Package com.atlassian.bonnie.index

Lucene indexing framework for objects.

See:
Description

Interface Summary

BatchOpIndexer	Performs batch operations on an index.
BatchOpIndexer.DocumentWritingScheme	Writes documents.
ObjectQueue	Queue of objects that need to be indexed.
ObjectQueue.Processor	Creates Documents from objects via ObjectToDocumentConverter, then passes the created Document to a Callback.
ObjectToDocumentConverter	Converts objects to Documents.
QueueProcessingRunnable
SingleObjectIndexer	Indexes/unindexes single objects.

Class Summary

BaseDocumentWritingScheme
BaseMultiThreadedIndexer	Multiple-thread implementation of a BatchOpIndexer
NoOpProgressWrapper
OnlineMultiThreadedIndexer	Performs batch indexing in multiple threads.
QueueProcessingRunnableImpl	Pops objects off an ObjectQueue, processes them (via ObjectQueue.Processor, then writes them using BatchOpIndexer.DocumentWritingScheme.
SingletonObjectQueue	A queue for indexing a single object.
TempDirectoryDocumentWritingScheme	Each thread has its own IndexWriter which writes to a temp directory.
TempIndexWriter	Index writer that is able to write to one or more temporary indices.

Package com.atlassian.bonnie.index Description

Lucene indexing framework for objects.

The 2 main interfaces are BatchOpIndexer and SingleObjectIndexer.

To perform batch indexing, clients must provide the following:

• ObjectQueue: the objects to be reindexed
• ObjectToDocumentConverter: converts objects to Lucene documents
• DocumentWritingScheme: provides the actual logic of how to write documents to index
• ProgressMeter: to track the progress of the indexing process

Possible implementations of the indexing include:

• writing documents to temporary index/indices and merging to the final index location
• writing documents to separate final indices depending on the nature of the documents
• obtaining references/handles to objects at the queue stage and actually loading from persistence on demand to conserve memory usage
• running indexing with multiple threads so that no single object/file seriously bottlenecks the indexing operation