Andrés, You are not the first one to run into this. "g" is an iterable of > relationships, and not a single relationship. Maybe we should make this a > valid query... > > Anyway, with Cypher as it works today, you have to do this: > START a = node(2) > MATCH p = (a -[g*0..5]->b) > WHERE ALL(r in g WHERE not(type(r) ="INSTANCE") > RETURN p > > Thinking about this, it would be trivial to translate your query to this > query under the hood. WDYT?
Thanks for the quick response! My only concern about translating the query is that it could prevent moments like this- if the query had "just worked" I would have continued to misunderstand 'g' as a collection. Maybe a few function aliases, like `ALL_OF_TYPE(g, 'INSTANCE')`, `NONE_OF_TYPE(g, 'INSTANCE')`, etc would make sense? They serve the same purpose, are easily discoverable in docs, and make it really clear that multi-hop relation identifiers refer to collections. > Then I realized that even if it were, this statement would still traverse > > across the reference node (potentially hitting supernodes in the > > type hierarchy) because where clauses are evaluated after traversals. > > > > No, the pattern matcher evaluates predicates as soon as the necessary parts > have been found, with a few exceptions. Filtering after the fact would be > much more expensive. I think the very first version of Cypher did do this, > but it's been changed for quite some time now. Ah! I saw some mention of subgraph generation and filtering being separate. I thought that was pretty dumb, but with all the talk about Cypher's mediocre performance I was worried it was true. Thanks for educating me! -- Matt Luongo Co-Founder, Scholr.ly _______________________________________________ NOTICE: THIS MAILING LIST IS BEING SWITCHED TO GOOGLE GROUPS, please register and consider posting at https://groups.google.com/forum/#!forum/neo4j Neo4j mailing list User@lists.neo4j.org https://lists.neo4j.org/mailman/listinfo/user
