Apache
/
lucene
mirror of https://github.com/apache/lucene.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

More documentation to help people pick the right PackedInts structure. 

git-svn-id: https://svn.apache.org/repos/asf/lucene/dev/trunk@1456647 13f79535-47bb-0310-9956-ffa450edef68
This commit is contained in:
Adrien Grand 2013-03-14 20:16:57 +00:00
parent 4e8f2b1749
commit 3c065d3922
1 changed files with 46 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

47
lucene/core/src/java/org/apache/lucene/util/packed/package.html
View File

@ -19,7 +19,7 @@
<head></head> <head></head>
<body bgcolor="white"> <body bgcolor="white">
<p>Packed integer arrays and streams.</p> <h2>Packed integer arrays and streams.</h2>
<p> <p>
The packed package provides The packed package provides
@ -35,5 +35,50 @@
The main access point is the {@link org.apache.lucene.util.packed.PackedInts} factory. The main access point is the {@link org.apache.lucene.util.packed.PackedInts} factory.
</p> </p>
<h3>In-memory structures</h3>
<ul>
<li><b>{@link org.apache.lucene.util.packed.PackedInts.Mutable}</b><ul>
<li>Only supports positive longs.</li>
<li>Requires the number of bits per value to be known in advance.</li>
<li>Random-access for both writing and reading.</li>
</ul></li>
<li><b>{@link org.apache.lucene.util.packed.GrowableWriter}</b><ul>
<li>Same as PackedInts.Mutable but grows the number of bits per values when needed.</li>
<li>Useful to build a PackedInts.Mutable from a read-once stream of longs.</li>
</ul></li>
<li><b>{@link org.apache.lucene.util.packed.AppendingLongBuffer}</b><ul>
<li>Can store any sequence of longs.</li>
<li>Compression is good when values are close to each other.</li>
<li>Supports random reads, but only sequential writes.</li>
<li>Can address up to 2^42 values.</li>
</ul></li>
<li><b>{@link org.apache.lucene.util.packed.MonotonicAppendingLongBuffer}</b><ul>
<li>Same as AppendingLongBuffer except that compression is good when the stream is a succession of affine functions.</li>
</ul></li>
</ul>
<h3>Disk-based structures</h3>
<ul>
<li><b>{@link org.apache.lucene.util.packed.PackedInts.Writer}, {@link org.apache.lucene.util.packed.PackedInts.Reader}, {@link org.apache.lucene.util.packed.PackedInts.ReaderIterator}</b><ul>
<li>Only supports positive longs.</li>
<li>Requires the number of bits per value to be known in advance.</li>
<li>Supports both fast sequential access with low memory footprint with ReaderIterator and random-access by either loading values in memory or leaving them on disk with Reader.</li>
</ul></li>
<li><b>{@link org.apache.lucene.util.packed.BlockPackedWriter}, {@link org.apache.lucene.util.packed.BlockPackedReader}, {@link org.apache.lucene.util.packed.BlockPackedReaderIterator}</b><ul>
<li>Splits the stream into fixed-size blocks.</li>
<li>Compression is good when values are close to each other.</li>
<li>Can address up to 2B * blockSize values.</li>
</ul></li>
<li><b>{@link org.apache.lucene.util.packed.MonotonicBlockPackedWriter}, {@link org.apache.lucene.util.packed.MonotonicBlockPackedReader}</b><ul>
<li>Same as the non-monotonic variants except that compression is good when the stream is a succession of affine functions.</li>
<li>The reason why there is no sequential access is that if you need sequential access, you should rather delta-encode and use BlockPackedWriter.</li>
</ul></li>
<li><b>{@link org.apache.lucene.util.packed.PackedDataOutput}, {@link org.apache.lucene.util.packed.PackedDataInput}</b><ul>
<li>Writes sequences of longs where each long can use any number of bits.</li>
</ul></li>
</ul>
</body> </body>
</html> </html>