Neo4j: full-text lucene legacy indexes (node_auto_index) does not work after migration

问题

After successful migration from Neo4j 2.2.8 to 3.0.4 using official faq, full text search does not work as expected. Fuzziness is not that fuzzy as it was before.

Example:

START n=node:node_auto_index('name:(+Target~0.85)') MATCH (n) RETURN n;

Should return nodes with field name that contain work like 85% similar to 'Target'.

Before it was matching the following:

Target
Target v2

After migration:

Target

Why and how to fix that?

回答1:

Reason was that after migration lucene node_auto_index wasn't configured properly. Probably migration tools does not respect its configuration or broken.

The solution was to setup indexes correctly and rebuild them.

Steps:

Check your /etc/neo4j/neo4j.conf that auto_index is enabled and keys are set to fields that you want to auto index:

dbms.auto_index.nodes.enabled=true                                                                                                                                                                                 
dbms.auto_index.nodes.keys=name

Check that node_auto_index configured correctly by running:

neo4j-shell -c 'index --get-config node_auto_index'
{
    "analyzer": "org.apache.lucene.analysis.standard.StandardAnalyzer",
    "provider": "lucene",
    "to_lower_case": "true",
    "type": "fulltext"
}

If it does not meet your requirements, for example type is not fulltext then you run following:

neo4j-shell -c 'index --set-config node_auto_index type fulltext'
neo4j-shell -c 'index --set-config node_auto_index to_lower_case true'
neo4j-shell -c 'index --set-config node_auto_index analyzer org.apache.lucene.analysis.standard.StandardAnalyzer'

After that you need to re-index your data. Based on dbms.auto_index.nodes.keys setting (name field in this example), run the following cypher on your data set:

MATCH (n) WHERE EXISTS(n.name) WITH (n) SKIP 0 LIMIT 50000 SET n.name=n.name;
MATCH (n) WHERE EXISTS(n.name) WITH (n) SKIP 50000 LIMIT 50000 SET n.name=n.name;
MATCH (n) WHERE EXISTS(n.name) WITH (n) SKIP 100000 LIMIT 50000 SET n.name=n.name;
// ...

The following steps will help you to setup full-text lucene indexes in Neo4j 3.0 and re-index your existing data.

来源：https://stackoverflow.com/questions/39268710/neo4j-full-text-lucene-legacy-indexes-node-auto-index-does-not-work-after-mig

标签

neo4j

lucene

migration

database-migration

fuzzy-search