Calling apoc.path.expandConfig twice override first result

[o9384496]

Hello,

I've tried using the apoc.path.expandConfig to find a path.
I've called it twice with different sequence and if the second one didn't find any path, it "deleted" the result of the first apoc.

For example - The graph

CREATE (n:Person:TrailStart)-[:FRIEND_OF]->(:Person)-[:FRIEND_OF]->(:Person)-[:FRIEND_OF]->(:Person:TrailEnd)

The queries -
This one returned the path as expected

MATCH (end_node:TrailEnd)
WITH collect(end_node) as end_nodes

MATCH (start_node:TrailStart)

CALL apoc.path.expandConfig(start_node, {
    endNodes: end_nodes,
    sequence: '+person, FRIEND_OF>'
    }) YIELD path as path1

RETURN path1

This one return nothing

MATCH (end_node:TrailEnd)
WITH collect(end_node) as end_nodes

MATCH (start_node:TrailStart)

CALL apoc.path.expandConfig(start_node, {
    endNodes: end_nodes,
    sequence: '+person, FRIEND_OF>'
    }) YIELD path as path1

CALL apoc.path.expandConfig(start_node, {
    endNodes: end_nodes,
    sequence: '+person, NOT_FRIEND_OF>'
    }) YIELD path as path2

RETURN path1

If the second apoc procedure did find something, the result of the first procedure wasn't deleted.

Thank you,
Michael

[michael_hunger]

Not really sure what you expect.
The first procedure call creates rows and for each of those rows the second procedure is called.
And the result is the combined product of both.

if the first doesn't find anything you will get no rows returned

[o9384496]

Oh I see, its like two matches.
So how can I add the two results of the procedures? like if the first found a trail, and the second didnt find any trail, how can I continue the query with the trail found in the first procedure? (trail + null)

Thank you,
Michael

[andrew_bowman]

You can use the optional config parameter when making the call which will keep the row even if no results are found (you'll get a null for path).

Secondly, you would want to collect the results of the first call before calling the second, so the second would only be called once total rather than once per every row yielded by the first call.

Then you can collect and combine the collections, similar to one of the approaches in this kb article on post-UNION processing.