Optimize cascading validation for large lists

Some notes from a meeting we had with and about this specific issue:

Solution #1 (Guillaume): detect statically that a tree doesn’t involve the same class twice. All cycle/duplicate detection can be disabled then.

Partially implemented, but for 6.2 only, and method validation is not implemented yet.

 

Solution #2 (Yoann): offer a mode switch (“huge tree mode”) to only track the nodes between the root and current node (for cycle detection). Disable duplicate detection during the walk; if we encounter the same node twice, we will validate it twice. If validation fails, before adding the violation we will go through previous violations: if a violation already exists for the same object and constraint, we ignore this new violation.

 

Solution #3: offer a mode where we only do cycle detection and disable duplicate detection entirely, without any attempt at replacing it.

Maybe this is already implemented, there seems to be some configuration flag?