org.apache.lucene.store.instantiated

InstantiatedIndex, alternative RAM store for small corpora.

@lucene.experimental

Abstract

Represented as a coupled graph of class instances, this all-in-memory index store implementation delivers search results up to a 100 times faster than the file-centric RAMDirectory at the cost of greater RAM consumption.

API

Just as the default store implementation, InstantiatedIndex comes with an IndexReader and IndexWriter. The latter share many method signatures with the file-centric IndexWriter.

It is also possible to load the content of another index by passing an IndexReader to the InstantiatedIndex constructor.

Performance

At a few thousand ~160 characters long documents InstantiatedIndex outperforms RAMDirectory some 50x, 15x at 100 documents of 2000 characters length, and is linear to RAMDirectory at 10,000 documents of 2000 characters length.

Mileage may vary depending on term saturation.

Populated with a single document InstantiatedIndex is almost, but not quite, as fast as MemoryIndex.

It takes more or less the same time to populate an InstantiatedIndex as it takes to populate a RAMDirectory. Hardly any effort has been put in to optimizing the InstantiatedIndexWriter, only minimizing the amount of time needed to write-lock the index has been considered.

Caveats

• No locks! Consider using InstantiatedIndex as if it was immutable.
• No documents with fields containing readers.
• No field selection when retrieving documents, as all stored field are available in memory.
• Any document returned must cloned if they are to be touched.
• Norms array returned must not be touched.

Use cases

Could replace any small index that could do with greater response time. spell check a priori index, the index of new documents exposed to user search agent queries, to compile classifiers in machine learning environments, etc.

Class diagram

Diagram rendered using UMLet 7.1.