TitanDB定义和使用索引

Question

Using this . I am constructing the following index:

TitanManagement management = graph.openManagement();

PropertyKey key = management.makePropertyKey("ITEM_IDENTIFIER").dataType(String.class).make();

management.buildIndex("byItemIdentifier", Vertex.class).addKey(key).buildCompositeIndex();

management.commit();

However when I run the following query:

graph.traversal().V().has("ITEM_IDENTIFIER", "Test");

I still get the warning:

Query requires iterating over all vertices [(ITEM_IDENTIFIER = Test)]. For better performance, use indexes

How can I get TitanDB to use the index in order to speed up the lookups ?

Answer 1

So I managed to figure out the issue. Thanks to this answer providing more info.

The problem is that I assumed that I would use a CompositeIndex but after more careful reading of Chapter 20 , Chapter 22 , and Chapter 28 I was actually wrong. String based indexes require a MixedIndex .

To get MisedIndexes working with strings I had to first install Elasticsearch to enable String based indexes. Then I was able to configure my index by replacing:

management.buildIndex("byItemIdentifier", Vertex.class).addKey(key).buildCompositeIndex();

with

management.buildIndex("byItemIdentifier", Vertex.class).addKey(key, Mapping.STRING.asParameter()).buildMixedIndex("search");

after doing this the warning disappeared. I am going to do more testing to confirm this worked but for now this seems to have solved my issue.

Answer 2

Your index definition looks reasonable, but there are a couple reasons the index might not be used you should check.

The index has to finish building.

mgmt.awaitGraphIndexStatus(graph, "byItemIdentifier").call()

Or you need to reindex your existing data

mgmt.updateIndex(mgmt.getGraphIndex("byItemIdentifier"), SchemaAction.REINDEX).get()

Note that this latter operation should be completed inside a transaction.