Titan: comparison between two query approach - titan

Question

What is the performance difference between
g.query().has("city","mumbai").vertices().iterator().next();
here each vertex will have a property city with city name mumbai
and
v.query().direction(Direction.IN).labels("belongTo").vertices();
here v is the vertex for mumbai city and all other vertex is connect to it through edge label belongTo.
I want to do query something like all vertex having city mumbai. Which approach will be better?
The problem is a user can enter anything as city name e.g mumbai or mummbai or mubai so its not possible to varify city name. So for mumbai i will create mumbai mummbai mubai vertex. its very inefficient.
How will you handle this kind of situation?

Answer 1

Titans ElasticSearch integration is great for those kind of fuzzy searches. Here's an example:
g = TitanFactory.open("conf/titan-cassandra-es.properties")
g.makeKey("city").dataType(String.class).indexed("search", Vertex.class).make()
g.makeKey("info").dataType(String.class).make()
g.makeLabel("belongsTo").make()
g.commit()
cities = ["washington", "mumbai", "phoenix", "uruguay", "pompeji"]
cities.each({ city ->
info = "belongs to ${city}"
g.addVertex(["info":info]).addEdge("belongsTo", g.addVertex(["city":city]))
}); g.commit()
info = { it.getElement().in("belongsTo").info.toList() }
userQueries = ["mumbai", "mummbai", "mubai", "phönix"]
userQueries.collectEntries({ userQuery ->
q = "v.city:${userQuery}~"
v = g.indexQuery("search", q).limit(1).vertices().collect(info).flatten()
[userQuery, v]
})
The last query will give you the following result:
==>mumbai=[belongs to mumbai]
==>mummbai=[belongs to mumbai]
==>mubai=[belongs to mumbai]
==>phönix=[belongs to phoenix]
Cheers,
Daniel