Access nodes in path

[neo4j]

Hi,
I’m working on a small buses lines dataset, I’m currently hitting a problem.
How can I check that, on a line with X buses, the previous bus isn’t after the next one using this kind of MATCH:

MATCH p=(a)-[*1..4]->(b)
 WHERE a.arrival_date + duration({ minutes: 5}) <= b.departure_date
 AND b.departure_date <= a.arrival_date + duration({ hours: 1})
RETURN nodes(p);

Because on this example a & b filtering are more like this if I’m right:
(a)-->()-->()-->(b)
So I can’t control this part:
()-->()
I’m sure it’s not that complicated but I can’t find the way, I think UNWIND can be what I’m looking for, but how?

[andrew_bowman · Accepted solution]

You can use nodes(path) to get the list of nodes that makes up the path, where the first node in the list is a and the last node in the list is b.

It may be a bit complicated, but if you want to work with adjacent pairs in the path, you can leverage APOC Procedures to get this, but you'll need to do a bit of aliasing and use UNWIND as well. Something like this:

MATCH p=(a)-[*1..4]->(b)
 WHERE a.arrival_date + duration({ minutes: 5}) <= b.departure_date
 AND b.departure_date <= a.arrival_date + duration({ hours: 1})
UNWIND apoc.coll.pairsMin(nodes(p)) as pair
WITH p, pair[0] as first, pair[1] as next
WHERE // add any predicates you want here for each adjacent pair of nodes in the path
RETURN nodes(p);

[neo4j]

Pretty interesting, thank you @andrew.bowman !
I had an issue as I need to check three nodes in row instead of two so I tried to use apoc.coll.partition instead of apoc.coll.pairsMin.
But apoc.coll.partition crashed saying :

Neo.ClientError.Statement.SyntaxError: Unknown function apoc.coll.partition

Which is strange as apoc.coll.pairsMin is correctly loaded.
Any idea @andrew.bowman ?

[neo4j]

It's not registred.

I'm using apoc version "3.5.0.1" but it seems to be released 2 years ago now.

[neo4j]

Even after an update to Neo4J 3.5.4 and APOC to 3.5.0.3 it seems that it's still not accessible.
But what is strange is that it seems it exist.

call apoc.help('coll');
"procedure" "apoc.coll.partition" "apoc.coll.partition(list,batchSize)" "apoc.coll.partition(values :: LIST? OF ANY?, batchSize :: INTEGER?) :: (value :: LIST? OF ANY?)" null null

My neo4j.conf says : dbms.security.procedures.unrestricted=apoc.*

[andrew_bowman]

This is because apoc.coll.partition() is a procedure, not a function. As such, you can't use it as an expression, but must use the CALL apoc.coll.partition() YIELD value syntax.

[neo4j]

Hi,
Thanks @andrew.bowman if I remember correctly that's not clearly written on the doc.
Thanks anyway, now it works
Have a pleasant day.

[andrew_bowman]

While it is noted that it's a procedure from apoc.help(), it could use a callout in the docs, as nearly all the other apoc.coll.* items here are functions. I'll see about addding that in.

[neo4j]

Hi @andrew.bowman,
It seems that if we have more than two nodes it does not works correctly.
For example, with two nodes, both will be treated by pairsMin but when I have another one (so 3 nodes) node 2 and 3 aren't unpacked together by pairsMin.
Do you have another solution ?
I'm a bit out of solutions right now

[neo4j]

BTW I just saw that it skip alone nodes.
If (a) is enough then pairsMin will skip it.

[andrew_bowman]

It seems that if we have more than two nodes it does not works correctly.
For example, with two nodes, both will be treated by pairsMin but when I have another one (so 3 nodes) node 2 and 3 aren't unpacked together by pairsMin.

I'm not sure what you mean by this.

As an example, with 3 elements:

return apoc.coll.pairsMin([1,2,3])

You'll correctly get two elements in the list, the pair of [1,2] and the pair of [2,3].

[neo4j]

Hi @andrew.bowman,

It seems that you are right, my bad.
The pairs are [1,2] then [2, 3].
But I still have the problem when I don't have a pair. Like [1, NULL] I will try to coalesce it.