cancel
Showing results for 
Search instead for 
Did you mean: 

GDS 2.1.12 Don't filter Nodes When Runs an Algorithm

ponceortiz
Node Clone

I've been using gds for native graph projection with directive ['*']  for nodeProjection parameter, the name of graph projected is  "compranet" the labels are "UC:Sospechoso","Contrato", "Provedor" 

1) the result of projection was  "nodes: {_ALL _

2) When I used de graph projected 'compranet' in gds.wcc.stream algorithm  using the same for labels ['*'] , then the comprobation is OK , returns all labels

3) Then using explicit node filter like ['UC','Proveedor']  in gds.wcc.stream algorithm ,the filter doesn't work , the results retrieve all nodes again.You can see at image with validation label "Contrato" doesn't have to be there ,  I had to review these because my rowcounts (validation) doesn't check 

4) At least I have to create graph projections explicit the labels name

The behavior expected for node filtering it's ok or I did understand something mistaken? or a bug in gds library ?

Thanks in advanced

 

3 REPLIES 3

Can you share the graph model and the exact code you're using in the projection and the algorithms in 

https://github.com/neo4j/graph-data-science/issues 

So that we can rule out it's a bug.

Also  you said the graph has UC:Sospechoso as label but you used UC ?? Which is not the same label?

Hi Michael , 

I received an email with "close" bug status , because an explicit graph projection can be created, instead using ['*'] It's worry about "the solution" of the bug.
Since I :

1.- I've noticed about the bug ,because my control numbers were don't  match.
2.- I shared with the people assigned to the bug solution, the reason why I can't create a graph projection for every need :memory allocation
3.-Then what about the documentation ?

At the moment I have to extend my testing to cover the GDS too 😐

Greetings 

ponceortiz
Node Clone

Hi, Michael

"UC:Sospechoso" is the same node, but "Contrato" node labels in the results  hasn't to be there, I've created the issue but with Fraud Risk Prevention Graph DataBase 

https://github.com/neo4j/graph-data-science/issues/222

Thanks