints within the same shape: what happens if one of them "passes" - does the engine still need to evaluate all others and maybe run into a failure.

Care needs to be taken as this may be a breaking change compared to 1.0 yet it may also be a poorly defined corner case.

[HolgerKnublauch]

Note this situation is quite different from sh:someValue. The loop in sh:someValue may visit many value nodes (perhaps thousands) while sh:or etc are usually about just a handful of shapes (and loop over those). So the gains of skipping validation are not as significant.

[afs]

and maybe even general validation order of constraints

Shape validation has no side-effects ¹ so it does not matter if they execute or not.
(errors look like failures)

sh:or should be "conforms" if one of shapes conforms (error handling included)

Footnotes

1. or maybe they do -- #517 comment ↩