org.terrier.structures.indexing.classical

Class BlockIndexer

java.lang.Object

org.terrier.structures.indexing.Indexer

org.terrier.structures.indexing.classical.BlockIndexer

public class BlockIndexer
extends Indexer

An indexer that saves block information for the indexed terms. Block information is usually recorded in terms of relative term positions (position 1, positions 2, etc), however, since 2.2, Terrier supports the presence of "marker terms" during indexing which are used to increment the block counter.

Properties:

• blocks.size - How many terms should be in one block. If you want to use phrasal search, this need to be 1 (default).
• blocks.max - Maximum number of blocks in a document. After this number of blocks, all subsequent terms will be in the same block. Default 100,000
• block.indexing - This class should only be used if the block.indexing property is set.
• indexing.max.encoded.documentindex.docs - how many docs before the DocumentIndexEncoded is dropped in favour of the DocumentIndex (on disk implementation).
• See Also: Properties in org.terrier.indexing.Indexer and org.terrier.indexing.BasicIndexer

Markered Blocks
Markers are terms (artificially inserted or otherwise into the term stream that are used to denote when the block counter should be incremented. This functionality is enabled using the block.delimiters.enabled property, while the terms are specified using a comma delimited fashion with the block.delimiters property. The following lists the properties:

• block.delimiters.enabled - enabled markered blocks. Defaults to false, set to true to enable.
• block.delimiters - comma delimited list of terms that are markers. Defaults to empty. Terms are lowercased is lowercase is set to true (default).
• block.delimiters.index.terms - set to true if markers terms should actually be indexed. Defaults to false.
• block.delimiters.index.doclength - set to true if markers terms should contribute to document length. Defaults to false, only has effect if block.delimiters.index.terms is set.

Author:

Craig Macdonald, Vassilis Plachouras, Rodrygo Santos

Nested Class Summary

Nested Classes

Modifier and Type	Class and Description
protected class	BlockIndexer.BasicTermProcessor This class implements an end of a TermPipeline that adds the term to the DocumentTree.
protected class	BlockIndexer.DelimFieldTermProcessor This class behaves in a similar fashion to FieldTermProcessor except that this one treats blocks bounded by delimiters instead of fixed-sized blocks.
protected class	BlockIndexer.DelimTermProcessor This class behaves in a similar fashion to BasicTermProcessor except that this one treats blocks bounded by delimiters instead of fixed-sized blocks.
protected class	BlockIndexer.FieldTermProcessor This class implements an end of a TermPipeline that adds the term to the DocumentTree.

Field Summary

Fields

Modifier and Type	Field and Description
protected int	BLOCK_SIZE The maximum number of terms allowed in a block.
protected int	blockId The block number of the current document.
protected CompressionFactory.CompressionConfiguration	compressionDirectConfig The compression configuration for the direct index
protected CompressionFactory.CompressionConfiguration	compressionInvertedConfig The compression configuration for the inverted index
protected int	MAX_BLOCKS The maximum number allowed number of blocks in a document.
protected int	numOfTokensInBlock The number of tokens in the current block of the current document.
protected int	numOfTokensInDocument The number of tokens in the current document so far.
protected Set<String>	termFields The fields that are set for the current term.
protected DocumentPostingList	termsInDocument The list of terms in this document, and for each, the block occurrences.

Fields inherited from class org.terrier.structures.indexing.Indexer

BUILDER_BOUNDARY_DOCUMENTS, currentIndex, directIndexBuilder, docIndexBuilder, emptyDocIndexEntry, fieldNames, fileNameNoExtension, IndexEmptyDocuments, invertedIndexBuilder, lexiconBuilder, logger, MAX_DOCS_PER_BUILDER, MAX_TOKENS_IN_DOCUMENT, metaBuilder, numFields, path, pipeline_first, prefix, useFieldInformation

Constructor Summary

Constructors

Constructor and Description
BlockIndexer(String pathname, String prefix) Constructs an instance of this class, where the created data structures are stored in the given path, with the given prefix on the filenames.

Method Summary

Methods

Modifier and Type	Method and Description
void	createDirectIndex(Collection[] collections) For the given collection, it iterates through the documents and creates the direct index, document index and lexicon, using information about blocks and possibly fields.
protected void	createDocumentPostings()
void	createInvertedIndex() Creates the inverted index from the already created direct index, document index and lexicon.
protected void	finishedInvertedIndexBuild() Hook method, called when the inverted index is finished - ie the lexicon is finished
protected TermPipeline	getEndOfPipeline() Returns the object that is to be the end of the TermPipeline.
protected void	indexDocument(Map<String,String> docProperties, DocumentPostingList _termsInDocument) This adds a document to the direct and document indexes, as well as it's terms to the lexicon.
protected void	load_indexer_properties()

Methods inherited from class org.terrier.structures.indexing.Indexer

createMetaIndexBuilder, finishedDirectIndexBuild, index, indexEmpty, init, load_builder_boundary_documents, load_field_ids, load_pipeline, main, merge, merge, mergeTwoIndices, parseInts, useFieldInformation

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Field Detail

numOfTokensInDocument

protected int numOfTokensInDocument

The number of tokens in the current document so far.

numOfTokensInBlock

protected int numOfTokensInBlock

The number of tokens in the current block of the current document.

blockId

protected int blockId

The block number of the current document.

termFields

protected Set<String> termFields

The fields that are set for the current term.

termsInDocument

protected DocumentPostingList termsInDocument

The list of terms in this document, and for each, the block occurrences.

BLOCK_SIZE

protected int BLOCK_SIZE

The maximum number of terms allowed in a block. See Property blocks.size

MAX_BLOCKS

protected int MAX_BLOCKS

The maximum number allowed number of blocks in a document. After this value, all the remaining terms are in the final block. See Property blocks.max.

compressionDirectConfig

protected CompressionFactory.CompressionConfiguration compressionDirectConfig

The compression configuration for the direct index

compressionInvertedConfig

protected CompressionFactory.CompressionConfiguration compressionInvertedConfig

The compression configuration for the inverted index

Constructor Detail

BlockIndexer

public BlockIndexer(String pathname,
            String prefix)

Constructs an instance of this class, where the created data structures are stored in the given path, with the given prefix on the filenames.

Parameters:

pathname - String the path in which the created data structures will be saved. This is assumed to be absolute.

prefix - String the prefix on the filenames of the created data structures, usually "data"

Method Detail

getEndOfPipeline

protected TermPipeline getEndOfPipeline()

Returns the object that is to be the end of the TermPipeline. This method is used at construction time of the parent object.

Specified by:

getEndOfPipeline in class Indexer

Returns:

TermPipeline the last component of the term pipeline.

createDirectIndex

public void createDirectIndex(Collection[] collections)

For the given collection, it iterates through the documents and creates the direct index, document index and lexicon, using information about blocks and possibly fields.

Specified by:

createDirectIndex in class Indexer

Parameters:

collections - Collection[] the collection to index.

See Also:

Indexer.createDirectIndex(org.terrier.indexing.Collection[])

indexDocument

protected void indexDocument(Map<String,String> docProperties,
                 DocumentPostingList _termsInDocument)
                      throws Exception

This adds a document to the direct and document indexes, as well as it's terms to the lexicon. Handled internally by the methods indexFieldDocument and indexNoFieldDocument.

Parameters:

docProperties - Map properties of the document

_termsInDocument - DocumentPostingList the terms in the document.

Throws:

Exception

createInvertedIndex

public void createInvertedIndex()

Creates the inverted index from the already created direct index, document index and lexicon. It saves block information and possibly field information as well.

Specified by:

createInvertedIndex in class Indexer

See Also:

Indexer.createInvertedIndex()

finishedInvertedIndexBuild

protected void finishedInvertedIndexBuild()

Hook method, called when the inverted index is finished - ie the lexicon is finished

Overrides:

finishedInvertedIndexBuild in class Indexer

createDocumentPostings

protected void createDocumentPostings()

load_indexer_properties

protected void load_indexer_properties()

Overrides:

load_indexer_properties in class Indexer