http://git-wip-us.apache.org/repos/asf/hbase-site/blob/7e7efd3e/devapidocs/src-html/org/apache/hadoop/hbase/mapreduce/MultiTableHFileOutputFormat.html
----------------------------------------------------------------------
diff --git
a/devapidocs/src-html/org/apache/hadoop/hbase/mapreduce/MultiTableHFileOutputFormat.html
b/devapidocs/src-html/org/apache/hadoop/hbase/mapreduce/MultiTableHFileOutputFormat.html
new file mode 100644
index 0000000..39a1977
--- /dev/null
+++
b/devapidocs/src-html/org/apache/hadoop/hbase/mapreduce/MultiTableHFileOutputFormat.html
@@ -0,0 +1,581 @@
+<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN"
"http://www.w3.org/TR/html4/loose.dtd";>
+<html lang="en">
+<head>
+<title>Source code</title>
+<link rel="stylesheet" type="text/css" href="../../../../../../stylesheet.css"
title="Style">
+</head>
+<body>
+<div class="sourceContainer">
+<pre><span class="sourceLineNo">001</span>/**<a name="line.1"></a>
+<span class="sourceLineNo">002</span> * Licensed to the Apache Software
Foundation (ASF) under one<a name="line.2"></a>
+<span class="sourceLineNo">003</span> * or more contributor license
agreements. See the NOTICE file<a name="line.3"></a>
+<span class="sourceLineNo">004</span> * distributed with this work for
additional information<a name="line.4"></a>
+<span class="sourceLineNo">005</span> * regarding copyright ownership. The
ASF licenses this file<a name="line.5"></a>
+<span class="sourceLineNo">006</span> * to you under the Apache License,
Version 2.0 (the<a name="line.6"></a>
+<span class="sourceLineNo">007</span> * "License"); you may not use this file
except in compliance<a name="line.7"></a>
+<span class="sourceLineNo">008</span> * with the License. You may obtain a
copy of the License at<a name="line.8"></a>
+<span class="sourceLineNo">009</span> *<a name="line.9"></a>
+<span class="sourceLineNo">010</span> *
http://www.apache.org/licenses/LICENSE-2.0<a name="line.10"></a>
+<span class="sourceLineNo">011</span> *<a name="line.11"></a>
+<span class="sourceLineNo">012</span> * Unless required by applicable law or
agreed to in writing, software<a name="line.12"></a>
+<span class="sourceLineNo">013</span> * distributed under the License is
distributed on an "AS IS" BASIS,<a name="line.13"></a>
+<span class="sourceLineNo">014</span> * WITHOUT WARRANTIES OR CONDITIONS OF
ANY KIND, either express or implied.<a name="line.14"></a>
+<span class="sourceLineNo">015</span> * See the License for the specific
language governing permissions and<a name="line.15"></a>
+<span class="sourceLineNo">016</span> * limitations under the License.<a
name="line.16"></a>
+<span class="sourceLineNo">017</span> */<a name="line.17"></a>
+<span class="sourceLineNo">018</span>package
org.apache.hadoop.hbase.mapreduce;<a name="line.18"></a>
+<span class="sourceLineNo">019</span><a name="line.19"></a>
+<span class="sourceLineNo">020</span>import java.io.IOException;<a
name="line.20"></a>
+<span class="sourceLineNo">021</span>import
java.io.UnsupportedEncodingException;<a name="line.21"></a>
+<span class="sourceLineNo">022</span>import java.util.HashMap;<a
name="line.22"></a>
+<span class="sourceLineNo">023</span>import java.util.List;<a
name="line.23"></a>
+<span class="sourceLineNo">024</span>import java.util.Map;<a
name="line.24"></a>
+<span class="sourceLineNo">025</span>import java.util.UUID;<a
name="line.25"></a>
+<span class="sourceLineNo">026</span>import java.util.TreeMap;<a
name="line.26"></a>
+<span class="sourceLineNo">027</span>import java.util.ArrayList;<a
name="line.27"></a>
+<span class="sourceLineNo">028</span>import java.util.TreeSet;<a
name="line.28"></a>
+<span class="sourceLineNo">029</span>import java.util.Collections;<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><a name="line.33"></a>
+<span class="sourceLineNo">034</span>import
org.apache.hadoop.hbase.classification.InterfaceAudience;<a name="line.34"></a>
+<span class="sourceLineNo">035</span>import
org.apache.hadoop.hbase.client.Admin;<a name="line.35"></a>
+<span class="sourceLineNo">036</span>import
org.apache.hadoop.hbase.client.Connection;<a name="line.36"></a>
+<span class="sourceLineNo">037</span>import
org.apache.hadoop.hbase.client.ConnectionFactory;<a name="line.37"></a>
+<span class="sourceLineNo">038</span>import
org.apache.hadoop.hbase.client.RegionLocator;<a name="line.38"></a>
+<span class="sourceLineNo">039</span>import
org.apache.hadoop.hbase.client.Table;<a name="line.39"></a>
+<span class="sourceLineNo">040</span>import
org.apache.hadoop.conf.Configurable;<a name="line.40"></a>
+<span class="sourceLineNo">041</span>import
org.apache.hadoop.conf.Configuration;<a name="line.41"></a>
+<span class="sourceLineNo">042</span>import org.apache.hadoop.fs.FileSystem;<a
name="line.42"></a>
+<span class="sourceLineNo">043</span>import org.apache.hadoop.fs.Path;<a
name="line.43"></a>
+<span class="sourceLineNo">044</span>import org.apache.hadoop.hbase.Cell;<a
name="line.44"></a>
+<span class="sourceLineNo">045</span>import
org.apache.hadoop.hbase.CellUtil;<a name="line.45"></a>
+<span class="sourceLineNo">046</span>import
org.apache.hadoop.hbase.HConstants;<a name="line.46"></a>
+<span class="sourceLineNo">047</span>import
org.apache.hadoop.hbase.HTableDescriptor;<a name="line.47"></a>
+<span class="sourceLineNo">048</span>import
org.apache.hadoop.hbase.KeyValue;<a name="line.48"></a>
+<span class="sourceLineNo">049</span>import
org.apache.hadoop.hbase.TableName;<a name="line.49"></a>
+<span class="sourceLineNo">050</span>import
org.apache.hadoop.hbase.io.ImmutableBytesWritable;<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.Pair;<a name="line.52"></a>
+<span class="sourceLineNo">053</span>import org.apache.hadoop.io.IOUtils;<a
name="line.53"></a>
+<span class="sourceLineNo">054</span>import
org.apache.hadoop.io.SequenceFile;<a name="line.54"></a>
+<span class="sourceLineNo">055</span>import
org.apache.hadoop.io.SequenceFile.Reader;<a name="line.55"></a>
+<span class="sourceLineNo">056</span>import
org.apache.hadoop.io.SequenceFile.Writer;<a name="line.56"></a>
+<span class="sourceLineNo">057</span>import org.apache.hadoop.mapreduce.Job;<a
name="line.57"></a>
+<span class="sourceLineNo">058</span>import
org.apache.hadoop.mapreduce.OutputFormat;<a name="line.58"></a>
+<span class="sourceLineNo">059</span>import
org.apache.hadoop.mapreduce.Partitioner;<a name="line.59"></a>
+<span class="sourceLineNo">060</span>import
org.apache.hadoop.mapreduce.RecordWriter;<a name="line.60"></a>
+<span class="sourceLineNo">061</span>import
org.apache.hadoop.mapreduce.TaskAttemptContext;<a name="line.61"></a>
+<span class="sourceLineNo">062</span>import
org.apache.hadoop.mapreduce.lib.output.FileOutputCommitter;<a
name="line.62"></a>
+<span class="sourceLineNo">063</span>import
org.apache.hadoop.mapreduce.lib.output.FileOutputFormat;<a name="line.63"></a>
+<span class="sourceLineNo">064</span>import
org.apache.hadoop.util.ReflectionUtils;<a name="line.64"></a>
+<span class="sourceLineNo">065</span><a name="line.65"></a>
+<span class="sourceLineNo">066</span>import
com.google.common.annotations.VisibleForTesting;<a name="line.66"></a>
+<span class="sourceLineNo">067</span><a name="line.67"></a>
+<span class="sourceLineNo">068</span>/**<a name="line.68"></a>
+<span class="sourceLineNo">069</span> * Create 3 level tree directory, first
level is using table name as parent directory and then use<a name="line.69"></a>
+<span class="sourceLineNo">070</span> * family name as child directory, and
all related HFiles for one family are under child directory<a
name="line.70"></a>
+<span class="sourceLineNo">071</span> * -tableName1<a name="line.71"></a>
+<span class="sourceLineNo">072</span> * -columnFamilyName1<a
name="line.72"></a>
+<span class="sourceLineNo">073</span> * -HFile (region1)<a
name="line.73"></a>
+<span class="sourceLineNo">074</span> * -columnFamilyName2<a
name="line.74"></a>
+<span class="sourceLineNo">075</span> * -HFile1 (region1)<a
name="line.75"></a>
+<span class="sourceLineNo">076</span> * -HFile2 (region2)<a
name="line.76"></a>
+<span class="sourceLineNo">077</span> * -HFile3 (region3)<a
name="line.77"></a>
+<span class="sourceLineNo">078</span> * -tableName2<a name="line.78"></a>
+<span class="sourceLineNo">079</span> * -columnFamilyName1<a
name="line.79"></a>
+<span class="sourceLineNo">080</span> * -HFile (region1)<a
name="line.80"></a>
+<span class="sourceLineNo">081</span> * family directory and its hfiles match
the output of HFileOutputFormat2<a name="line.81"></a>
+<span class="sourceLineNo">082</span> * @see
org.apache.hadoop.hbase.mapreduce.HFileOutputFormat2<a name="line.82"></a>
+<span class="sourceLineNo">083</span> */<a name="line.83"></a>
+<span class="sourceLineNo">084</span><a name="line.84"></a>
+<span class="sourceLineNo">085</span>@InterfaceAudience.Public<a
name="line.85"></a>
+<span class="sourceLineNo">086</span>@VisibleForTesting<a name="line.86"></a>
+<span class="sourceLineNo">087</span>public class MultiTableHFileOutputFormat
extends FileOutputFormat<ImmutableBytesWritable, Cell> {<a
name="line.87"></a>
+<span class="sourceLineNo">088</span> private static final Log LOG =
LogFactory.getLog(MultiTableHFileOutputFormat.class);<a name="line.88"></a>
+<span class="sourceLineNo">089</span><a name="line.89"></a>
+<span class="sourceLineNo">090</span> @Override<a name="line.90"></a>
+<span class="sourceLineNo">091</span> public
RecordWriter<ImmutableBytesWritable, Cell><a name="line.91"></a>
+<span class="sourceLineNo">092</span> getRecordWriter(final
TaskAttemptContext context) throws IOException, InterruptedException {<a
name="line.92"></a>
+<span class="sourceLineNo">093</span> return
createMultiHFileRecordWriter(context);<a name="line.93"></a>
+<span class="sourceLineNo">094</span> }<a name="line.94"></a>
+<span class="sourceLineNo">095</span><a name="line.95"></a>
+<span class="sourceLineNo">096</span> static <V extends Cell>
RecordWriter<ImmutableBytesWritable, V><a name="line.96"></a>
+<span class="sourceLineNo">097</span> createMultiHFileRecordWriter(final
TaskAttemptContext context) throws IOException {<a name="line.97"></a>
+<span class="sourceLineNo">098</span><a name="line.98"></a>
+<span class="sourceLineNo">099</span> // Get the path of the output
directory<a name="line.99"></a>
+<span class="sourceLineNo">100</span> final Path outputPath =
FileOutputFormat.getOutputPath(context);<a name="line.100"></a>
+<span class="sourceLineNo">101</span> final Path outputDir = new
FileOutputCommitter(outputPath, context).getWorkPath();<a name="line.101"></a>
+<span class="sourceLineNo">102</span> final Configuration conf =
context.getConfiguration();<a name="line.102"></a>
+<span class="sourceLineNo">103</span> final FileSystem fs =
outputDir.getFileSystem(conf);<a name="line.103"></a>
+<span class="sourceLineNo">104</span><a name="line.104"></a>
+<span class="sourceLineNo">105</span> Connection conn =
ConnectionFactory.createConnection(conf);<a name="line.105"></a>
+<span class="sourceLineNo">106</span> Admin admin = conn.getAdmin();<a
name="line.106"></a>
+<span class="sourceLineNo">107</span><a name="line.107"></a>
+<span class="sourceLineNo">108</span> // Map of existing tables, avoid
calling getTable() everytime<a name="line.108"></a>
+<span class="sourceLineNo">109</span> final Map<ImmutableBytesWritable,
Table> tables = new HashMap<>();<a name="line.109"></a>
+<span class="sourceLineNo">110</span><a name="line.110"></a>
+<span class="sourceLineNo">111</span> // Map of tables to writers<a
name="line.111"></a>
+<span class="sourceLineNo">112</span> final Map<ImmutableBytesWritable,
RecordWriter<ImmutableBytesWritable, V>> tableWriters = new
HashMap<>();<a name="line.112"></a>
+<span class="sourceLineNo">113</span><a name="line.113"></a>
+<span class="sourceLineNo">114</span> return new
RecordWriter<ImmutableBytesWritable, V>() {<a name="line.114"></a>
+<span class="sourceLineNo">115</span> @Override<a name="line.115"></a>
+<span class="sourceLineNo">116</span> public void
write(ImmutableBytesWritable tableName, V cell)<a name="line.116"></a>
+<span class="sourceLineNo">117</span> throws IOException,
InterruptedException {<a name="line.117"></a>
+<span class="sourceLineNo">118</span>
RecordWriter<ImmutableBytesWritable, V> tableWriter =
tableWriters.get(tableName);<a name="line.118"></a>
+<span class="sourceLineNo">119</span> // if there is new table, verify
that table directory exists<a name="line.119"></a>
+<span class="sourceLineNo">120</span> if (tableWriter == null) {<a
name="line.120"></a>
+<span class="sourceLineNo">121</span> // using table name as
directory name<a name="line.121"></a>
+<span class="sourceLineNo">122</span> final Path tableOutputDir = new
Path(outputDir, Bytes.toString(tableName.copyBytes()));<a name="line.122"></a>
+<span class="sourceLineNo">123</span> fs.mkdirs(tableOutputDir);<a
name="line.123"></a>
+<span class="sourceLineNo">124</span> LOG.info("Writing Table '" +
tableName.toString() + "' data into following directory"<a name="line.124"></a>
+<span class="sourceLineNo">125</span> +
tableOutputDir.toString());<a name="line.125"></a>
+<span class="sourceLineNo">126</span> // Configure for tableWriter,
if table exist, write configuration of table into conf<a name="line.126"></a>
+<span class="sourceLineNo">127</span> Table table = null;<a
name="line.127"></a>
+<span class="sourceLineNo">128</span> if
(tables.containsKey(tableName)) {<a name="line.128"></a>
+<span class="sourceLineNo">129</span> table =
tables.get(tableName);<a name="line.129"></a>
+<span class="sourceLineNo">130</span> } else {<a name="line.130"></a>
+<span class="sourceLineNo">131</span> table =
getTable(tableName.copyBytes(), conn, admin);<a name="line.131"></a>
+<span class="sourceLineNo">132</span> tables.put(tableName,
table);<a name="line.132"></a>
+<span class="sourceLineNo">133</span> }<a name="line.133"></a>
+<span class="sourceLineNo">134</span> if (table != null) {<a
name="line.134"></a>
+<span class="sourceLineNo">135</span> configureForOneTable(conf,
table.getTableDescriptor());<a name="line.135"></a>
+<span class="sourceLineNo">136</span> }<a name="line.136"></a>
+<span class="sourceLineNo">137</span> // Create writer for one
specific table<a name="line.137"></a>
+<span class="sourceLineNo">138</span> tableWriter = new
HFileOutputFormat2.HFileRecordWriter<>(context, tableOutputDir);<a
name="line.138"></a>
+<span class="sourceLineNo">139</span> // Put table into map<a
name="line.139"></a>
+<span class="sourceLineNo">140</span> tableWriters.put(tableName,
tableWriter);<a name="line.140"></a>
+<span class="sourceLineNo">141</span> }<a name="line.141"></a>
+<span class="sourceLineNo">142</span> // Write <Row, Cell> into
tableWriter<a name="line.142"></a>
+<span class="sourceLineNo">143</span> // in the original code, it does
not use Row<a name="line.143"></a>
+<span class="sourceLineNo">144</span> tableWriter.write(null, cell);<a
name="line.144"></a>
+<span class="sourceLineNo">145</span> }<a name="line.145"></a>
+<span class="sourceLineNo">146</span><a name="line.146"></a>
+<span class="sourceLineNo">147</span> @Override<a name="line.147"></a>
+<span class="sourceLineNo">148</span> public void
close(TaskAttemptContext c) throws IOException, InterruptedException {<a
name="line.148"></a>
+<span class="sourceLineNo">149</span> for
(RecordWriter<ImmutableBytesWritable, V> writer : tableWriters.values())
{<a name="line.149"></a>
+<span class="sourceLineNo">150</span> writer.close(c);<a
name="line.150"></a>
+<span class="sourceLineNo">151</span> }<a name="line.151"></a>
+<span class="sourceLineNo">152</span> if (conn != null) {<a
name="line.152"></a>
+<span class="sourceLineNo">153</span> conn.close();<a
name="line.153"></a>
+<span class="sourceLineNo">154</span> }<a name="line.154"></a>
+<span class="sourceLineNo">155</span> if (admin != null) {<a
name="line.155"></a>
+<span class="sourceLineNo">156</span> admin.close();<a
name="line.156"></a>
+<span class="sourceLineNo">157</span> }<a name="line.157"></a>
+<span class="sourceLineNo">158</span> }<a name="line.158"></a>
+<span class="sourceLineNo">159</span> };<a name="line.159"></a>
+<span class="sourceLineNo">160</span> }<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> * Configure for one table, should be
used before creating a new HFileRecordWriter,<a name="line.163"></a>
+<span class="sourceLineNo">164</span> * Set compression algorithms and
related configuration based on column families<a name="line.164"></a>
+<span class="sourceLineNo">165</span> */<a name="line.165"></a>
+<span class="sourceLineNo">166</span> private static void
configureForOneTable(Configuration conf, final HTableDescriptor
tableDescriptor)<a name="line.166"></a>
+<span class="sourceLineNo">167</span> throws UnsupportedEncodingException
{<a name="line.167"></a>
+<span class="sourceLineNo">168</span>
HFileOutputFormat2.configureCompression(conf, tableDescriptor);<a
name="line.168"></a>
+<span class="sourceLineNo">169</span>
HFileOutputFormat2.configureBlockSize(tableDescriptor, conf);<a
name="line.169"></a>
+<span class="sourceLineNo">170</span>
HFileOutputFormat2.configureBloomType(tableDescriptor, conf);<a
name="line.170"></a>
+<span class="sourceLineNo">171</span>
HFileOutputFormat2.configureDataBlockEncoding(tableDescriptor, conf);<a
name="line.171"></a>
+<span class="sourceLineNo">172</span> }<a name="line.172"></a>
+<span class="sourceLineNo">173</span><a name="line.173"></a>
+<span class="sourceLineNo">174</span> /**<a name="line.174"></a>
+<span class="sourceLineNo">175</span> * Configure a MapReduce Job to output
HFiles for performing an incremental load into<a name="line.175"></a>
+<span class="sourceLineNo">176</span> * the multiple tables.<a
name="line.176"></a>
+<span class="sourceLineNo">177</span> * <ul><a name="line.177"></a>
+<span class="sourceLineNo">178</span> * <li>Inspects the tables to
configure a partitioner based on their region boundaries</li><a
name="line.178"></a>
+<span class="sourceLineNo">179</span> * <li>Writes the partitions
file and configures the partitioner</li><a name="line.179"></a>
+<span class="sourceLineNo">180</span> * <li>Sets the number of
reduce tasks to match the total number of all tables' regions</li><a
name="line.180"></a>
+<span class="sourceLineNo">181</span> * <li>Sets the reducer up to
perform the appropriate sorting (KeyValueSortReducer)</li><a
name="line.181"></a>
+<span class="sourceLineNo">182</span> * </ul><a name="line.182"></a>
+<span class="sourceLineNo">183</span> *<a name="line.183"></a>
+<span class="sourceLineNo">184</span> * ConfigureIncrementalLoad has set up
partitioner and reducer for mapreduce job.<a name="line.184"></a>
+<span class="sourceLineNo">185</span> * Caller needs to setup input path,
output path and mapper<a name="line.185"></a>
+<span class="sourceLineNo">186</span> *<a name="line.186"></a>
+<span class="sourceLineNo">187</span> * @param job<a name="line.187"></a>
+<span class="sourceLineNo">188</span> * @param tables A list of tables to
inspects<a name="line.188"></a>
+<span class="sourceLineNo">189</span> * @throws IOException<a
name="line.189"></a>
+<span class="sourceLineNo">190</span> */<a name="line.190"></a>
+<span class="sourceLineNo">191</span> public static void
configureIncrementalLoad(Job job, List<TableName> tables) throws
IOException {<a name="line.191"></a>
+<span class="sourceLineNo">192</span> configureIncrementalLoad(job, tables,
MultiTableHFileOutputFormat.class);<a name="line.192"></a>
+<span class="sourceLineNo">193</span> }<a name="line.193"></a>
+<span class="sourceLineNo">194</span><a name="line.194"></a>
+<span class="sourceLineNo">195</span> public static void
configureIncrementalLoad(Job job, List<TableName> tables,<a
name="line.195"></a>
+<span class="sourceLineNo">196</span> Class<? extends
OutputFormat<?, ?>> cls) throws IOException {<a name="line.196"></a>
+<span class="sourceLineNo">197</span><a name="line.197"></a>
+<span class="sourceLineNo">198</span> Configuration conf =
job.getConfiguration();<a name="line.198"></a>
+<span class="sourceLineNo">199</span> Map<ImmutableBytesWritable,
List<ImmutableBytesWritable>> tableSplitKeys =<a name="line.199"></a>
+<span class="sourceLineNo">200</span>
MultiHFilePartitioner.getTablesRegionStartKeys(conf, tables);<a
name="line.200"></a>
+<span class="sourceLineNo">201</span> configureIncrementalLoad(job,
tableSplitKeys, cls);<a name="line.201"></a>
+<span class="sourceLineNo">202</span> }<a name="line.202"></a>
+<span class="sourceLineNo">203</span><a name="line.203"></a>
+<span class="sourceLineNo">204</span> /**<a name="line.204"></a>
+<span class="sourceLineNo">205</span> * Same purpose as
configureIncrementalLoad(Job job, List<TableName> tables)<a
name="line.205"></a>
+<span class="sourceLineNo">206</span> * Used when region startKeys of each
table is available, input as <TableName, List<RegionStartKey>><a
name="line.206"></a>
+<span class="sourceLineNo">207</span> *<a name="line.207"></a>
+<span class="sourceLineNo">208</span> * Caller needs to transfer TableName
and byte[] to ImmutableBytesWritable<a name="line.208"></a>
+<span class="sourceLineNo">209</span> */<a name="line.209"></a>
+<span class="sourceLineNo">210</span> public static void
configureIncrementalLoad(Job job, Map<ImmutableBytesWritable,<a
name="line.210"></a>
+<span class="sourceLineNo">211</span>
List<ImmutableBytesWritable>> tableSplitKeys) throws IOException {<a
name="line.211"></a>
+<span class="sourceLineNo">212</span> configureIncrementalLoad(job,
tableSplitKeys, MultiTableHFileOutputFormat.class);<a name="line.212"></a>
+<span class="sourceLineNo">213</span> }<a name="line.213"></a>
+<span class="sourceLineNo">214</span><a name="line.214"></a>
+<span class="sourceLineNo">215</span> public static void
configureIncrementalLoad(Job job, Map<ImmutableBytesWritable,<a
name="line.215"></a>
+<span class="sourceLineNo">216</span>
List<ImmutableBytesWritable>> tableSplitKeys, Class<? extends
OutputFormat<?, ?>> cls) throws IOException {<a name="line.216"></a>
+<span class="sourceLineNo">217</span> Configuration conf =
job.getConfiguration();<a name="line.217"></a>
+<span class="sourceLineNo">218</span><a name="line.218"></a>
+<span class="sourceLineNo">219</span> // file path to store <table,
splitKey><a name="line.219"></a>
+<span class="sourceLineNo">220</span> String hbaseTmpFsDir =
conf.get(HConstants.TEMPORARY_FS_DIRECTORY_KEY,<a name="line.220"></a>
+<span class="sourceLineNo">221</span>
HConstants.DEFAULT_TEMPORARY_HDFS_DIRECTORY);<a name="line.221"></a>
+<span class="sourceLineNo">222</span> final Path partitionsPath = new
Path(hbaseTmpFsDir, "partitions_" + UUID.randomUUID());<a name="line.222"></a>
+<span class="sourceLineNo">223</span> LOG.info("Writing partition info into
dir: " + partitionsPath.toString());<a name="line.223"></a>
+<span class="sourceLineNo">224</span>
job.setPartitionerClass(MultiHFilePartitioner.class);<a name="line.224"></a>
+<span class="sourceLineNo">225</span> // get split keys for all the tables,
and write them into partition file<a name="line.225"></a>
+<span class="sourceLineNo">226</span>
MultiHFilePartitioner.writeTableSplitKeys(conf, partitionsPath,
tableSplitKeys);<a name="line.226"></a>
+<span class="sourceLineNo">227</span>
MultiHFilePartitioner.setPartitionFile(conf, partitionsPath);<a
name="line.227"></a>
+<span class="sourceLineNo">228</span>
partitionsPath.getFileSystem(conf).makeQualified(partitionsPath);<a
name="line.228"></a>
+<span class="sourceLineNo">229</span>
partitionsPath.getFileSystem(conf).deleteOnExit(partitionsPath);<a
name="line.229"></a>
+<span class="sourceLineNo">230</span><a name="line.230"></a>
+<span class="sourceLineNo">231</span> // now only support Mapper output
<ImmutableBytesWritable, KeyValue><a name="line.231"></a>
+<span class="sourceLineNo">232</span> // we can use KeyValueSortReducer
directly to sort Mapper output<a name="line.232"></a>
+<span class="sourceLineNo">233</span> if
(KeyValue.class.equals(job.getMapOutputValueClass())) {<a name="line.233"></a>
+<span class="sourceLineNo">234</span>
job.setReducerClass(KeyValueSortReducer.class);<a name="line.234"></a>
+<span class="sourceLineNo">235</span> } else {<a name="line.235"></a>
+<span class="sourceLineNo">236</span> LOG.warn("Unknown map output value
type:" + job.getMapOutputValueClass());<a name="line.236"></a>
+<span class="sourceLineNo">237</span> }<a name="line.237"></a>
+<span class="sourceLineNo">238</span> int reducerNum =
getReducerNumber(tableSplitKeys);<a name="line.238"></a>
+<span class="sourceLineNo">239</span> job.setNumReduceTasks(reducerNum);<a
name="line.239"></a>
+<span class="sourceLineNo">240</span> LOG.info("Configuring " + reducerNum
+ " reduce partitions " + "to match current region count");<a
name="line.240"></a>
+<span class="sourceLineNo">241</span><a name="line.241"></a>
+<span class="sourceLineNo">242</span> // setup output format<a
name="line.242"></a>
+<span class="sourceLineNo">243</span> job.setOutputFormatClass(cls);<a
name="line.243"></a>
+<span class="sourceLineNo">244</span>
job.setOutputKeyClass(ImmutableBytesWritable.class);<a name="line.244"></a>
+<span class="sourceLineNo">245</span>
job.setOutputValueClass(KeyValue.class);<a name="line.245"></a>
+<span class="sourceLineNo">246</span><a name="line.246"></a>
+<span class="sourceLineNo">247</span>
job.getConfiguration().setStrings("io.serializations",
conf.get("io.serializations"),<a name="line.247"></a>
+<span class="sourceLineNo">248</span>
MutationSerialization.class.getName(), ResultSerialization.class.getName(),<a
name="line.248"></a>
+<span class="sourceLineNo">249</span>
KeyValueSerialization.class.getName());<a name="line.249"></a>
+<span class="sourceLineNo">250</span>
TableMapReduceUtil.addDependencyJars(job);<a name="line.250"></a>
+<span class="sourceLineNo">251</span>
TableMapReduceUtil.initCredentials(job);<a name="line.251"></a>
+<span class="sourceLineNo">252</span> }<a name="line.252"></a>
+<span class="sourceLineNo">253</span><a name="line.253"></a>
+<span class="sourceLineNo">254</span> /**<a name="line.254"></a>
+<span class="sourceLineNo">255</span> * Check if table exist, should not
dependent on HBase instance<a name="line.255"></a>
+<span class="sourceLineNo">256</span> * @return instance of table, if it
exist<a name="line.256"></a>
+<span class="sourceLineNo">257</span> */<a name="line.257"></a>
+<span class="sourceLineNo">258</span> private static Table getTable(final
byte[] tableName, Connection conn, Admin admin) {<a name="line.258"></a>
+<span class="sourceLineNo">259</span> if (conn == null || admin == null)
{<a name="line.259"></a>
+<span class="sourceLineNo">260</span> LOG.info("can not get Connection or
Admin");<a name="line.260"></a>
+<span class="sourceLineNo">261</span> return null;<a name="line.261"></a>
+<span class="sourceLineNo">262</span> }<a name="line.262"></a>
+<span class="sourceLineNo">263</span><a name="line.263"></a>
+<span class="sourceLineNo">264</span> try {<a name="line.264"></a>
+<span class="sourceLineNo">265</span> TableName table =
TableName.valueOf(tableName);<a name="line.265"></a>
+<span class="sourceLineNo">266</span> if (admin.tableExists(table)) {<a
name="line.266"></a>
+<span class="sourceLineNo">267</span> return conn.getTable(table);<a
name="line.267"></a>
+<span class="sourceLineNo">268</span> }<a name="line.268"></a>
+<span class="sourceLineNo">269</span> } catch (IOException e) {<a
name="line.269"></a>
+<span class="sourceLineNo">270</span> LOG.info("Exception found in
getTable()" + e.toString());<a name="line.270"></a>
+<span class="sourceLineNo">271</span> return null;<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> LOG.warn("Table: '" +
TableName.valueOf(tableName) + "' does not exist");<a name="line.274"></a>
+<span class="sourceLineNo">275</span> return null;<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> /**<a name="line.278"></a>
+<span class="sourceLineNo">279</span> * Get the number of reducers by
tables' split keys<a name="line.279"></a>
+<span class="sourceLineNo">280</span> */<a name="line.280"></a>
+<span class="sourceLineNo">281</span> private static int getReducerNumber(<a
name="line.281"></a>
+<span class="sourceLineNo">282</span> Map<ImmutableBytesWritable,
List<ImmutableBytesWritable>> tableSplitKeys) {<a name="line.282"></a>
+<span class="sourceLineNo">283</span> int reducerNum = 0;<a
name="line.283"></a>
+<span class="sourceLineNo">284</span> for
(Map.Entry<ImmutableBytesWritable, List<ImmutableBytesWritable>>
entry : tableSplitKeys.entrySet()) {<a name="line.284"></a>
+<span class="sourceLineNo">285</span> reducerNum +=
entry.getValue().size();<a name="line.285"></a>
+<span class="sourceLineNo">286</span> }<a name="line.286"></a>
+<span class="sourceLineNo">287</span> return reducerNum;<a
name="line.287"></a>
+<span class="sourceLineNo">288</span> }<a name="line.288"></a>
+<span class="sourceLineNo">289</span><a name="line.289"></a>
+<span class="sourceLineNo">290</span> /**<a name="line.290"></a>
+<span class="sourceLineNo">291</span> * MultiTableHFileOutputFormat writes
files based on partitions created by MultiHFilePartitioner<a
name="line.291"></a>
+<span class="sourceLineNo">292</span> * The input is partitioned based on
table's name and its region boundaries with the table.<a name="line.292"></a>
+<span class="sourceLineNo">293</span> * Two records are in the same
partition if they have same table name and the their cells are<a
name="line.293"></a>
+<span class="sourceLineNo">294</span> * in the same region<a
name="line.294"></a>
+<span class="sourceLineNo">295</span> */<a name="line.295"></a>
+<span class="sourceLineNo">296</span> static class MultiHFilePartitioner
extends Partitioner<ImmutableBytesWritable, Cell><a name="line.296"></a>
+<span class="sourceLineNo">297</span> implements Configurable {<a
name="line.297"></a>
+<span class="sourceLineNo">298</span><a name="line.298"></a>
+<span class="sourceLineNo">299</span> public static final String
DEFAULT_PATH = "_partition_multihfile.lst";<a name="line.299"></a>
+<span class="sourceLineNo">300</span> public static final String
PARTITIONER_PATH = "mapreduce.multihfile.partitioner.path";<a
name="line.300"></a>
+<span class="sourceLineNo">301</span> private Configuration conf;<a
name="line.301"></a>
+<span class="sourceLineNo">302</span> // map to receive <table,
splitKeys> from file<a name="line.302"></a>
+<span class="sourceLineNo">303</span> private
Map<ImmutableBytesWritable, List<ImmutableBytesWritable>>
table_SplitKeys;<a name="line.303"></a>
+<span class="sourceLineNo">304</span> // each <table,splitKey> pair
is map to one unique integer<a name="line.304"></a>
+<span class="sourceLineNo">305</span> private TreeMap<TableSplitKeyPair,
Integer> partitionMap;<a name="line.305"></a>
+<span class="sourceLineNo">306</span><a name="line.306"></a>
+<span class="sourceLineNo">307</span> @Override<a name="line.307"></a>
+<span class="sourceLineNo">308</span> public void setConf(Configuration
conf) {<a name="line.308"></a>
+<span class="sourceLineNo">309</span> try {<a name="line.309"></a>
+<span class="sourceLineNo">310</span> this.conf = conf;<a
name="line.310"></a>
+<span class="sourceLineNo">311</span> partitionMap = new
TreeMap<>();<a name="line.311"></a>
+<span class="sourceLineNo">312</span> table_SplitKeys =
readTableSplitKeys(conf);<a name="line.312"></a>
+<span class="sourceLineNo">313</span><a name="line.313"></a>
+<span class="sourceLineNo">314</span> // initiate partitionMap by
table_SplitKeys map<a name="line.314"></a>
+<span class="sourceLineNo">315</span> int splitNum = 0;<a
name="line.315"></a>
+<span class="sourceLineNo">316</span> for
(Map.Entry<ImmutableBytesWritable, List<ImmutableBytesWritable>>
entry : table_SplitKeys.entrySet()) {<a name="line.316"></a>
+<span class="sourceLineNo">317</span> ImmutableBytesWritable table =
entry.getKey();<a name="line.317"></a>
+<span class="sourceLineNo">318</span>
List<ImmutableBytesWritable> list = entry.getValue();<a
name="line.318"></a>
+<span class="sourceLineNo">319</span> for (ImmutableBytesWritable
splitKey : list) {<a name="line.319"></a>
+<span class="sourceLineNo">320</span> partitionMap.put(new
TableSplitKeyPair(table, splitKey), splitNum++);<a name="line.320"></a>
+<span class="sourceLineNo">321</span> }<a name="line.321"></a>
+<span class="sourceLineNo">322</span> }<a name="line.322"></a>
+<span class="sourceLineNo">323</span> } catch (IOException e) {<a
name="line.323"></a>
+<span class="sourceLineNo">324</span> throw new
IllegalArgumentException("Can't read partitions file", e);<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> @Override<a name="line.328"></a>
+<span class="sourceLineNo">329</span> public Configuration getConf() {<a
name="line.329"></a>
+<span class="sourceLineNo">330</span> return conf;<a name="line.330"></a>
+<span class="sourceLineNo">331</span> }<a name="line.331"></a>
+<span class="sourceLineNo">332</span><a name="line.332"></a>
+<span class="sourceLineNo">333</span> /**<a name="line.333"></a>
+<span class="sourceLineNo">334</span> * Set the path to the SequenceFile
storing the sorted <table, splitkey>. It must be the case<a
name="line.334"></a>
+<span class="sourceLineNo">335</span> * that for <tt>R</tt>
reduces, there are <tt>R-1</tt> keys in the SequenceFile.<a
name="line.335"></a>
+<span class="sourceLineNo">336</span> */<a name="line.336"></a>
+<span class="sourceLineNo">337</span> public static void
setPartitionFile(Configuration conf, Path p) {<a name="line.337"></a>
+<span class="sourceLineNo">338</span> conf.set(PARTITIONER_PATH,
p.toString());<a name="line.338"></a>
+<span class="sourceLineNo">339</span> }<a name="line.339"></a>
+<span class="sourceLineNo">340</span><a name="line.340"></a>
+<span class="sourceLineNo">341</span> /**<a name="line.341"></a>
+<span class="sourceLineNo">342</span> * Get the path to the SequenceFile
storing the sorted <table, splitkey>.<a name="line.342"></a>
+<span class="sourceLineNo">343</span> * @see
#setPartitionFile(Configuration, Path)<a name="line.343"></a>
+<span class="sourceLineNo">344</span> */<a name="line.344"></a>
+<span class="sourceLineNo">345</span> public static String
getPartitionFile(Configuration conf) {<a name="line.345"></a>
+<span class="sourceLineNo">346</span> return conf.get(PARTITIONER_PATH,
DEFAULT_PATH);<a name="line.346"></a>
+<span class="sourceLineNo">347</span> }<a name="line.347"></a>
+<span class="sourceLineNo">348</span><a name="line.348"></a>
+<span class="sourceLineNo">349</span><a name="line.349"></a>
+<span class="sourceLineNo">350</span> /**<a name="line.350"></a>
+<span class="sourceLineNo">351</span> * Return map of <tableName, the
start keys of all of the regions in this table><a name="line.351"></a>
+<span class="sourceLineNo">352</span> */<a name="line.352"></a>
+<span class="sourceLineNo">353</span> public static
Map<ImmutableBytesWritable, List<ImmutableBytesWritable>>
getTablesRegionStartKeys(<a name="line.353"></a>
+<span class="sourceLineNo">354</span> Configuration conf,
List<TableName> tables) throws IOException {<a name="line.354"></a>
+<span class="sourceLineNo">355</span> final
TreeMap<ImmutableBytesWritable, List<ImmutableBytesWritable>> ret =
new TreeMap<>();<a name="line.355"></a>
+<span class="sourceLineNo">356</span><a name="line.356"></a>
+<span class="sourceLineNo">357</span> try (Connection conn =
ConnectionFactory.createConnection(conf);<a name="line.357"></a>
+<span class="sourceLineNo">358</span> Admin admin = conn.getAdmin())
{<a name="line.358"></a>
+<span class="sourceLineNo">359</span> LOG.info("Looking up current
regions for tables");<a name="line.359"></a>
+<span class="sourceLineNo">360</span> for (TableName tName : tables)
{<a name="line.360"></a>
+<span class="sourceLineNo">361</span> RegionLocator table =
conn.getRegionLocator(tName);<a name="line.361"></a>
+<span class="sourceLineNo">362</span> // if table not exist, use
default split keys for this table<a name="line.362"></a>
+<span class="sourceLineNo">363</span> byte[][] byteKeys = {
HConstants.EMPTY_BYTE_ARRAY };<a name="line.363"></a>
+<span class="sourceLineNo">364</span> if (admin.tableExists(tName))
{<a name="line.364"></a>
+<span class="sourceLineNo">365</span> byteKeys =
table.getStartKeys();<a name="line.365"></a>
+<span class="sourceLineNo">366</span> }<a name="line.366"></a>
+<span class="sourceLineNo">367</span>
List<ImmutableBytesWritable> tableStartKeys = new
ArrayList<>(byteKeys.length);<a name="line.367"></a>
+<span class="sourceLineNo">368</span> for (byte[] byteKey : byteKeys)
{<a name="line.368"></a>
+<span class="sourceLineNo">369</span> tableStartKeys.add(new
ImmutableBytesWritable(byteKey));<a name="line.369"></a>
+<span class="sourceLineNo">370</span> }<a name="line.370"></a>
+<span class="sourceLineNo">371</span> ret.put(new
ImmutableBytesWritable(tName.toBytes()), tableStartKeys);<a name="line.371"></a>
+<span class="sourceLineNo">372</span><a name="line.372"></a>
+<span class="sourceLineNo">373</span> }<a name="line.373"></a>
+<span class="sourceLineNo">374</span> return ret;<a name="line.374"></a>
+<span class="sourceLineNo">375</span> }<a name="line.375"></a>
+<span class="sourceLineNo">376</span> }<a name="line.376"></a>
+<span class="sourceLineNo">377</span><a name="line.377"></a>
+<span class="sourceLineNo">378</span> /**<a name="line.378"></a>
+<span class="sourceLineNo">379</span> * write <tableName, start key of
each region in table> into sequence file in order,<a name="line.379"></a>
+<span class="sourceLineNo">380</span> * and this format can be parsed by
MultiHFilePartitioner<a name="line.380"></a>
+<span class="sourceLineNo">381</span> */<a name="line.381"></a>
+<span class="sourceLineNo">382</span> public static void
writeTableSplitKeys(Configuration conf, Path partitionsPath,<a
name="line.382"></a>
+<span class="sourceLineNo">383</span> Map<ImmutableBytesWritable,
List<ImmutableBytesWritable>> map) throws IOException {<a
name="line.383"></a>
+<span class="sourceLineNo">384</span> LOG.info("Writing partition
information to " + partitionsPath);<a name="line.384"></a>
+<span class="sourceLineNo">385</span><a name="line.385"></a>
+<span class="sourceLineNo">386</span> if (map == null || map.isEmpty())
{<a name="line.386"></a>
+<span class="sourceLineNo">387</span> throw new
IllegalArgumentException("No regions passed for all tables");<a
name="line.387"></a>
+<span class="sourceLineNo">388</span> }<a name="line.388"></a>
+<span class="sourceLineNo">389</span><a name="line.389"></a>
+<span class="sourceLineNo">390</span> SequenceFile.Writer writer =
SequenceFile.createWriter(conf, Writer.file(partitionsPath),<a
name="line.390"></a>
+<span class="sourceLineNo">391</span>
Writer.keyClass(ImmutableBytesWritable.class),<a name="line.391"></a>
+<span class="sourceLineNo">392</span>
Writer.valueClass(ImmutableBytesWritable.class));<a name="line.392"></a>
+<span class="sourceLineNo">393</span><a name="line.393"></a>
+<span class="sourceLineNo">394</span> try {<a name="line.394"></a>
+<span class="sourceLineNo">395</span> for
(Map.Entry<ImmutableBytesWritable, List<ImmutableBytesWritable>>
entry : map.entrySet()) {<a name="line.395"></a>
+<span class="sourceLineNo">396</span> ImmutableBytesWritable table =
entry.getKey();<a name="line.396"></a>
+<span class="sourceLineNo">397</span>
List<ImmutableBytesWritable> list = entry.getValue();<a
name="line.397"></a>
+<span class="sourceLineNo">398</span> if (list == null) {<a
name="line.398"></a>
+<span class="sourceLineNo">399</span> throw new IOException("Split
keys for a table can not be null");<a name="line.399"></a>
+<span class="sourceLineNo">400</span> }<a name="line.400"></a>
+<span class="sourceLineNo">401</span><a name="line.401"></a>
+<span class="sourceLineNo">402</span>
TreeSet<ImmutableBytesWritable> sorted = new TreeSet<>(list);<a
name="line.402"></a>
+<span class="sourceLineNo">403</span><a name="line.403"></a>
+<span class="sourceLineNo">404</span> ImmutableBytesWritable first =
sorted.first();<a name="line.404"></a>
+<span class="sourceLineNo">405</span> if
(!first.equals(HConstants.EMPTY_BYTE_ARRAY)) {<a name="line.405"></a>
+<span class="sourceLineNo">406</span> throw new
IllegalArgumentException(<a name="line.406"></a>
+<span class="sourceLineNo">407</span> "First region of table
should have empty start key. Instead has: "<a name="line.407"></a>
+<span class="sourceLineNo">408</span> +
Bytes.toStringBinary(first.get()));<a name="line.408"></a>
+<span class="sourceLineNo">409</span> }<a name="line.409"></a>
+<span class="sourceLineNo">410</span><a name="line.410"></a>
+<span class="sourceLineNo">411</span> for (ImmutableBytesWritable
startKey : sorted) {<a name="line.411"></a>
+<span class="sourceLineNo">412</span> writer.append(table,
startKey);<a name="line.412"></a>
+<span class="sourceLineNo">413</span> }<a name="line.413"></a>
+<span class="sourceLineNo">414</span> }<a name="line.414"></a>
+<span class="sourceLineNo">415</span> } finally {<a name="line.415"></a>
+<span class="sourceLineNo">416</span> writer.close();<a
name="line.416"></a>
+<span class="sourceLineNo">417</span> }<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> * read partition file into map
<table, splitKeys of this table><a name="line.421"></a>
+<span class="sourceLineNo">422</span> */<a name="line.422"></a>
+<span class="sourceLineNo">423</span> private
Map<ImmutableBytesWritable, List<ImmutableBytesWritable>>
readTableSplitKeys(<a name="line.423"></a>
+<span class="sourceLineNo">424</span> Configuration conf) throws
IOException {<a name="line.424"></a>
+<span class="sourceLineNo">425</span> String parts =
getPartitionFile(conf);<a name="line.425"></a>
+<span class="sourceLineNo">426</span> LOG.info("Read partition info from
file: " + parts);<a name="line.426"></a>
+<span class="sourceLineNo">427</span> final Path partFile = new
Path(parts);<a name="line.427"></a>
+<span class="sourceLineNo">428</span><a name="line.428"></a>
+<span class="sourceLineNo">429</span> SequenceFile.Reader reader = new
SequenceFile.Reader(conf, Reader.file(partFile));<a name="line.429"></a>
+<span class="sourceLineNo">430</span> // values are already sorted in
file, so use list<a name="line.430"></a>
+<span class="sourceLineNo">431</span> final
Map<ImmutableBytesWritable, List<ImmutableBytesWritable>> map =<a
name="line.431"></a>
+<span class="sourceLineNo">432</span> new TreeMap<>();<a
name="line.432"></a>
+<span class="sourceLineNo">433</span> // key and value have same type<a
name="line.433"></a>
+<span class="sourceLineNo">434</span> ImmutableBytesWritable key =
ReflectionUtils.newInstance(ImmutableBytesWritable.class, conf);<a
name="line.434"></a>
+<span class="sourceLineNo">435</span> ImmutableBytesWritable value =<a
name="line.435"></a>
+<span class="sourceLineNo">436</span>
ReflectionUtils.newInstance(ImmutableBytesWritable.class, conf);<a
name="line.436"></a>
+<span class="sourceLineNo">437</span> try {<a name="line.437"></a>
+<span class="sourceLineNo">438</span> while (reader.next(key, value))
{<a name="line.438"></a>
+<span class="sourceLineNo">439</span><a name="line.439"></a>
+<span class="sourceLineNo">440</span>
List<ImmutableBytesWritable> list = map.get(key);<a name="line.440"></a>
+<span class="sourceLineNo">441</span> if (list == null) {<a
name="line.441"></a>
+<span class="sourceLineNo">442</span> list = new
ArrayList<>();<a name="line.442"></a>
+<span class="sourceLineNo">443</span> }<a name="line.443"></a>
+<span class="sourceLineNo">444</span> list.add(value);<a
name="line.444"></a>
+<span class="sourceLineNo">445</span> map.put(key, list);<a
name="line.445"></a>
+<span class="sourceLineNo">446</span><a name="line.446"></a>
+<span class="sourceLineNo">447</span> key =
ReflectionUtils.newInstance(ImmutableBytesWritable.class, conf);<a
name="line.447"></a>
+<span class="sourceLineNo">448</span> value =
ReflectionUtils.newInstance(ImmutableBytesWritable.class, conf);<a
name="line.448"></a>
+<span class="sourceLineNo">449</span> }<a name="line.449"></a>
+<span class="sourceLineNo">450</span> } finally {<a name="line.450"></a>
+<span class="sourceLineNo">451</span> IOUtils.cleanup(LOG, reader);<a
name="line.451"></a>
+<span class="sourceLineNo">452</span> }<a name="line.452"></a>
+<span class="sourceLineNo">453</span> return map;<a name="line.453"></a>
+<span class="sourceLineNo">454</span> }<a name="line.454"></a>
+<span class="sourceLineNo">455</span><a name="line.455"></a>
+<span class="sourceLineNo">456</span> @Override<a name="line.456"></a>
+<span class="sourceLineNo">457</span> public int
getPartition(ImmutableBytesWritable table, Cell value, int numPartitions) {<a
name="line.457"></a>
+<span class="sourceLineNo">458</span> byte[] row =
CellUtil.cloneRow(value);<a name="line.458"></a>
+<span class="sourceLineNo">459</span> final ImmutableBytesWritable rowKey
= new ImmutableBytesWritable(row);<a name="line.459"></a>
+<span class="sourceLineNo">460</span> ImmutableBytesWritable splitId =
new ImmutableBytesWritable(HConstants.EMPTY_BYTE_ARRAY);<a name="line.460"></a>
+<span class="sourceLineNo">461</span> //find splitKey by input rowKey<a
name="line.461"></a>
+<span class="sourceLineNo">462</span> if
(table_SplitKeys.containsKey(table)) {<a name="line.462"></a>
+<span class="sourceLineNo">463</span>
List<ImmutableBytesWritable> list = table_SplitKeys.get(table);<a
name="line.463"></a>
+<span class="sourceLineNo">464</span> int index =
Collections.binarySearch(list, rowKey, new
ImmutableBytesWritable.Comparator());<a name="line.464"></a>
+<span class="sourceLineNo">465</span> if (index < 0) {<a
name="line.465"></a>
+<span class="sourceLineNo">466</span> index = (index + 1) * (-1) -
1;<a name="line.466"></a>
+<span class="sourceLineNo">467</span> } else if (index == list.size())
{<a name="line.467"></a>
+<span class="sourceLineNo">468</span> index -= 1;<a
name="line.468"></a>
+<span class="sourceLineNo">469</span> }<a name="line.469"></a>
+<span class="sourceLineNo">470</span> if (index < 0) {<a
name="line.470"></a>
+<span class="sourceLineNo">471</span> index = 0;<a
name="line.471"></a>
+<span class="sourceLineNo">472</span> LOG.error("row key can not less
than HConstants.EMPTY_BYTE_ARRAY ");<a name="line.472"></a>
+<span class="sourceLineNo">473</span> }<a name="line.473"></a>
+<span class="sourceLineNo">474</span> splitId = list.get(index);<a
name="line.474"></a>
+<span class="sourceLineNo">475</span> }<a name="line.475"></a>
+<span class="sourceLineNo">476</span><a name="line.476"></a>
+<span class="sourceLineNo">477</span> // find the id of the reducer for
the input<a name="line.477"></a>
+<span class="sourceLineNo">478</span> Integer id = partitionMap.get(new
TableSplitKeyPair(table, splitId));<a name="line.478"></a>
+<span class="sourceLineNo">479</span> if (id == null) {<a
name="line.479"></a>
+<span class="sourceLineNo">480</span> LOG.warn("Can not get reducer
id for input record");<a name="line.480"></a>
+<span class="sourceLineNo">481</span> return -1;<a
name="line.481"></a>
+<span class="sourceLineNo">482</span> }<a name="line.482"></a>
+<span class="sourceLineNo">483</span> return id.intValue() %
numPartitions;<a name="line.483"></a>
+<span class="sourceLineNo">484</span> }<a name="line.484"></a>
+<span class="sourceLineNo">485</span><a name="line.485"></a>
+<span class="sourceLineNo">486</span> /**<a name="line.486"></a>
+<span class="sourceLineNo">487</span> * A class store pair<TableName,
SplitKey>, has two main usage<a name="line.487"></a>
+<span class="sourceLineNo">488</span> * 1. store tableName and one of its
splitKey as a pair<a name="line.488"></a>
+<span class="sourceLineNo">489</span> * 2. implement comparable, so that
partitioner can find splitKey of its input cell<a name="line.489"></a>
+<span class="sourceLineNo">490</span> */<a name="line.490"></a>
+<span class="sourceLineNo">491</span> static class TableSplitKeyPair
extends Pair<ImmutableBytesWritable, ImmutableBytesWritable><a
name="line.491"></a>
+<span class="sourceLineNo">492</span> implements
Comparable<TableSplitKeyPair> {<a name="line.492"></a>
+<span class="sourceLineNo">493</span><a name="line.493"></a>
+<span class="sourceLineNo">494</span> private static final long
serialVersionUID = -6485999667666325594L;<a name="line.494"></a>
+<span class="sourceLineNo">495</span><a name="line.495"></a>
+<span class="sourceLineNo">496</span> public
TableSplitKeyPair(ImmutableBytesWritable a, ImmutableBytesWritable b) {<a
name="line.496"></a>
+<span class="sourceLineNo">497</span> super(a, b);<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> @Override<a name="line.500"></a>
+<span class="sourceLineNo">501</span> public int
compareTo(TableSplitKeyPair other) {<a name="line.501"></a>
+<span class="sourceLineNo">502</span> if
(this.getFirst().equals(other.getFirst())) {<a name="line.502"></a>
+<span class="sourceLineNo">503</span> return
this.getSecond().compareTo(other.getSecond());<a name="line.503"></a>
+<span class="sourceLineNo">504</span> }<a name="line.504"></a>
+<span class="sourceLineNo">505</span> return
this.getFirst().compareTo(other.getFirst());<a name="line.505"></a>
+<span class="sourceLineNo">506</span> }<a name="line.506"></a>
+<span class="sourceLineNo">507</span> }<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>
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+</pre>
+</div>
+</body>
+</html>
