lucene/xdocs/fileformats.xml

1350 lines
56 KiB
XML

<?xml version="1.0"?>
<document>
<properties>
<title>Index File Formats</title>
<authors>
<person email="cutting@apache.org" name="Doug Cutting"/>
</authors>
</properties>
<body>
<section name="Index File Formats">
<p>
This document defines the index file formats used
in Lucene version 1.4 and above.
</p>
<p>
Apache Lucene is written in Java, but several
efforts are underway to write
<a href="http://wiki.apache.org/jakarta-lucene/LuceneImplementations">versions
of Lucene in other programming
languages</a>. If these versions are to remain compatible with Apache
Lucene, then a language-independent definition of the Lucene index
format is required. This document thus attempts to provide a
complete and independent definition of the Apache Lucene 1.4 file
formats.
</p>
<p>
As Lucene evolves, this document should evolve.
Versions of Lucene in different programming languages should endeavor
to agree on file formats, and generate new versions of this document.
</p>
<p>
Compatibility notes are provided in this document,
describing how file formats have changed from prior versions.
</p>
</section>
<section name="Definitions">
<p>
The fundamental concepts in Lucene are index,
document, field and term.
</p>
<p>
An index contains a sequence of documents.
</p>
<ul>
<li>
<p>
A document is a sequence of fields.
</p>
</li>
<li>
<p>
A field is a named sequence of terms.
</p>
</li>
<li>
A term is a string.
</li>
</ul>
<p>
The same string in two different fields is
considered a different term. Thus terms are represented as a pair of
strings, the first naming the field, and the second naming text
within the field.
</p>
<subsection name="Inverted Indexing">
<p>
The index stores statistics about terms in order
to make term-based search more efficient. Lucene's
index falls into the family of indexes known as an <i>inverted
index.</i> This is because it can list, for a term, the documents that contain
it. This is the inverse of the natural relationship, in which
documents list terms.
</p>
</subsection>
<subsection name="Types of Fields">
<p>
In Lucene, fields may be <i>stored</i>, in which
case their text is stored in the index literally, in a non-inverted
manner. Fields that are inverted are called <i>indexed</i>. A field
may be both stored and indexed.</p>
<p>The text of a field may be <i>tokenized</i> into terms to be
indexed, or the text of a field may be used literally as a term to be indexed.
Most fields are
tokenized, but sometimes it is useful for certain identifier fields
to be indexed literally.
</p>
</subsection>
<subsection name="Segments">
<p>
Lucene indexes may be composed of multiple sub-indexes, or<i>
segments</i>. Each segment is a fully independent index, which could be searched
separately. Indexes evolve by:
</p>
<ol>
<li><p>Creating new segments for newly added documents.</p>
</li>
<li><p>Merging existing segments.</p>
</li>
</ol>
<p>
Searches may involve multiple segments and/or multiple indexes, each
index potentially composed of a set of segments.
</p>
</subsection>
<subsection name="Document Numbers">
<p>
Internally, Lucene refers to documents by an integer <i>document
number</i>. The first document added to an index is numbered zero, and each
subsequent document added gets a number one greater than the previous.
</p>
<p>
<br/>
</p>
<p>
Note that a document's number may change, so caution should be taken
when storing these numbers outside of Lucene. In particular, numbers may
change in the following situations:
</p>
<ul>
<li>
<p>
The
numbers stored in each segment are unique only within the segment,
and must be converted before they can be used in a larger context.
The standard technique is to allocate each segment a range of
values, based on the range of numbers used in that segment. To
convert a document number from a segment to an external value, the
segment's <i>base</i> document
number is added. To convert an external value back to a
segment-specific value, the segment is identified by the range that
the external value is in, and the segment's base value is
subtracted. For example two five document segments might be
combined, so that the first segment has a base value of zero, and
the second of five. Document three from the second segment would
have an external value of eight.
</p>
</li>
<li>
<p>
When documents are deleted, gaps are created
in the numbering. These are eventually removed as the index evolves
through merging. Deleted documents are dropped when segments are
merged. A freshly-merged segment thus has no gaps in its numbering.
</p>
</li>
</ul>
</subsection>
</section>
<section name="Overview">
<p>
Each segment index maintains the following:
</p>
<ul>
<li><p>Field names. This
contains the set of field names used in the index.
</p>
</li>
<li><p>Stored Field
values. This contains, for each document, a list of attribute-value
pairs, where the attributes are field names. These are used to
store auxiliary information about the document, such as its title,
url, or an identifier to access a
database. The set of stored fields are what is returned for each hit
when searching. This is keyed by document number.
</p>
</li>
<li><p>Term dictionary.
A dictionary containing all of the terms used in all of the indexed
fields of all of the documents. The dictionary also contains the
number of documents which contain the term, and pointers to the
term's frequency and proximity data.
</p>
</li>
<li><p>Term Frequency
data. For each term in the dictionary, the numbers of all the
documents that contain that term, and the frequency of the term in
that document.
</p>
</li>
<li><p>Term Proximity
data. For each term in the dictionary, the positions that the term
occurs in each document.
</p>
</li>
<li><p>Normalization
factors. For each field in each document, a value is stored that is
multiplied into the score for hits on that field.
</p>
</li>
<li><p>Term Vectors. For each field in each document, the term vector
(sometimes called document vector) is stored. A term vector consists
of term text and term frequency.
</p>
</li>
<li><p>Deleted documents.
An optional file indicating which documents are deleted.
</p>
</li>
</ul>
<p>Details on each of these are provided in subsequent sections.
</p>
</section>
<section name="File Naming">
<p>
All files belonging to a segment have the same name with varying
extensions. The extensions correspond to the different file formats
described below.
</p>
<p>
Typically, all segments
in an index are stored in a single directory, although this is not
required.
</p>
</section>
<section name="Primitive Types">
<subsection name="Byte">
<p>
The most primitive type
is an eight-bit byte. Files are accessed as sequences of bytes. All
other data types are defined as sequences
of bytes, so file formats are byte-order independent.
</p>
</subsection>
<subsection name="UInt32">
<p>
32-bit unsigned integers are written as four
bytes, high-order bytes first.
</p>
<p>
UInt32 --&gt; &lt;Byte&gt;<sup>4</sup>
</p>
</subsection>
<subsection name="Uint64">
<p>
64-bit unsigned integers are written as eight
bytes, high-order bytes first.
</p>
<p>UInt64 --&gt; &lt;Byte&gt;<sup>8</sup>
</p>
</subsection>
<subsection name="VInt">
<p>
A variable-length format for positive integers is
defined where the high-order bit of each byte indicates whether more
bytes remain to be read. The low-order seven bits are appended as
increasingly more significant bits in the resulting integer value.
Thus values from zero to 127 may be stored in a single byte, values
from 128 to 16,383 may be stored in two bytes, and so on.
</p>
<p><b>VInt Encoding Example</b></p>
<table width="100%" border="0" cellpadding="4" cellspacing="0">
<col width="64*" />
<col width="64*" />
<col width="64*" />
<col width="64*" />
<tr valign="TOP">
<td width="25%">
<p align="RIGHT"><b>Value</b>
</p>
</td>
<td width="25%">
<p align="RIGHT"><b>First byte</b>
</p>
</td>
<td width="25%">
<p align="RIGHT"><b>Second byte</b>
</p>
</td>
<td width="25%">
<p align="RIGHT"><b>Third byte</b>
</p>
</td>
</tr>
<tr valign="BOTTOM">
<td width="25%" sdval="0" sdnum="1033;0;#,##0">
<p align="RIGHT">0
</p>
</td>
<td width="25%" sdval="0" sdnum="1033;0;00000000">
<p class="western" align="RIGHT" style="margin-left: 0.11cm;
margin-right: 0.01cm">
00000000
</p>
</td>
<td width="25%" sdnum="1033;0;00000000">
<p align="RIGHT" style="margin-left: -0.07cm; margin-right:
0.01cm"><br/>
</p>
</td>
<td width="25%" sdnum="1033;0;00000000">
<p align="RIGHT" style="margin-left: -0.47cm; margin-right:
0.01cm"><br/>
</p>
</td>
</tr>
<tr valign="BOTTOM">
<td width="25%" sdval="1" sdnum="1033;0;#,##0">
<p align="RIGHT">1
</p>
</td>
<td width="25%" sdval="1" sdnum="1033;0;00000000">
<p class="western" align="RIGHT" style="margin-left: 0.11cm;
margin-right: 0.01cm">
00000001
</p>
</td>
<td width="25%" sdnum="1033;0;00000000">
<p align="RIGHT" style="margin-left: -0.07cm; margin-right:
0.01cm"><br/>
</p>
</td>
<td width="25%" sdnum="1033;0;00000000">
<p align="RIGHT" style="margin-left: -0.47cm; margin-right:
0.01cm"><br/>
</p>
</td>
</tr>
<tr valign="BOTTOM">
<td width="25%" sdval="2" sdnum="1033;0;#,##0">
<p align="RIGHT">2
</p>
</td>
<td width="25%" sdval="10" sdnum="1033;0;00000000">
<p class="western" align="RIGHT" style="margin-left: 0.11cm;
margin-right: 0.01cm">
00000010
</p>
</td>
<td width="25%" sdnum="1033;0;00000000">
<p align="RIGHT" style="margin-left: -0.07cm; margin-right:
0.01cm"><br/>
</p>
</td>
<td width="25%" sdnum="1033;0;00000000">
<p align="RIGHT" style="margin-left: -0.47cm; margin-right:
0.01cm"><br/>
</p>
</td>
</tr>
<tr>
<td width="25%" valign="TOP">
<p align="RIGHT">...
</p>
</td>
<td width="25%" valign="BOTTOM" sdnum="1033;0;00000000">
<p align="RIGHT" style="margin-left: 0.11cm; margin-right:
0.01cm"><br/>
</p>
</td>
<td width="25%" valign="BOTTOM" sdnum="1033;0;00000000">
<p align="RIGHT" style="margin-left: -0.07cm; margin-right:
0.01cm"><br/>
</p>
</td>
<td width="25%" valign="BOTTOM" sdnum="1033;0;00000000">
<p align="RIGHT" style="margin-left: -0.47cm; margin-right:
0.01cm"><br/>
</p>
</td>
</tr>
<tr valign="BOTTOM">
<td width="25%" sdval="127" sdnum="1033;0;#,##0">
<p align="RIGHT">127
</p>
</td>
<td width="25%" sdval="1111111" sdnum="1033;0;00000000">
<p class="western" align="RIGHT" style="margin-left: 0.11cm;
margin-right: 0.01cm">
01111111
</p>
</td>
<td width="25%" sdnum="1033;0;00000000">
<p align="RIGHT" style="margin-left: -0.07cm; margin-right:
0.01cm"><br/>
</p>
</td>
<td width="25%" sdnum="1033;0;00000000">
<p align="RIGHT" style="margin-left: -0.47cm; margin-right:
0.01cm"><br/>
</p>
</td>
</tr>
<tr valign="BOTTOM">
<td width="25%" sdval="128" sdnum="1033;0;#,##0">
<p align="RIGHT">128
</p>
</td>
<td width="25%" sdval="10000000" sdnum="1033;0;00000000">
<p class="western" align="RIGHT" style="margin-left: 0.11cm;
margin-right: 0.01cm">
10000000
</p>
</td>
<td width="25%" sdval="1" sdnum="1033;0;00000000">
<p class="western" align="RIGHT" style="margin-left: -0.07cm;
margin-right: 0.01cm">
00000001
</p>
</td>
<td width="25%" sdnum="1033;0;00000000">
<p align="RIGHT" style="margin-left: -0.47cm; margin-right:
0.01cm"><br/>
</p>
</td>
</tr>
<tr valign="BOTTOM">
<td width="25%" sdval="129" sdnum="1033;0;#,##0">
<p align="RIGHT">129
</p>
</td>
<td width="25%" sdval="10000001" sdnum="1033;0;00000000">
<p class="western" align="RIGHT" style="margin-left: 0.11cm;
margin-right: 0.01cm">
10000001
</p>
</td>
<td width="25%" sdval="1" sdnum="1033;0;00000000">
<p class="western" align="RIGHT" style="margin-left: -0.07cm;
margin-right: 0.01cm">
00000001
</p>
</td>
<td width="25%" sdnum="1033;0;00000000">
<p align="RIGHT" style="margin-left: -0.47cm; margin-right:
0.01cm"><br/>
</p>
</td>
</tr>
<tr valign="BOTTOM">
<td width="25%" sdval="130" sdnum="1033;0;#,##0">
<p align="RIGHT">130
</p>
</td>
<td width="25%" sdval="10000010" sdnum="1033;0;00000000">
<p class="western" align="RIGHT" style="margin-left: 0.11cm;
margin-right: 0.01cm">
10000010
</p>
</td>
<td width="25%" sdval="1" sdnum="1033;0;00000000">
<p class="western" align="RIGHT" style="margin-left: -0.07cm;
margin-right: 0.01cm">
00000001
</p>
</td>
<td width="25%" sdnum="1033;0;00000000">
<p align="RIGHT" style="margin-left: -0.47cm; margin-right:
0.01cm"><br/>
</p>
</td>
</tr>
<tr>
<td width="25%" valign="TOP">
<p align="RIGHT">...
</p>
</td>
<td width="25%" valign="BOTTOM" sdnum="1033;0;00000000">
<p align="RIGHT" style="margin-left: 0.11cm; margin-right:
0.01cm"><br/>
</p>
</td>
<td width="25%" valign="BOTTOM" sdnum="1033;0;00000000">
<p align="RIGHT" style="margin-left: -0.07cm; margin-right:
0.01cm"><br/>
</p>
</td>
<td width="25%" valign="BOTTOM" sdnum="1033;0;00000000">
<p align="RIGHT" style="margin-left: -0.47cm; margin-right:
0.01cm"><br/>
</p>
</td>
</tr>
<tr valign="BOTTOM">
<td width="25%" sdval="16383" sdnum="1033;0;#,##0">
<p align="RIGHT">16,383
</p>
</td>
<td width="25%" sdval="11111111" sdnum="1033;0;00000000">
<p class="western" align="RIGHT" style="margin-left: 0.11cm;
margin-right: 0.01cm">
11111111
</p>
</td>
<td width="25%" sdval="1111111" sdnum="1033;0;00000000">
<p class="western" align="RIGHT" style="margin-left: -0.07cm;
margin-right: 0.01cm">
01111111
</p>
</td>
<td width="25%" sdnum="1033;0;00000000">
<p align="RIGHT" style="margin-left: -0.47cm; margin-right:
0.01cm"><br/>
</p>
</td>
</tr>
<tr valign="BOTTOM">
<td width="25%" sdval="16384" sdnum="1033;0;#,##0">
<p align="RIGHT">16,384
</p>
</td>
<td width="25%" sdval="10000000" sdnum="1033;0;00000000">
<p class="western" align="RIGHT" style="margin-left: 0.11cm;
margin-right: 0.01cm">
10000000
</p>
</td>
<td width="25%" sdval="10000000" sdnum="1033;0;00000000">
<p class="western" align="RIGHT" style="margin-left: -0.07cm;
margin-right: 0.01cm">
10000000
</p>
</td>
<td width="25%" sdval="1" sdnum="1033;0;00000000">
<p class="western" align="RIGHT" style="margin-left: -0.47cm;
margin-right: 0.01cm">
00000001
</p>
</td>
</tr>
<tr valign="BOTTOM">
<td width="25%" sdval="16385" sdnum="1033;0;#,##0">
<p align="RIGHT">16,385
</p>
</td>
<td width="25%" sdval="10000001" sdnum="1033;0;00000000">
<p class="western" align="RIGHT" style="margin-left: 0.11cm;
margin-right: 0.01cm">
10000001
</p>
</td>
<td width="25%" sdval="10000000" sdnum="1033;0;00000000">
<p class="western" align="RIGHT" style="margin-left: -0.07cm;
margin-right: 0.01cm">
10000000
</p>
</td>
<td width="25%" sdval="1" sdnum="1033;0;00000000">
<p class="western" align="RIGHT" style="margin-left: -0.47cm;
margin-right: 0.01cm">
00000001
</p>
</td>
</tr>
<tr>
<td width="25%" valign="TOP">
<p align="RIGHT">...
</p>
</td>
<td width="25%" valign="BOTTOM" sdnum="1033;0;00000000">
<p class="western" align="RIGHT" style="margin-left: 0.11cm;
margin-right: 0.01cm">
<br/>
</p>
</td>
<td width="25%" valign="BOTTOM" sdnum="1033;0;00000000">
<p class="western" align="RIGHT" style="margin-left: -0.07cm;
margin-right: 0.01cm">
<br/>
</p>
</td>
<td width="25%" valign="BOTTOM" sdnum="1033;0;00000000">
<p class="western" align="RIGHT" style="margin-left: -0.47cm;
margin-right: 0.01cm">
<br/>
</p>
</td>
</tr>
</table>
<p>
This provides compression while still being
efficient to decode.
</p>
</subsection>
<subsection name="Chars">
<p>
Lucene writes unicode
character sequences using Java's
<a href="http://en.wikipedia.org/wiki/UTF-8#Modified_UTF-8">"modified
UTF-8 encoding"</a>.
</p>
</subsection>
<subsection name="String">
<p>
Lucene writes strings as a VInt representing the length, followed by
the character data.
</p>
<p>
String --&gt; VInt, Chars
</p>
</subsection>
</section>
<section name="Per-Index Files">
<p>
The files in this section exist one-per-index.
</p>
<subsection name="Segments File">
<p>
The active segments in the index are stored in the
segment info file. An index only has
a single file in this format, and it is named "segments".
This lists each segment by name, and also contains the size of each
segment.
</p>
<p>
Segments --&gt; Format, Version, NameCounter, SegCount, &lt;SegName, SegSize&gt;<sup>SegCount</sup>
</p>
<p>
Format, NameCounter, SegCount, SegSize --&gt; UInt32
</p>
<p>
Version --&gt; UInt64
</p>
<p>
SegName --&gt; String
</p>
<p>
Format is -1 in Lucene 1.4.
</p>
<p>
Version counts how often the index has been
changed by adding or deleting documents.
</p>
<p>
NameCounter is used to generate names for new segment files.
</p>
<p>
SegName is the name of the segment, and is used as the file name prefix
for all of the files that compose the segment's index.
</p>
<p>
SegSize is the number of documents contained in the segment index.
</p>
</subsection>
<subsection name="Lock Files">
<p>
Several files are used to indicate that another
process is using an index. Note that these files are not
stored in the index directory itself, but rather in the
system's temporary directory, as indicated in the Java
system property "java.io.tmpdir".
</p>
<ul>
<li>
<p>
When a file named "commit.lock"
is present, a process is currently re-writing the "segments"
file and deleting outdated segment index files, or a process is
reading the "segments"
file and opening the files of the segments it names. This lock file
prevents files from being deleted by another process after a process
has read the "segments"
file but before it has managed to open all of the files of the
segments named therein.
</p>
</li>
<li>
<p>
When a file named "write.lock"
is present, a process is currently adding documents to an index, or
removing files from that index. This lock file prevents several
processes from attempting to modify an index at the same time.
</p>
</li>
</ul>
</subsection>
<subsection name="Deletable File">
<p>
A file named "deletable"
contains the names of files that are no longer used by the index, but
which could not be deleted. This is only used on Win32, where a
file may not be deleted while it is still open. On other platforms
the file contains only null bytes.
</p>
<p>
Deletable --&gt; DeletableCount,
&lt;DelableName&gt;<sup>DeletableCount</sup>
</p>
<p>DeletableCount --&gt; UInt32
</p>
<p>DeletableName --&gt;
String
</p>
</subsection>
<subsection name="Compound Files">
<p>Starting with Lucene 1.4 the compound file format became default. This
is simply a container for all files described in the next section.</p>
<p>Compound (.cfs) --&gt; FileCount, &lt;DataOffset, FileName&gt;<sup>FileCount</sup>,
FileData<sup>FileCount</sup></p>
<p>FileCount --&gt; VInt</p>
<p>DataOffset --&gt; Long</p>
<p>FileName --&gt; String</p>
<p>FileData --&gt; raw file data</p>
</subsection>
</section>
<section name="Per-Segment Files">
<p>
The remaining files are all per-segment, and are
thus defined by suffix.
</p>
<subsection name="Fields">
<p><br/><b>Field Info</b><br/></p>
<p>
Field names are
stored in the field info file, with suffix .fnm.
</p>
<p>
FieldInfos
(.fnm) --&gt; FieldsCount, &lt;FieldName,
FieldBits&gt;<sup>FieldsCount</sup>
</p>
<p>
FieldsCount --&gt; VInt
</p>
<p>
FieldName --&gt; String
</p>
<p>
FieldBits --&gt; Byte
</p>
<p>
The low-order bit is one for
indexed fields, and zero for non-indexed fields. The second lowest-order
bit is one for fields that have term vectors stored, and zero for fields
without term vectors.
</p>
<p>
Fields are numbered by their order in this file. Thus field zero is
the
first field in the file, field one the next, and so on. Note that,
like document numbers, field numbers are segment relative.
</p>
<p><br/><b>Stored Fields</b><br/></p>
<p>
Stored fields are represented by two files:
</p>
<ol>
<li>
<p>
The field index, or .fdx file.
</p>
<p>
This contains, for each document, a pointer to
its field data, as follows:
</p>
<p>
FieldIndex
(.fdx) --&gt;
&lt;FieldValuesPosition&gt;<sup>SegSize</sup>
</p>
<p>FieldValuesPosition
--&gt; Uint64
</p>
<p>This
is used to find the location within the field data file of the
fields of a particular document. Because it contains fixed-length
data, this file may be easily randomly accessed. The position of
document<i> n</i>'s<i> </i>field data is the Uint64 at <i>n*8</i> in
this file.
</p>
</li>
<li>
<p>
The field data, or .fdt file.
</p>
<p>
This contains the stored fields of each document,
as follows:
</p>
<p>
FieldData (.fdt) --&gt;
&lt;DocFieldData&gt;<sup>SegSize</sup>
</p>
<p>DocFieldData --&gt;
FieldCount, &lt;FieldNum, Bits, Value&gt;<sup>FieldCount</sup>
</p>
<p>FieldCount --&gt;
VInt
</p>
<p>FieldNum --&gt;
VInt
</p>
<p><b>Lucene &lt;= 1.4:</b></p>
<p>Bits --&gt;
Byte
</p>
<p>Value --&gt;
String
</p>
<p>Only the low-order bit of Bits is used. It is one for
tokenized fields, and zero for non-tokenized fields.
</p>
<p><b>Lucene &gt;= 1.9:</b></p>
<p>Bits --&gt;
Byte
</p>
<p>
<ul>
<li>low order bit is one for tokenized fields</li>
<li>second bit is one for fields containing binary data</li>
<li>third bit is one for fields with compression option enabled
(if compression is enabled, the algorithm used is ZLIB)</li>
</ul>
</p>
<p>Value --&gt;
String | BinaryValue (depending on Bits)
</p>
<p>BinaryValue --&gt;
ValueSize, &lt;Byte&gt;^ValueSize
</p>
<p>ValueSize --&gt;
VInt
</p>
</li>
</ol>
</subsection>
<subsection name="Term Dictionary">
<p>
The term dictionary is represented as two files:
</p>
<ol>
<li>
<p>
The term infos, or tis file.
</p>
<p>
TermInfoFile (.tis)--&gt;
TIVersion, TermCount, IndexInterval, SkipInterval, TermInfos
</p>
<p>TIVersion --&gt;
UInt32
</p>
<p>TermCount --&gt;
UInt64
</p>
<p>IndexInterval --&gt;
UInt32
</p>
<p>SkipInterval --&gt;
UInt32
</p>
<p>TermInfos --&gt;
&lt;TermInfo&gt;<sup>TermCount</sup>
</p>
<p>TermInfo --&gt;
&lt;Term, DocFreq, FreqDelta, ProxDelta, SkipDelta&gt;
</p>
<p>Term --&gt;
&lt;PrefixLength, Suffix, FieldNum&gt;
</p>
<p>Suffix --&gt;
String
</p>
<p>PrefixLength,
DocFreq, FreqDelta, ProxDelta, SkipDelta<br/> --&gt; VInt
</p>
<p>This
file is sorted by Term. Terms are ordered first lexicographically
by the term's field name, and within that lexicographically by the
term's text.
</p>
<p>TIVersion names the version of the format
of this file and is -2 in Lucene 1.4.
</p>
<p>Term
text prefixes are shared. The PrefixLength is the number of initial
characters from the previous term which must be pre-pended to a
term's suffix in order to form the term's text. Thus, if the
previous term's text was "bone" and the term is "boy",
the PrefixLength is two and the suffix is "y".
</p>
<p>FieldNumber
determines the term's field, whose name is stored in the .fdt file.
</p>
<p>DocFreq
is the count of documents which contain the term.
</p>
<p>FreqDelta
determines the position of this term's TermFreqs within the .frq
file. In particular, it is the difference between the position of
this term's data in that file and the position of the previous
term's data (or zero, for the first term in the file).
</p>
<p>ProxDelta
determines the position of this term's TermPositions within the .prx
file. In particular, it is the difference between the position of
this term's data in that file and the position of the previous
term's data (or zero, for the first term in the file.
</p>
<p>SkipDelta determines the position of this
term's SkipData within the .frq file. In
particular, it is the number of bytes
after TermFreqs that the SkipData starts.
In other words, it is the length of the
TermFreq data.
</p>
</li>
<li>
<p>
The term info index, or .tii file.
</p>
<p>
This contains every IndexInterval<sup>th</sup> entry from the .tis
file, along with its location in the &quot;tis&quot; file. This is
designed to be read entirely into memory and used to provide random
access to the &quot;tis&quot; file.
</p>
<p>
The structure of this file is very similar to the
.tis file, with the addition of one item per record, the IndexDelta.
</p>
<p>
TermInfoIndex (.tii)--&gt;
TIVersion, IndexTermCount, IndexInterval, SkipInterval, TermIndices
</p>
<p>TIVersion --&gt;
UInt32
</p>
<p>IndexTermCount --&gt;
UInt64
</p>
<p>IndexInterval --&gt;
UInt32
</p>
<p>SkipInterval --&gt;
UInt32
</p>
<p>TermIndices --&gt;
&lt;TermInfo, IndexDelta&gt;<sup>IndexTermCount</sup>
</p>
<p>IndexDelta --&gt;
VLong
</p>
<p>IndexDelta
determines the position of this term's TermInfo within the .tis file. In
particular, it is the difference between the position of this term's
entry in that file and the position of the previous term's entry.
</p>
<p>TODO: document skipInterval information</p>
</li>
</ol>
</subsection>
<subsection name="Frequencies">
<p>
The .frq file contains the lists of documents
which contain each term, along with the frequency of the term in that
document.
</p>
<p>FreqFile (.frq) --&gt;
&lt;TermFreqs, SkipData&gt;<sup>TermCount</sup>
</p>
<p>TermFreqs --&gt;
&lt;TermFreq&gt;<sup>DocFreq</sup>
</p>
<p>TermFreq --&gt;
DocDelta, Freq?
</p>
<p>SkipData --&gt;
&lt;SkipDatum&gt;<sup>DocFreq/SkipInterval</sup>
</p>
<p>SkipDatum --&gt;
DocSkip,FreqSkip,ProxSkip
</p>
<p>DocDelta,Freq,DocSkip,FreqSkip,ProxSkip --&gt;
VInt
</p>
<p>TermFreqs
are ordered by term (the term is implicit, from the .tis file).
</p>
<p>TermFreq
entries are ordered by increasing document number.
</p>
<p>DocDelta
determines both the document number and the frequency. In
particular, DocDelta/2 is the difference between this document number
and the previous document number (or zero when this is the first
document in a TermFreqs). When DocDelta is odd, the frequency is
one. When DocDelta is even, the frequency is read as another VInt.
</p>
<p>For
example, the TermFreqs for a term which occurs once in document seven
and three times in document eleven would be the following sequence of
VInts:
</p>
<p> 15,
22, 3
</p>
<p>DocSkip records the document number before every
SkipInterval<sup>th</sup> document in TermFreqs.
Document numbers are represented as differences
from the previous value in the sequence. FreqSkip
and ProxSkip record the position of every
SkipInterval<sup>th</sup> entry in FreqFile and
ProxFile, respectively. File positions are
relative to the start of TermFreqs and Positions,
to the previous SkipDatum in the sequence.
</p>
<p>For example, if DocFreq=35 and SkipInterval=16,
then there are two SkipData entries, containing
the 15<sup>th</sup> and 31<sup>st</sup> document
numbers in TermFreqs. The first FreqSkip names
the number of bytes after the beginning of
TermFreqs that the 16<sup>th</sup> SkipDatum
starts, and the second the number of bytes after
that that the 32<sup>nd</sup> starts. The first
ProxSkip names the number of bytes after the
beginning of Positions that the 16<sup>th</sup>
SkipDatum starts, and the second the number of
bytes after that that the 32<sup>nd</sup> starts.
</p>
</subsection>
<subsection name="Positions">
<p>
The .prx file contains the lists of positions that
each term occurs at within documents.
</p>
<p>ProxFile (.prx) --&gt;
&lt;TermPositions&gt;<sup>TermCount</sup>
</p>
<p>TermPositions --&gt;
&lt;Positions&gt;<sup>DocFreq</sup>
</p>
<p>Positions --&gt;
&lt;PositionDelta&gt;<sup>Freq</sup>
</p>
<p>PositionDelta --&gt;
VInt
</p>
<p>TermPositions
are ordered by term (the term is implicit, from the .tis file).
</p>
<p>Positions
entries are ordered by increasing document number (the document
number is implicit from the .frq file).
</p>
<p>PositionDelta
is the difference between the position of the current occurrence in
the document and the previous occurrence (or zero, if this is the
first occurrence in this document).
</p>
<p>
For example, the TermPositions for a
term which occurs as the fourth term in one document, and as the
fifth and ninth term in a subsequent document, would be the following
sequence of VInts:
</p>
<p> 4,
5, 4
</p>
</subsection>
<subsection name="Normalization Factors">
<p>There's a norm file for each indexed field with a byte for
each document. The .f[0-9]* file contains,
for each document, a byte that encodes a value that is multiplied
into the score for hits on that field:
</p>
<p>Norms
(.f[0-9]*) --&gt; &lt;Byte&gt;<sup>SegSize</sup>
</p>
<p>Each
byte encodes a floating point value. Bits 0-2 contain the 3-bit
mantissa, and bits 3-8 contain the 5-bit exponent.
</p>
<p>These
are converted to an IEEE single float value as follows:
</p>
<ol>
<li><p>If
the byte is zero, use a zero float.
</p>
</li>
<li><p>Otherwise,
set the sign bit of the float to zero;
</p>
</li>
<li><p>add
48 to the exponent and use this as the float's exponent;
</p>
</li>
<li><p>map
the mantissa to the high-order 3 bits of the float's mantissa; and
</p>
</li>
<li><p>set
the low-order 21 bits of the float's mantissa to zero.
</p>
</li>
</ol>
</subsection>
<subsection name="Term Vectors">
Term Vector support is an optional on a field by field basis. It consists of 4
files.
<ol>
<li>
<p>The Document Index or .tvx file.</p>
<p>This contains, for each document, a pointer to the document data in the Document
(.tvd) file.
</p>
<p>DocumentIndex (.tvx) --&gt; TVXVersion&lt;DocumentPosition&gt;<sup>NumDocs</sup></p>
<p>TVXVersion --&gt; Int</p>
<p>DocumentPosition --&gt; UInt64</p>
<p>This is used to find the position of the Document in the .tvd file.</p>
</li>
<li>
<p>The Document or .tvd file.</p>
<p>This contains, for each document, the number of fields, a list of the fields with
term vector info and finally a list of pointers to the field information in the .tvf
(Term Vector Fields) file.</p>
<p>
Document (.tvd) --&gt; TVDVersion&lt;NumFields, FieldNums, FieldPositions,&gt;<sup>NumDocs</sup>
</p>
<p>TVDVersion --&gt; Int</p>
<p>NumFields --&gt; VInt</p>
<p>FieldNums --&gt; &lt;FieldNumDelta&gt;<sup>NumFields</sup></p>
<p>FieldNumDelta --&gt; VInt</p>
<p>FieldPositions --&gt; &lt;FieldPosition&gt;<sup>NumFields</sup></p>
<p>FieldPosition --&gt; VLong</p>
<p>The .tvd file is used to map out the fields that have term vectors stored and
where the field information is in the .tvf file.</p>
</li>
<li>
<p>The Field or .tvf file.</p>
<p>This file contains, for each field that has a term vector stored, a list of
the terms and their frequencies.</p>
<p>Field (.tvf) --&gt; TVFVersion&lt;NumTerms, NumDistinct, TermFreqs&gt;<sup>NumFields</sup></p>
<p>TVFVersion --&gt; Int</p>
<p>NumTerms --&gt; VInt</p>
<p>NumDistinct --&gt; VInt -- Future Use</p>
<p>TermFreqs --&gt; &lt;TermText, TermFreq&gt;<sup>NumTerms</sup></p>
<p>TermText --&gt; &lt;PrefixLength, Suffix&gt;</p>
<p>PrefixLength --&gt; VInt</p>
<p>Suffix --&gt; String</p>
<p>TermFreq --&gt; VInt</p>
<p>Term
text prefixes are shared. The PrefixLength is the number of initial
characters from the previous term which must be pre-pended to a
term's suffix in order to form the term's text. Thus, if the
previous term's text was "bone" and the term is "boy",
the PrefixLength is two and the suffix is "y".
</p>
</li>
</ol>
</subsection>
<subsection name="Deleted Documents">
<p>The .del file is
optional, and only exists when a segment contains deletions:
</p>
<p>Deletions
(.del) --&gt; ByteCount,BitCount,Bits
</p>
<p>ByteSize,BitCount --&gt;
Uint32
</p>
<p>Bits --&gt;
&lt;Byte&gt;<sup>ByteCount</sup>
</p>
<p>ByteCount
indicates the number of bytes in Bits. It is typically
(SegSize/8)+1.
</p>
<p>
BitCount
indicates the number of bits that are currently set in Bits.
</p>
<p>Bits
contains one bit for each document indexed. When the bit
corresponding to a document number is set, that document is marked as
deleted. Bit ordering is from least to most significant. Thus, if
Bits contains two bytes, 0x00 and 0x02, then document 9 is marked as
deleted.
</p>
</subsection>
</section>
<section name="Limitations">
<p>There
are a few places where these file formats limit the maximum number of
terms and documents to a 32-bit quantity, or to approximately 4
billion. This is not today a problem, but, in the long term,
probably will be. These should therefore be replaced with either
UInt64 values, or better yet, with VInt values which have no limit.
</p>
</section>
</body>
</document>