http://git-wip-us.apache.org/repos/asf/hbase-site/blob/d0f1a9f6/devapidocs/src-html/org/apache/hadoop/hbase/mapreduce/TableSnapshotInputFormatImpl.InputSplit.html
----------------------------------------------------------------------
diff --git
a/devapidocs/src-html/org/apache/hadoop/hbase/mapreduce/TableSnapshotInputFormatImpl.InputSplit.html
b/devapidocs/src-html/org/apache/hadoop/hbase/mapreduce/TableSnapshotInputFormatImpl.InputSplit.html
index b1892fa..a11100d 100644
---
a/devapidocs/src-html/org/apache/hadoop/hbase/mapreduce/TableSnapshotInputFormatImpl.InputSplit.html
+++
b/devapidocs/src-html/org/apache/hadoop/hbase/mapreduce/TableSnapshotInputFormatImpl.InputSplit.html
@@ -78,7 +78,7 @@
<span class="sourceLineNo">070</span> // key for specifying the root dir of
the restored snapshot<a name="line.70"></a>
<span class="sourceLineNo">071</span> protected static final String
RESTORE_DIR_KEY = "hbase.TableSnapshotInputFormat.restore.dir";<a
name="line.71"></a>
<span class="sourceLineNo">072</span><a name="line.72"></a>
-<span class="sourceLineNo">073</span> /** See {@link
#getBestLocations(Configuration, HDFSBlocksDistribution)} */<a
name="line.73"></a>
+<span class="sourceLineNo">073</span> /** See {@link
#getBestLocations(Configuration, HDFSBlocksDistribution, int)} */<a
name="line.73"></a>
<span class="sourceLineNo">074</span> private static final String
LOCALITY_CUTOFF_MULTIPLIER =<a name="line.74"></a>
<span class="sourceLineNo">075</span>
"hbase.tablesnapshotinputformat.locality.cutoff.multiplier";<a
name="line.75"></a>
<span class="sourceLineNo">076</span> private static final float
DEFAULT_LOCALITY_CUTOFF_MULTIPLIER = 0.8f;<a name="line.76"></a>
@@ -95,417 +95,456 @@
<span class="sourceLineNo">087</span> public static final String
NUM_SPLITS_PER_REGION = "hbase.mapreduce.splits.per.region";<a
name="line.87"></a>
<span class="sourceLineNo">088</span><a name="line.88"></a>
<span class="sourceLineNo">089</span> /**<a name="line.89"></a>
-<span class="sourceLineNo">090</span> * Implementation class for InputSplit
logic common between mapred and mapreduce.<a name="line.90"></a>
-<span class="sourceLineNo">091</span> */<a name="line.91"></a>
-<span class="sourceLineNo">092</span> public static class InputSplit
implements Writable {<a name="line.92"></a>
-<span class="sourceLineNo">093</span><a name="line.93"></a>
-<span class="sourceLineNo">094</span> private TableDescriptor htd;<a
name="line.94"></a>
-<span class="sourceLineNo">095</span> private HRegionInfo regionInfo;<a
name="line.95"></a>
-<span class="sourceLineNo">096</span> private String[] locations;<a
name="line.96"></a>
-<span class="sourceLineNo">097</span> private String scan;<a
name="line.97"></a>
-<span class="sourceLineNo">098</span> private String restoreDir;<a
name="line.98"></a>
-<span class="sourceLineNo">099</span><a name="line.99"></a>
-<span class="sourceLineNo">100</span> // constructor for mapreduce
framework / Writable<a name="line.100"></a>
-<span class="sourceLineNo">101</span> public InputSplit() {}<a
name="line.101"></a>
-<span class="sourceLineNo">102</span><a name="line.102"></a>
-<span class="sourceLineNo">103</span> public InputSplit(TableDescriptor
htd, HRegionInfo regionInfo, List<String> locations,<a
name="line.103"></a>
-<span class="sourceLineNo">104</span> Scan scan, Path restoreDir) {<a
name="line.104"></a>
-<span class="sourceLineNo">105</span> this.htd = htd;<a
name="line.105"></a>
-<span class="sourceLineNo">106</span> this.regionInfo = regionInfo;<a
name="line.106"></a>
-<span class="sourceLineNo">107</span> if (locations == null ||
locations.isEmpty()) {<a name="line.107"></a>
-<span class="sourceLineNo">108</span> this.locations = new String[0];<a
name="line.108"></a>
-<span class="sourceLineNo">109</span> } else {<a name="line.109"></a>
-<span class="sourceLineNo">110</span> this.locations =
locations.toArray(new String[locations.size()]);<a name="line.110"></a>
-<span class="sourceLineNo">111</span> }<a name="line.111"></a>
-<span class="sourceLineNo">112</span> try {<a name="line.112"></a>
-<span class="sourceLineNo">113</span> this.scan = scan != null ?
TableMapReduceUtil.convertScanToString(scan) : "";<a name="line.113"></a>
-<span class="sourceLineNo">114</span> } catch (IOException e) {<a
name="line.114"></a>
-<span class="sourceLineNo">115</span> LOG.warn("Failed to convert Scan
to String", e);<a name="line.115"></a>
-<span class="sourceLineNo">116</span> }<a name="line.116"></a>
-<span class="sourceLineNo">117</span><a name="line.117"></a>
-<span class="sourceLineNo">118</span> this.restoreDir =
restoreDir.toString();<a name="line.118"></a>
-<span class="sourceLineNo">119</span> }<a name="line.119"></a>
-<span class="sourceLineNo">120</span><a name="line.120"></a>
-<span class="sourceLineNo">121</span> public TableDescriptor getHtd() {<a
name="line.121"></a>
-<span class="sourceLineNo">122</span> return htd;<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> public String getScan() {<a
name="line.125"></a>
-<span class="sourceLineNo">126</span> return scan;<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> public String getRestoreDir() {<a
name="line.129"></a>
-<span class="sourceLineNo">130</span> return restoreDir;<a
name="line.130"></a>
-<span class="sourceLineNo">131</span> }<a name="line.131"></a>
-<span class="sourceLineNo">132</span><a name="line.132"></a>
-<span class="sourceLineNo">133</span> public long getLength() {<a
name="line.133"></a>
-<span class="sourceLineNo">134</span> //TODO: We can obtain the file
sizes of the snapshot here.<a name="line.134"></a>
-<span class="sourceLineNo">135</span> return 0;<a name="line.135"></a>
+<span class="sourceLineNo">090</span> * Whether to calculate the block
location for splits. Default to true.<a name="line.90"></a>
+<span class="sourceLineNo">091</span> * If the computing layer runs outside
of HBase cluster, the block locality does not master.<a name="line.91"></a>
+<span class="sourceLineNo">092</span> * Setting this value to false could
skip the calculation and save some time.<a name="line.92"></a>
+<span class="sourceLineNo">093</span> *<a name="line.93"></a>
+<span class="sourceLineNo">094</span> * Set access modifier to "public" so
that these could be accessed by test classes of<a name="line.94"></a>
+<span class="sourceLineNo">095</span> * both
org.apache.hadoop.hbase.mapred<a name="line.95"></a>
+<span class="sourceLineNo">096</span> * and
org.apache.hadoop.hbase.mapreduce.<a name="line.96"></a>
+<span class="sourceLineNo">097</span> */<a name="line.97"></a>
+<span class="sourceLineNo">098</span> public static final String
SNAPSHOT_INPUTFORMAT_LOCALITY_ENABLED_KEY =<a name="line.98"></a>
+<span class="sourceLineNo">099</span>
"hbase.TableSnapshotInputFormat.locality.enabled";<a name="line.99"></a>
+<span class="sourceLineNo">100</span> public static final boolean
SNAPSHOT_INPUTFORMAT_LOCALITY_ENABLED_DEFAULT = true;<a name="line.100"></a>
+<span class="sourceLineNo">101</span><a name="line.101"></a>
+<span class="sourceLineNo">102</span> /**<a name="line.102"></a>
+<span class="sourceLineNo">103</span> * Implementation class for InputSplit
logic common between mapred and mapreduce.<a name="line.103"></a>
+<span class="sourceLineNo">104</span> */<a name="line.104"></a>
+<span class="sourceLineNo">105</span> public static class InputSplit
implements Writable {<a name="line.105"></a>
+<span class="sourceLineNo">106</span><a name="line.106"></a>
+<span class="sourceLineNo">107</span> private TableDescriptor htd;<a
name="line.107"></a>
+<span class="sourceLineNo">108</span> private HRegionInfo regionInfo;<a
name="line.108"></a>
+<span class="sourceLineNo">109</span> private String[] locations;<a
name="line.109"></a>
+<span class="sourceLineNo">110</span> private String scan;<a
name="line.110"></a>
+<span class="sourceLineNo">111</span> private String restoreDir;<a
name="line.111"></a>
+<span class="sourceLineNo">112</span><a name="line.112"></a>
+<span class="sourceLineNo">113</span> // constructor for mapreduce
framework / Writable<a name="line.113"></a>
+<span class="sourceLineNo">114</span> public InputSplit() {}<a
name="line.114"></a>
+<span class="sourceLineNo">115</span><a name="line.115"></a>
+<span class="sourceLineNo">116</span> public InputSplit(TableDescriptor
htd, HRegionInfo regionInfo, List<String> locations,<a
name="line.116"></a>
+<span class="sourceLineNo">117</span> Scan scan, Path restoreDir) {<a
name="line.117"></a>
+<span class="sourceLineNo">118</span> this.htd = htd;<a
name="line.118"></a>
+<span class="sourceLineNo">119</span> this.regionInfo = regionInfo;<a
name="line.119"></a>
+<span class="sourceLineNo">120</span> if (locations == null ||
locations.isEmpty()) {<a name="line.120"></a>
+<span class="sourceLineNo">121</span> this.locations = new String[0];<a
name="line.121"></a>
+<span class="sourceLineNo">122</span> } else {<a name="line.122"></a>
+<span class="sourceLineNo">123</span> this.locations =
locations.toArray(new String[locations.size()]);<a name="line.123"></a>
+<span class="sourceLineNo">124</span> }<a name="line.124"></a>
+<span class="sourceLineNo">125</span> try {<a name="line.125"></a>
+<span class="sourceLineNo">126</span> this.scan = scan != null ?
TableMapReduceUtil.convertScanToString(scan) : "";<a name="line.126"></a>
+<span class="sourceLineNo">127</span> } catch (IOException e) {<a
name="line.127"></a>
+<span class="sourceLineNo">128</span> LOG.warn("Failed to convert Scan
to String", e);<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> this.restoreDir =
restoreDir.toString();<a name="line.131"></a>
+<span class="sourceLineNo">132</span> }<a name="line.132"></a>
+<span class="sourceLineNo">133</span><a name="line.133"></a>
+<span class="sourceLineNo">134</span> public TableDescriptor getHtd() {<a
name="line.134"></a>
+<span class="sourceLineNo">135</span> return htd;<a name="line.135"></a>
<span class="sourceLineNo">136</span> }<a name="line.136"></a>
<span class="sourceLineNo">137</span><a name="line.137"></a>
-<span class="sourceLineNo">138</span> public String[] getLocations() {<a
name="line.138"></a>
-<span class="sourceLineNo">139</span> return locations;<a
name="line.139"></a>
+<span class="sourceLineNo">138</span> public String getScan() {<a
name="line.138"></a>
+<span class="sourceLineNo">139</span> return scan;<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> public TableDescriptor
getTableDescriptor() {<a name="line.142"></a>
-<span class="sourceLineNo">143</span> return htd;<a name="line.143"></a>
+<span class="sourceLineNo">142</span> public String getRestoreDir() {<a
name="line.142"></a>
+<span class="sourceLineNo">143</span> return restoreDir;<a
name="line.143"></a>
<span class="sourceLineNo">144</span> }<a name="line.144"></a>
<span class="sourceLineNo">145</span><a name="line.145"></a>
-<span class="sourceLineNo">146</span> public HRegionInfo getRegionInfo()
{<a name="line.146"></a>
-<span class="sourceLineNo">147</span> return regionInfo;<a
name="line.147"></a>
-<span class="sourceLineNo">148</span> }<a name="line.148"></a>
-<span class="sourceLineNo">149</span><a name="line.149"></a>
-<span class="sourceLineNo">150</span> // TODO: We should have
ProtobufSerialization in Hadoop, and directly use PB objects instead of<a
name="line.150"></a>
-<span class="sourceLineNo">151</span> // doing this wrapping with
Writables.<a name="line.151"></a>
-<span class="sourceLineNo">152</span> @Override<a name="line.152"></a>
-<span class="sourceLineNo">153</span> public void write(DataOutput out)
throws IOException {<a name="line.153"></a>
-<span class="sourceLineNo">154</span> TableSnapshotRegionSplit.Builder
builder = TableSnapshotRegionSplit.newBuilder()<a name="line.154"></a>
-<span class="sourceLineNo">155</span>
.setTable(ProtobufUtil.toTableSchema(htd))<a name="line.155"></a>
-<span class="sourceLineNo">156</span>
.setRegion(HRegionInfo.convert(regionInfo));<a name="line.156"></a>
-<span class="sourceLineNo">157</span><a name="line.157"></a>
-<span class="sourceLineNo">158</span> for (String location : locations)
{<a name="line.158"></a>
-<span class="sourceLineNo">159</span> builder.addLocations(location);<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> TableSnapshotRegionSplit split =
builder.build();<a name="line.162"></a>
-<span class="sourceLineNo">163</span><a name="line.163"></a>
-<span class="sourceLineNo">164</span> ByteArrayOutputStream baos = new
ByteArrayOutputStream();<a name="line.164"></a>
-<span class="sourceLineNo">165</span> split.writeTo(baos);<a
name="line.165"></a>
-<span class="sourceLineNo">166</span> baos.close();<a name="line.166"></a>
-<span class="sourceLineNo">167</span> byte[] buf = baos.toByteArray();<a
name="line.167"></a>
-<span class="sourceLineNo">168</span> out.writeInt(buf.length);<a
name="line.168"></a>
-<span class="sourceLineNo">169</span> out.write(buf);<a
name="line.169"></a>
+<span class="sourceLineNo">146</span> public long getLength() {<a
name="line.146"></a>
+<span class="sourceLineNo">147</span> //TODO: We can obtain the file
sizes of the snapshot here.<a name="line.147"></a>
+<span class="sourceLineNo">148</span> return 0;<a name="line.148"></a>
+<span class="sourceLineNo">149</span> }<a name="line.149"></a>
+<span class="sourceLineNo">150</span><a name="line.150"></a>
+<span class="sourceLineNo">151</span> public String[] getLocations() {<a
name="line.151"></a>
+<span class="sourceLineNo">152</span> return locations;<a
name="line.152"></a>
+<span class="sourceLineNo">153</span> }<a name="line.153"></a>
+<span class="sourceLineNo">154</span><a name="line.154"></a>
+<span class="sourceLineNo">155</span> public TableDescriptor
getTableDescriptor() {<a name="line.155"></a>
+<span class="sourceLineNo">156</span> return htd;<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> public HRegionInfo getRegionInfo()
{<a name="line.159"></a>
+<span class="sourceLineNo">160</span> return regionInfo;<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> // TODO: We should have
ProtobufSerialization in Hadoop, and directly use PB objects instead of<a
name="line.163"></a>
+<span class="sourceLineNo">164</span> // doing this wrapping with
Writables.<a name="line.164"></a>
+<span class="sourceLineNo">165</span> @Override<a name="line.165"></a>
+<span class="sourceLineNo">166</span> public void write(DataOutput out)
throws IOException {<a name="line.166"></a>
+<span class="sourceLineNo">167</span> TableSnapshotRegionSplit.Builder
builder = TableSnapshotRegionSplit.newBuilder()<a name="line.167"></a>
+<span class="sourceLineNo">168</span>
.setTable(ProtobufUtil.toTableSchema(htd))<a name="line.168"></a>
+<span class="sourceLineNo">169</span>
.setRegion(HRegionInfo.convert(regionInfo));<a name="line.169"></a>
<span class="sourceLineNo">170</span><a name="line.170"></a>
-<span class="sourceLineNo">171</span> Bytes.writeByteArray(out,
Bytes.toBytes(scan));<a name="line.171"></a>
-<span class="sourceLineNo">172</span> Bytes.writeByteArray(out,
Bytes.toBytes(restoreDir));<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><a name="line.175"></a>
-<span class="sourceLineNo">176</span> @Override<a name="line.176"></a>
-<span class="sourceLineNo">177</span> public void readFields(DataInput in)
throws IOException {<a name="line.177"></a>
-<span class="sourceLineNo">178</span> int len = in.readInt();<a
name="line.178"></a>
-<span class="sourceLineNo">179</span> byte[] buf = new byte[len];<a
name="line.179"></a>
-<span class="sourceLineNo">180</span> in.readFully(buf);<a
name="line.180"></a>
-<span class="sourceLineNo">181</span> TableSnapshotRegionSplit split =
TableSnapshotRegionSplit.PARSER.parseFrom(buf);<a name="line.181"></a>
-<span class="sourceLineNo">182</span> this.htd =
ProtobufUtil.toTableDescriptor(split.getTable());<a name="line.182"></a>
-<span class="sourceLineNo">183</span> this.regionInfo =
HRegionInfo.convert(split.getRegion());<a name="line.183"></a>
-<span class="sourceLineNo">184</span> List<String> locationsList =
split.getLocationsList();<a name="line.184"></a>
-<span class="sourceLineNo">185</span> this.locations =
locationsList.toArray(new String[locationsList.size()]);<a name="line.185"></a>
+<span class="sourceLineNo">171</span> for (String location : locations)
{<a name="line.171"></a>
+<span class="sourceLineNo">172</span> builder.addLocations(location);<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> TableSnapshotRegionSplit split =
builder.build();<a name="line.175"></a>
+<span class="sourceLineNo">176</span><a name="line.176"></a>
+<span class="sourceLineNo">177</span> ByteArrayOutputStream baos = new
ByteArrayOutputStream();<a name="line.177"></a>
+<span class="sourceLineNo">178</span> split.writeTo(baos);<a
name="line.178"></a>
+<span class="sourceLineNo">179</span> baos.close();<a name="line.179"></a>
+<span class="sourceLineNo">180</span> byte[] buf = baos.toByteArray();<a
name="line.180"></a>
+<span class="sourceLineNo">181</span> out.writeInt(buf.length);<a
name="line.181"></a>
+<span class="sourceLineNo">182</span> out.write(buf);<a
name="line.182"></a>
+<span class="sourceLineNo">183</span><a name="line.183"></a>
+<span class="sourceLineNo">184</span> Bytes.writeByteArray(out,
Bytes.toBytes(scan));<a name="line.184"></a>
+<span class="sourceLineNo">185</span> Bytes.writeByteArray(out,
Bytes.toBytes(restoreDir));<a name="line.185"></a>
<span class="sourceLineNo">186</span><a name="line.186"></a>
-<span class="sourceLineNo">187</span> this.scan =
Bytes.toString(Bytes.readByteArray(in));<a name="line.187"></a>
-<span class="sourceLineNo">188</span> this.restoreDir =
Bytes.toString(Bytes.readByteArray(in));<a name="line.188"></a>
-<span class="sourceLineNo">189</span> }<a name="line.189"></a>
-<span class="sourceLineNo">190</span> }<a name="line.190"></a>
-<span class="sourceLineNo">191</span><a name="line.191"></a>
-<span class="sourceLineNo">192</span> /**<a name="line.192"></a>
-<span class="sourceLineNo">193</span> * Implementation class for
RecordReader logic common between mapred and mapreduce.<a name="line.193"></a>
-<span class="sourceLineNo">194</span> */<a name="line.194"></a>
-<span class="sourceLineNo">195</span> public static class RecordReader {<a
name="line.195"></a>
-<span class="sourceLineNo">196</span> private InputSplit split;<a
name="line.196"></a>
-<span class="sourceLineNo">197</span> private Scan scan;<a
name="line.197"></a>
-<span class="sourceLineNo">198</span> private Result result = null;<a
name="line.198"></a>
-<span class="sourceLineNo">199</span> private ImmutableBytesWritable row =
null;<a name="line.199"></a>
-<span class="sourceLineNo">200</span> private ClientSideRegionScanner
scanner;<a name="line.200"></a>
-<span class="sourceLineNo">201</span><a name="line.201"></a>
-<span class="sourceLineNo">202</span> public ClientSideRegionScanner
getScanner() {<a name="line.202"></a>
-<span class="sourceLineNo">203</span> return scanner;<a
name="line.203"></a>
-<span class="sourceLineNo">204</span> }<a name="line.204"></a>
-<span class="sourceLineNo">205</span><a name="line.205"></a>
-<span class="sourceLineNo">206</span> public void initialize(InputSplit
split, Configuration conf) throws IOException {<a name="line.206"></a>
-<span class="sourceLineNo">207</span> this.scan =
TableMapReduceUtil.convertStringToScan(split.getScan());<a name="line.207"></a>
-<span class="sourceLineNo">208</span> this.split = split;<a
name="line.208"></a>
-<span class="sourceLineNo">209</span> TableDescriptor htd = split.htd;<a
name="line.209"></a>
-<span class="sourceLineNo">210</span> HRegionInfo hri =
this.split.getRegionInfo();<a name="line.210"></a>
-<span class="sourceLineNo">211</span> FileSystem fs =
FSUtils.getCurrentFileSystem(conf);<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> // region is immutable, this should
be fine,<a name="line.214"></a>
-<span class="sourceLineNo">215</span> // otherwise we have to set the
thread read point<a name="line.215"></a>
-<span class="sourceLineNo">216</span>
scan.setIsolationLevel(IsolationLevel.READ_UNCOMMITTED);<a name="line.216"></a>
-<span class="sourceLineNo">217</span> // disable caching of data blocks<a
name="line.217"></a>
-<span class="sourceLineNo">218</span> scan.setCacheBlocks(false);<a
name="line.218"></a>
-<span class="sourceLineNo">219</span> scan.setScanMetricsEnabled(true);<a
name="line.219"></a>
-<span class="sourceLineNo">220</span><a name="line.220"></a>
-<span class="sourceLineNo">221</span> scanner =<a name="line.221"></a>
-<span class="sourceLineNo">222</span> new
ClientSideRegionScanner(conf, fs, new Path(split.restoreDir), htd, hri, scan,
null);<a name="line.222"></a>
-<span class="sourceLineNo">223</span> }<a name="line.223"></a>
-<span class="sourceLineNo">224</span><a name="line.224"></a>
-<span class="sourceLineNo">225</span> public boolean nextKeyValue() throws
IOException {<a name="line.225"></a>
-<span class="sourceLineNo">226</span> result = scanner.next();<a
name="line.226"></a>
-<span class="sourceLineNo">227</span> if (result == null) {<a
name="line.227"></a>
-<span class="sourceLineNo">228</span> //we are done<a
name="line.228"></a>
-<span class="sourceLineNo">229</span> return false;<a
name="line.229"></a>
-<span class="sourceLineNo">230</span> }<a name="line.230"></a>
-<span class="sourceLineNo">231</span><a name="line.231"></a>
-<span class="sourceLineNo">232</span> if (this.row == null) {<a
name="line.232"></a>
-<span class="sourceLineNo">233</span> this.row = new
ImmutableBytesWritable();<a name="line.233"></a>
-<span class="sourceLineNo">234</span> }<a name="line.234"></a>
-<span class="sourceLineNo">235</span> this.row.set(result.getRow());<a
name="line.235"></a>
-<span class="sourceLineNo">236</span> return true;<a name="line.236"></a>
-<span class="sourceLineNo">237</span> }<a name="line.237"></a>
-<span class="sourceLineNo">238</span><a name="line.238"></a>
-<span class="sourceLineNo">239</span> public ImmutableBytesWritable
getCurrentKey() {<a name="line.239"></a>
-<span class="sourceLineNo">240</span> return row;<a name="line.240"></a>
-<span class="sourceLineNo">241</span> }<a name="line.241"></a>
-<span class="sourceLineNo">242</span><a name="line.242"></a>
-<span class="sourceLineNo">243</span> public Result getCurrentValue() {<a
name="line.243"></a>
-<span class="sourceLineNo">244</span> return result;<a
name="line.244"></a>
-<span class="sourceLineNo">245</span> }<a name="line.245"></a>
-<span class="sourceLineNo">246</span><a name="line.246"></a>
-<span class="sourceLineNo">247</span> public long getPos() {<a
name="line.247"></a>
-<span class="sourceLineNo">248</span> return 0;<a name="line.248"></a>
-<span class="sourceLineNo">249</span> }<a name="line.249"></a>
-<span class="sourceLineNo">250</span><a name="line.250"></a>
-<span class="sourceLineNo">251</span> public float getProgress() {<a
name="line.251"></a>
-<span class="sourceLineNo">252</span> return 0; // TODO: use total bytes
to estimate<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> public void close() {<a
name="line.255"></a>
-<span class="sourceLineNo">256</span> if (this.scanner != null) {<a
name="line.256"></a>
-<span class="sourceLineNo">257</span> this.scanner.close();<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> }<a name="line.260"></a>
-<span class="sourceLineNo">261</span><a name="line.261"></a>
-<span class="sourceLineNo">262</span> public static List<InputSplit>
getSplits(Configuration conf) throws IOException {<a name="line.262"></a>
-<span class="sourceLineNo">263</span> String snapshotName =
getSnapshotName(conf);<a name="line.263"></a>
-<span class="sourceLineNo">264</span><a name="line.264"></a>
-<span class="sourceLineNo">265</span> Path rootDir =
FSUtils.getRootDir(conf);<a name="line.265"></a>
-<span class="sourceLineNo">266</span> FileSystem fs =
rootDir.getFileSystem(conf);<a name="line.266"></a>
+<span class="sourceLineNo">187</span> }<a name="line.187"></a>
+<span class="sourceLineNo">188</span><a name="line.188"></a>
+<span class="sourceLineNo">189</span> @Override<a name="line.189"></a>
+<span class="sourceLineNo">190</span> public void readFields(DataInput in)
throws IOException {<a name="line.190"></a>
+<span class="sourceLineNo">191</span> int len = in.readInt();<a
name="line.191"></a>
+<span class="sourceLineNo">192</span> byte[] buf = new byte[len];<a
name="line.192"></a>
+<span class="sourceLineNo">193</span> in.readFully(buf);<a
name="line.193"></a>
+<span class="sourceLineNo">194</span> TableSnapshotRegionSplit split =
TableSnapshotRegionSplit.PARSER.parseFrom(buf);<a name="line.194"></a>
+<span class="sourceLineNo">195</span> this.htd =
ProtobufUtil.toTableDescriptor(split.getTable());<a name="line.195"></a>
+<span class="sourceLineNo">196</span> this.regionInfo =
HRegionInfo.convert(split.getRegion());<a name="line.196"></a>
+<span class="sourceLineNo">197</span> List<String> locationsList =
split.getLocationsList();<a name="line.197"></a>
+<span class="sourceLineNo">198</span> this.locations =
locationsList.toArray(new String[locationsList.size()]);<a name="line.198"></a>
+<span class="sourceLineNo">199</span><a name="line.199"></a>
+<span class="sourceLineNo">200</span> this.scan =
Bytes.toString(Bytes.readByteArray(in));<a name="line.200"></a>
+<span class="sourceLineNo">201</span> this.restoreDir =
Bytes.toString(Bytes.readByteArray(in));<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> /**<a name="line.205"></a>
+<span class="sourceLineNo">206</span> * Implementation class for
RecordReader logic common between mapred and mapreduce.<a name="line.206"></a>
+<span class="sourceLineNo">207</span> */<a name="line.207"></a>
+<span class="sourceLineNo">208</span> public static class RecordReader {<a
name="line.208"></a>
+<span class="sourceLineNo">209</span> private InputSplit split;<a
name="line.209"></a>
+<span class="sourceLineNo">210</span> private Scan scan;<a
name="line.210"></a>
+<span class="sourceLineNo">211</span> private Result result = null;<a
name="line.211"></a>
+<span class="sourceLineNo">212</span> private ImmutableBytesWritable row =
null;<a name="line.212"></a>
+<span class="sourceLineNo">213</span> private ClientSideRegionScanner
scanner;<a name="line.213"></a>
+<span class="sourceLineNo">214</span><a name="line.214"></a>
+<span class="sourceLineNo">215</span> public ClientSideRegionScanner
getScanner() {<a name="line.215"></a>
+<span class="sourceLineNo">216</span> return scanner;<a
name="line.216"></a>
+<span class="sourceLineNo">217</span> }<a name="line.217"></a>
+<span class="sourceLineNo">218</span><a name="line.218"></a>
+<span class="sourceLineNo">219</span> public void initialize(InputSplit
split, Configuration conf) throws IOException {<a name="line.219"></a>
+<span class="sourceLineNo">220</span> this.scan =
TableMapReduceUtil.convertStringToScan(split.getScan());<a name="line.220"></a>
+<span class="sourceLineNo">221</span> this.split = split;<a
name="line.221"></a>
+<span class="sourceLineNo">222</span> TableDescriptor htd = split.htd;<a
name="line.222"></a>
+<span class="sourceLineNo">223</span> HRegionInfo hri =
this.split.getRegionInfo();<a name="line.223"></a>
+<span class="sourceLineNo">224</span> FileSystem fs =
FSUtils.getCurrentFileSystem(conf);<a name="line.224"></a>
+<span class="sourceLineNo">225</span><a name="line.225"></a>
+<span class="sourceLineNo">226</span><a name="line.226"></a>
+<span class="sourceLineNo">227</span> // region is immutable, this should
be fine,<a name="line.227"></a>
+<span class="sourceLineNo">228</span> // otherwise we have to set the
thread read point<a name="line.228"></a>
+<span class="sourceLineNo">229</span>
scan.setIsolationLevel(IsolationLevel.READ_UNCOMMITTED);<a name="line.229"></a>
+<span class="sourceLineNo">230</span> // disable caching of data blocks<a
name="line.230"></a>
+<span class="sourceLineNo">231</span> scan.setCacheBlocks(false);<a
name="line.231"></a>
+<span class="sourceLineNo">232</span> scan.setScanMetricsEnabled(true);<a
name="line.232"></a>
+<span class="sourceLineNo">233</span><a name="line.233"></a>
+<span class="sourceLineNo">234</span> scanner =<a name="line.234"></a>
+<span class="sourceLineNo">235</span> new
ClientSideRegionScanner(conf, fs, new Path(split.restoreDir), htd, hri, scan,
null);<a name="line.235"></a>
+<span class="sourceLineNo">236</span> }<a name="line.236"></a>
+<span class="sourceLineNo">237</span><a name="line.237"></a>
+<span class="sourceLineNo">238</span> public boolean nextKeyValue() throws
IOException {<a name="line.238"></a>
+<span class="sourceLineNo">239</span> result = scanner.next();<a
name="line.239"></a>
+<span class="sourceLineNo">240</span> if (result == null) {<a
name="line.240"></a>
+<span class="sourceLineNo">241</span> //we are done<a
name="line.241"></a>
+<span class="sourceLineNo">242</span> return false;<a
name="line.242"></a>
+<span class="sourceLineNo">243</span> }<a name="line.243"></a>
+<span class="sourceLineNo">244</span><a name="line.244"></a>
+<span class="sourceLineNo">245</span> if (this.row == null) {<a
name="line.245"></a>
+<span class="sourceLineNo">246</span> this.row = new
ImmutableBytesWritable();<a name="line.246"></a>
+<span class="sourceLineNo">247</span> }<a name="line.247"></a>
+<span class="sourceLineNo">248</span> this.row.set(result.getRow());<a
name="line.248"></a>
+<span class="sourceLineNo">249</span> return true;<a name="line.249"></a>
+<span class="sourceLineNo">250</span> }<a name="line.250"></a>
+<span class="sourceLineNo">251</span><a name="line.251"></a>
+<span class="sourceLineNo">252</span> public ImmutableBytesWritable
getCurrentKey() {<a name="line.252"></a>
+<span class="sourceLineNo">253</span> return row;<a name="line.253"></a>
+<span class="sourceLineNo">254</span> }<a name="line.254"></a>
+<span class="sourceLineNo">255</span><a name="line.255"></a>
+<span class="sourceLineNo">256</span> public Result getCurrentValue() {<a
name="line.256"></a>
+<span class="sourceLineNo">257</span> return result;<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> public long getPos() {<a
name="line.260"></a>
+<span class="sourceLineNo">261</span> return 0;<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> public float getProgress() {<a
name="line.264"></a>
+<span class="sourceLineNo">265</span> return 0; // TODO: use total bytes
to estimate<a name="line.265"></a>
+<span class="sourceLineNo">266</span> }<a name="line.266"></a>
<span class="sourceLineNo">267</span><a name="line.267"></a>
-<span class="sourceLineNo">268</span> SnapshotManifest manifest =
getSnapshotManifest(conf, snapshotName, rootDir, fs);<a name="line.268"></a>
-<span class="sourceLineNo">269</span><a name="line.269"></a>
-<span class="sourceLineNo">270</span> List<HRegionInfo> regionInfos =
getRegionInfosFromManifest(manifest);<a name="line.270"></a>
-<span class="sourceLineNo">271</span><a name="line.271"></a>
-<span class="sourceLineNo">272</span> // TODO: mapred does not support scan
as input API. Work around for now.<a name="line.272"></a>
-<span class="sourceLineNo">273</span> Scan scan =
extractScanFromConf(conf);<a name="line.273"></a>
-<span class="sourceLineNo">274</span> // the temp dir where the snapshot is
restored<a name="line.274"></a>
-<span class="sourceLineNo">275</span> Path restoreDir = new
Path(conf.get(RESTORE_DIR_KEY));<a name="line.275"></a>
-<span class="sourceLineNo">276</span><a name="line.276"></a>
-<span class="sourceLineNo">277</span> RegionSplitter.SplitAlgorithm
splitAlgo = getSplitAlgo(conf);<a name="line.277"></a>
-<span class="sourceLineNo">278</span><a name="line.278"></a>
-<span class="sourceLineNo">279</span> int numSplits =
conf.getInt(NUM_SPLITS_PER_REGION, 1);<a name="line.279"></a>
+<span class="sourceLineNo">268</span> public void close() {<a
name="line.268"></a>
+<span class="sourceLineNo">269</span> if (this.scanner != null) {<a
name="line.269"></a>
+<span class="sourceLineNo">270</span> this.scanner.close();<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><a name="line.274"></a>
+<span class="sourceLineNo">275</span> public static List<InputSplit>
getSplits(Configuration conf) throws IOException {<a name="line.275"></a>
+<span class="sourceLineNo">276</span> String snapshotName =
getSnapshotName(conf);<a name="line.276"></a>
+<span class="sourceLineNo">277</span><a name="line.277"></a>
+<span class="sourceLineNo">278</span> Path rootDir =
FSUtils.getRootDir(conf);<a name="line.278"></a>
+<span class="sourceLineNo">279</span> FileSystem fs =
rootDir.getFileSystem(conf);<a name="line.279"></a>
<span class="sourceLineNo">280</span><a name="line.280"></a>
-<span class="sourceLineNo">281</span> return getSplits(scan, manifest,
regionInfos, restoreDir, conf, splitAlgo, numSplits);<a name="line.281"></a>
-<span class="sourceLineNo">282</span> }<a name="line.282"></a>
-<span class="sourceLineNo">283</span><a name="line.283"></a>
-<span class="sourceLineNo">284</span> public static
RegionSplitter.SplitAlgorithm getSplitAlgo(Configuration conf) throws
IOException{<a name="line.284"></a>
-<span class="sourceLineNo">285</span> String splitAlgoClassName =
conf.get(SPLIT_ALGO);<a name="line.285"></a>
-<span class="sourceLineNo">286</span> if (splitAlgoClassName == null)<a
name="line.286"></a>
-<span class="sourceLineNo">287</span> return null;<a name="line.287"></a>
-<span class="sourceLineNo">288</span> try {<a name="line.288"></a>
-<span class="sourceLineNo">289</span> return ((Class<? extends
RegionSplitter.SplitAlgorithm>)<a name="line.289"></a>
-<span class="sourceLineNo">290</span>
Class.forName(splitAlgoClassName)).newInstance();<a name="line.290"></a>
-<span class="sourceLineNo">291</span> } catch (ClassNotFoundException e)
{<a name="line.291"></a>
-<span class="sourceLineNo">292</span> throw new IOException("SplitAlgo
class " + splitAlgoClassName +<a name="line.292"></a>
-<span class="sourceLineNo">293</span> " is not found", e);<a
name="line.293"></a>
-<span class="sourceLineNo">294</span> } catch (InstantiationException e)
{<a name="line.294"></a>
-<span class="sourceLineNo">295</span> throw new IOException("SplitAlgo
class " + splitAlgoClassName +<a name="line.295"></a>
-<span class="sourceLineNo">296</span> " is not instantiable",
e);<a name="line.296"></a>
-<span class="sourceLineNo">297</span> } catch (IllegalAccessException e)
{<a name="line.297"></a>
-<span class="sourceLineNo">298</span> throw new IOException("SplitAlgo
class " + splitAlgoClassName +<a name="line.298"></a>
-<span class="sourceLineNo">299</span> " is not instantiable",
e);<a name="line.299"></a>
-<span class="sourceLineNo">300</span> }<a name="line.300"></a>
-<span class="sourceLineNo">301</span> }<a name="line.301"></a>
-<span class="sourceLineNo">302</span><a name="line.302"></a>
-<span class="sourceLineNo">303</span><a name="line.303"></a>
-<span class="sourceLineNo">304</span> public static List<HRegionInfo>
getRegionInfosFromManifest(SnapshotManifest manifest) {<a name="line.304"></a>
-<span class="sourceLineNo">305</span> List<SnapshotRegionManifest>
regionManifests = manifest.getRegionManifests();<a name="line.305"></a>
-<span class="sourceLineNo">306</span> if (regionManifests == null) {<a
name="line.306"></a>
-<span class="sourceLineNo">307</span> throw new
IllegalArgumentException("Snapshot seems empty");<a name="line.307"></a>
-<span class="sourceLineNo">308</span> }<a name="line.308"></a>
-<span class="sourceLineNo">309</span><a name="line.309"></a>
-<span class="sourceLineNo">310</span> List<HRegionInfo> regionInfos =
Lists.newArrayListWithCapacity(regionManifests.size());<a name="line.310"></a>
-<span class="sourceLineNo">311</span><a name="line.311"></a>
-<span class="sourceLineNo">312</span> for (SnapshotRegionManifest
regionManifest : regionManifests) {<a name="line.312"></a>
-<span class="sourceLineNo">313</span> HRegionInfo hri =
HRegionInfo.convert(regionManifest.getRegionInfo());<a name="line.313"></a>
-<span class="sourceLineNo">314</span> if (hri.isOffline() &&
(hri.isSplit() || hri.isSplitParent())) {<a name="line.314"></a>
-<span class="sourceLineNo">315</span> continue;<a name="line.315"></a>
-<span class="sourceLineNo">316</span> }<a name="line.316"></a>
-<span class="sourceLineNo">317</span> regionInfos.add(hri);<a
name="line.317"></a>
-<span class="sourceLineNo">318</span> }<a name="line.318"></a>
-<span class="sourceLineNo">319</span> return regionInfos;<a
name="line.319"></a>
-<span class="sourceLineNo">320</span> }<a name="line.320"></a>
-<span class="sourceLineNo">321</span><a name="line.321"></a>
-<span class="sourceLineNo">322</span> public static SnapshotManifest
getSnapshotManifest(Configuration conf, String snapshotName,<a
name="line.322"></a>
-<span class="sourceLineNo">323</span> Path rootDir, FileSystem fs) throws
IOException {<a name="line.323"></a>
-<span class="sourceLineNo">324</span> Path snapshotDir =
SnapshotDescriptionUtils.getCompletedSnapshotDir(snapshotName, rootDir);<a
name="line.324"></a>
-<span class="sourceLineNo">325</span> SnapshotDescription snapshotDesc =
SnapshotDescriptionUtils.readSnapshotInfo(fs, snapshotDir);<a
name="line.325"></a>
-<span class="sourceLineNo">326</span> return SnapshotManifest.open(conf,
fs, snapshotDir, snapshotDesc);<a name="line.326"></a>
-<span class="sourceLineNo">327</span> }<a name="line.327"></a>
-<span class="sourceLineNo">328</span><a name="line.328"></a>
-<span class="sourceLineNo">329</span> public static Scan
extractScanFromConf(Configuration conf) throws IOException {<a
name="line.329"></a>
-<span class="sourceLineNo">330</span> Scan scan = null;<a
name="line.330"></a>
-<span class="sourceLineNo">331</span> if (conf.get(TableInputFormat.SCAN)
!= null) {<a name="line.331"></a>
-<span class="sourceLineNo">332</span> scan =
TableMapReduceUtil.convertStringToScan(conf.get(TableInputFormat.SCAN));<a
name="line.332"></a>
-<span class="sourceLineNo">333</span> } else if
(conf.get(org.apache.hadoop.hbase.mapred.TableInputFormat.COLUMN_LIST) != null)
{<a name="line.333"></a>
-<span class="sourceLineNo">334</span> String[] columns =<a
name="line.334"></a>
-<span class="sourceLineNo">335</span>
conf.get(org.apache.hadoop.hbase.mapred.TableInputFormat.COLUMN_LIST).split("
");<a name="line.335"></a>
-<span class="sourceLineNo">336</span> scan = new Scan();<a
name="line.336"></a>
-<span class="sourceLineNo">337</span> for (String col : columns) {<a
name="line.337"></a>
-<span class="sourceLineNo">338</span>
scan.addFamily(Bytes.toBytes(col));<a name="line.338"></a>
-<span class="sourceLineNo">339</span> }<a name="line.339"></a>
-<span class="sourceLineNo">340</span> } else {<a name="line.340"></a>
-<span class="sourceLineNo">341</span> throw new
IllegalArgumentException("Unable to create scan");<a name="line.341"></a>
-<span class="sourceLineNo">342</span> }<a name="line.342"></a>
-<span class="sourceLineNo">343</span> return scan;<a name="line.343"></a>
-<span class="sourceLineNo">344</span> }<a name="line.344"></a>
-<span class="sourceLineNo">345</span><a name="line.345"></a>
-<span class="sourceLineNo">346</span> public static List<InputSplit>
getSplits(Scan scan, SnapshotManifest manifest,<a name="line.346"></a>
-<span class="sourceLineNo">347</span> List<HRegionInfo>
regionManifests, Path restoreDir, Configuration conf) throws IOException {<a
name="line.347"></a>
-<span class="sourceLineNo">348</span> return getSplits(scan, manifest,
regionManifests, restoreDir, conf, null, 1);<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> public static List<InputSplit>
getSplits(Scan scan, SnapshotManifest manifest,<a name="line.351"></a>
-<span class="sourceLineNo">352</span> List<HRegionInfo>
regionManifests, Path restoreDir,<a name="line.352"></a>
-<span class="sourceLineNo">353</span> Configuration conf,
RegionSplitter.SplitAlgorithm sa, int numSplits) throws IOException {<a
name="line.353"></a>
-<span class="sourceLineNo">354</span> // load table descriptor<a
name="line.354"></a>
-<span class="sourceLineNo">355</span> TableDescriptor htd =
manifest.getTableDescriptor();<a name="line.355"></a>
-<span class="sourceLineNo">356</span><a name="line.356"></a>
-<span class="sourceLineNo">357</span> Path tableDir =
FSUtils.getTableDir(restoreDir, htd.getTableName());<a name="line.357"></a>
+<span class="sourceLineNo">281</span> SnapshotManifest manifest =
getSnapshotManifest(conf, snapshotName, rootDir, fs);<a name="line.281"></a>
+<span class="sourceLineNo">282</span><a name="line.282"></a>
+<span class="sourceLineNo">283</span> List<HRegionInfo> regionInfos =
getRegionInfosFromManifest(manifest);<a name="line.283"></a>
+<span class="sourceLineNo">284</span><a name="line.284"></a>
+<span class="sourceLineNo">285</span> // TODO: mapred does not support scan
as input API. Work around for now.<a name="line.285"></a>
+<span class="sourceLineNo">286</span> Scan scan =
extractScanFromConf(conf);<a name="line.286"></a>
+<span class="sourceLineNo">287</span> // the temp dir where the snapshot is
restored<a name="line.287"></a>
+<span class="sourceLineNo">288</span> Path restoreDir = new
Path(conf.get(RESTORE_DIR_KEY));<a name="line.288"></a>
+<span class="sourceLineNo">289</span><a name="line.289"></a>
+<span class="sourceLineNo">290</span> RegionSplitter.SplitAlgorithm
splitAlgo = getSplitAlgo(conf);<a name="line.290"></a>
+<span class="sourceLineNo">291</span><a name="line.291"></a>
+<span class="sourceLineNo">292</span> int numSplits =
conf.getInt(NUM_SPLITS_PER_REGION, 1);<a name="line.292"></a>
+<span class="sourceLineNo">293</span><a name="line.293"></a>
+<span class="sourceLineNo">294</span> return getSplits(scan, manifest,
regionInfos, restoreDir, conf, splitAlgo, numSplits);<a name="line.294"></a>
+<span class="sourceLineNo">295</span> }<a name="line.295"></a>
+<span class="sourceLineNo">296</span><a name="line.296"></a>
+<span class="sourceLineNo">297</span> public static
RegionSplitter.SplitAlgorithm getSplitAlgo(Configuration conf) throws
IOException{<a name="line.297"></a>
+<span class="sourceLineNo">298</span> String splitAlgoClassName =
conf.get(SPLIT_ALGO);<a name="line.298"></a>
+<span class="sourceLineNo">299</span> if (splitAlgoClassName == null)<a
name="line.299"></a>
+<span class="sourceLineNo">300</span> return null;<a name="line.300"></a>
+<span class="sourceLineNo">301</span> try {<a name="line.301"></a>
+<span class="sourceLineNo">302</span> return ((Class<? extends
RegionSplitter.SplitAlgorithm>)<a name="line.302"></a>
+<span class="sourceLineNo">303</span>
Class.forName(splitAlgoClassName)).newInstance();<a name="line.303"></a>
+<span class="sourceLineNo">304</span> } catch (ClassNotFoundException e)
{<a name="line.304"></a>
+<span class="sourceLineNo">305</span> throw new IOException("SplitAlgo
class " + splitAlgoClassName +<a name="line.305"></a>
+<span class="sourceLineNo">306</span> " is not found", e);<a
name="line.306"></a>
+<span class="sourceLineNo">307</span> } catch (InstantiationException e)
{<a name="line.307"></a>
+<span class="sourceLineNo">308</span> throw new IOException("SplitAlgo
class " + splitAlgoClassName +<a name="line.308"></a>
+<span class="sourceLineNo">309</span> " is not instantiable",
e);<a name="line.309"></a>
+<span class="sourceLineNo">310</span> } catch (IllegalAccessException e)
{<a name="line.310"></a>
+<span class="sourceLineNo">311</span> throw new IOException("SplitAlgo
class " + splitAlgoClassName +<a name="line.311"></a>
+<span class="sourceLineNo">312</span> " is not instantiable",
e);<a name="line.312"></a>
+<span class="sourceLineNo">313</span> }<a name="line.313"></a>
+<span class="sourceLineNo">314</span> }<a name="line.314"></a>
+<span class="sourceLineNo">315</span><a name="line.315"></a>
+<span class="sourceLineNo">316</span><a name="line.316"></a>
+<span class="sourceLineNo">317</span> public static List<HRegionInfo>
getRegionInfosFromManifest(SnapshotManifest manifest) {<a name="line.317"></a>
+<span class="sourceLineNo">318</span> List<SnapshotRegionManifest>
regionManifests = manifest.getRegionManifests();<a name="line.318"></a>
+<span class="sourceLineNo">319</span> if (regionManifests == null) {<a
name="line.319"></a>
+<span class="sourceLineNo">320</span> throw new
IllegalArgumentException("Snapshot seems empty");<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> List<HRegionInfo> regionInfos =
Lists.newArrayListWithCapacity(regionManifests.size());<a name="line.323"></a>
+<span class="sourceLineNo">324</span><a name="line.324"></a>
+<span class="sourceLineNo">325</span> for (SnapshotRegionManifest
regionManifest : regionManifests) {<a name="line.325"></a>
+<span class="sourceLineNo">326</span> HRegionInfo hri =
HRegionInfo.convert(regionManifest.getRegionInfo());<a name="line.326"></a>
+<span class="sourceLineNo">327</span> if (hri.isOffline() &&
(hri.isSplit() || hri.isSplitParent())) {<a name="line.327"></a>
+<span class="sourceLineNo">328</span> continue;<a name="line.328"></a>
+<span class="sourceLineNo">329</span> }<a name="line.329"></a>
+<span class="sourceLineNo">330</span> regionInfos.add(hri);<a
name="line.330"></a>
+<span class="sourceLineNo">331</span> }<a name="line.331"></a>
+<span class="sourceLineNo">332</span> return regionInfos;<a
name="line.332"></a>
+<span class="sourceLineNo">333</span> }<a name="line.333"></a>
+<span class="sourceLineNo">334</span><a name="line.334"></a>
+<span class="sourceLineNo">335</span> public static SnapshotManifest
getSnapshotManifest(Configuration conf, String snapshotName,<a
name="line.335"></a>
+<span class="sourceLineNo">336</span> Path rootDir, FileSystem fs) throws
IOException {<a name="line.336"></a>
+<span class="sourceLineNo">337</span> Path snapshotDir =
SnapshotDescriptionUtils.getCompletedSnapshotDir(snapshotName, rootDir);<a
name="line.337"></a>
+<span class="sourceLineNo">338</span> SnapshotDescription snapshotDesc =
SnapshotDescriptionUtils.readSnapshotInfo(fs, snapshotDir);<a
name="line.338"></a>
+<span class="sourceLineNo">339</span> return SnapshotManifest.open(conf,
fs, snapshotDir, snapshotDesc);<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> public static Scan
extractScanFromConf(Configuration conf) throws IOException {<a
name="line.342"></a>
+<span class="sourceLineNo">343</span> Scan scan = null;<a
name="line.343"></a>
+<span class="sourceLineNo">344</span> if (conf.get(TableInputFormat.SCAN)
!= null) {<a name="line.344"></a>
+<span class="sourceLineNo">345</span> scan =
TableMapReduceUtil.convertStringToScan(conf.get(TableInputFormat.SCAN));<a
name="line.345"></a>
+<span class="sourceLineNo">346</span> } else if
(conf.get(org.apache.hadoop.hbase.mapred.TableInputFormat.COLUMN_LIST) != null)
{<a name="line.346"></a>
+<span class="sourceLineNo">347</span> String[] columns =<a
name="line.347"></a>
+<span class="sourceLineNo">348</span>
conf.get(org.apache.hadoop.hbase.mapred.TableInputFormat.COLUMN_LIST).split("
");<a name="line.348"></a>
+<span class="sourceLineNo">349</span> scan = new Scan();<a
name="line.349"></a>
+<span class="sourceLineNo">350</span> for (String col : columns) {<a
name="line.350"></a>
+<span class="sourceLineNo">351</span>
scan.addFamily(Bytes.toBytes(col));<a name="line.351"></a>
+<span class="sourceLineNo">352</span> }<a name="line.352"></a>
+<span class="sourceLineNo">353</span> } else {<a name="line.353"></a>
+<span class="sourceLineNo">354</span> throw new
IllegalArgumentException("Unable to create scan");<a name="line.354"></a>
+<span class="sourceLineNo">355</span> }<a name="line.355"></a>
+<span class="sourceLineNo">356</span> return scan;<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> List<InputSplit> splits = new
ArrayList<>();<a name="line.359"></a>
-<span class="sourceLineNo">360</span> for (HRegionInfo hri :
regionManifests) {<a name="line.360"></a>
-<span class="sourceLineNo">361</span> // load region descriptor<a
name="line.361"></a>
-<span class="sourceLineNo">362</span><a name="line.362"></a>
-<span class="sourceLineNo">363</span> if (numSplits > 1) {<a
name="line.363"></a>
-<span class="sourceLineNo">364</span> byte[][] sp =
sa.split(hri.getStartKey(), hri.getEndKey(), numSplits, true);<a
name="line.364"></a>
-<span class="sourceLineNo">365</span> for (int i = 0; i < sp.length
- 1; i++) {<a name="line.365"></a>
-<span class="sourceLineNo">366</span> if
(PrivateCellUtil.overlappingKeys(scan.getStartRow(), scan.getStopRow(),
sp[i],<a name="line.366"></a>
-<span class="sourceLineNo">367</span> sp[i + 1])) {<a
name="line.367"></a>
-<span class="sourceLineNo">368</span> // compute HDFS locations
from snapshot files (which will get the locations for<a name="line.368"></a>
-<span class="sourceLineNo">369</span> // referred hfiles)<a
name="line.369"></a>
-<span class="sourceLineNo">370</span> List<String> hosts =
getBestLocations(conf,<a name="line.370"></a>
-<span class="sourceLineNo">371</span>
HRegion.computeHDFSBlocksDistribution(conf, htd, hri, tableDir));<a
name="line.371"></a>
-<span class="sourceLineNo">372</span><a name="line.372"></a>
-<span class="sourceLineNo">373</span> int len = Math.min(3,
hosts.size());<a name="line.373"></a>
-<span class="sourceLineNo">374</span> hosts = hosts.subList(0,
len);<a name="line.374"></a>
-<span class="sourceLineNo">375</span> Scan boundedScan = new
Scan(scan);<a name="line.375"></a>
-<span class="sourceLineNo">376</span>
boundedScan.setStartRow(sp[i]);<a name="line.376"></a>
-<span class="sourceLineNo">377</span> boundedScan.setStopRow(sp[i +
1]);<a name="line.377"></a>
-<span class="sourceLineNo">378</span> splits.add(new
InputSplit(htd, hri, hosts, boundedScan, restoreDir));<a name="line.378"></a>
-<span class="sourceLineNo">379</span> }<a name="line.379"></a>
-<span class="sourceLineNo">380</span> }<a name="line.380"></a>
-<span class="sourceLineNo">381</span> } else {<a name="line.381"></a>
-<span class="sourceLineNo">382</span> if
(PrivateCellUtil.overlappingKeys(scan.getStartRow(), scan.getStopRow(),<a
name="line.382"></a>
-<span class="sourceLineNo">383</span> hri.getStartKey(),
hri.getEndKey())) {<a name="line.383"></a>
-<span class="sourceLineNo">384</span> // compute HDFS locations from
snapshot files (which will get the locations for<a name="line.384"></a>
-<span class="sourceLineNo">385</span> // referred hfiles)<a
name="line.385"></a>
-<span class="sourceLineNo">386</span> List<String> hosts =
getBestLocations(conf,<a name="line.386"></a>
-<span class="sourceLineNo">387</span>
HRegion.computeHDFSBlocksDistribution(conf, htd, hri, tableDir));<a
name="line.387"></a>
-<span class="sourceLineNo">388</span><a name="line.388"></a>
-<span class="sourceLineNo">389</span> int len = Math.min(3,
hosts.size());<a name="line.389"></a>
-<span class="sourceLineNo">390</span> hosts = hosts.subList(0,
len);<a name="line.390"></a>
-<span class="sourceLineNo">391</span> splits.add(new InputSplit(htd,
hri, hosts, scan, restoreDir));<a name="line.391"></a>
-<span class="sourceLineNo">392</span> }<a name="line.392"></a>
-<span class="sourceLineNo">393</span> }<a name="line.393"></a>
-<span class="sourceLineNo">394</span> }<a name="line.394"></a>
-<span class="sourceLineNo">395</span><a name="line.395"></a>
-<span class="sourceLineNo">396</span> return splits;<a name="line.396"></a>
-<span class="sourceLineNo">397</span><a name="line.397"></a>
-<span class="sourceLineNo">398</span> }<a name="line.398"></a>
-<span class="sourceLineNo">399</span><a name="line.399"></a>
-<span class="sourceLineNo">400</span> /**<a name="line.400"></a>
-<span class="sourceLineNo">401</span> * This computes the locations to be
passed from the InputSplit. MR/Yarn schedulers does not take<a
name="line.401"></a>
-<span class="sourceLineNo">402</span> * weights into account, thus will
treat every location passed from the input split as equal. We<a
name="line.402"></a>
-<span class="sourceLineNo">403</span> * do not want to blindly pass all the
locations, since we are creating one split per region, and<a
name="line.403"></a>
-<span class="sourceLineNo">404</span> * the region's blocks are all
distributed throughout the cluster unless favorite node assignment<a
name="line.404"></a>
-<span class="sourceLineNo">405</span> * is used. On the expected stable
case, only one location will contain most of the blocks as<a
name="line.405"></a>
-<span class="sourceLineNo">406</span> * local.<a name="line.406"></a>
-<span class="sourceLineNo">407</span> * On the other hand, in favored node
assignment, 3 nodes will contain highly local blocks. Here<a
name="line.407"></a>
-<span class="sourceLineNo">408</span> * we are doing a simple heuristic,
where we will pass all hosts which have at least 80%<a name="line.408"></a>
-<span class="sourceLineNo">409</span> *
(hbase.tablesnapshotinputformat.locality.cutoff.multiplier) as much block
locality as the top<a name="line.409"></a>
-<span class="sourceLineNo">410</span> * host with the best locality.<a
name="line.410"></a>
-<span class="sourceLineNo">411</span> */<a name="line.411"></a>
-<span class="sourceLineNo">412</span> public static List<String>
getBestLocations(<a name="line.412"></a>
-<span class="sourceLineNo">413</span> Configuration conf,
HDFSBlocksDistribution blockDistribution) {<a name="line.413"></a>
-<span class="sourceLineNo">414</span> List<String> locations = new
ArrayList<>(3);<a name="line.414"></a>
-<span class="sourceLineNo">415</span><a name="line.415"></a>
-<span class="sourceLineNo">416</span> HostAndWeight[] hostAndWeights =
blockDistribution.getTopHostsWithWeights();<a name="line.416"></a>
-<span class="sourceLineNo">417</span><a name="line.417"></a>
-<span class="sourceLineNo">418</span> if (hostAndWeights.length == 0) {<a
name="line.418"></a>
-<span class="sourceLineNo">419</span> return locations;<a
name="line.419"></a>
-<span class="sourceLineNo">420</span> }<a name="line.420"></a>
+<span class="sourceLineNo">359</span> public static List<InputSplit>
getSplits(Scan scan, SnapshotManifest manifest,<a name="line.359"></a>
+<span class="sourceLineNo">360</span> List<HRegionInfo>
regionManifests, Path restoreDir, Configuration conf) throws IOException {<a
name="line.360"></a>
+<span class="sourceLineNo">361</span> return getSplits(scan, manifest,
regionManifests, restoreDir, conf, null, 1);<a name="line.361"></a>
+<span class="sourceLineNo">362</span> }<a name="line.362"></a>
+<span class="sourceLineNo">363</span><a name="line.363"></a>
+<span class="sourceLineNo">364</span> public static List<InputSplit>
getSplits(Scan scan, SnapshotManifest manifest,<a name="line.364"></a>
+<span class="sourceLineNo">365</span> List<HRegionInfo>
regionManifests, Path restoreDir,<a name="line.365"></a>
+<span class="sourceLineNo">366</span> Configuration conf,
RegionSplitter.SplitAlgorithm sa, int numSplits) throws IOException {<a
name="line.366"></a>
+<span class="sourceLineNo">367</span> // load table descriptor<a
name="line.367"></a>
+<span class="sourceLineNo">368</span> TableDescriptor htd =
manifest.getTableDescriptor();<a name="line.368"></a>
+<span class="sourceLineNo">369</span><a name="line.369"></a>
+<span class="sourceLineNo">370</span> Path tableDir =
FSUtils.getTableDir(restoreDir, htd.getTableName());<a name="line.370"></a>
+<span class="sourceLineNo">371</span><a name="line.371"></a>
+<span class="sourceLineNo">372</span> boolean localityEnabled =
conf.getBoolean(SNAPSHOT_INPUTFORMAT_LOCALITY_ENABLED_KEY,<a
name="line.372"></a>
+<span class="sourceLineNo">373</span>
SNAPSHOT_INPUTFORMAT_LOCALITY_ENABLED_DEFAULT);<a name="line.373"></a>
+<span class="sourceLineNo">374</span><a name="line.374"></a>
+<span class="sourceLineNo">375</span> List<InputSplit> splits = new
ArrayList<>();<a name="line.375"></a>
+<span class="sourceLineNo">376</span> for (HRegionInfo hri :
regionManifests) {<a name="line.376"></a>
+<span class="sourceLineNo">377</span> // load region descriptor<a
name="line.377"></a>
+<span class="sourceLineNo">378</span><a name="line.378"></a>
+<span class="sourceLineNo">379</span> if (numSplits > 1) {<a
name="line.379"></a>
+<span class="sourceLineNo">380</span> byte[][] sp =
sa.split(hri.getStartKey(), hri.getEndKey(), numSplits, true);<a
name="line.380"></a>
+<span class="sourceLineNo">381</span> for (int i = 0; i < sp.length
- 1; i++) {<a name="line.381"></a>
+<span class="sourceLineNo">382</span> if
(PrivateCellUtil.overlappingKeys(scan.getStartRow(), scan.getStopRow(),
sp[i],<a name="line.382"></a>
+<span class="sourceLineNo">383</span> sp[i + 1])) {<a
name="line.383"></a>
+<span class="sourceLineNo">384</span> List<String> hosts =<a
name="line.384"></a>
+<span class="sourceLineNo">385</span>
calculateLocationsForInputSplit(conf, htd, hri, tableDir, localityEnabled);<a
name="line.385"></a>
+<span class="sourceLineNo">386</span><a name="line.386"></a>
+<span class="sourceLineNo">387</span> Scan boundedScan = new
Scan(scan);<a name="line.387"></a>
+<span class="sourceLineNo">388</span>
boundedScan.setStartRow(sp[i]);<a name="line.388"></a>
+<span class="sourceLineNo">389</span> boundedScan.setStopRow(sp[i +
1]);<a name="line.389"></a>
+<span class="sourceLineNo">390</span><a name="line.390"></a>
+<span class="sourceLineNo">391</span> splits.add(new
InputSplit(htd, hri, hosts, boundedScan, restoreDir));<a name="line.391"></a>
+<span class="sourceLineNo">392</span> }<a name="line.392"></a>
+<span class="sourceLineNo">393</span> }<a name="line.393"></a>
+<span class="sourceLineNo">394</span> } else {<a name="line.394"></a>
+<span class="sourceLineNo">395</span> if
(PrivateCellUtil.overlappingKeys(scan.getStartRow(), scan.getStopRow(),<a
name="line.395"></a>
+<span class="sourceLineNo">396</span> hri.getStartKey(),
hri.getEndKey())) {<a name="line.396"></a>
+<span class="sourceLineNo">397</span> List<String> hosts =<a
name="line.397"></a>
+<span class="sourceLineNo">398</span>
calculateLocationsForInputSplit(conf, htd, hri, tableDir, localityEnabled);<a
name="line.398"></a>
+<span class="sourceLineNo">399</span> splits.add(new InputSplit(htd,
hri, hosts, scan, restoreDir));<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> }<a name="line.402"></a>
+<span class="sourceLineNo">403</span><a name="line.403"></a>
+<span class="sourceLineNo">404</span> return splits;<a name="line.404"></a>
+<span class="sourceLineNo">405</span> }<a name="line.405"></a>
+<span class="sourceLineNo">406</span><a name="line.406"></a>
+<span class="sourceLineNo">407</span> /**<a name="line.407"></a>
+<span class="sourceLineNo">408</span> * Compute block locations for snapshot
files (which will get the locations for referred hfiles)<a name="line.408"></a>
+<span class="sourceLineNo">409</span> * only when localityEnabled is true.<a
name="line.409"></a>
+<span class="sourceLineNo">410</span> */<a name="line.410"></a>
+<span class="sourceLineNo">411</span> private static List<String>
calculateLocationsForInputSplit(Configuration conf,<a name="line.411"></a>
+<span class="sourceLineNo">412</span> TableDescriptor htd, HRegionInfo
hri, Path tableDir, boolean localityEnabled)<a name="line.412"></a>
+<span class="sourceLineNo">413</span> throws IOException {<a
name="line.413"></a>
+<span class="sourceLineNo">414</span> if (localityEnabled) { // care block
locality<a name="line.414"></a>
+<span class="sourceLineNo">415</span> return getBestLocations(conf,<a
name="line.415"></a>
+<span class="sourceLineNo">416</span>
HRegion.computeHDFSBlocksDistribution(conf, htd, hri, tableDir));<a
name="line.416"></a>
+<span class="sourceLineNo">417</span> } else { // do not care block
locality<a name="line.417"></a>
+<span class="sourceLineNo">418</span> return null;<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><a name="line.421"></a>
-<span class="sourceLineNo">422</span> HostAndWeight topHost =
hostAndWeights[0];<a name="line.422"></a>
-<span class="sourceLineNo">423</span> locations.add(topHost.getHost());<a
name="line.423"></a>
-<span class="sourceLineNo">424</span><a name="line.424"></a>
-<span class="sourceLineNo">425</span> // Heuristic: filter all hosts which
have at least cutoffMultiplier % of block locality<a name="line.425"></a>
-<span class="sourceLineNo">426</span> double cutoffMultiplier<a
name="line.426"></a>
-<span class="sourceLineNo">427</span> =
conf.getFloat(LOCALITY_CUTOFF_MULTIPLIER,
DEFAULT_LOCALITY_CUTOFF_MULTIPLIER);<a name="line.427"></a>
-<span class="sourceLineNo">428</span><a name="line.428"></a>
-<span class="sourceLineNo">429</span> double filterWeight =
topHost.getWeight() * cutoffMultiplier;<a name="line.429"></a>
-<span class="sourceLineNo">430</span><a name="line.430"></a>
-<span class="sourceLineNo">431</span> for (int i = 1; i <
hostAndWeights.length; i++) {<a name="line.431"></a>
-<span class="sourceLineNo">432</span> if (hostAndWeights[i].getWeight()
>= filterWeight) {<a name="line.432"></a>
-<span class="sourceLineNo">433</span>
locations.add(hostAndWeights[i].getHost());<a name="line.433"></a>
-<span class="sourceLineNo">434</span> } else {<a name="line.434"></a>
-<span class="sourceLineNo">435</span> break;<a name="line.435"></a>
-<span class="sourceLineNo">436</span> }<a name="line.436"></a>
-<span class="sourceLineNo">437</span> }<a name="line.437"></a>
+<span class="sourceLineNo">422</span> /**<a name="line.422"></a>
+<span class="sourceLineNo">423</span> * This computes the locations to be
passed from the InputSplit. MR/Yarn schedulers does not take<a
name="line.423"></a>
+<span class="sourceLineNo">424</span> * weights into account, thus will
treat every location passed from the input split as equal. We<a
name="line.424"></a>
+<span class="sourceLineNo">425</span> * do not want to blindly pass all the
locations, since we are creating one split per region, and<a
name="line.425"></a>
+<span class="sourceLineNo">426</span> * the region's blocks are all
distributed throughout the cluster unless favorite node assignment<a
name="line.426"></a>
+<span class="sourceLineNo">427</span> * is used. On the expected stable
case, only one location will contain most of the blocks as<a
name="line.427"></a>
+<span class="sourceLineNo">428</span> * local.<a name="line.428"></a>
+<span class="sourceLineNo">429</span> * On the other hand, in favored node
assignment, 3 nodes will contain highly local blocks. Here<a
name="line.429"></a>
+<span class="sourceLineNo">430</span> * we are doing a simple heuristic,
where we will pass all hosts which have at least 80%<a name="line.430"></a>
+<span class="sourceLineNo">431</span> *
(hbase.tablesnapshotinputformat.locality.cutoff.multiplier) as much block
locality as the top<a name="line.431"></a>
+<span class="sourceLineNo">432</span> * host with the best locality.<a
name="line.432"></a>
+<span class="sourceLineNo">433</span> * Return at most numTopsAtMost
locations if there are more than that.<a name="line.433"></a>
+<span class="sourceLineNo">434</span> */<a name="line.434"></a>
+<span class="sourceLineNo">435</span> private static List<String>
getBestLocations(Configuration conf,<a name="line.435"></a>
+<span class="sourceLineNo">436</span> HDFSBlocksDistribution
blockDistribution, int numTopsAtMost) {<a name="line.436"></a>
+<span class="sourceLineNo">437</span> HostAndWeight[] hostAndWeights =
blockDistribution.getTopHostsWithWeights();<a name="line.437"></a>
<span class="sourceLineNo">438</span><a name="line.438"></a>
-<span class="sourceLineNo">439</span> return locations;<a
name="line.439"></a>
-<span class="sourceLineNo">440</span> }<a name="line.440"></a>
-<span class="sourceLineNo">441</span><a name="line.441"></a>
-<span class="sourceLineNo">442</span> private static String
getSnapshotName(Configuration conf) {<a name="line.442"></a>
-<span class="sourceLineNo">443</span> String snapshotName =
conf.get(SNAPSHOT_NAME_KEY);<a name="line.443"></a>
-<span class="sourceLineNo">444</span> if (snapshotName == null) {<a
name="line.444"></a>
-<span class="sourceLineNo">445</span> throw new
IllegalArgumentException("Snapshot name must be provided");<a
name="line.445"></a>
-<span class="sourceLineNo">446</span> }<a name="line.446"></a>
-<span class="sourceLineNo">447</span> return snapshotName;<a
name="line.447"></a>
-<span class="sourceLineNo">448</span> }<a name="line.448"></a>
-<span class="sourceLineNo">449</span><a name="line.449"></a>
-<span class="sourceLineNo">450</span> /**<a name="line.450"></a>
-<span class="sourceLineNo">451</span> * Configures the job to use
TableSnapshotInputFormat to read from a snapshot.<a name="line.451"></a>
-<span class="sourceLineNo">452</span> * @param conf the job to
configuration<a name="line.452"></a>
-<span class="sourceLineNo">453</span> * @param snapshotName the name of the
snapshot to read from<a name="line.453"></a>
-<span class="sourceLineNo">454</span> * @param restoreDir a temporary
directory to restore the snapshot into. Current user should<a
name="line.454"></a>
-<span class="sourceLineNo">455</span> * have write permissions to this
directory, and this should not be a subdirectory of rootdir.<a
name="line.455"></a>
-<span class="sourceLineNo">456</span> * After the job is finished,
restoreDir can be deleted.<a name="line.456"></a>
-<span class="sourceLineNo">457</span> * @throws IOException if an error
occurs<a name="line.457"></a>
-<span class="sourceLineNo">458</span> */<a name="line.458"></a>
-<span class="sourceLineNo">459</span> public static void
setInput(Configuration conf, String snapshotName, Path restoreDir)<a
name="line.459"></a>
-<span class="sourceLineNo">460</span> throws IOException {<a
name="line.460"></a>
-<span class="sourceLineNo">461</span> setInput(conf, snapshotName,
restoreDir, null, 1);<a name="line.461"></a>
-<span class="sourceLineNo">462</span> }<a name="line.462"></a>
-<span class="sourceLineNo">463</span><a name="line.463"></a>
-<span class="sourceLineNo">464</span> /**<a name="line.464"></a>
-<span class="sourceLineNo">465</span> * Configures the job to use
TableSnapshotInputFormat to read from a snapshot.<a name="line.465"></a>
-<span class="sourceLineNo">466</span> * @param conf the job to configure<a
name="line.466"></a>
-<span class="sourceLineNo">467</span> * @param snapshotName the name of the
snapshot to read from<a name="line.467"></a>
-<span class="sourceLineNo">468</span> * @param restoreDir a temporary
directory to restore the snapshot into. Current user should<a
name="line.468"></a>
-<span class="sourceLineNo">469</span> * have write permissions to this
directory, and this should not be a subdirectory of rootdir.<a
name="line.469"></a>
-<span class="sourceLineNo">470</span> * After the job is finished,
restoreDir can be deleted.<a name="line.470"></a>
-<span class="sourceLineNo">471</span> * @param numSplitsPerRegion how many
input splits to generate per one region<a name="line.471"></a>
-<span class="sourceLineNo">472</span> * @param splitAlgo SplitAlgorithm to
be used when generating InputSplits<a name="line.472"></a>
-<span class="sourceLineNo">473</span> * @throws IOException if an error
occurs<a name="line.473"></a>
-<span class="sourceLineNo">474</span> */<a name="line.474"></a>
-<span class="sourceLineNo">475</span> public static void
setInput(Configuration conf, String snapshotName, Path restoreDir,<a
name="line.475"></a>
-<span class="sourceLineNo">476</span>
RegionSplitter.SplitAlgorithm splitAlgo, int numSplitsPerRegion)<a
name="line.476"></a>
-<span class="sourceLineNo">477</span> throws IOException {<a
name="line.477"></a>
-<span class="sourceLineNo">478</span> conf.set(SNAPSHOT_NAME_KEY,
snapshotName);<a name="line.478"></a>
-<span class="sourceLineNo">479</span> if (numSplitsPerRegion < 1) {<a
name="line.479"></a>
-<span class="sourceLineNo">480</span> throw new
IllegalArgumentException("numSplits must be >= 1, " +<a name="line.480"></a>
-<span class="sourceLineNo">481</span> "illegal numSplits : " +
numSplitsPerRegion);<a name="line.481"></a>
-<span class="sourceLineNo">482</span> }<a name="line.482"></a>
-<span class="sourceLineNo">483</span> if (splitAlgo == null &&
numSplitsPerRegion > 1) {<a name="line.483"></a>
-<span class="sourceLineNo">484</span> throw new
IllegalArgumentException("Split algo can't be null when numSplits > 1");<a
name="line.484"></a>
+<span class="sourceLineNo">439</span> if (hostAndWeights.length == 0) { //
no matter what numTopsAtMost is<a name="line.439"></a>
+<span class="sourceLineNo">440</span> return null;<a name="line.440"></a>
+<span class="sourceLineNo">441</span> }<a name="line.441"></a>
+<span class="sourceLineNo">442</span><a name="line.442"></a>
+<span class="sourceLineNo">443</span> if (numTopsAtMost < 1) { //
invalid if numTopsAtMost < 1, correct it to be 1<a name="line.443"></a>
+<span class="sourceLineNo">444</span> numTopsAtMost = 1;<a
name="line.444"></a>
+<span class="sourceLineNo">445</span> }<a name="line.445"></a>
+<span class="sourceLineNo">446</span> int top = Math.min(numTopsAtMost,
hostAndWeights.length);<a name="line.446"></a>
+<span class="sourceLineNo">447</span> List<String> locations = new
ArrayList<>(top);<a name="line.447"></a>
+<span class="sourceLineNo">448</span> HostAndWeight topHost =
hostAndWeights[0];<a name="line.448"></a>
+<span class="sourceLineNo">449</span> locations.add(topHost.getHost());<a
name="line.449"></a>
+<span class="sourceLineNo">450</span><a name="line.450"></a>
+<span class="sourceLineNo">451</span> if (top == 1) { // only care about
the top host<a name="line.451"></a>
+<span class="sourceLineNo">452</span> return locations;<a
name="line.452"></a>
+<span class="sourceLineNo">453</span> }<a name="line.453"></a>
+<span class="sourceLineNo">454</span><a name="line.454"></a>
+<span class="sourceLineNo">455</span> // When top >= 2,<a
name="line.455"></a>
+<span class="sourceLineNo">456</span> // do the heuristic: filter all hosts
which have at least cutoffMultiplier % of block locality<a name="line.456"></a>
+<span class="sourceLineNo">457</span> double cutoffMultiplier<a
name="line.457"></a>
+<span class="sourceLineNo">458</span> =
conf.getFloat(LOCALITY_CUTOFF_MULTIPLIER,
DEFAULT_LOCALITY_CUTOFF_MULTIPLIER);<a name="line.458"></a>
+<span class="sourceLineNo">459</span><a name="line.459"></a>
+<span class="sourceLineNo">460</span> double filterWeight =
topHost.getWeight() * cutoffMultiplier;<a name="line.460"></a>
+<span class="sourceLineNo">461</span><a name="line.461"></a>
+<span class="sourceLineNo">462</span> for (int i = 1; i <= top - 1; i++)
{<a name="line.462"></a>
+<span class="sourceLineNo">463</span> if (hostAndWeights[i].getWeight()
>= filterWeight) {<a name="line.463"></a>
+<span class="sourceLineNo">464</span>
locations.add(hostAndWeights[i].getHost());<a name="line.464"></a>
+<span class="sourceLineNo">465</span> } else {<a name="line.465"></a>
+<span class="sourceLineNo">466</span> // As hostAndWeights is in
descending order,<a name="line.466"></a>
+<span class="sourceLineNo">467</span> // we could break the loop as
long as we meet a weight which is less than filterWeight.<a name="line.467"></a>
+<span class="sourceLineNo">468</span> break;<a name="line.468"></a>
+<span class="sourceLineNo">469</span> }<a name="line.469"></a>
+<span class="sourceLineNo">470</span> }<a name="line.470"></a>
+<span class="sourceLineNo">471</span><a name="line.471"></a>
+<span class="sourceLineNo">472</span> return locations;<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> public static List<String>
getBestLocations(Configuration conf,<a name="line.475"></a>
+<span class="sourceLineNo">476</span> HDFSBlocksDistribution
blockDistribution) {<a name="line.476"></a>
+<span class="sourceLineNo">477</span> // 3 nodes will contain highly local
blocks. So default to 3.<a name="line.477"></a>
+<span class="sourceLineNo">478</span> return getBestLocations(conf,
blockDistribution, 3);<a name="line.478"></a>
+<span class="sourceLineNo">479</span> }<a name="line.479"></a>
+<span class="sourceLineNo">480</span><a name="line.480"></a>
+<span class="sourceLineNo">481</span> private static String
getSnapshotName(Configuration conf) {<a name="line.481"></a>
+<span class="sourceLineNo">482</span> String snapshotName =
conf.get(SNAPSHOT_NAME_KEY);<a name="line.482"></a>
+<span class="sourceLineNo">483</span> if (snapshotName == null) {<a
name="line.483"></a>
+<span class="sourceLineNo">484</span> throw new
IllegalArgumentException("Snapshot name must be provided");<a
name="line.484"></a>
<span class="sourceLineNo">485</span> }<a name="line.485"></a>
-<span class="sourceLineNo">486</span> if (splitAlgo != null) {<a
name="line.486"></a>
-<span class="sourceLineNo">487</span> conf.set(SPLIT_ALGO,
splitAlgo.getClass().getName());<a name="line.487"></a>
-<span class="sourceLineNo">488</span> }<a name="line.488"></a>
-<span class="sourceLineNo">489</span> conf.setInt(NUM_SPLITS_PER_REGION,
numSplitsPerRegion);<a name="line.489"></a>
-<span class="sourceLineNo">490</span> Path rootDir =
FSUtils.getRootDir(conf);<a name="line.490"></a>
-<span class="sourceLineNo">491</span> FileSystem fs =
rootDir.getFileSystem(conf);<a name="line.491"></a>
-<span class="sourceLineNo">492</span><a name="line.492"></a>
-<span class="sourceLineNo">493</span> restoreDir = new Path(restoreDir,
UUID.randomUUID().toString());<a name="line.493"></a>
-<span class="sourceLineNo">494</span><a name="line.494"></a>
-<span class="sourceLineNo">495</span> // TODO: restore from record readers
to parallelize.<a name="line.495"></a>
-<span class="sourceLineNo">496</span>
RestoreSnapshotHelper.copySnapshotForScanner(conf, fs, rootDir, restoreDir,
snapshotName);<a name="line.496"></a>
-<span class="sourceLineNo">497</span><a name="line.497"></a>
-<span class="sourceLineNo">498</span> conf.set(RESTORE_DIR_KEY,
restoreDir.toString());<a name="line.498"></a>
-<span class="sourceLineNo">499</span> }<a name="line.499"></a>
-<span class="sourceLineNo">500</span>}<a name="line.500"></a>
+<span class="sourceLineNo">486</span> return snapshotName;<a
name="line.486"></a>
+<span class="sourceLineNo">487</span> }<a name="line.487"></a>
+<span class="sourceLineNo">488</span><a name="line.488"></a>
+<span class="sourceLineNo">489</span> /**<a name="line.489"></a>
+<span class="sourceLineNo">490</span> * Configures the job to use
TableSnapshotInputFormat to read from a snapshot.<a name="line.490"></a>
+<span class="sourceLineNo">491</span> * @param conf the job to
configuration<a name="line.491"></a>
+<span class="sourceLineNo">492</span> * @param snapshotName the name of the
snapshot to read from<a name="line.492"></a>
+<span class="sourceLineNo">493</span> * @param restoreDir a temporary
directory to restore the snapshot into. Current user should<a
name="line.493"></a>
+<span class="sourceLineNo">494</span> * have write permissions to this
directory, and this should not be a subdirectory of rootdir.<a
name="line.494"></a>
+<span class="sourceLineNo">495</span> * After the job is finished,
restoreDir can be deleted.<a name="line.495"></a>
+<span class="sourceLineNo">496</span> * @throws IOException if an error
occurs<a name="line.496"></a>
+<span class="sourceLineNo">497</span> */<a name="line.497"></a>
+<span class="sourceLineNo">498</span> public static void
setInput(Configuration conf, String snapshotName, Path restoreDir)<a
name="line.498"></a>
+<span class="sourceLineNo">499</span> throws IOException {<a
name="line.499"></a>
+<span class="sourceLineNo">500</span> setInput(conf, snapshotName,
restoreDir, null, 1);<a name="line.500"></a>
+<span class="sourceLineNo">501</span> }<a name="line.501"></a>
+<span class="sourceLineNo">502</span><a name="line.502"></a>
+<span class="sourceLineNo">503</span> /**<a name="line.503"></a>
+<span class="sourceLineNo">504</span> * Configures the job to use
TableSnapshotInputFormat to read from a snapshot.<a name="line.504"></a>
+<span class="sourceLineNo">505</span> * @param conf the job to configure<a
name="line.505"></a>
+<span class="sourceLineNo">506</span> * @param snapshotName the name of the
snapshot to read from<a name="line.506"></a>
+<span class="sourceLineNo">507</span> * @param restoreDir a temporary
directory to restore the snapshot into. Current user should<a
name="line.507"></a>
+<span class="sourceLineNo">508</span> * have write permissions to this
directory, and this should not be a subdirectory of rootdir.<a
name="line.508"></a>
+<span class="sourceLineNo">509</span> * After the job is finished,
restoreDir can be deleted.<a name="line.509"></a>
+<span class="sourceLineNo">510</span> * @param numSplitsPerRegion how many
input splits to generate per one region<a name="line.510"></a>
+<span class="sourceLineNo">511</span> * @param splitAlgo SplitAlgorithm to
be used when generating InputSplits<a name="line.511"></a>
+<span class="sourceLineNo">512</span> * @throws IOException if an error
occurs<a name="line.512"></a>
+<span class="sourceLineNo">513</span> */<a name="line.513"></a>
+<span class="sourceLineNo">514</span> public static void
setInput(Configuration conf, String snapshotName, Path restoreDir,<a
name="line.514"></a>
+<span class="sourceLineNo">515</span>
RegionSplitter.SplitAlgorithm splitAlgo, int numSplitsPerRegion)<a
name="line.515"></a>
+<span class="sourceLineNo">516</span> throws IOException {<a
name="line.516"></a>
+<span class="sourceLineNo">517</span> conf.set(SNAPSHOT_NAME_KEY,
snapshotName);<a name="line.517"></a>
+<span class="sourceLineNo">518</span> if (numSplitsPerRegion < 1) {<a
name="line.518"></a>
+<span class="sourceLineNo">519</span> throw new
IllegalArgumentException("numSplits must be >= 1, " +<a name="line.519"></a>
+<span class="sourceLineNo">520</span> "illegal numSplits : " +
numSplitsPerRegion);<a name="line.520"></a>
+<span class="sourceLineNo">521</span> }<a name="line.521"></a>
+<span class="sourceLineNo">522</span> if (splitAlgo == null &&
numSplitsPerRegion > 1) {<a name="line.522"></a>
+<span class="sourceLineNo">523</span> throw new
IllegalArgumentException("Split algo can't be null when numSplits > 1");<a
name="line.523"></a>
+<span class="sourceLineNo">524</span> }<a name="line.524"></a>
+<span class="sourceLineNo">525</span> if (splitAlgo != null) {<a
name="line.525"></a>
+<span class="sourceLineNo">526</span> conf.set(SPLIT_ALGO,
splitAlgo.getClass().getName());<a name="line.526"></a>
+<span class="sourceLineNo">527</span> }<a name="line.527"></a>
+<span class="sourceLineNo">528</span> conf.setInt(NUM_SPLITS_PER_REGION,
numSplitsPerRegion);<a name="line.528"></a>
+<span class="sourceLineNo">529</span> Path rootDir =
FSUtils.getRootDir(conf);<a name="line.529"></a>
+<span class="sourceLineNo">530</span> FileSystem fs =
rootDir.getFileSystem(conf);<a name="line.530"></a>
+<span class="sourceLineNo">531</span><a name="line.531"></a>
+<span class="sourceLineNo">532</span> restoreDir = new Path(restoreDir,
UUID.randomUUID().toString());<a name="line.532"></a>
+<span class="sourceLineNo">533</span><a name="line.533"></a>
+<span class="sourceLineNo">534</span> // TODO: restore from record readers
to parallelize.<a name="line.534"></a>
+<span class="sourceLineNo">535</span>
RestoreSnapshotHelper.copySnapshotForScanner(conf, fs, rootDir, restoreDir,
snapshotName);<a name="line.535"></a>
+<span class="sourceLineNo">536</span><a name="line.536"></a>
+<span class="sourceLineNo">537</span> conf.set(RESTORE_DIR_KEY,
restoreDir.toString());<a name="line.537"></a>
+<span class="sourceLineNo">538</span> }<a name="line.538"></a>
+<span class="sourceLineNo">539</span>}<a name="line.539"></a>
