关于二次排序主要涉及到这么几个东西:
在0.20.0 以前使用的是
setPartitionerClass
setOutputkeyComparatorClass
setOutputValueGroupingComparator
在0.20.0以后使用是
job.setPartitionerClass(Partitioner p);
job.setSortComparatorClass(RawComparator c);
job.setGroupingComparatorClass(RawComparator c);
下面的例子里面只用到了 setGroupingComparatorClass
mr自带的例子中的源码SecondarySort,我重新写了一下,基本没变。
这个例子中定义的map和reduce如下,关键是它对输入输出类型的定义:(java泛型编程)
public static class Map extends Mapper<LongWritable, Text, IntPair, IntWritable>
public static class Reduce extends Reducer<IntPair, NullWritable, IntWritable, IntWritable>
就是首先按照第一字段排序,然后再对第一字段相同的行按照第二字段排序,注意不能破坏第一次排序 的结果 。例如 :
echo "3 b
1 c
2 a
1 d
3 a"|sort -k1 -k2
1 c
1 d
2 a
3 a
3 b
2.1 分区函数类。这是key的第一次比较。
public static class FirstPartitioner extends Partitioner<IntPair,IntWritable>这个例子中没有使用key比较函数类,而是使用key的实现的compareTo方法:
package SecondarySort; import java.io.DataInput; import java.io.DataOutput; import java.io.IOException; import org.apache.hadoop.conf.Configuration; import org.apache.hadoop.fs.FileSystem; import org.apache.hadoop.fs.Path; import org.apache.hadoop.io.LongWritable; import org.apache.hadoop.io.Text; import org.apache.hadoop.io.WritableComparable; import org.apache.hadoop.io.WritableComparator; import org.apache.hadoop.mapreduce.Job; import org.apache.hadoop.mapreduce.Mapper; import org.apache.hadoop.mapreduce.Partitioner; import org.apache.hadoop.mapreduce.Reducer; import org.apache.hadoop.mapreduce.lib.input.FileInputFormat; import org.apache.hadoop.mapreduce.lib.input.TextInputFormat; import org.apache.hadoop.mapreduce.lib.output.FileOutputFormat; import org.apache.hadoop.mapreduce.lib.output.TextOutputFormat; public class SecondarySort { //自己定义的key类应该实现WritableComparable接口 public static class IntPair implements WritableComparable<IntPair> { String first; String second; /** * Set the left and right values. */ public void set(String left, String right) { first = left; second = right; } public String getFirst() { return first; } public String getSecond() { return second; } //反序列化,从流中的二进制转换成IntPair public void readFields(DataInput in) throws IOException { first = in.readUTF(); second = in.readUTF(); } //序列化,将IntPair转化成使用流传送的二进制 public void write(DataOutput out) throws IOException { out.writeUTF(first); out.writeUTF(second); } //重载 compareTo 方法,进行组合键 key 的比较,该过程是默认行为。 //分组后的二次排序会隐式调用该方法。 public int compareTo(IntPair o) { if (!first.equals(o.first) ) { return first.compareTo(o.first); } else if (!second.equals(o.second)) { return second.compareTo(o.second); } else { return 0; } } //新定义类应该重写的两个方法 //The hashCode() method is used by the HashPartitioner (the default partitioner in MapReduce) public int hashCode() { return first.hashCode() * 157 + second.hashCode(); } public boolean equals(Object right) { if (right == null) return false; if (this == right) return true; if (right instanceof IntPair) { IntPair r = (IntPair) right; return r.first.equals(first) && r.second.equals(second) ; } else { return false; } } } /** * 分区函数类。根据first确定Partition。 */ public static class FirstPartitioner extends Partitioner<IntPair, Text> { public int getPartition(IntPair key, Text value,int numPartitions) { return Math.abs(key.getFirst().hashCode() * 127) % numPartitions; } } /** * 分组函数类。只要first相同就属于同一个组。 */ /*//第一种方法,实现接口RawComparator public static class GroupingComparator implements RawComparator<IntPair> { public int compare(IntPair o1, IntPair o2) { int l = o1.getFirst(); int r = o2.getFirst(); return l == r ? 0 : (l < r ? -1 : 1); } //一个字节一个字节的比,直到找到一个不相同的字节,然后比这个字节的大小作为两个字节流的大小比较结果。 public int compare(byte[] b1, int s1, int l1, byte[] b2, int s2, int l2){ return WritableComparator.compareBytes(b1, s1, Integer.SIZE/8, b2, s2, Integer.SIZE/8); } }*/ //第二种方法,继承WritableComparator public static class GroupingComparator extends WritableComparator { protected GroupingComparator() { super(IntPair.class, true); } //Compare two WritableComparables. // 重载 compare:对组合键按第一个自然键排序分组 public int compare(WritableComparable w1, WritableComparable w2) { IntPair ip1 = (IntPair) w1; IntPair ip2 = (IntPair) w2; String l = ip1.getFirst(); String r = ip2.getFirst(); return l.compareTo(r); } } // 自定义map public static class Map extends Mapper<LongWritable, Text, IntPair, Text> { private final IntPair keyPair = new IntPair(); String[] lineArr = null; public void map(LongWritable key, Text value, Context context) throws IOException, InterruptedException { String line = value.toString(); lineArr = line.split("\t", -1); keyPair.set(lineArr[0], lineArr[1]); context.write(keyPair, value); } } // 自定义reduce // public static class Reduce extends Reducer<IntPair, Text, Text, Text> { private static final Text SEPARATOR = new Text("------------------------------------------------"); public void reduce(IntPair key, Iterable<Text> values,Context context) throws IOException, InterruptedException { context.write(SEPARATOR, null); for (Text val : values) { context.write(null, val); } } } public static void main(String[] args) throws IOException, InterruptedException, ClassNotFoundException { // 读取hadoop配置 Configuration conf = new Configuration(); // 实例化一道作业 Job job = new Job(conf, "secondarysort"); job.setJarByClass(SecondarySort.class); // Mapper类型 job.setMapperClass(Map.class); // 不再需要Combiner类型,因为Combiner的输出类型<Text, IntWritable>对Reduce的输入类型<IntPair, IntWritable>不适用 //job.setCombinerClass(Reduce.class); // Reducer类型 job.setReducerClass(Reduce.class); // 分区函数 job.setPartitionerClass(FirstPartitioner.class); // 分组函数 job.setGroupingComparatorClass(GroupingComparator.class); // map 输出Key的类型 job.setMapOutputKeyClass(IntPair.class); // map输出Value的类型 job.setMapOutputValueClass(Text.class); // rduce输出Key的类型,是Text,因为使用的OutputFormatClass是TextOutputFormat job.setOutputKeyClass(Text.class); // rduce输出Value的类型 job.setOutputValueClass(Text.class); // 将输入的数据集分割成小数据块splites,同时提供一个RecordReder的实现。 job.setInputFormatClass(TextInputFormat.class); // 提供一个RecordWriter的实现,负责数据输出。 job.setOutputFormatClass(TextOutputFormat.class); // 输入hdfs路径 FileInputFormat.setInputPaths(job, new Path(args[0])); // 输出hdfs路径 FileSystem.get(conf).delete(new Path(args[1]), true); FileOutputFormat.setOutputPath(job, new Path(args[1])); // 提交job System.exit(job.waitForCompletion(true) ? 0 : 1); } }
假如我们现在的需求是先按 cookieId 排序,然后按 time 排序,以便按 session 切分日志
cookieId time url 2 12:12:34 2_hao123 3 09:10:34 3_baidu 1 15:02:41 1_google 3 22:11:34 3_sougou 1 19:10:34 1_baidu 2 15:02:41 2_google 1 12:12:34 1_hao123 3 23:10:34 3_soso 2 05:02:41 2_google 结果: ------------------------------------------------ 1 12:12:34 1_hao123 1 15:02:41 1_google 1 19:10:34 1_baidu ------------------------------------------------ 2 05:02:41 2_google 2 12:12:34 2_hao123 2 15:02:41 2_google ------------------------------------------------ 3 09:10:34 3_baidu 3 22:11:34 3_sougou 3 23:10:34 3_soso
hive中使用标准sql实现分组内排序
http://superlxw1234.iteye.com/blog/1869612
Pig、Hive、MapReduce 解决分组 Top K 问题
http://my.oschina.net/leejun2005/blog/85187
mapreduce的二次排序 SecondarySort
http://blog.csdn.net/zyj8170/article/details/7530728
学会定制MapReduce里的partition,sort和grouping,Secondary Sort Made Easy 进行二次排序
http://blog.sina.com.cn/s/blog_9bf980ad0100zk7r.html
Simple Moving Average, Secondary Sort, and MapReduce (Part 3)
http://blog.cloudera.com/blog/2011/04/simple-moving-average-secondary-sort-and-mapreduce-part-3/
https://github.com/jpatanooga/Caduceus/tree/master/src/tv/floe/caduceus/hadoop/movingaverage
MapReduce的排序和二次排序原理总结
http://hugh-wangp.iteye.com/blog/1491175
泛型value的二次排序
http://wenku.baidu.com/view/a3826a235901020207409c47.html
http://vangjee.wordpress.com/2012/03/20/secondary-sorting-aka-sorting-values-in-hadoops-mapreduce-programming-paradigm/