Is there an approach where an Accum can be attached to an edge?

GSQL
#1

Here is the gist of what I’m trying to do, for obvious reasons following query returns an error on Line 4. “Error: Accumulators cannot be attached to the edge variable ‘e’. Thus, reading and writing edge accumulators are not supported.”

I know we can’t attach an accumulator to the edge but is there any alternate approach I can try to get similar results?

CREATE QUERY Temp(SET<VERTEX<User>> users, INT a, INT b) FOR GRAPH social { 
      
    	1. SumAccum<INT> @testCount;
    	2. USER_SET = {users};
    	
    	3. X = SELECT t FROM USER_SET:s -(User_SENT_MESSAGE_TO_User:e)-> :t
        4.                         ACCUM e.@testCount += e.countsByMonths.get(a) + e.countsByMonths.get(b);

        5. Y = SELECT s FROM USER_SET:s -(User_SENT_MESSAGE_TO_User:e)-> :t
	    6.                         WHERE t IN users AND e.@testCount > 0 ACCUM ...;

    }
1 Like
#2

How about a vertex attached accumulator which includes the target vertex, and the edge count?

Something like this:

MapAccum<VERTEX, INT> @testCount;

X = SELECT t FROM USER_SET:s -(User_SENT_MESSAGE_TO_User:e)-> :t
ACCUM s.@testCount += (t -> e.countsByMonths.get(a) + e.countsByMonths.get(b));

2 Likes
#3

Thank you for the quick response!

That was my initial instinct but I need to iterate through those edges multiple times. Would this be a best approach in that case?

Suppose I wanted to implement Louvain https://github.com/tigergraph/gsql-graph-algorithms/blob/master/algorithms/examples/Community/louvain_parallel.gsql
where Person is User, Coworker is User_SENT_MESSAGE_TO_User and e.weight is e.@testCount here.

How does MapAccum<VERTEX, INT> affect the complexity and performance of Louvian Parallel query?

1 Like
#4

@dhruva
Hi, have you found an alternative?

I’m running into a similar problem now, trying to do a local accumulator attached to the edge, but I can’t find a better way to do it than defining an ArrayList and representing the graph with a matrix