9.4. Fast memory-based indices

In section 2.10 we showed you how to use RAMDirectory to load an index entirely in RAM. This is especially convenient if you have a prebuilt index living on disk and you’d like to slurp the whole thing into RAM for faster searching. But because RAMDirectory still treats all data from the index as files, there’s significant overhead during searching for Lucene to decode this file structure for every query. This is where two interesting contrib modules come in: MemoryIndex and InstantiatedIndex.

MemoryIndex, contributed by Wolfgang Hoschek, is a fast RAM-only index designed to test whether a single document matches a query. It’s only able to index and search a single document. You instantiate the MemoryIndex, then use its addField method to add the document’s fields into it. Then, use its search methods to search with an arbitrary Lucene query. This method returns a float relevance score; 0.0 means there was no match.