org.terrier.structures.indexing.singlepass

Class Inverted2DirectIndexBuilder

java.lang.Object

org.terrier.structures.indexing.singlepass.Inverted2DirectIndexBuilder

Direct Known Subclasses:

BlockInverted2DirectIndexBuilder

public class Inverted2DirectIndexBuilder
extends Object

Create a direct index from an InvertedIndex. The algorithm is similar to that followed by InvertedIndexBuilder. To summarise, InvertedIndexBuilder builds an InvertedIndex from a DirectIndex. This class does the opposite, building a DirectIndex from an InvertedIndex.

Algorithm:
For a selection of document ids (Scan the inverted index looking for postings with these document ids) For each term in the inverted index Select required postings from all the postings of that term Add these to posting objects that represents each document &nsbp;For each posting object Write out the postings for that document

Notes:
This algorithm assumes that termids start at 0 and are strictly increasing. This assumption holds true only for inverted indices generated by the single pass indexing method.

Properties:

1. inverted2direct.processtokens - total number of tokens to attempt each iteration. Defaults to 100000000. Memory usage would more likely be linked to the number of pointers, however as the document index does not contain the number of unique terms in each document, the pointers calculation is impossible to make.

Since:

2.0

Author:

Craig Macdonald

Field Summary

Fields

Modifier and Type	Field and Description
protected String	basicDirectIndexPostingIteratorClass
protected String	destinationStructure
protected String	directIndexClass Class to read the generated direct index
protected String	directIndexInputStreamClass Class to read the generated inverted index
protected int	fieldCount The number of different fields that are used for indexing field information.
protected String	fieldDirectIndexPostingIteratorClass
protected IndexOnDisk	index index currently being used
protected static org.apache.log4j.Logger	logger The logger used
protected long	processTokens number of tokens limit per iteration
protected boolean	saveTagInformation Indicates whether field information is used.
protected String	sourceStructure

Constructor Summary

Constructors

Constructor and Description
Inverted2DirectIndexBuilder(IndexOnDisk i) Construct a new instance of this builder class

Method Summary

Methods

Modifier and Type	Method and Description
void	createDirectIndex() create the direct index when the collection contains an existing inverted index
protected PostingInRun	getPostingReader() returns the SPIR implementation that should be used for reading the postings written earlier
protected Posting[]	getPostings(int count) get an array of posting object of the specified size.
static void	main(String[] args) main
protected int	scanDocumentIndexForTokens(long _processTokens, Iterator<DocumentIndexEntry> docidStream) Iterates through the document index, until it has reached the given number of terms
protected long	traverseInvertedFile(PostingIndexInputStream iiis, int firstDocid, int lastDocid, Posting[] directPostings) traverse the inverted file, looking for all occurrences of documents in the given range

Methods inherited from class java.lang.Object

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

Field Detail

logger

protected static final org.apache.log4j.Logger logger

The logger used

index

protected IndexOnDisk index

index currently being used

fieldCount

protected final int fieldCount

The number of different fields that are used for indexing field information.

saveTagInformation

protected final boolean saveTagInformation

Indicates whether field information is used.

directIndexClass

protected String directIndexClass

Class to read the generated direct index

directIndexInputStreamClass

protected String directIndexInputStreamClass

Class to read the generated inverted index

basicDirectIndexPostingIteratorClass

protected String basicDirectIndexPostingIteratorClass

fieldDirectIndexPostingIteratorClass

protected String fieldDirectIndexPostingIteratorClass

processTokens

protected long processTokens

number of tokens limit per iteration

sourceStructure

protected String sourceStructure

destinationStructure

protected String destinationStructure

Constructor Detail

Inverted2DirectIndexBuilder

public Inverted2DirectIndexBuilder(IndexOnDisk i)

Construct a new instance of this builder class

Method Detail

createDirectIndex

public void createDirectIndex()

create the direct index when the collection contains an existing inverted index

getPostings

protected Posting[] getPostings(int count)

get an array of posting object of the specified size. These will be used to hold the postings for a range of documents

getPostingReader

protected PostingInRun getPostingReader()

returns the SPIR implementation that should be used for reading the postings written earlier

traverseInvertedFile

protected long traverseInvertedFile(PostingIndexInputStream iiis,
                        int firstDocid,
                        int lastDocid,
                        Posting[] directPostings)
                             throws IOException

traverse the inverted file, looking for all occurrences of documents in the given range

Returns:

the number of tokens found in all of the document.

Throws:

IOException

scanDocumentIndexForTokens

protected int scanDocumentIndexForTokens(long _processTokens,
                             Iterator<DocumentIndexEntry> docidStream)
                                  throws IOException

Iterates through the document index, until it has reached the given number of terms

Parameters:

_processTokens - Number of tokens to stop reading the lexicon after

docidStream - the document index stream to read

Returns:

the number of documents to process

Throws:

IOException

main

public static void main(String[] args)
                 throws Exception

main

Parameters:

args -

Throws:

Exception