I had written up my problem here too:
http://assimilationsystems.com/2016/07/03/assimilation-subgraph-queries/
Here's a thought which occurred to me while sleeping last night...
The two ways I've done it are:
1. shortest paths from initial node set
2. all paths
As noted in the original email thread below, the problems I ran into are:
1. Shortest paths produces all nodes, but misses some relationships I
care about.
2. All paths produces all nodes and relationships but takes a really
long time and often times out.
A way which occurred to me while sleeping tonight was to create a 2-step
query like this:
Compute the set of shortest paths to get the set of nodes (as in (1)
above). Ignore the relationships produced.
>From that set of nodes (n) compute (n)-[:a|b|c|d]-(n) and return 'n',
relationships.
If Cypher won't let me compute from 'n' to 'n', then compute the from
n-[:a|b|c|d]-m WHERE m.nodetype is in the same set of node types as was
in the original constraint. Then return n, relationships. [Although it's
obvious, I'll mention that the difference is that the second query only
follows a single level of relationship].
It still might need a little experimentation...
Any thoughts?
I'm going back to bed now. Will look into it some more in the morning... ;-)
On 07/21/2016 02:09 PM, Alan Robertson wrote:
> On 04/28/2016 03:32 PM, Alan Robertson wrote:
>> The query isn't quite right. It has an error in it (variable n
>> already used) and something about the order of operations causes it
>> to return more than one row - with duplicates ;-)
>>
>> But this one seems to do the trick:
>> |match path = shortestPath( (n)-[*]->(m) ) return collect(distinct
>> unwind(nodes(path))), collect(distinct unwind(rels(path)))|
> That query is quite fast -- but...
>
> I don't actually want /only/ the shortest paths, I want all the paths
> that involve those relationships.
>
> But if I delete the shortestPath() function call it runs so long I get
> an HTTP timeout on many of these queries. I restrict the types of
> relationships with [:rel1type|:rel2type|:etctype*]. And even though
> the result only includes one or two additional paths, it runs like 100
> times slower - and often just times out via REST (via Py2neo).
>
> That's really disappointing.
>
> Does anyone have a suggestion on how to make this faster - or at least
> not time out?
>
--
Alan Robertson / CTO
[email protected] <mailto:[email protected]>/ +1
303.947.7999
Assimilation Systems Limited
http://AssimilationSystems.com
Twitter <https://twitter.com/ossalanr> Linkedin
<https://www.linkedin.com/in/alanr> skype
<https://htmlsig.com/skype?username=alanr_unix.sh>
--
You received this message because you are subscribed to the Google Groups
"Neo4j" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
For more options, visit https://groups.google.com/d/optout.
