hbase-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From git-site-r...@apache.org
Subject [29/51] [partial] hbase-site git commit: Published site at 82d554e3783372cc6b05489452c815b57c06f6cd.
Date Thu, 11 May 2017 14:59:45 GMT
http://git-wip-us.apache.org/repos/asf/hbase-site/blob/1241ee85/devapidocs/src-html/org/apache/hadoop/hbase/io/hfile/HFileBlock.html
----------------------------------------------------------------------
diff --git a/devapidocs/src-html/org/apache/hadoop/hbase/io/hfile/HFileBlock.html b/devapidocs/src-html/org/apache/hadoop/hbase/io/hfile/HFileBlock.html
index 3fedd0b..6414009 100644
--- a/devapidocs/src-html/org/apache/hadoop/hbase/io/hfile/HFileBlock.html
+++ b/devapidocs/src-html/org/apache/hadoop/hbase/io/hfile/HFileBlock.html
@@ -25,2075 +25,2063 @@
 <span class="sourceLineNo">017</span> */<a name="line.17"></a>
 <span class="sourceLineNo">018</span>package org.apache.hadoop.hbase.io.hfile;<a name="line.18"></a>
 <span class="sourceLineNo">019</span><a name="line.19"></a>
-<span class="sourceLineNo">020</span>import java.io.DataInputStream;<a name="line.20"></a>
-<span class="sourceLineNo">021</span>import java.io.DataOutput;<a name="line.21"></a>
-<span class="sourceLineNo">022</span>import java.io.DataOutputStream;<a name="line.22"></a>
-<span class="sourceLineNo">023</span>import java.io.IOException;<a name="line.23"></a>
-<span class="sourceLineNo">024</span>import java.io.InputStream;<a name="line.24"></a>
-<span class="sourceLineNo">025</span>import java.nio.ByteBuffer;<a name="line.25"></a>
-<span class="sourceLineNo">026</span>import java.util.concurrent.atomic.AtomicReference;<a name="line.26"></a>
-<span class="sourceLineNo">027</span>import java.util.concurrent.locks.Lock;<a name="line.27"></a>
-<span class="sourceLineNo">028</span>import java.util.concurrent.locks.ReentrantLock;<a name="line.28"></a>
-<span class="sourceLineNo">029</span><a name="line.29"></a>
-<span class="sourceLineNo">030</span>import org.apache.commons.logging.Log;<a name="line.30"></a>
-<span class="sourceLineNo">031</span>import org.apache.commons.logging.LogFactory;<a name="line.31"></a>
-<span class="sourceLineNo">032</span>import org.apache.hadoop.fs.FSDataInputStream;<a name="line.32"></a>
-<span class="sourceLineNo">033</span>import org.apache.hadoop.fs.FSDataOutputStream;<a name="line.33"></a>
-<span class="sourceLineNo">034</span>import org.apache.hadoop.fs.Path;<a name="line.34"></a>
-<span class="sourceLineNo">035</span>import org.apache.hadoop.hbase.Cell;<a name="line.35"></a>
-<span class="sourceLineNo">036</span>import org.apache.hadoop.hbase.HConstants;<a name="line.36"></a>
-<span class="sourceLineNo">037</span>import org.apache.hadoop.hbase.classification.InterfaceAudience;<a name="line.37"></a>
-<span class="sourceLineNo">038</span>import org.apache.hadoop.hbase.fs.HFileSystem;<a name="line.38"></a>
-<span class="sourceLineNo">039</span>import org.apache.hadoop.hbase.io.ByteArrayOutputStream;<a name="line.39"></a>
-<span class="sourceLineNo">040</span>import org.apache.hadoop.hbase.io.ByteBuffInputStream;<a name="line.40"></a>
-<span class="sourceLineNo">041</span>import org.apache.hadoop.hbase.io.ByteBufferWriterDataOutputStream;<a name="line.41"></a>
-<span class="sourceLineNo">042</span>import org.apache.hadoop.hbase.io.FSDataInputStreamWrapper;<a name="line.42"></a>
-<span class="sourceLineNo">043</span>import org.apache.hadoop.hbase.io.encoding.DataBlockEncoding;<a name="line.43"></a>
-<span class="sourceLineNo">044</span>import org.apache.hadoop.hbase.io.encoding.HFileBlockDecodingContext;<a name="line.44"></a>
-<span class="sourceLineNo">045</span>import org.apache.hadoop.hbase.io.encoding.HFileBlockDefaultDecodingContext;<a name="line.45"></a>
-<span class="sourceLineNo">046</span>import org.apache.hadoop.hbase.io.encoding.HFileBlockDefaultEncodingContext;<a name="line.46"></a>
-<span class="sourceLineNo">047</span>import org.apache.hadoop.hbase.io.encoding.HFileBlockEncodingContext;<a name="line.47"></a>
-<span class="sourceLineNo">048</span>import org.apache.hadoop.hbase.nio.ByteBuff;<a name="line.48"></a>
-<span class="sourceLineNo">049</span>import org.apache.hadoop.hbase.nio.MultiByteBuff;<a name="line.49"></a>
-<span class="sourceLineNo">050</span>import org.apache.hadoop.hbase.nio.SingleByteBuff;<a name="line.50"></a>
-<span class="sourceLineNo">051</span>import org.apache.hadoop.hbase.util.Bytes;<a name="line.51"></a>
-<span class="sourceLineNo">052</span>import org.apache.hadoop.hbase.util.ChecksumType;<a name="line.52"></a>
-<span class="sourceLineNo">053</span>import org.apache.hadoop.hbase.util.ClassSize;<a name="line.53"></a>
-<span class="sourceLineNo">054</span>import org.apache.hadoop.io.IOUtils;<a name="line.54"></a>
-<span class="sourceLineNo">055</span><a name="line.55"></a>
-<span class="sourceLineNo">056</span>import com.google.common.annotations.VisibleForTesting;<a name="line.56"></a>
-<span class="sourceLineNo">057</span>import com.google.common.base.Preconditions;<a name="line.57"></a>
-<span class="sourceLineNo">058</span><a name="line.58"></a>
-<span class="sourceLineNo">059</span>/**<a name="line.59"></a>
-<span class="sourceLineNo">060</span> * Reads {@link HFile} version 2 blocks to HFiles and via {@link Cacheable} Interface to caches.<a name="line.60"></a>
-<span class="sourceLineNo">061</span> * Version 2 was introduced in hbase-0.92.0. No longer has support for version 1 blocks since<a name="line.61"></a>
-<span class="sourceLineNo">062</span> * hbase-1.3.0.<a name="line.62"></a>
-<span class="sourceLineNo">063</span> *<a name="line.63"></a>
-<span class="sourceLineNo">064</span> * &lt;p&gt;Version 1 was the original file block. Version 2 was introduced when we changed the hbase file<a name="line.64"></a>
-<span class="sourceLineNo">065</span> * format to support multi-level block indexes and compound bloom filters (HBASE-3857).<a name="line.65"></a>
-<span class="sourceLineNo">066</span> *<a name="line.66"></a>
-<span class="sourceLineNo">067</span> * &lt;h3&gt;HFileBlock: Version 2&lt;/h3&gt;<a name="line.67"></a>
-<span class="sourceLineNo">068</span> * In version 2, a block is structured as follows:<a name="line.68"></a>
-<span class="sourceLineNo">069</span> * &lt;ul&gt;<a name="line.69"></a>
-<span class="sourceLineNo">070</span> * &lt;li&gt;&lt;b&gt;Header:&lt;/b&gt; See Writer#putHeader() for where header is written; header total size is<a name="line.70"></a>
-<span class="sourceLineNo">071</span> * HFILEBLOCK_HEADER_SIZE<a name="line.71"></a>
-<span class="sourceLineNo">072</span> * &lt;ul&gt;<a name="line.72"></a>
-<span class="sourceLineNo">073</span> * &lt;li&gt;0. blockType: Magic record identifying the {@link BlockType} (8 bytes):<a name="line.73"></a>
-<span class="sourceLineNo">074</span> * e.g. &lt;code&gt;DATABLK*&lt;/code&gt;<a name="line.74"></a>
-<span class="sourceLineNo">075</span> * &lt;li&gt;1. onDiskSizeWithoutHeader: Compressed -- a.k.a 'on disk' -- block size, excluding header,<a name="line.75"></a>
-<span class="sourceLineNo">076</span> * but including tailing checksum bytes (4 bytes)<a name="line.76"></a>
-<span class="sourceLineNo">077</span> * &lt;li&gt;2. uncompressedSizeWithoutHeader: Uncompressed block size, excluding header, and excluding<a name="line.77"></a>
-<span class="sourceLineNo">078</span> * checksum bytes (4 bytes)<a name="line.78"></a>
-<span class="sourceLineNo">079</span> * &lt;li&gt;3. prevBlockOffset: The offset of the previous block of the same type (8 bytes). This is<a name="line.79"></a>
-<span class="sourceLineNo">080</span> * used to navigate to the previous block without having to go to the block index<a name="line.80"></a>
-<span class="sourceLineNo">081</span> * &lt;li&gt;4: For minorVersions &amp;gt;=1, the ordinal describing checksum type (1 byte)<a name="line.81"></a>
-<span class="sourceLineNo">082</span> * &lt;li&gt;5: For minorVersions &amp;gt;=1, the number of data bytes/checksum chunk (4 bytes)<a name="line.82"></a>
-<span class="sourceLineNo">083</span> * &lt;li&gt;6: onDiskDataSizeWithHeader: For minorVersions &amp;gt;=1, the size of data 'on disk', including<a name="line.83"></a>
-<span class="sourceLineNo">084</span> * header, excluding checksums (4 bytes)<a name="line.84"></a>
-<span class="sourceLineNo">085</span> * &lt;/ul&gt;<a name="line.85"></a>
-<span class="sourceLineNo">086</span> * &lt;/li&gt;<a name="line.86"></a>
-<span class="sourceLineNo">087</span> * &lt;li&gt;&lt;b&gt;Raw/Compressed/Encrypted/Encoded data:&lt;/b&gt; The compression<a name="line.87"></a>
-<span class="sourceLineNo">088</span> * algorithm is the same for all the blocks in an {@link HFile}. If compression is NONE, this is<a name="line.88"></a>
-<span class="sourceLineNo">089</span> * just raw, serialized Cells.<a name="line.89"></a>
-<span class="sourceLineNo">090</span> * &lt;li&gt;&lt;b&gt;Tail:&lt;/b&gt; For minorVersions &amp;gt;=1, a series of 4 byte checksums, one each for<a name="line.90"></a>
-<span class="sourceLineNo">091</span> * the number of bytes specified by bytesPerChecksum.<a name="line.91"></a>
-<span class="sourceLineNo">092</span> * &lt;/ul&gt;<a name="line.92"></a>
-<span class="sourceLineNo">093</span> *<a name="line.93"></a>
-<span class="sourceLineNo">094</span> * &lt;h3&gt;Caching&lt;/h3&gt;<a name="line.94"></a>
-<span class="sourceLineNo">095</span> * Caches cache whole blocks with trailing checksums if any. We then tag on some metadata, the<a name="line.95"></a>
-<span class="sourceLineNo">096</span> * content of BLOCK_METADATA_SPACE which will be flag on if we are doing 'hbase'<a name="line.96"></a>
-<span class="sourceLineNo">097</span> * checksums and then the offset into the file which is needed when we re-make a cache key<a name="line.97"></a>
-<span class="sourceLineNo">098</span> * when we return the block to the cache as 'done'. See {@link Cacheable#serialize(ByteBuffer)} and<a name="line.98"></a>
-<span class="sourceLineNo">099</span> * {@link Cacheable#getDeserializer()}.<a name="line.99"></a>
-<span class="sourceLineNo">100</span> *<a name="line.100"></a>
-<span class="sourceLineNo">101</span> * &lt;p&gt;TODO: Should we cache the checksums? Down in Writer#getBlockForCaching(CacheConfig) where<a name="line.101"></a>
-<span class="sourceLineNo">102</span> * we make a block to cache-on-write, there is an attempt at turning off checksums. This is not the<a name="line.102"></a>
-<span class="sourceLineNo">103</span> * only place we get blocks to cache. We also will cache the raw return from an hdfs read. In this<a name="line.103"></a>
-<span class="sourceLineNo">104</span> * case, the checksums may be present. If the cache is backed by something that doesn't do ECC,<a name="line.104"></a>
-<span class="sourceLineNo">105</span> * say an SSD, we might want to preserve checksums. For now this is open question.<a name="line.105"></a>
-<span class="sourceLineNo">106</span> * &lt;p&gt;TODO: Over in BucketCache, we save a block allocation by doing a custom serialization.<a name="line.106"></a>
-<span class="sourceLineNo">107</span> * Be sure to change it if serialization changes in here. Could we add a method here that takes an<a name="line.107"></a>
-<span class="sourceLineNo">108</span> * IOEngine and that then serializes to it rather than expose our internals over in BucketCache?<a name="line.108"></a>
-<span class="sourceLineNo">109</span> * IOEngine is in the bucket subpackage. Pull it up? Then this class knows about bucketcache. Ugh.<a name="line.109"></a>
-<span class="sourceLineNo">110</span> */<a name="line.110"></a>
-<span class="sourceLineNo">111</span>@InterfaceAudience.Private<a name="line.111"></a>
-<span class="sourceLineNo">112</span>public class HFileBlock implements Cacheable {<a name="line.112"></a>
-<span class="sourceLineNo">113</span>  private static final Log LOG = LogFactory.getLog(HFileBlock.class);<a name="line.113"></a>
-<span class="sourceLineNo">114</span><a name="line.114"></a>
-<span class="sourceLineNo">115</span>  /** Type of block. Header field 0. */<a name="line.115"></a>
-<span class="sourceLineNo">116</span>  private BlockType blockType;<a name="line.116"></a>
-<span class="sourceLineNo">117</span><a name="line.117"></a>
-<span class="sourceLineNo">118</span>  /**<a name="line.118"></a>
-<span class="sourceLineNo">119</span>   * Size on disk excluding header, including checksum. Header field 1.<a name="line.119"></a>
-<span class="sourceLineNo">120</span>   * @see Writer#putHeader(byte[], int, int, int, int)<a name="line.120"></a>
-<span class="sourceLineNo">121</span>   */<a name="line.121"></a>
-<span class="sourceLineNo">122</span>  private int onDiskSizeWithoutHeader;<a name="line.122"></a>
-<span class="sourceLineNo">123</span><a name="line.123"></a>
-<span class="sourceLineNo">124</span>  /**<a name="line.124"></a>
-<span class="sourceLineNo">125</span>   * Size of pure data. Does not include header or checksums. Header field 2.<a name="line.125"></a>
-<span class="sourceLineNo">126</span>   * @see Writer#putHeader(byte[], int, int, int, int)<a name="line.126"></a>
-<span class="sourceLineNo">127</span>   */<a name="line.127"></a>
-<span class="sourceLineNo">128</span>  private int uncompressedSizeWithoutHeader;<a name="line.128"></a>
-<span class="sourceLineNo">129</span><a name="line.129"></a>
-<span class="sourceLineNo">130</span>  /**<a name="line.130"></a>
-<span class="sourceLineNo">131</span>   * The offset of the previous block on disk. Header field 3.<a name="line.131"></a>
-<span class="sourceLineNo">132</span>   * @see Writer#putHeader(byte[], int, int, int, int)<a name="line.132"></a>
-<span class="sourceLineNo">133</span>   */<a name="line.133"></a>
-<span class="sourceLineNo">134</span>  private long prevBlockOffset;<a name="line.134"></a>
-<span class="sourceLineNo">135</span><a name="line.135"></a>
-<span class="sourceLineNo">136</span>  /**<a name="line.136"></a>
-<span class="sourceLineNo">137</span>   * Size on disk of header + data. Excludes checksum. Header field 6,<a name="line.137"></a>
-<span class="sourceLineNo">138</span>   * OR calculated from {@link #onDiskSizeWithoutHeader} when using HDFS checksum.<a name="line.138"></a>
-<span class="sourceLineNo">139</span>   * @see Writer#putHeader(byte[], int, int, int, int)<a name="line.139"></a>
-<span class="sourceLineNo">140</span>   */<a name="line.140"></a>
-<span class="sourceLineNo">141</span>  private int onDiskDataSizeWithHeader;<a name="line.141"></a>
-<span class="sourceLineNo">142</span><a name="line.142"></a>
-<span class="sourceLineNo">143</span><a name="line.143"></a>
-<span class="sourceLineNo">144</span>  /**<a name="line.144"></a>
-<span class="sourceLineNo">145</span>   * The in-memory representation of the hfile block. Can be on or offheap. Can be backed by<a name="line.145"></a>
-<span class="sourceLineNo">146</span>   * a single ByteBuffer or by many. Make no assumptions.<a name="line.146"></a>
-<span class="sourceLineNo">147</span>   *<a name="line.147"></a>
-<span class="sourceLineNo">148</span>   * &lt;p&gt;Be careful reading from this &lt;code&gt;buf&lt;/code&gt;. Duplicate and work on the duplicate or if<a name="line.148"></a>
-<span class="sourceLineNo">149</span>   * not, be sure to reset position and limit else trouble down the road.<a name="line.149"></a>
+<span class="sourceLineNo">020</span>import com.google.common.annotations.VisibleForTesting;<a name="line.20"></a>
+<span class="sourceLineNo">021</span>import com.google.common.base.Preconditions;<a name="line.21"></a>
+<span class="sourceLineNo">022</span><a name="line.22"></a>
+<span class="sourceLineNo">023</span>import java.io.DataInputStream;<a name="line.23"></a>
+<span class="sourceLineNo">024</span>import java.io.DataOutput;<a name="line.24"></a>
+<span class="sourceLineNo">025</span>import java.io.DataOutputStream;<a name="line.25"></a>
+<span class="sourceLineNo">026</span>import java.io.IOException;<a name="line.26"></a>
+<span class="sourceLineNo">027</span>import java.io.InputStream;<a name="line.27"></a>
+<span class="sourceLineNo">028</span>import java.nio.ByteBuffer;<a name="line.28"></a>
+<span class="sourceLineNo">029</span>import java.util.concurrent.atomic.AtomicReference;<a name="line.29"></a>
+<span class="sourceLineNo">030</span><a name="line.30"></a>
+<span class="sourceLineNo">031</span>import org.apache.commons.logging.Log;<a name="line.31"></a>
+<span class="sourceLineNo">032</span>import org.apache.commons.logging.LogFactory;<a name="line.32"></a>
+<span class="sourceLineNo">033</span>import org.apache.hadoop.fs.FSDataInputStream;<a name="line.33"></a>
+<span class="sourceLineNo">034</span>import org.apache.hadoop.fs.FSDataOutputStream;<a name="line.34"></a>
+<span class="sourceLineNo">035</span>import org.apache.hadoop.fs.Path;<a name="line.35"></a>
+<span class="sourceLineNo">036</span>import org.apache.hadoop.hbase.Cell;<a name="line.36"></a>
+<span class="sourceLineNo">037</span>import org.apache.hadoop.hbase.HConstants;<a name="line.37"></a>
+<span class="sourceLineNo">038</span>import org.apache.hadoop.hbase.classification.InterfaceAudience;<a name="line.38"></a>
+<span class="sourceLineNo">039</span>import org.apache.hadoop.hbase.fs.HFileSystem;<a name="line.39"></a>
+<span class="sourceLineNo">040</span>import org.apache.hadoop.hbase.io.ByteArrayOutputStream;<a name="line.40"></a>
+<span class="sourceLineNo">041</span>import org.apache.hadoop.hbase.io.ByteBuffInputStream;<a name="line.41"></a>
+<span class="sourceLineNo">042</span>import org.apache.hadoop.hbase.io.ByteBufferWriterDataOutputStream;<a name="line.42"></a>
+<span class="sourceLineNo">043</span>import org.apache.hadoop.hbase.io.FSDataInputStreamWrapper;<a name="line.43"></a>
+<span class="sourceLineNo">044</span>import org.apache.hadoop.hbase.io.encoding.DataBlockEncoding;<a name="line.44"></a>
+<span class="sourceLineNo">045</span>import org.apache.hadoop.hbase.io.encoding.HFileBlockDecodingContext;<a name="line.45"></a>
+<span class="sourceLineNo">046</span>import org.apache.hadoop.hbase.io.encoding.HFileBlockDefaultDecodingContext;<a name="line.46"></a>
+<span class="sourceLineNo">047</span>import org.apache.hadoop.hbase.io.encoding.HFileBlockDefaultEncodingContext;<a name="line.47"></a>
+<span class="sourceLineNo">048</span>import org.apache.hadoop.hbase.io.encoding.HFileBlockEncodingContext;<a name="line.48"></a>
+<span class="sourceLineNo">049</span>import org.apache.hadoop.hbase.nio.ByteBuff;<a name="line.49"></a>
+<span class="sourceLineNo">050</span>import org.apache.hadoop.hbase.nio.MultiByteBuff;<a name="line.50"></a>
+<span class="sourceLineNo">051</span>import org.apache.hadoop.hbase.nio.SingleByteBuff;<a name="line.51"></a>
+<span class="sourceLineNo">052</span>import org.apache.hadoop.hbase.util.Bytes;<a name="line.52"></a>
+<span class="sourceLineNo">053</span>import org.apache.hadoop.hbase.util.ChecksumType;<a name="line.53"></a>
+<span class="sourceLineNo">054</span>import org.apache.hadoop.hbase.util.ClassSize;<a name="line.54"></a>
+<span class="sourceLineNo">055</span>import org.apache.hadoop.io.IOUtils;<a name="line.55"></a>
+<span class="sourceLineNo">056</span><a name="line.56"></a>
+<span class="sourceLineNo">057</span>/**<a name="line.57"></a>
+<span class="sourceLineNo">058</span> * Reads {@link HFile} version 2 blocks to HFiles and via {@link Cacheable} Interface to caches.<a name="line.58"></a>
+<span class="sourceLineNo">059</span> * Version 2 was introduced in hbase-0.92.0. No longer has support for version 1 blocks since<a name="line.59"></a>
+<span class="sourceLineNo">060</span> * hbase-1.3.0.<a name="line.60"></a>
+<span class="sourceLineNo">061</span> *<a name="line.61"></a>
+<span class="sourceLineNo">062</span> * &lt;p&gt;Version 1 was the original file block. Version 2 was introduced when we changed the hbase file<a name="line.62"></a>
+<span class="sourceLineNo">063</span> * format to support multi-level block indexes and compound bloom filters (HBASE-3857).<a name="line.63"></a>
+<span class="sourceLineNo">064</span> *<a name="line.64"></a>
+<span class="sourceLineNo">065</span> * &lt;h3&gt;HFileBlock: Version 2&lt;/h3&gt;<a name="line.65"></a>
+<span class="sourceLineNo">066</span> * In version 2, a block is structured as follows:<a name="line.66"></a>
+<span class="sourceLineNo">067</span> * &lt;ul&gt;<a name="line.67"></a>
+<span class="sourceLineNo">068</span> * &lt;li&gt;&lt;b&gt;Header:&lt;/b&gt; See Writer#putHeader() for where header is written; header total size is<a name="line.68"></a>
+<span class="sourceLineNo">069</span> * HFILEBLOCK_HEADER_SIZE<a name="line.69"></a>
+<span class="sourceLineNo">070</span> * &lt;ul&gt;<a name="line.70"></a>
+<span class="sourceLineNo">071</span> * &lt;li&gt;0. blockType: Magic record identifying the {@link BlockType} (8 bytes):<a name="line.71"></a>
+<span class="sourceLineNo">072</span> * e.g. &lt;code&gt;DATABLK*&lt;/code&gt;<a name="line.72"></a>
+<span class="sourceLineNo">073</span> * &lt;li&gt;1. onDiskSizeWithoutHeader: Compressed -- a.k.a 'on disk' -- block size, excluding header,<a name="line.73"></a>
+<span class="sourceLineNo">074</span> * but including tailing checksum bytes (4 bytes)<a name="line.74"></a>
+<span class="sourceLineNo">075</span> * &lt;li&gt;2. uncompressedSizeWithoutHeader: Uncompressed block size, excluding header, and excluding<a name="line.75"></a>
+<span class="sourceLineNo">076</span> * checksum bytes (4 bytes)<a name="line.76"></a>
+<span class="sourceLineNo">077</span> * &lt;li&gt;3. prevBlockOffset: The offset of the previous block of the same type (8 bytes). This is<a name="line.77"></a>
+<span class="sourceLineNo">078</span> * used to navigate to the previous block without having to go to the block index<a name="line.78"></a>
+<span class="sourceLineNo">079</span> * &lt;li&gt;4: For minorVersions &amp;gt;=1, the ordinal describing checksum type (1 byte)<a name="line.79"></a>
+<span class="sourceLineNo">080</span> * &lt;li&gt;5: For minorVersions &amp;gt;=1, the number of data bytes/checksum chunk (4 bytes)<a name="line.80"></a>
+<span class="sourceLineNo">081</span> * &lt;li&gt;6: onDiskDataSizeWithHeader: For minorVersions &amp;gt;=1, the size of data 'on disk', including<a name="line.81"></a>
+<span class="sourceLineNo">082</span> * header, excluding checksums (4 bytes)<a name="line.82"></a>
+<span class="sourceLineNo">083</span> * &lt;/ul&gt;<a name="line.83"></a>
+<span class="sourceLineNo">084</span> * &lt;/li&gt;<a name="line.84"></a>
+<span class="sourceLineNo">085</span> * &lt;li&gt;&lt;b&gt;Raw/Compressed/Encrypted/Encoded data:&lt;/b&gt; The compression<a name="line.85"></a>
+<span class="sourceLineNo">086</span> * algorithm is the same for all the blocks in an {@link HFile}. If compression is NONE, this is<a name="line.86"></a>
+<span class="sourceLineNo">087</span> * just raw, serialized Cells.<a name="line.87"></a>
+<span class="sourceLineNo">088</span> * &lt;li&gt;&lt;b&gt;Tail:&lt;/b&gt; For minorVersions &amp;gt;=1, a series of 4 byte checksums, one each for<a name="line.88"></a>
+<span class="sourceLineNo">089</span> * the number of bytes specified by bytesPerChecksum.<a name="line.89"></a>
+<span class="sourceLineNo">090</span> * &lt;/ul&gt;<a name="line.90"></a>
+<span class="sourceLineNo">091</span> *<a name="line.91"></a>
+<span class="sourceLineNo">092</span> * &lt;h3&gt;Caching&lt;/h3&gt;<a name="line.92"></a>
+<span class="sourceLineNo">093</span> * Caches cache whole blocks with trailing checksums if any. We then tag on some metadata, the<a name="line.93"></a>
+<span class="sourceLineNo">094</span> * content of BLOCK_METADATA_SPACE which will be flag on if we are doing 'hbase'<a name="line.94"></a>
+<span class="sourceLineNo">095</span> * checksums and then the offset into the file which is needed when we re-make a cache key<a name="line.95"></a>
+<span class="sourceLineNo">096</span> * when we return the block to the cache as 'done'. See {@link Cacheable#serialize(ByteBuffer)} and<a name="line.96"></a>
+<span class="sourceLineNo">097</span> * {@link Cacheable#getDeserializer()}.<a name="line.97"></a>
+<span class="sourceLineNo">098</span> *<a name="line.98"></a>
+<span class="sourceLineNo">099</span> * &lt;p&gt;TODO: Should we cache the checksums? Down in Writer#getBlockForCaching(CacheConfig) where<a name="line.99"></a>
+<span class="sourceLineNo">100</span> * we make a block to cache-on-write, there is an attempt at turning off checksums. This is not the<a name="line.100"></a>
+<span class="sourceLineNo">101</span> * only place we get blocks to cache. We also will cache the raw return from an hdfs read. In this<a name="line.101"></a>
+<span class="sourceLineNo">102</span> * case, the checksums may be present. If the cache is backed by something that doesn't do ECC,<a name="line.102"></a>
+<span class="sourceLineNo">103</span> * say an SSD, we might want to preserve checksums. For now this is open question.<a name="line.103"></a>
+<span class="sourceLineNo">104</span> * &lt;p&gt;TODO: Over in BucketCache, we save a block allocation by doing a custom serialization.<a name="line.104"></a>
+<span class="sourceLineNo">105</span> * Be sure to change it if serialization changes in here. Could we add a method here that takes an<a name="line.105"></a>
+<span class="sourceLineNo">106</span> * IOEngine and that then serializes to it rather than expose our internals over in BucketCache?<a name="line.106"></a>
+<span class="sourceLineNo">107</span> * IOEngine is in the bucket subpackage. Pull it up? Then this class knows about bucketcache. Ugh.<a name="line.107"></a>
+<span class="sourceLineNo">108</span> */<a name="line.108"></a>
+<span class="sourceLineNo">109</span>@InterfaceAudience.Private<a name="line.109"></a>
+<span class="sourceLineNo">110</span>public class HFileBlock implements Cacheable {<a name="line.110"></a>
+<span class="sourceLineNo">111</span>  private static final Log LOG = LogFactory.getLog(HFileBlock.class);<a name="line.111"></a>
+<span class="sourceLineNo">112</span><a name="line.112"></a>
+<span class="sourceLineNo">113</span>  /** Type of block. Header field 0. */<a name="line.113"></a>
+<span class="sourceLineNo">114</span>  private BlockType blockType;<a name="line.114"></a>
+<span class="sourceLineNo">115</span><a name="line.115"></a>
+<span class="sourceLineNo">116</span>  /**<a name="line.116"></a>
+<span class="sourceLineNo">117</span>   * Size on disk excluding header, including checksum. Header field 1.<a name="line.117"></a>
+<span class="sourceLineNo">118</span>   * @see Writer#putHeader(byte[], int, int, int, int)<a name="line.118"></a>
+<span class="sourceLineNo">119</span>   */<a name="line.119"></a>
+<span class="sourceLineNo">120</span>  private int onDiskSizeWithoutHeader;<a name="line.120"></a>
+<span class="sourceLineNo">121</span><a name="line.121"></a>
+<span class="sourceLineNo">122</span>  /**<a name="line.122"></a>
+<span class="sourceLineNo">123</span>   * Size of pure data. Does not include header or checksums. Header field 2.<a name="line.123"></a>
+<span class="sourceLineNo">124</span>   * @see Writer#putHeader(byte[], int, int, int, int)<a name="line.124"></a>
+<span class="sourceLineNo">125</span>   */<a name="line.125"></a>
+<span class="sourceLineNo">126</span>  private int uncompressedSizeWithoutHeader;<a name="line.126"></a>
+<span class="sourceLineNo">127</span><a name="line.127"></a>
+<span class="sourceLineNo">128</span>  /**<a name="line.128"></a>
+<span class="sourceLineNo">129</span>   * The offset of the previous block on disk. Header field 3.<a name="line.129"></a>
+<span class="sourceLineNo">130</span>   * @see Writer#putHeader(byte[], int, int, int, int)<a name="line.130"></a>
+<span class="sourceLineNo">131</span>   */<a name="line.131"></a>
+<span class="sourceLineNo">132</span>  private long prevBlockOffset;<a name="line.132"></a>
+<span class="sourceLineNo">133</span><a name="line.133"></a>
+<span class="sourceLineNo">134</span>  /**<a name="line.134"></a>
+<span class="sourceLineNo">135</span>   * Size on disk of header + data. Excludes checksum. Header field 6,<a name="line.135"></a>
+<span class="sourceLineNo">136</span>   * OR calculated from {@link #onDiskSizeWithoutHeader} when using HDFS checksum.<a name="line.136"></a>
+<span class="sourceLineNo">137</span>   * @see Writer#putHeader(byte[], int, int, int, int)<a name="line.137"></a>
+<span class="sourceLineNo">138</span>   */<a name="line.138"></a>
+<span class="sourceLineNo">139</span>  private int onDiskDataSizeWithHeader;<a name="line.139"></a>
+<span class="sourceLineNo">140</span><a name="line.140"></a>
+<span class="sourceLineNo">141</span><a name="line.141"></a>
+<span class="sourceLineNo">142</span>  /**<a name="line.142"></a>
+<span class="sourceLineNo">143</span>   * The in-memory representation of the hfile block. Can be on or offheap. Can be backed by<a name="line.143"></a>
+<span class="sourceLineNo">144</span>   * a single ByteBuffer or by many. Make no assumptions.<a name="line.144"></a>
+<span class="sourceLineNo">145</span>   *<a name="line.145"></a>
+<span class="sourceLineNo">146</span>   * &lt;p&gt;Be careful reading from this &lt;code&gt;buf&lt;/code&gt;. Duplicate and work on the duplicate or if<a name="line.146"></a>
+<span class="sourceLineNo">147</span>   * not, be sure to reset position and limit else trouble down the road.<a name="line.147"></a>
+<span class="sourceLineNo">148</span>   *<a name="line.148"></a>
+<span class="sourceLineNo">149</span>   * &lt;p&gt;TODO: Make this read-only once made.<a name="line.149"></a>
 <span class="sourceLineNo">150</span>   *<a name="line.150"></a>
-<span class="sourceLineNo">151</span>   * &lt;p&gt;TODO: Make this read-only once made.<a name="line.151"></a>
-<span class="sourceLineNo">152</span>   *<a name="line.152"></a>
-<span class="sourceLineNo">153</span>   * &lt;p&gt;We are using the ByteBuff type. ByteBuffer is not extensible yet we need to be able to have<a name="line.153"></a>
-<span class="sourceLineNo">154</span>   * a ByteBuffer-like API across multiple ByteBuffers reading from a cache such as BucketCache.<a name="line.154"></a>
-<span class="sourceLineNo">155</span>   * So, we have this ByteBuff type. Unfortunately, it is spread all about HFileBlock. Would be<a name="line.155"></a>
-<span class="sourceLineNo">156</span>   * good if could be confined to cache-use only but hard-to-do.<a name="line.156"></a>
-<span class="sourceLineNo">157</span>   */<a name="line.157"></a>
-<span class="sourceLineNo">158</span>  private ByteBuff buf;<a name="line.158"></a>
-<span class="sourceLineNo">159</span><a name="line.159"></a>
-<span class="sourceLineNo">160</span>  /** Meta data that holds meta information on the hfileblock.<a name="line.160"></a>
-<span class="sourceLineNo">161</span>   */<a name="line.161"></a>
-<span class="sourceLineNo">162</span>  private HFileContext fileContext;<a name="line.162"></a>
-<span class="sourceLineNo">163</span><a name="line.163"></a>
-<span class="sourceLineNo">164</span>  /**<a name="line.164"></a>
-<span class="sourceLineNo">165</span>   * The offset of this block in the file. Populated by the reader for<a name="line.165"></a>
-<span class="sourceLineNo">166</span>   * convenience of access. This offset is not part of the block header.<a name="line.166"></a>
-<span class="sourceLineNo">167</span>   */<a name="line.167"></a>
-<span class="sourceLineNo">168</span>  private long offset = UNSET;<a name="line.168"></a>
+<span class="sourceLineNo">151</span>   * &lt;p&gt;We are using the ByteBuff type. ByteBuffer is not extensible yet we need to be able to have<a name="line.151"></a>
+<span class="sourceLineNo">152</span>   * a ByteBuffer-like API across multiple ByteBuffers reading from a cache such as BucketCache.<a name="line.152"></a>
+<span class="sourceLineNo">153</span>   * So, we have this ByteBuff type. Unfortunately, it is spread all about HFileBlock. Would be<a name="line.153"></a>
+<span class="sourceLineNo">154</span>   * good if could be confined to cache-use only but hard-to-do.<a name="line.154"></a>
+<span class="sourceLineNo">155</span>   */<a name="line.155"></a>
+<span class="sourceLineNo">156</span>  private ByteBuff buf;<a name="line.156"></a>
+<span class="sourceLineNo">157</span><a name="line.157"></a>
+<span class="sourceLineNo">158</span>  /** Meta data that holds meta information on the hfileblock.<a name="line.158"></a>
+<span class="sourceLineNo">159</span>   */<a name="line.159"></a>
+<span class="sourceLineNo">160</span>  private HFileContext fileContext;<a name="line.160"></a>
+<span class="sourceLineNo">161</span><a name="line.161"></a>
+<span class="sourceLineNo">162</span>  /**<a name="line.162"></a>
+<span class="sourceLineNo">163</span>   * The offset of this block in the file. Populated by the reader for<a name="line.163"></a>
+<span class="sourceLineNo">164</span>   * convenience of access. This offset is not part of the block header.<a name="line.164"></a>
+<span class="sourceLineNo">165</span>   */<a name="line.165"></a>
+<span class="sourceLineNo">166</span>  private long offset = UNSET;<a name="line.166"></a>
+<span class="sourceLineNo">167</span><a name="line.167"></a>
+<span class="sourceLineNo">168</span>  private MemoryType memType = MemoryType.EXCLUSIVE;<a name="line.168"></a>
 <span class="sourceLineNo">169</span><a name="line.169"></a>
-<span class="sourceLineNo">170</span>  private MemoryType memType = MemoryType.EXCLUSIVE;<a name="line.170"></a>
-<span class="sourceLineNo">171</span><a name="line.171"></a>
-<span class="sourceLineNo">172</span>  /**<a name="line.172"></a>
-<span class="sourceLineNo">173</span>   * The on-disk size of the next block, including the header and checksums if present, obtained by<a name="line.173"></a>
-<span class="sourceLineNo">174</span>   * peeking into the first {@link HConstants#HFILEBLOCK_HEADER_SIZE} bytes of the next block's<a name="line.174"></a>
-<span class="sourceLineNo">175</span>   * header, or UNSET if unknown.<a name="line.175"></a>
-<span class="sourceLineNo">176</span>   *<a name="line.176"></a>
-<span class="sourceLineNo">177</span>   * Blocks try to carry the size of the next block to read in this data member. They will even have<a name="line.177"></a>
-<span class="sourceLineNo">178</span>   * this value when served from cache. Could save a seek in the case where we are iterating through<a name="line.178"></a>
-<span class="sourceLineNo">179</span>   * a file and some of the blocks come from cache. If from cache, then having this info to hand<a name="line.179"></a>
-<span class="sourceLineNo">180</span>   * will save us doing a seek to read the header so we can read the body of a block.<a name="line.180"></a>
-<span class="sourceLineNo">181</span>   * TODO: see how effective this is at saving seeks.<a name="line.181"></a>
-<span class="sourceLineNo">182</span>   */<a name="line.182"></a>
-<span class="sourceLineNo">183</span>  private int nextBlockOnDiskSize = UNSET;<a name="line.183"></a>
-<span class="sourceLineNo">184</span><a name="line.184"></a>
-<span class="sourceLineNo">185</span>  /**<a name="line.185"></a>
-<span class="sourceLineNo">186</span>   * On a checksum failure, do these many succeeding read requests using hdfs checksums before<a name="line.186"></a>
-<span class="sourceLineNo">187</span>   * auto-reenabling hbase checksum verification.<a name="line.187"></a>
-<span class="sourceLineNo">188</span>   */<a name="line.188"></a>
-<span class="sourceLineNo">189</span>  static final int CHECKSUM_VERIFICATION_NUM_IO_THRESHOLD = 3;<a name="line.189"></a>
-<span class="sourceLineNo">190</span><a name="line.190"></a>
-<span class="sourceLineNo">191</span>  private static int UNSET = -1;<a name="line.191"></a>
-<span class="sourceLineNo">192</span>  public static final boolean FILL_HEADER = true;<a name="line.192"></a>
-<span class="sourceLineNo">193</span>  public static final boolean DONT_FILL_HEADER = false;<a name="line.193"></a>
-<span class="sourceLineNo">194</span><a name="line.194"></a>
-<span class="sourceLineNo">195</span>  // How to get the estimate correctly? if it is a singleBB?<a name="line.195"></a>
-<span class="sourceLineNo">196</span>  public static final int MULTI_BYTE_BUFFER_HEAP_SIZE =<a name="line.196"></a>
-<span class="sourceLineNo">197</span>      (int)ClassSize.estimateBase(MultiByteBuff.class, false);<a name="line.197"></a>
-<span class="sourceLineNo">198</span><a name="line.198"></a>
-<span class="sourceLineNo">199</span>  /**<a name="line.199"></a>
-<span class="sourceLineNo">200</span>   * Space for metadata on a block that gets stored along with the block when we cache it.<a name="line.200"></a>
-<span class="sourceLineNo">201</span>   * There are a few bytes stuck on the end of the HFileBlock that we pull in from HDFS (note,<a name="line.201"></a>
-<span class="sourceLineNo">202</span>   * when we read from HDFS, we pull in an HFileBlock AND the header of the next block if one).<a name="line.202"></a>
-<span class="sourceLineNo">203</span>   * 8 bytes are offset of this block (long) in the file. Offset is important because<a name="line.203"></a>
-<span class="sourceLineNo">204</span>   * used when we remake the CacheKey when we return the block to cache when done. There is also<a name="line.204"></a>
-<span class="sourceLineNo">205</span>   * a flag on whether checksumming is being done by hbase or not. See class comment for note on<a name="line.205"></a>
-<span class="sourceLineNo">206</span>   * uncertain state of checksumming of blocks that come out of cache (should we or should we not?).<a name="line.206"></a>
-<span class="sourceLineNo">207</span>   * Finally there 4 bytes to hold the length of the next block which can save a seek on occasion.<a name="line.207"></a>
-<span class="sourceLineNo">208</span>   * &lt;p&gt;This EXTRA came in with original commit of the bucketcache, HBASE-7404. Was formerly<a name="line.208"></a>
-<span class="sourceLineNo">209</span>   * known as EXTRA_SERIALIZATION_SPACE.<a name="line.209"></a>
-<span class="sourceLineNo">210</span>   */<a name="line.210"></a>
-<span class="sourceLineNo">211</span>  static final int BLOCK_METADATA_SPACE = Bytes.SIZEOF_BYTE + Bytes.SIZEOF_LONG + Bytes.SIZEOF_INT;<a name="line.211"></a>
-<span class="sourceLineNo">212</span><a name="line.212"></a>
-<span class="sourceLineNo">213</span>  /**<a name="line.213"></a>
-<span class="sourceLineNo">214</span>   * Each checksum value is an integer that can be stored in 4 bytes.<a name="line.214"></a>
-<span class="sourceLineNo">215</span>   */<a name="line.215"></a>
-<span class="sourceLineNo">216</span>  static final int CHECKSUM_SIZE = Bytes.SIZEOF_INT;<a name="line.216"></a>
-<span class="sourceLineNo">217</span><a name="line.217"></a>
-<span class="sourceLineNo">218</span>  static final byte[] DUMMY_HEADER_NO_CHECKSUM =<a name="line.218"></a>
-<span class="sourceLineNo">219</span>      new byte[HConstants.HFILEBLOCK_HEADER_SIZE_NO_CHECKSUM];<a name="line.219"></a>
-<span class="sourceLineNo">220</span><a name="line.220"></a>
-<span class="sourceLineNo">221</span>  /**<a name="line.221"></a>
-<span class="sourceLineNo">222</span>   * Used deserializing blocks from Cache.<a name="line.222"></a>
-<span class="sourceLineNo">223</span>   *<a name="line.223"></a>
-<span class="sourceLineNo">224</span>   * &lt;code&gt;<a name="line.224"></a>
+<span class="sourceLineNo">170</span>  /**<a name="line.170"></a>
+<span class="sourceLineNo">171</span>   * The on-disk size of the next block, including the header and checksums if present, obtained by<a name="line.171"></a>
+<span class="sourceLineNo">172</span>   * peeking into the first {@link HConstants#HFILEBLOCK_HEADER_SIZE} bytes of the next block's<a name="line.172"></a>
+<span class="sourceLineNo">173</span>   * header, or UNSET if unknown.<a name="line.173"></a>
+<span class="sourceLineNo">174</span>   *<a name="line.174"></a>
+<span class="sourceLineNo">175</span>   * Blocks try to carry the size of the next block to read in this data member. They will even have<a name="line.175"></a>
+<span class="sourceLineNo">176</span>   * this value when served from cache. Could save a seek in the case where we are iterating through<a name="line.176"></a>
+<span class="sourceLineNo">177</span>   * a file and some of the blocks come from cache. If from cache, then having this info to hand<a name="line.177"></a>
+<span class="sourceLineNo">178</span>   * will save us doing a seek to read the header so we can read the body of a block.<a name="line.178"></a>
+<span class="sourceLineNo">179</span>   * TODO: see how effective this is at saving seeks.<a name="line.179"></a>
+<span class="sourceLineNo">180</span>   */<a name="line.180"></a>
+<span class="sourceLineNo">181</span>  private int nextBlockOnDiskSize = UNSET;<a name="line.181"></a>
+<span class="sourceLineNo">182</span><a name="line.182"></a>
+<span class="sourceLineNo">183</span>  /**<a name="line.183"></a>
+<span class="sourceLineNo">184</span>   * On a checksum failure, do these many succeeding read requests using hdfs checksums before<a name="line.184"></a>
+<span class="sourceLineNo">185</span>   * auto-reenabling hbase checksum verification.<a name="line.185"></a>
+<span class="sourceLineNo">186</span>   */<a name="line.186"></a>
+<span class="sourceLineNo">187</span>  static final int CHECKSUM_VERIFICATION_NUM_IO_THRESHOLD = 3;<a name="line.187"></a>
+<span class="sourceLineNo">188</span><a name="line.188"></a>
+<span class="sourceLineNo">189</span>  private static int UNSET = -1;<a name="line.189"></a>
+<span class="sourceLineNo">190</span>  public static final boolean FILL_HEADER = true;<a name="line.190"></a>
+<span class="sourceLineNo">191</span>  public static final boolean DONT_FILL_HEADER = false;<a name="line.191"></a>
+<span class="sourceLineNo">192</span><a name="line.192"></a>
+<span class="sourceLineNo">193</span>  // How to get the estimate correctly? if it is a singleBB?<a name="line.193"></a>
+<span class="sourceLineNo">194</span>  public static final int MULTI_BYTE_BUFFER_HEAP_SIZE =<a name="line.194"></a>
+<span class="sourceLineNo">195</span>      (int)ClassSize.estimateBase(MultiByteBuff.class, false);<a name="line.195"></a>
+<span class="sourceLineNo">196</span><a name="line.196"></a>
+<span class="sourceLineNo">197</span>  /**<a name="line.197"></a>
+<span class="sourceLineNo">198</span>   * Space for metadata on a block that gets stored along with the block when we cache it.<a name="line.198"></a>
+<span class="sourceLineNo">199</span>   * There are a few bytes stuck on the end of the HFileBlock that we pull in from HDFS (note,<a name="line.199"></a>
+<span class="sourceLineNo">200</span>   * when we read from HDFS, we pull in an HFileBlock AND the header of the next block if one).<a name="line.200"></a>
+<span class="sourceLineNo">201</span>   * 8 bytes are offset of this block (long) in the file. Offset is important because<a name="line.201"></a>
+<span class="sourceLineNo">202</span>   * used when we remake the CacheKey when we return the block to cache when done. There is also<a name="line.202"></a>
+<span class="sourceLineNo">203</span>   * a flag on whether checksumming is being done by hbase or not. See class comment for note on<a name="line.203"></a>
+<span class="sourceLineNo">204</span>   * uncertain state of checksumming of blocks that come out of cache (should we or should we not?).<a name="line.204"></a>
+<span class="sourceLineNo">205</span>   * Finally there 4 bytes to hold the length of the next block which can save a seek on occasion.<a name="line.205"></a>
+<span class="sourceLineNo">206</span>   * &lt;p&gt;This EXTRA came in with original commit of the bucketcache, HBASE-7404. Was formerly<a name="line.206"></a>
+<span class="sourceLineNo">207</span>   * known as EXTRA_SERIALIZATION_SPACE.<a name="line.207"></a>
+<span class="sourceLineNo">208</span>   */<a name="line.208"></a>
+<span class="sourceLineNo">209</span>  static final int BLOCK_METADATA_SPACE = Bytes.SIZEOF_BYTE + Bytes.SIZEOF_LONG + Bytes.SIZEOF_INT;<a name="line.209"></a>
+<span class="sourceLineNo">210</span><a name="line.210"></a>
+<span class="sourceLineNo">211</span>  /**<a name="line.211"></a>
+<span class="sourceLineNo">212</span>   * Each checksum value is an integer that can be stored in 4 bytes.<a name="line.212"></a>
+<span class="sourceLineNo">213</span>   */<a name="line.213"></a>
+<span class="sourceLineNo">214</span>  static final int CHECKSUM_SIZE = Bytes.SIZEOF_INT;<a name="line.214"></a>
+<span class="sourceLineNo">215</span><a name="line.215"></a>
+<span class="sourceLineNo">216</span>  static final byte[] DUMMY_HEADER_NO_CHECKSUM =<a name="line.216"></a>
+<span class="sourceLineNo">217</span>      new byte[HConstants.HFILEBLOCK_HEADER_SIZE_NO_CHECKSUM];<a name="line.217"></a>
+<span class="sourceLineNo">218</span><a name="line.218"></a>
+<span class="sourceLineNo">219</span>  /**<a name="line.219"></a>
+<span class="sourceLineNo">220</span>   * Used deserializing blocks from Cache.<a name="line.220"></a>
+<span class="sourceLineNo">221</span>   *<a name="line.221"></a>
+<span class="sourceLineNo">222</span>   * &lt;code&gt;<a name="line.222"></a>
+<span class="sourceLineNo">223</span>   * ++++++++++++++<a name="line.223"></a>
+<span class="sourceLineNo">224</span>   * + HFileBlock +<a name="line.224"></a>
 <span class="sourceLineNo">225</span>   * ++++++++++++++<a name="line.225"></a>
-<span class="sourceLineNo">226</span>   * + HFileBlock +<a name="line.226"></a>
+<span class="sourceLineNo">226</span>   * + Checksums  + &lt;= Optional<a name="line.226"></a>
 <span class="sourceLineNo">227</span>   * ++++++++++++++<a name="line.227"></a>
-<span class="sourceLineNo">228</span>   * + Checksums  + &lt;= Optional<a name="line.228"></a>
+<span class="sourceLineNo">228</span>   * + Metadata!  +<a name="line.228"></a>
 <span class="sourceLineNo">229</span>   * ++++++++++++++<a name="line.229"></a>
-<span class="sourceLineNo">230</span>   * + Metadata!  +<a name="line.230"></a>
-<span class="sourceLineNo">231</span>   * ++++++++++++++<a name="line.231"></a>
-<span class="sourceLineNo">232</span>   * &lt;/code&gt;<a name="line.232"></a>
-<span class="sourceLineNo">233</span>   * @see #serialize(ByteBuffer)<a name="line.233"></a>
-<span class="sourceLineNo">234</span>   */<a name="line.234"></a>
-<span class="sourceLineNo">235</span>  static final CacheableDeserializer&lt;Cacheable&gt; BLOCK_DESERIALIZER =<a name="line.235"></a>
-<span class="sourceLineNo">236</span>      new CacheableDeserializer&lt;Cacheable&gt;() {<a name="line.236"></a>
-<span class="sourceLineNo">237</span>        public HFileBlock deserialize(ByteBuff buf, boolean reuse, MemoryType memType)<a name="line.237"></a>
-<span class="sourceLineNo">238</span>        throws IOException {<a name="line.238"></a>
-<span class="sourceLineNo">239</span>          // The buf has the file block followed by block metadata.<a name="line.239"></a>
-<span class="sourceLineNo">240</span>          // Set limit to just before the BLOCK_METADATA_SPACE then rewind.<a name="line.240"></a>
-<span class="sourceLineNo">241</span>          buf.limit(buf.limit() - BLOCK_METADATA_SPACE).rewind();<a name="line.241"></a>
-<span class="sourceLineNo">242</span>          // Get a new buffer to pass the HFileBlock for it to 'own'.<a name="line.242"></a>
-<span class="sourceLineNo">243</span>          ByteBuff newByteBuff;<a name="line.243"></a>
-<span class="sourceLineNo">244</span>          if (reuse) {<a name="line.244"></a>
-<span class="sourceLineNo">245</span>            newByteBuff = buf.slice();<a name="line.245"></a>
-<span class="sourceLineNo">246</span>          } else {<a name="line.246"></a>
-<span class="sourceLineNo">247</span>            int len = buf.limit();<a name="line.247"></a>
-<span class="sourceLineNo">248</span>            newByteBuff = new SingleByteBuff(ByteBuffer.allocate(len));<a name="line.248"></a>
-<span class="sourceLineNo">249</span>            newByteBuff.put(0, buf, buf.position(), len);<a name="line.249"></a>
-<span class="sourceLineNo">250</span>          }<a name="line.250"></a>
-<span class="sourceLineNo">251</span>          // Read out the BLOCK_METADATA_SPACE content and shove into our HFileBlock.<a name="line.251"></a>
-<span class="sourceLineNo">252</span>          buf.position(buf.limit());<a name="line.252"></a>
-<span class="sourceLineNo">253</span>          buf.limit(buf.limit() + HFileBlock.BLOCK_METADATA_SPACE);<a name="line.253"></a>
-<span class="sourceLineNo">254</span>          boolean usesChecksum = buf.get() == (byte)1;<a name="line.254"></a>
-<span class="sourceLineNo">255</span>          long offset = buf.getLong();<a name="line.255"></a>
-<span class="sourceLineNo">256</span>          int nextBlockOnDiskSize = buf.getInt();<a name="line.256"></a>
-<span class="sourceLineNo">257</span>          HFileBlock hFileBlock =<a name="line.257"></a>
-<span class="sourceLineNo">258</span>              new HFileBlock(newByteBuff, usesChecksum, memType, offset, nextBlockOnDiskSize, null);<a name="line.258"></a>
-<span class="sourceLineNo">259</span>          return hFileBlock;<a name="line.259"></a>
-<span class="sourceLineNo">260</span>        }<a name="line.260"></a>
-<span class="sourceLineNo">261</span><a name="line.261"></a>
-<span class="sourceLineNo">262</span>        @Override<a name="line.262"></a>
-<span class="sourceLineNo">263</span>        public int getDeserialiserIdentifier() {<a name="line.263"></a>
-<span class="sourceLineNo">264</span>          return DESERIALIZER_IDENTIFIER;<a name="line.264"></a>
-<span class="sourceLineNo">265</span>        }<a name="line.265"></a>
-<span class="sourceLineNo">266</span><a name="line.266"></a>
-<span class="sourceLineNo">267</span>        @Override<a name="line.267"></a>
-<span class="sourceLineNo">268</span>        public HFileBlock deserialize(ByteBuff b) throws IOException {<a name="line.268"></a>
-<span class="sourceLineNo">269</span>          // Used only in tests<a name="line.269"></a>
-<span class="sourceLineNo">270</span>          return deserialize(b, false, MemoryType.EXCLUSIVE);<a name="line.270"></a>
-<span class="sourceLineNo">271</span>        }<a name="line.271"></a>
-<span class="sourceLineNo">272</span>      };<a name="line.272"></a>
-<span class="sourceLineNo">273</span><a name="line.273"></a>
-<span class="sourceLineNo">274</span>  private static final int DESERIALIZER_IDENTIFIER;<a name="line.274"></a>
-<span class="sourceLineNo">275</span>  static {<a name="line.275"></a>
-<span class="sourceLineNo">276</span>    DESERIALIZER_IDENTIFIER =<a name="line.276"></a>
-<span class="sourceLineNo">277</span>        CacheableDeserializerIdManager.registerDeserializer(BLOCK_DESERIALIZER);<a name="line.277"></a>
-<span class="sourceLineNo">278</span>  }<a name="line.278"></a>
-<span class="sourceLineNo">279</span><a name="line.279"></a>
-<span class="sourceLineNo">280</span>  // Todo: encapsulate Header related logic in this inner class.<a name="line.280"></a>
-<span class="sourceLineNo">281</span>  static class Header {<a name="line.281"></a>
-<span class="sourceLineNo">282</span>    // Format of header is:<a name="line.282"></a>
-<span class="sourceLineNo">283</span>    // 8 bytes - block magic<a name="line.283"></a>
-<span class="sourceLineNo">284</span>    // 4 bytes int - onDiskSizeWithoutHeader<a name="line.284"></a>
-<span class="sourceLineNo">285</span>    // 4 bytes int - uncompressedSizeWithoutHeader<a name="line.285"></a>
-<span class="sourceLineNo">286</span>    // 8 bytes long - prevBlockOffset<a name="line.286"></a>
-<span class="sourceLineNo">287</span>    // The following 3 are only present if header contains checksum information<a name="line.287"></a>
-<span class="sourceLineNo">288</span>    // 1 byte - checksum type<a name="line.288"></a>
-<span class="sourceLineNo">289</span>    // 4 byte int - bytes per checksum<a name="line.289"></a>
-<span class="sourceLineNo">290</span>    // 4 byte int - onDiskDataSizeWithHeader<a name="line.290"></a>
-<span class="sourceLineNo">291</span>    static int BLOCK_MAGIC_INDEX = 0;<a name="line.291"></a>
-<span class="sourceLineNo">292</span>    static int ON_DISK_SIZE_WITHOUT_HEADER_INDEX = 8;<a name="line.292"></a>
-<span class="sourceLineNo">293</span>    static int UNCOMPRESSED_SIZE_WITHOUT_HEADER_INDEX = 12;<a name="line.293"></a>
-<span class="sourceLineNo">294</span>    static int PREV_BLOCK_OFFSET_INDEX = 16;<a name="line.294"></a>
-<span class="sourceLineNo">295</span>    static int CHECKSUM_TYPE_INDEX = 24;<a name="line.295"></a>
-<span class="sourceLineNo">296</span>    static int BYTES_PER_CHECKSUM_INDEX = 25;<a name="line.296"></a>
-<span class="sourceLineNo">297</span>    static int ON_DISK_DATA_SIZE_WITH_HEADER_INDEX = 29;<a name="line.297"></a>
-<span class="sourceLineNo">298</span>  }<a name="line.298"></a>
-<span class="sourceLineNo">299</span><a name="line.299"></a>
-<span class="sourceLineNo">300</span>  /**<a name="line.300"></a>
-<span class="sourceLineNo">301</span>   * Copy constructor. Creates a shallow copy of {@code that}'s buffer.<a name="line.301"></a>
-<span class="sourceLineNo">302</span>   */<a name="line.302"></a>
-<span class="sourceLineNo">303</span>  private HFileBlock(HFileBlock that) {<a name="line.303"></a>
-<span class="sourceLineNo">304</span>    this(that, false);<a name="line.304"></a>
-<span class="sourceLineNo">305</span>  }<a name="line.305"></a>
-<span class="sourceLineNo">306</span><a name="line.306"></a>
-<span class="sourceLineNo">307</span>  /**<a name="line.307"></a>
-<span class="sourceLineNo">308</span>   * Copy constructor. Creates a shallow/deep copy of {@code that}'s buffer as per the boolean<a name="line.308"></a>
-<span class="sourceLineNo">309</span>   * param.<a name="line.309"></a>
-<span class="sourceLineNo">310</span>   */<a name="line.310"></a>
-<span class="sourceLineNo">311</span>  private HFileBlock(HFileBlock that,boolean bufCopy) {<a name="line.311"></a>
-<span class="sourceLineNo">312</span>    this.blockType = that.blockType;<a name="line.312"></a>
-<span class="sourceLineNo">313</span>    this.onDiskSizeWithoutHeader = that.onDiskSizeWithoutHeader;<a name="line.313"></a>
-<span class="sourceLineNo">314</span>    this.uncompressedSizeWithoutHeader = that.uncompressedSizeWithoutHeader;<a name="line.314"></a>
-<span class="sourceLineNo">315</span>    this.prevBlockOffset = that.prevBlockOffset;<a name="line.315"></a>
-<span class="sourceLineNo">316</span>    if (bufCopy) {<a name="line.316"></a>
-<span class="sourceLineNo">317</span>      this.buf = new SingleByteBuff(ByteBuffer.wrap(that.buf.toBytes(0, that.buf.limit())));<a name="line.317"></a>
-<span class="sourceLineNo">318</span>    } else {<a name="line.318"></a>
-<span class="sourceLineNo">319</span>      this.buf = that.buf.duplicate();<a name="line.319"></a>
-<span class="sourceLineNo">320</span>    }<a name="line.320"></a>
-<span class="sourceLineNo">321</span>    this.offset = that.offset;<a name="line.321"></a>
-<span class="sourceLineNo">322</span>    this.onDiskDataSizeWithHeader = that.onDiskDataSizeWithHeader;<a name="line.322"></a>
-<span class="sourceLineNo">323</span>    this.fileContext = that.fileContext;<a name="line.323"></a>
-<span class="sourceLineNo">324</span>    this.nextBlockOnDiskSize = that.nextBlockOnDiskSize;<a name="line.324"></a>
-<span class="sourceLineNo">325</span>  }<a name="line.325"></a>
-<span class="sourceLineNo">326</span><a name="line.326"></a>
-<span class="sourceLineNo">327</span>  /**<a name="line.327"></a>
-<span class="sourceLineNo">328</span>   * Creates a new {@link HFile} block from the given fields. This constructor<a name="line.328"></a>
-<span class="sourceLineNo">329</span>   * is used only while writing blocks and caching,<a name="line.329"></a>
-<span class="sourceLineNo">330</span>   * and is sitting in a byte buffer and we want to stuff the block into cache.<a name="line.330"></a>
-<span class="sourceLineNo">331</span>   * See {@link Writer#getBlockForCaching(CacheConfig)}.<a name="line.331"></a>
-<span class="sourceLineNo">332</span>   *<a name="line.332"></a>
-<span class="sourceLineNo">333</span>   * &lt;p&gt;TODO: The caller presumes no checksumming<a name="line.333"></a>
-<span class="sourceLineNo">334</span>   * required of this block instance since going into cache; checksum already verified on<a name="line.334"></a>
-<span class="sourceLineNo">335</span>   * underlying block data pulled in from filesystem. Is that correct? What if cache is SSD?<a name="line.335"></a>
-<span class="sourceLineNo">336</span>   *<a name="line.336"></a>
-<span class="sourceLineNo">337</span>   * @param blockType the type of this block, see {@link BlockType}<a name="line.337"></a>
-<span class="sourceLineNo">338</span>   * @param onDiskSizeWithoutHeader see {@link #onDiskSizeWithoutHeader}<a name="line.338"></a>
-<span class="sourceLineNo">339</span>   * @param uncompressedSizeWithoutHeader see {@link #uncompressedSizeWithoutHeader}<a name="line.339"></a>
-<span class="sourceLineNo">340</span>   * @param prevBlockOffset see {@link #prevBlockOffset}<a name="line.340"></a>
-<span class="sourceLineNo">341</span>   * @param b block header ({@link HConstants#HFILEBLOCK_HEADER_SIZE} bytes)<a name="line.341"></a>
-<span class="sourceLineNo">342</span>   * @param fillHeader when true, write the first 4 header fields into passed buffer.<a name="line.342"></a>
-<span class="sourceLineNo">343</span>   * @param offset the file offset the block was read from<a name="line.343"></a>
-<span class="sourceLineNo">344</span>   * @param onDiskDataSizeWithHeader see {@link #onDiskDataSizeWithHeader}<a name="line.344"></a>
-<span class="sourceLineNo">345</span>   * @param fileContext HFile meta data<a name="line.345"></a>
-<span class="sourceLineNo">346</span>   */<a name="line.346"></a>
-<span class="sourceLineNo">347</span>  HFileBlock(BlockType blockType, int onDiskSizeWithoutHeader, int uncompressedSizeWithoutHeader,<a name="line.347"></a>
-<span class="sourceLineNo">348</span>      long prevBlockOffset, ByteBuffer b, boolean fillHeader, long offset,<a name="line.348"></a>
-<span class="sourceLineNo">349</span>      final int nextBlockOnDiskSize, int onDiskDataSizeWithHeader, HFileContext fileContext) {<a name="line.349"></a>
-<span class="sourceLineNo">350</span>    init(blockType, onDiskSizeWithoutHeader, uncompressedSizeWithoutHeader,<a name="line.350"></a>
-<span class="sourceLineNo">351</span>        prevBlockOffset, offset, onDiskDataSizeWithHeader, nextBlockOnDiskSize, fileContext);<a name="line.351"></a>
-<span class="sourceLineNo">352</span>    this.buf = new SingleByteBuff(b);<a name="line.352"></a>
-<span class="sourceLineNo">353</span>    if (fillHeader) {<a name="line.353"></a>
-<span class="sourceLineNo">354</span>      overwriteHeader();<a name="line.354"></a>
-<span class="sourceLineNo">355</span>    }<a name="line.355"></a>
-<span class="sourceLineNo">356</span>    this.buf.rewind();<a name="line.356"></a>
-<span class="sourceLineNo">357</span>  }<a name="line.357"></a>
-<span class="sourceLineNo">358</span><a name="line.358"></a>
-<span class="sourceLineNo">359</span>  /**<a name="line.359"></a>
-<span class="sourceLineNo">360</span>   * Creates a block from an existing buffer starting with a header. Rewinds<a name="line.360"></a>
-<span class="sourceLineNo">361</span>   * and takes ownership of the buffer. By definition of rewind, ignores the<a name="line.361"></a>
-<span class="sourceLineNo">362</span>   * buffer position, but if you slice the buffer beforehand, it will rewind<a name="line.362"></a>
-<span class="sourceLineNo">363</span>   * to that point.<a name="line.363"></a>
-<span class="sourceLineNo">364</span>   * @param buf Has header, content, and trailing checksums if present.<a name="line.364"></a>
-<span class="sourceLineNo">365</span>   */<a name="line.365"></a>
-<span class="sourceLineNo">366</span>  HFileBlock(ByteBuff buf, boolean usesHBaseChecksum, MemoryType memType, final long offset,<a name="line.366"></a>
-<span class="sourceLineNo">367</span>      final int nextBlockOnDiskSize, HFileContext fileContext) throws IOException {<a name="line.367"></a>
-<span class="sourceLineNo">368</span>    buf.rewind();<a name="line.368"></a>
-<span class="sourceLineNo">369</span>    final BlockType blockType = BlockType.read(buf);<a name="line.369"></a>
-<span class="sourceLineNo">370</span>    final int onDiskSizeWithoutHeader = buf.getInt(Header.ON_DISK_SIZE_WITHOUT_HEADER_INDEX);<a name="line.370"></a>
-<span class="sourceLineNo">371</span>    final int uncompressedSizeWithoutHeader =<a name="line.371"></a>
-<span class="sourceLineNo">372</span>        buf.getInt(Header.UNCOMPRESSED_SIZE_WITHOUT_HEADER_INDEX);<a name="line.372"></a>
-<span class="sourceLineNo">373</span>    final long prevBlockOffset = buf.getLong(Header.PREV_BLOCK_OFFSET_INDEX);<a name="line.373"></a>
-<span class="sourceLineNo">374</span>    // This constructor is called when we deserialize a block from cache and when we read a block in<a name="line.374"></a>
-<span class="sourceLineNo">375</span>    // from the fs. fileCache is null when deserialized from cache so need to make up one.<a name="line.375"></a>
-<span class="sourceLineNo">376</span>    HFileContextBuilder fileContextBuilder = fileContext != null?<a name="line.376"></a>
-<span class="sourceLineNo">377</span>        new HFileContextBuilder(fileContext): new HFileContextBuilder();<a name="line.377"></a>
-<span class="sourceLineNo">378</span>    fileContextBuilder.withHBaseCheckSum(usesHBaseChecksum);<a name="line.378"></a>
-<span class="sourceLineNo">379</span>    int onDiskDataSizeWithHeader;<a name="line.379"></a>
-<span class="sourceLineNo">380</span>    if (usesHBaseChecksum) {<a name="line.380"></a>
-<span class="sourceLineNo">381</span>      byte checksumType = buf.get(Header.CHECKSUM_TYPE_INDEX);<a name="line.381"></a>
-<span class="sourceLineNo">382</span>      int bytesPerChecksum = buf.getInt(Header.BYTES_PER_CHECKSUM_INDEX);<a name="line.382"></a>
-<span class="sourceLineNo">383</span>      onDiskDataSizeWithHeader = buf.getInt(Header.ON_DISK_DATA_SIZE_WITH_HEADER_INDEX);<a name="line.383"></a>
-<span class="sourceLineNo">384</span>      // Use the checksum type and bytes per checksum from header, not from filecontext.<a name="line.384"></a>
-<span class="sourceLineNo">385</span>      fileContextBuilder.withChecksumType(ChecksumType.codeToType(checksumType));<a name="line.385"></a>
-<span class="sourceLineNo">386</span>      fileContextBuilder.withBytesPerCheckSum(bytesPerChecksum);<a name="line.386"></a>
-<span class="sourceLineNo">387</span>    } else {<a name="line.387"></a>
-<span class="sourceLineNo">388</span>      fileContextBuilder.withChecksumType(ChecksumType.NULL);<a name="line.388"></a>
-<span class="sourceLineNo">389</span>      fileContextBuilder.withBytesPerCheckSum(0);<a name="line.389"></a>
-<span class="sourceLineNo">390</span>      // Need to fix onDiskDataSizeWithHeader; there are not checksums after-block-data<a name="line.390"></a>
-<span class="sourceLineNo">391</span>      onDiskDataSizeWithHeader = onDiskSizeWithoutHeader + headerSize(usesHBaseChecksum);<a name="line.391"></a>
-<span class="sourceLineNo">392</span>    }<a name="line.392"></a>
-<span class="sourceLineNo">393</span>    fileContext = fileContextBuilder.build();<a name="line.393"></a>
-<span class="sourceLineNo">394</span>    assert usesHBaseChecksum == fileContext.isUseHBaseChecksum();<a name="line.394"></a>
-<span class="sourceLineNo">395</span>    init(blockType, onDiskSizeWithoutHeader, uncompressedSizeWithoutHeader,<a name="line.395"></a>
-<span class="sourceLineNo">396</span>        prevBlockOffset, offset, onDiskDataSizeWithHeader, nextBlockOnDiskSize, fileContext);<a name="line.396"></a>
-<span class="sourceLineNo">397</span>    this.memType = memType;<a name="line.397"></a>
-<span class="sourceLineNo">398</span>    this.offset = offset;<a name="line.398"></a>
-<span class="sourceLineNo">399</span>    this.buf = buf;<a name="line.399"></a>
-<span class="sourceLineNo">400</span>    this.buf.rewind();<a name="line.400"></a>
-<span class="sourceLineNo">401</span>  }<a name="line.401"></a>
-<span class="sourceLineNo">402</span><a name="line.402"></a>
-<span class="sourceLineNo">403</span>  /**<a name="line.403"></a>
-<span class="sourceLineNo">404</span>   * Called from constructors.<a name="line.404"></a>
-<span class="sourceLineNo">405</span>   */<a name="line.405"></a>
-<span class="sourceLineNo">406</span>  private void init(BlockType blockType, int onDiskSizeWithoutHeader,<a name="line.406"></a>
-<span class="sourceLineNo">407</span>      int uncompressedSizeWithoutHeader, long prevBlockOffset,<a name="line.407"></a>
-<span class="sourceLineNo">408</span>      long offset, int onDiskDataSizeWithHeader, final int nextBlockOnDiskSize,<a name="line.408"></a>
-<span class="sourceLineNo">409</span>      HFileContext fileContext) {<a name="line.409"></a>
-<span class="sourceLineNo">410</span>    this.blockType = blockType;<a name="line.410"></a>
-<span class="sourceLineNo">411</span>    this.onDiskSizeWithoutHeader = onDiskSizeWithoutHeader;<a name="line.411"></a>
-<span class="sourceLineNo">412</span>    this.uncompressedSizeWithoutHeader = uncompressedSizeWithoutHeader;<a name="line.412"></a>
-<span class="sourceLineNo">413</span>    this.prevBlockOffset = prevBlockOffset;<a name="line.413"></a>
-<span class="sourceLineNo">414</span>    this.offset = offset;<a name="line.414"></a>
-<span class="sourceLineNo">415</span>    this.onDiskDataSizeWithHeader = onDiskDataSizeWithHeader;<a name="line.415"></a>
-<span class="sourceLineNo">416</span>    this.nextBlockOnDiskSize = nextBlockOnDiskSize;<a name="line.416"></a>
-<span class="sourceLineNo">417</span>    this.fileContext = fileContext;<a name="line.417"></a>
-<span class="sourceLineNo">418</span>  }<a name="line.418"></a>
-<span class="sourceLineNo">419</span><a name="line.419"></a>
-<span class="sourceLineNo">420</span>  /**<a name="line.420"></a>
-<span class="sourceLineNo">421</span>   * Parse total ondisk size including header and checksum.<a name="line.421"></a>
-<span class="sourceLineNo">422</span>   * @param headerBuf Header ByteBuffer. Presumed exact size of header.<a name="line.422"></a>
-<span class="sourceLineNo">423</span>   * @param verifyChecksum true if checksum verification is in use.<a name="line.423"></a>
-<span class="sourceLineNo">424</span>   * @return Size of the block with header included.<a name="line.424"></a>
-<span class="sourceLineNo">425</span>   */<a name="line.425"></a>
-<span class="sourceLineNo">426</span>  private static int getOnDiskSizeWithHeader(final ByteBuffer headerBuf, boolean verifyChecksum) {<a name="line.426"></a>
-<span class="sourceLineNo">427</span>    return headerBuf.getInt(Header.ON_DISK_SIZE_WITHOUT_HEADER_INDEX) +<a name="line.427"></a>
-<span class="sourceLineNo">428</span>      headerSize(verifyChecksum);<a name="line.428"></a>
-<span class="sourceLineNo">429</span>  }<a name="line.429"></a>
-<span class="sourceLineNo">430</span><a name="line.430"></a>
-<span class="sourceLineNo">431</span>  /**<a name="line.431"></a>
-<span class="sourceLineNo">432</span>   * @return the on-disk size of the next block (including the header size and any checksums if<a name="line.432"></a>
-<span class="sourceLineNo">433</span>   * present) read by peeking into the next block's header; use as a hint when doing<a name="line.433"></a>
-<span class="sourceLineNo">434</span>   * a read of the next block when scanning or running over a file.<a name="line.434"></a>
-<span class="sourceLineNo">435</span>   */<a name="line.435"></a>
-<span class="sourceLineNo">436</span>  public int getNextBlockOnDiskSize() {<a name="line.436"></a>
-<span class="sourceLineNo">437</span>    return nextBlockOnDiskSize;<a name="line.437"></a>
-<span class="sourceLineNo">438</span>  }<a name="line.438"></a>
-<span class="sourceLineNo">439</span><a name="line.439"></a>
-<span class="sourceLineNo">440</span>  public BlockType getBlockType() {<a name="line.440"></a>
-<span class="sourceLineNo">441</span>    return blockType;<a name="line.441"></a>
-<span class="sourceLineNo">442</span>  }<a name="line.442"></a>
-<span class="sourceLineNo">443</span><a name="line.443"></a>
-<span class="sourceLineNo">444</span>  /** @return get data block encoding id that was used to encode this block */<a name="line.444"></a>
-<span class="sourceLineNo">445</span>  public short getDataBlockEncodingId() {<a name="line.445"></a>
-<span class="sourceLineNo">446</span>    if (blockType != BlockType.ENCODED_DATA) {<a name="line.446"></a>
-<span class="sourceLineNo">447</span>      throw new IllegalArgumentException("Querying encoder ID of a block " +<a name="line.447"></a>
-<span class="sourceLineNo">448</span>          "of type other than " + BlockType.ENCODED_DATA + ": " + blockType);<a name="line.448"></a>
-<span class="sourceLineNo">449</span>    }<a name="line.449"></a>
-<span class="sourceLineNo">450</span>    return buf.getShort(headerSize());<a name="line.450"></a>
-<span class="sourceLineNo">451</span>  }<a name="line.451"></a>
-<span class="sourceLineNo">452</span><a name="line.452"></a>
-<span class="sourceLineNo">453</span>  /**<a name="line.453"></a>
-<span class="sourceLineNo">454</span>   * @return the on-disk size of header + data part + checksum.<a name="line.454"></a>
-<span class="sourceLineNo">455</span>   */<a name="line.455"></a>
-<span class="sourceLineNo">456</span>  public int getOnDiskSizeWithHeader() {<a name="line.456"></a>
-<span class="sourceLineNo">457</span>    return onDiskSizeWithoutHeader + headerSize();<a name="line.457"></a>
-<span class="sourceLineNo">458</span>  }<a name="line.458"></a>
-<span class="sourceLineNo">459</span><a name="line.459"></a>
-<span class="sourceLineNo">460</span>  /**<a name="line.460"></a>
-<span class="sourceLineNo">461</span>   * @return the on-disk size of the data part + checksum (header excluded).<a name="line.461"></a>
-<span class="sourceLineNo">462</span>   */<a name="line.462"></a>
-<span class="sourceLineNo">463</span>  int getOnDiskSizeWithoutHeader() {<a name="line.463"></a>
-<span class="sourceLineNo">464</span>    return onDiskSizeWithoutHeader;<a name="line.464"></a>
-<span class="sourceLineNo">465</span>  }<a name="line.465"></a>
-<span class="sourceLineNo">466</span><a name="line.466"></a>
-<span class="sourceLineNo">467</span>  /**<a name="line.467"></a>
-<span class="sourceLineNo">468</span>   * @return the uncompressed size of data part (header and checksum excluded).<a name="line.468"></a>
-<span class="sourceLineNo">469</span>   */<a name="line.469"></a>
-<span class="sourceLineNo">470</span>   int getUncompressedSizeWithoutHeader() {<a name="line.470"></a>
-<span class="sourceLineNo">471</span>    return uncompressedSizeWithoutHeader;<a name="line.471"></a>
-<span class="sourceLineNo">472</span>  }<a name="line.472"></a>
-<span class="sourceLineNo">473</span><a name="line.473"></a>
-<span class="sourceLineNo">474</span>  /**<a name="line.474"></a>
-<span class="sourceLineNo">475</span>   * @return the offset of the previous block of the same type in the file, or<a name="line.475"></a>
-<span class="sourceLineNo">476</span>   *         -1 if unknown<a name="line.476"></a>
-<span class="sourceLineNo">477</span>   */<a name="line.477"></a>
-<span class="sourceLineNo">478</span>  long getPrevBlockOffset() {<a name="line.478"></a>
-<span class="sourceLineNo">479</span>    return prevBlockOffset;<a name="line.479"></a>
-<span class="sourceLineNo">480</span>  }<a name="line.480"></a>
-<span class="sourceLineNo">481</span><a name="line.481"></a>
-<span class="sourceLineNo">482</span>  /**<a name="line.482"></a>
-<span class="sourceLineNo">483</span>   * Rewinds {@code buf} and writes first 4 header fields. {@code buf} position<a name="line.483"></a>
-<span class="sourceLineNo">484</span>   * is modified as side-effect.<a name="line.484"></a>
-<span class="sourceLineNo">485</span>   */<a name="line.485"></a>
-<span class="sourceLineNo">486</span>  private void overwriteHeader() {<a name="line.486"></a>
-<span class="sourceLineNo">487</span>    buf.rewind();<a name="line.487"></a>
-<span class="sourceLineNo">488</span>    blockType.write(buf);<a name="line.488"></a>
-<span class="sourceLineNo">489</span>    buf.putInt(onDiskSizeWithoutHeader);<a name="line.489"></a>
-<span class="sourceLineNo">490</span>    buf.putInt(uncompressedSizeWithoutHeader);<a name="line.490"></a>
-<span class="sourceLineNo">491</span>    buf.putLong(prevBlockOffset);<a name="line.491"></a>
-<span class="sourceLineNo">492</span>    if (this.fileContext.isUseHBaseChecksum()) {<a name="line.492"></a>
-<span class="sourceLineNo">493</span>      buf.put(fileContext.getChecksumType().getCode());<a name="line.493"></a>
-<span class="sourceLineNo">494</span>      buf.putInt(fileContext.getBytesPerChecksum());<a name="line.494"></a>
-<span class="sourceLineNo">495</span>      buf.putInt(onDiskDataSizeWithHeader);<a name="line.495"></a>
-<span class="sourceLineNo">496</span>    }<a name="line.496"></a>
-<span class="sourceLineNo">497</span>  }<a name="line.497"></a>
-<span class="sourceLineNo">498</span><a name="line.498"></a>
-<span class="sourceLineNo">499</span>  /**<a name="line.499"></a>
-<span class="sourceLineNo">500</span>   * Returns a buffer that does not include the header or checksum.<a name="line.500"></a>
-<span class="sourceLineNo">501</span>   *<a name="line.501"></a>
-<span class="sourceLineNo">502</span>   * @return the buffer with header skipped and checksum omitted.<a name="line.502"></a>
-<span class="sourceLineNo">503</span>   */<a name="line.503"></a>
-<span class="sourceLineNo">504</span>  public ByteBuff getBufferWithoutHeader() {<a name="line.504"></a>
-<span class="sourceLineNo">505</span>    ByteBuff dup = getBufferReadOnly();<a name="line.505"></a>
-<span class="sourceLineNo">506</span>    // Now set it up so Buffer spans content only -- no header or no checksums.<a name="line.506"></a>
-<span class="sourceLineNo">507</span>    return dup.position(headerSize()).limit(buf.limit() - totalChecksumBytes()).slice();<a name="line.507"></a>
-<span class="sourceLineNo">508</span>  }<a name="line.508"></a>
-<span class="sourceLineNo">509</span><a name="line.509"></a>
-<span class="sourceLineNo">510</span>  /**<a name="line.510"></a>
-<span class="sourceLineNo">511</span>   * Returns a read-only duplicate of the buffer this block stores internally ready to be read.<a name="line.511"></a>
-<span class="sourceLineNo">512</span>   * Clients must not modify the buffer object though they may set position and limit on the<a name="line.512"></a>
-<span class="sourceLineNo">513</span>   * returned buffer since we pass back a duplicate. This method has to be public because it is used<a name="line.513"></a>
-<span class="sourceLineNo">514</span>   * in {@link CompoundBloomFilter} to avoid object creation on every Bloom<a name="line.514"></a>
-<span class="sourceLineNo">515</span>   * filter lookup, but has to be used with caution. Buffer holds header, block content,<a name="line.515"></a>
-<span class="sourceLineNo">516</span>   * and any follow-on checksums if present.<a name="line.516"></a>
-<span class="sourceLineNo">517</span>   *<a name="line.517"></a>
-<span class="sourceLineNo">518</span>   * @return the buffer of this block for read-only operations<a name="line.518"></a>
-<span class="sourceLineNo">519</span>   */<a name="line.519"></a>
-<span class="sourceLineNo">520</span>  public ByteBuff getBufferReadOnly() {<a name="line.520"></a>
-<span class="sourceLineNo">521</span>    // TODO: ByteBuf does not support asReadOnlyBuffer(). Fix.<a name="line.521"></a>
-<span class="sourceLineNo">522</span>    ByteBuff dup = this.buf.duplicate();<a name="line.522"></a>
-<span class="sourceLineNo">523</span>    assert dup.position() == 0;<a name="line.523"></a>
-<span class="sourceLineNo">524</span>    return dup;<a name="line.524"></a>
-<span class="sourceLineNo">525</span>  }<a name="line.525"></a>
-<span class="sourceLineNo">526</span><a name="line.526"></a>
-<span class="sourceLineNo">527</span>  private void sanityCheckAssertion(long valueFromBuf, long valueFromField,<a name="line.527"></a>
-<span class="sourceLineNo">528</span>      String fieldName) throws IOException {<a name="line.528"></a>
-<span class="sourceLineNo">529</span>    if (valueFromBuf != valueFromField) {<a name="line.529"></a>
-<span class="sourceLineNo">530</span>      throw new AssertionError(fieldName + " in the buffer (" + valueFromBuf<a name="line.530"></a>
-<span class="sourceLineNo">531</span>          + ") is different from that in the field (" + valueFromField + ")");<a name="line.531"></a>
-<span class="sourceLineNo">532</span>    }<a name="line.532"></a>
-<span class="sourceLineNo">533</span>  }<a name="line.533"></a>
-<span class="sourceLineNo">534</span><a name="line.534"></a>
-<span class="sourceLineNo">535</span>  private void sanityCheckAssertion(BlockType valueFromBuf, BlockType valueFromField)<a name="line.535"></a>
-<span class="sourceLineNo">536</span>      throws IOException {<a name="line.536"></a>
-<span class="sourceLineNo">537</span>    if (valueFromBuf != valueFromField) {<a name="line.537"></a>
-<span class="sourceLineNo">538</span>      throw new IOException("Block type stored in the buffer: " +<a name="line.538"></a>
-<span class="sourceLineNo">539</span>        valueFromBuf + ", block type field: " + valueFromField);<a name="line.539"></a>
-<span class="sourceLineNo">540</span>    }<a name="line.540"></a>
-<span class="sourceLineNo">541</span>  }<a name="line.541"></a>
-<span class="sourceLineNo">542</span><a name="line.542"></a>
-<span class="sourceLineNo">543</span>  /**<a name="line.543"></a>
-<span class="sourceLineNo">544</span>   * Checks if the block is internally consistent, i.e. the first<a name="line.544"></a>
-<span class="sourceLineNo">545</span>   * {@link HConstants#HFILEBLOCK_HEADER_SIZE} bytes of the buffer contain a<a name="line.545"></a>
-<span class="sourceLineNo">546</span>   * valid header consistent with the fields. Assumes a packed block structure.<a name="line.546"></a>
-<span class="sourceLineNo">547</span>   * This function is primary for testing and debugging, and is not<a name="line.547"></a>
-<span class="sourceLineNo">548</span>   * thread-safe, because it alters the internal buffer pointer.<a name="line.548"></a>
-<span class="sourceLineNo">549</span>   * Used by tests only.<a name="line.549"></a>
-<span class="sourceLineNo">550</span>   */<a name="line.550"></a>
-<span class="sourceLineNo">551</span>  @VisibleForTesting<a name="line.551"></a>
-<span class="sourceLineNo">552</span>  void sanityCheck() throws IOException {<a name="line.552"></a>
-<span class="sourceLineNo">553</span>    // Duplicate so no side-effects<a name="line.553"></a>
-<span class="sourceLineNo">554</span>    ByteBuff dup = this.buf.duplicate().rewind();<a name="line.554"></a>
-<span class="sourceLineNo">555</span>    sanityCheckAssertion(BlockType.read(dup), blockType);<a name="line.555"></a>
+<span class="sourceLineNo">230</span>   * &lt;/code&gt;<a name="line.230"></a>
+<span class="sourceLineNo">231</span>   * @see #serialize(ByteBuffer)<a name="line.231"></a>
+<span class="sourceLineNo">232</span>   */<a name="line.232"></a>
+<span class="sourceLineNo">233</span>  static final CacheableDeserializer&lt;Cacheable&gt; BLOCK_DESERIALIZER =<a name="line.233"></a>
+<span class="sourceLineNo">234</span>      new CacheableDeserializer&lt;Cacheable&gt;() {<a name="line.234"></a>
+<span class="sourceLineNo">235</span>        public HFileBlock deserialize(ByteBuff buf, boolean reuse, MemoryType memType)<a name="line.235"></a>
+<span class="sourceLineNo">236</span>        throws IOException {<a name="line.236"></a>
+<span class="sourceLineNo">237</span>          // The buf has the file block followed by block metadata.<a name="line.237"></a>
+<span class="sourceLineNo">238</span>          // Set limit to just before the BLOCK_METADATA_SPACE then rewind.<a name="line.238"></a>
+<span class="sourceLineNo">239</span>          buf.limit(buf.limit() - BLOCK_METADATA_SPACE).rewind();<a name="line.239"></a>
+<span class="sourceLineNo">240</span>          // Get a new buffer to pass the HFileBlock for it to 'own'.<a name="line.240"></a>
+<span class="sourceLineNo">241</span>          ByteBuff newByteBuff;<a name="line.241"></a>
+<span class="sourceLineNo">242</span>          if (reuse) {<a name="line.242"></a>
+<span class="sourceLineNo">243</span>            newByteBuff = buf.slice();<a name="line.243"></a>
+<span class="sourceLineNo">244</span>          } else {<a name="line.244"></a>
+<span class="sourceLineNo">245</span>            int len = buf.limit();<a name="line.245"></a>
+<span class="sourceLineNo">246</span>            newByteBuff = new SingleByteBuff(ByteBuffer.allocate(len));<a name="line.246"></a>
+<span class="sourceLineNo">247</span>            newByteBuff.put(0, buf, buf.position(), len);<a name="line.247"></a>
+<span class="sourceLineNo">248</span>          }<a name="line.248"></a>
+<span class="sourceLineNo">249</span>          // Read out the BLOCK_METADATA_SPACE content and shove into our HFileBlock.<a name="line.249"></a>
+<span class="sourceLineNo">250</span>          buf.position(buf.limit());<a name="line.250"></a>
+<span class="sourceLineNo">251</span>          buf.limit(buf.limit() + HFileBlock.BLOCK_METADATA_SPACE);<a name="line.251"></a>
+<span class="sourceLineNo">252</span>          boolean usesChecksum = buf.get() == (byte)1;<a name="line.252"></a>
+<span class="sourceLineNo">253</span>          long offset = buf.getLong();<a name="line.253"></a>
+<span class="sourceLineNo">254</span>          int nextBlockOnDiskSize = buf.getInt();<a name="line.254"></a>
+<span class="sourceLineNo">255</span>          HFileBlock hFileBlock =<a name="line.255"></a>
+<span class="sourceLineNo">256</span>              new HFileBlock(newByteBuff, usesChecksum, memType, offset, nextBlockOnDiskSize, null);<a name="line.256"></a>
+<span class="sourceLineNo">257</span>          return hFileBlock;<a name="line.257"></a>
+<span class="sourceLineNo">258</span>        }<a name="line.258"></a>
+<span class="sourceLineNo">259</span><a name="line.259"></a>
+<span class="sourceLineNo">260</span>        @Override<a name="line.260"></a>
+<span class="sourceLineNo">261</span>        public int getDeserialiserIdentifier() {<a name="line.261"></a>
+<span class="sourceLineNo">262</span>          return DESERIALIZER_IDENTIFIER;<a name="line.262"></a>
+<span class="sourceLineNo">263</span>        }<a name="line.263"></a>
+<span class="sourceLineNo">264</span><a name="line.264"></a>
+<span class="sourceLineNo">265</span>        @Override<a name="line.265"></a>
+<span class="sourceLineNo">266</span>        public HFileBlock deserialize(ByteBuff b) throws IOException {<a name="line.266"></a>
+<span class="sourceLineNo">267</span>          // Used only in tests<a name="line.267"></a>
+<span class="sourceLineNo">268</span>          return deserialize(b, false, MemoryType.EXCLUSIVE);<a name="line.268"></a>
+<span class="sourceLineNo">269</span>        }<a name="line.269"></a>
+<span class="sourceLineNo">270</span>      };<a name="line.270"></a>
+<span class="sourceLineNo">271</span><a name="line.271"></a>
+<span class="sourceLineNo">272</span>  private static final int DESERIALIZER_IDENTIFIER;<a name="line.272"></a>
+<span class="sourceLineNo">273</span>  static {<a name="line.273"></a>
+<span class="sourceLineNo">274</span>    DESERIALIZER_IDENTIFIER =<a name="line.274"></a>
+<span class="sourceLineNo">275</span>        CacheableDeserializerIdManager.registerDeserializer(BLOCK_DESERIALIZER);<a name="line.275"></a>
+<span class="sourceLineNo">276</span>  }<a name="line.276"></a>
+<span class="sourceLineNo">277</span><a name="line.277"></a>
+<span class="sourceLineNo">278</span>  // Todo: encapsulate Header related logic in this inner class.<a name="line.278"></a>
+<span class="sourceLineNo">279</span>  static class Header {<a name="line.279"></a>
+<span class="sourceLineNo">280</span>    // Format of header is:<a name="line.280"></a>
+<span class="sourceLineNo">281</span>    // 8 bytes - block magic<a name="line.281"></a>
+<span class="sourceLineNo">282</span>    // 4 bytes int - onDiskSizeWithoutHeader<a name="line.282"></a>
+<span class="sourceLineNo">283</span>    // 4 bytes int - uncompressedSizeWithoutHeader<a name="line.283"></a>
+<span class="sourceLineNo">284</span>    // 8 bytes long - prevBlockOffset<a name="line.284"></a>
+<span class="sourceLineNo">285</span>    // The following 3 are only present if header contains checksum information<a name="line.285"></a>
+<span class="sourceLineNo">286</span>    // 1 byte - checksum type<a name="line.286"></a>
+<span class="sourceLineNo">287</span>    // 4 byte int - bytes per checksum<a name="line.287"></a>
+<span class="sourceLineNo">288</span>    // 4 byte int - onDiskDataSizeWithHeader<a name="line.288"></a>
+<span class="sourceLineNo">289</span>    static int BLOCK_MAGIC_INDEX = 0;<a name="line.289"></a>
+<span class="sourceLineNo">290</span>    static int ON_DISK_SIZE_WITHOUT_HEADER_INDEX = 8;<a name="line.290"></a>
+<span class="sourceLineNo">291</span>    static int UNCOMPRESSED_SIZE_WITHOUT_HEADER_INDEX = 12;<a name="line.291"></a>
+<span class="sourceLineNo">292</span>    static int PREV_BLOCK_OFFSET_INDEX = 16;<a name="line.292"></a>
+<span class="sourceLineNo">293</span>    static int CHECKSUM_TYPE_INDEX = 24;<a name="line.293"></a>
+<span class="sourceLineNo">294</span>    static int BYTES_PER_CHECKSUM_INDEX = 25;<a name="line.294"></a>
+<span class="sourceLineNo">295</span>    static int ON_DISK_DATA_SIZE_WITH_HEADER_INDEX = 29;<a name="line.295"></a>
+<span class="sourceLineNo">296</span>  }<a name="line.296"></a>
+<span class="sourceLineNo">297</span><a name="line.297"></a>
+<span class="sourceLineNo">298</span>  /**<a name="line.298"></a>
+<span class="sourceLineNo">299</span>   * Copy constructor. Creates a shallow copy of {@code that}'s buffer.<a name="line.299"></a>
+<span class="sourceLineNo">300</span>   */<a name="line.300"></a>
+<span class="sourceLineNo">301</span>  private HFileBlock(HFileBlock that) {<a name="line.301"></a>
+<span class="sourceLineNo">302</span>    this(that, false);<a name="line.302"></a>
+<span class="sourceLineNo">303</span>  }<a name="line.303"></a>
+<span class="sourceLineNo">304</span><a name="line.304"></a>
+<span class="sourceLineNo">305</span>  /**<a name="line.305"></a>
+<span class="sourceLineNo">306</span>   * Copy constructor. Creates a shallow/deep copy of {@code that}'s buffer as per the boolean<a name="line.306"></a>
+<span class="sourceLineNo">307</span>   * param.<a name="line.307"></a>
+<span class="sourceLineNo">308</span>   */<a name="line.308"></a>
+<span class="sourceLineNo">309</span>  private HFileBlock(HFileBlock that,boolean bufCopy) {<a name="line.309"></a>
+<span class="sourceLineNo">310</span>    this.blockType = that.blockType;<a name="line.310"></a>
+<span class="sourceLineNo">311</span>    this.onDiskSizeWithoutHeader = that.onDiskSizeWithoutHeader;<a name="line.311"></a>
+<span class="sourceLineNo">312</span>    this.uncompressedSizeWithoutHeader = that.uncompressedSizeWithoutHeader;<a name="line.312"></a>
+<span class="sourceLineNo">313</span>    this.prevBlockOffset = that.prevBlockOffset;<a name="line.313"></a>
+<span class="sourceLineNo">314</span>    if (bufCopy) {<a name="line.314"></a>
+<span class="sourceLineNo">315</span>      this.buf = new SingleByteBuff(ByteBuffer.wrap(that.buf.toBytes(0, that.buf.limit())));<a name="line.315"></a>
+<span class="sourceLineNo">316</span>    } else {<a name="line.316"></a>
+<span class="sourceLineNo">317</span>      this.buf = that.buf.duplicate();<a name="line.317"></a>
+<span class="sourceLineNo">318</span>    }<a name="line.318"></a>
+<span class="sourceLineNo">319</span>    this.offset = that.offset;<a name="line.319"></a>
+<span class="sourceLineNo">320</span>    this.onDiskDataSizeWithHeader = that.onDiskDataSizeWithHeader;<a name="line.320"></a>
+<span class="sourceLineNo">321</span>    this.fileContext = that.fileContext;<a name="line.321"></a>
+<span class="sourceLineNo">322</span>    this.nextBlockOnDiskSize = that.nextBlockOnDiskSize;<a name="line.322"></a>
+<span class="sourceLineNo">323</span>  }<a name="line.323"></a>
+<span class="sourceLineNo">324</span><a name="line.324"></a>
+<span class="sourceLineNo">325</span>  /**<a name="line.325"></a>
+<span class="sourceLineNo">326</span>   * Creates a new {@link HFile} block from the given fields. This constructor<a name="line.326"></a>
+<span class="sourceLineNo">327</span>   * is used only while writing blocks and caching,<a name="line.327"></a>
+<span class="sourceLineNo">328</span>   * and is sitting in a byte buffer and we want to stuff the block into cache.<a name="line.328"></a>
+<span class="sourceLineNo">329</span>   * See {@link Writer#getBlockForCaching(CacheConfig)}.<a name="line.329"></a>
+<span class="sourceLineNo">330</span>   *<a name="line.330"></a>
+<span class="sourceLineNo">331</span>   * &lt;p&gt;TODO: The caller presumes no checksumming<a name="line.331"></a>
+<span class="sourceLineNo">332</span>   * required of this block instance since going into cache; checksum already verified on<a name="line.332"></a>
+<s

<TRUNCATED>

Mime
View raw message