Re: Neo4j - How to speed up Cypher search query

[Max De Marzi Jr.]

START n = node:Destination('name:*') <-- bad

Take a look at the search example here=>

http://maxdemarzi.com/2012/11/14/crunchbase-on-neo4j/

  query = "name:*#{params[:term]}* OR name2:*#{params[:term]}*"

get '/search' do 
  content_type :json
  neo = Neography::Rest.new    

  cypher = "START me=node:node_index({query}) 
            RETURN ID(me), me.name
            ORDER BY me.name
            LIMIT 15"
  query = "permalink:*#{params[:term]}* OR name:*#{params[:term]}*"
  neo.execute_query(cypher, 
                    {:query => query })["data"].
                      map{|x| 
                           { label: x[1], 
                             value: x[0] }
                         }.to_json   
end

On Wednesday, December 5, 2012 11:31:48 AM UTC-6, Kobulniczky Csongor wrote:

I'm working on a site that uses Neo4j through a php library. The thing that bothers me is that my suggestions show up quite slowly ( 1-3 seconds ). They should appear almoast instantaneously. I am using Jquery ui Autocomplete for showing results, and the following cypher query:

$query = "START n = node:Destination('name:*') 
          WHERE ( n.name =~ '(?i).*".$input.".*' OR n.name2 =~ '(?i).*".$input.".*' ) AND has( n.published ) 
          RETURN n 
          LIMIT 6";

Does anyone have any optimization ideas, or can someone tell me what am i doing wrong?

[Michael Hunger]

In general you're doing a full index scan and filter afterwards, not a graph query but just a big index lookup.

Max' approach does two things different:

- it moves your filter into the index query itself

- it creates a parameter for the index-query so that cypher can utilize its query cache and doesn't have to re-parse the query time and again

--
 
 

[Kobulniczky Csongor]

It looks great but i never worked with Ruby. I am working with PHP. Also is this case insensitive search?

I did try index lookup with lucene query ( it seems to be much faster ), but i didn`t find any way to do incasesensitive search.

[Kobulniczky Csongor]

Yes, you are right.

I did try to move filter into index query, but i couldn`t do case insensitive search that way.

Tried to use fuzzy search but that didn`t give good results.

Is there any way to do incasesensitive search in index query? Something like 'name:*term*'

[Michael Hunger]

If you configure your index to be a fulltext-index it is case insensitive.

Michael

See: 

http://docs.neo4j.org/chunked/milestone/rest-api-indexes.html#rest-api-create-node-index-with-configuration

http://docs.neo4j.org/chunked/milestone/indexing-create-advanced.html

--