Multiple types for relationships about pypher HOT 3 CLOSED

I will definitely limit Node to and (:) and Relationship to or (|) when instantiated with lists(set, Tuple) of strings, but I do think that allowing the user to write incorrect (for now maybe?) logical operators for Labels is fine. I actually wouldn’t be able to control it without either deep object inspection during init or creating NodeLabel etc objects.

Pypher allows users to write incorrect Cypher all over the place and to check here would be a little inconsistent. There could definitely be a layer on top of it that could guarantee “correctness” that would be a hefty project though

from pypher.

Great catch.l, I need more scrutinization like this as I’m fairly new to Cypher/Neo4j. Thanks.

I think that a potential solution is two fold. First the Entity would be updated to use actual Label objects for label representations. Second would be to update that Label class to allow for ‘and’ or ‘or’ chaining (with Relationship objects only allow for or)

With these two updates one could potentially write something like:

p.relationship(labels=Label(‘Person’) | Label(‘Animal’))

# or
p.relationship(labels=(‘Person’, ‘Animal’))

Both would work the same under the hood. I’ll have to test that interface later

(Excuse the typos I’m on my phone)

from pypher.

Yes I thought the same myself.

But I think Pypher would beneficiate from it only if Neo4j adds the possibility to match nodes with OR'ed labels, and/or allow rels to have more than one type.

Currently we can only do AND for node labels, and OR for rel types. So implementing __and__ / __or__ in a Label class might be a bit overpowered, since under the hood only AND will be used for nodes and only OR will be used for rels. Besides, users might get it wrong and think they can do complex logic like Label('Pokémon') & (Label('Fire') | Label('Thunder'))

from pypher.