MapReduce常见算法
•单词计数
•数据去重
•排序
•Top K
•选择　　以求最值为例，从100万数据中选出一行最小值
•投影　　以求处理手机上网日志为例，从其11个字段选出了五个字段(列)来显示我们的手机上网流量
•分组　　相当于分区，以求处理手机上网日志为例，喊手机号和非手机号分为两组
•多表连接
•单表关联

使用TopK算法找出文件中的最大数：

复制代码
1 package suanfa;
2
3 import java.io.IOException;
4 import java.net.URI;
5
6 import org.apache.hadoop.conf.Configuration;
7 import org.apache.hadoop.fs.FileSystem;
8 import org.apache.hadoop.fs.Path;
9 import org.apache.hadoop.io.LongWritable;
10 import org.apache.hadoop.io.NullWritable;
11 import org.apache.hadoop.io.Text;
12 import org.apache.hadoop.mapreduce.Job;
13 import org.apache.hadoop.mapreduce.Mapper;
14 import org.apache.hadoop.mapreduce.Reducer;
15 import org.apache.hadoop.mapreduce.lib.input.FileInputFormat;
16 import org.apache.hadoop.mapreduce.lib.output.FileOutputFormat;
17
18 /**
19 * 找出文件中的最大数
20 *
21 * @author ahu_lichang
22 *
23 */
24 public class TopKApp {
25 static final String INPUT_PATH = “hdfs://chaoren:9000/input”;
26 static final String OUT_PATH = “hdfs://chaoren:9000/out”;
27
28 public static void main(String[] args) throws Exception {
29 Configuration conf = new Configuration();
30 FileSystem fileSystem = FileSystem.get(new URI(INPUT_PATH), conf);
31 Path outPath = new Path(OUT_PATH);
32 if (fileSystem.exists(outPath)) {
33 fileSystem.delete(outPath, true);
34 }
35 Job job = new Job(conf, TopKApp.class.getSimpleName());
36 FileInputFormat.setInputPaths(job, INPUT_PATH);
37 job.setMapperClass(MyMapper.class);
38 job.setReducerClass(MyReducer.class);
39 job.setOutputKeyClass(LongWritable.class);
40 job.setOutputValueClass(NullWritable.class);
41 FileOutputFormat.setOutputPath(job, outPath);
42 job.waitForCompletion(true);
43 }
44
45 static class MyMapper extends
46 Mapper<LongWritable, Text, LongWritable, NullWritable> {
47 long max = Long.MIN_VALUE;
48
49 protected void map(LongWritable k1, Text v1, Context context)
50 throws java.io.IOException, InterruptedException {
51 long temp = Long.parseLong(v1.toString());
52 if (temp > max) {
53 max = temp;
54 }
55 }
56
57 @Override
58 protected void cleanup(
59 org.apache.hadoop.mapreduce.Mapper<LongWritable, Text, LongWritable, NullWritable>.Context context)
60 throws IOException, InterruptedException {
61 context.write(new LongWritable(max), NullWritable.get());
62 }
63 }
64
65 static class MyReducer extends
66 Reducer<LongWritable, NullWritable, LongWritable, NullWritable> {
67 long max = Long.MIN_VALUE;
68
69 protected void reduce(
70 LongWritable k2,
71 Iterable v2s,
72 org.apache.hadoop.mapreduce.Reducer<LongWritable, NullWritable, LongWritable, NullWritable>.Context context)
73 throws IOException, InterruptedException {
74 long temp = k2.get();
75 if (temp > max) {
76 max = temp;
77 }
78 }
79
80 @Override
81 protected void cleanup(
82 org.apache.hadoop.mapreduce.Reducer<LongWritable, NullWritable, LongWritable, NullWritable>.Context context)
83 throws IOException, InterruptedException {
84 context.write(new LongWritable(max), NullWritable.get());
85 }
86 }
87 }

复制代码
遇见一个问题：在删除HDFS中的文件的时候，说文件时安全模式下，无法删除？

这时候要想删除该文件，必须退出安全模式，Hadoop退出安全模式的命令是：hadoop dfsadmin -safemode leave

自定义排序

将两列数据进行排序，第一列按照升序排列，当第一列相同时，第二列升序排列。

在map和reduce阶段进行排序时，比较的是k2。v2是不参与排序比较的。如果要想让v2也进行排序，需要把k2和v2组装成新的类，作为k2，才能参与比较。

复制代码
1 package sort;
2
3 import java.io.DataInput;
4 import java.io.DataOutput;
5 import java.io.IOException;
6 import java.net.URI;
7
8 import org.apache.hadoop.conf.Configuration;
9 import org.apache.hadoop.fs.FileSystem;
10 import org.apache.hadoop.fs.Path;
11 import org.apache.hadoop.io.LongWritable;
12 import org.apache.hadoop.io.Text;
13 import org.apache.hadoop.io.WritableComparable;
14 import org.apache.hadoop.mapreduce.Job;
15 import org.apache.hadoop.mapreduce.Mapper;
16 import org.apache.hadoop.mapreduce.Reducer;
17 import org.apache.hadoop.mapreduce.lib.input.FileInputFormat;
18 import org.apache.hadoop.mapreduce.lib.input.TextInputFormat;
19 import org.apache.hadoop.mapreduce.lib.output.FileOutputFormat;
20 import org.apache.hadoop.mapreduce.lib.output.TextOutputFormat;
21 import org.apache.hadoop.mapreduce.lib.partition.HashPartitioner;
22
23 public class SortApp {
24 static final String INPUT_PATH = “hdfs://chaoren:9000/input”;
25 static final String OUT_PATH = “hdfs://chaoren:9000/out”;
26
27 public static void main(String[] args) throws Exception {
28 final Configuration configuration = new Configuration();
29
30 final FileSystem fileSystem = FileSystem.get(new URI(INPUT_PATH),
31 configuration);
32 if (fileSystem.exists(new Path(OUT_PATH))) {
33 fileSystem.delete(new Path(OUT_PATH), true);
34 }
35
36 final Job job = new Job(configuration, SortApp.class.getSimpleName());
37
38 // 1.1 指定输入文件路径
39 FileInputFormat.setInputPaths(job, INPUT_PATH);
40 // 指定哪个类用来格式化输入文件
41 job.setInputFormatClass(TextInputFormat.class);
42
43 // 1.2指定自定义的Mapper类
44 job.setMapperClass(MyMapper.class);
45 // 指定输出<k2,v2>的类型
46 job.setMapOutputKeyClass(NewK2.class);
47 job.setMapOutputValueClass(LongWritable.class);
48
49 // 1.3 指定分区类
50 job.setPartitionerClass(HashPartitioner.class);
51 job.setNumReduceTasks(1);
52
53 // 1.4 TODO 排序、分区
54
55 // 1.5 TODO （可选）合并
56
57 // 2.2 指定自定义的reduce类
58 job.setReducerClass(MyReducer.class);
59 // 指定输出<k3,v3>的类型
60 job.setOutputKeyClass(LongWritable.class);
61 job.setOutputValueClass(LongWritable.class);
62
63 // 2.3 指定输出到哪里
64 FileOutputFormat.setOutputPath(job, new Path(OUT_PATH));
65 // 设定输出文件的格式化类
66 job.setOutputFormatClass(TextOutputFormat.class);
67
68 // 把代码提交给JobTracker执行
69 job.waitForCompletion(true);
70 }
71
72 static class MyMapper extends
73 Mapper<LongWritable, Text, NewK2, LongWritable> {
74 protected void map(
75 LongWritable key,
76 Text value,
77 org.apache.hadoop.mapreduce.Mapper<LongWritable, Text, NewK2, LongWritable>.Context context)
78 throws java.io.IOException, InterruptedException {
79 final String[] splited = value.toString().split("\t");
80 final NewK2 k2 = new NewK2(Long.parseLong(splited[0]),
81 Long.parseLong(splited[1]));
82 final LongWritable v2 = new LongWritable(Long.parseLong(splited[1]));
83 context.write(k2, v2);
84 };
85 }
86
87 static class MyReducer extends
88 Reducer<NewK2, LongWritable, LongWritable, LongWritable> {
89 protected void reduce(
90 NewK2 k2,
91 java.lang.Iterable v2s,
92 org.apache.hadoop.mapreduce.Reducer<NewK2, LongWritable, LongWritable, LongWritable>.Context context)
93 throws java.io.IOException, InterruptedException {
94 context.write(new LongWritable(k2.first), new LongWritable(
95 k2.second));
96 };
97 }
98
99 /**
100 * 问：为什么实现该类？ 答：因为原来的v2不能参与排序，把原来的k2和v2封装到一个类中，作为新的k2
101 *
102 /
103 // WritableComparable：Hadoop的序列化
104 static class NewK2 implements WritableComparable {
105 Long first;
106 Long second;
107
108 public NewK2() {
109 }
110
111 public NewK2(long first, long second) {
112 this.first = first;
113 this.second = second;
114 }
115
116 public void readFields(DataInput in) throws IOException {
117 this.first = in.readLong();
118 this.second = in.readLong();
119 }
120
121 public void write(DataOutput out) throws IOException {
122 out.writeLong(first);
123 out.writeLong(second);
124 }
125
126 /*
127 * 当k2进行排序时，会调用该方法. 当第一列不同时，升序；当第一列相同时，第二列升序
128 */
129 public int compareTo(NewK2 o) {
130 final long minus = this.first - o.first;
131 if (minus != 0) {
132 return (int) minus;
133 }
134 return (int) (this.second - o.second);
135 }
136
137 @Override
138 public int hashCode() {
139 return this.first.hashCode() + this.second.hashCode();
140 }
141
142 @Override
143 public boolean equals(Object obj) {
144 if (!(obj instanceof NewK2)) {
145 return false;
146 }
147 NewK2 oK2 = (NewK2) obj;
148 return (this.first == oK2.first) && (this.second == oK2.second);
149 }
150 }
151
152 }

复制代码

Hadoop序列化

序列化概念：

序列化：把结构化对象转化为字节流。

反序列化：是序列化的逆过程。即把字节流转回结构化对象。

Hadoop序列化的特点：

1、紧凑：高效使用存储空间。

2、快速：读写数据的额外开销小。

3、可扩展：可透明的读取老格式的数据。

4、互操作：支持多语言的交互。

Hadoop的序列化格式：Writable

Hadoop序列化的作用：

序列化在分布式环境的两大作用：进程间通信，永久存储。

Hadoop节点间通信：

Writable接口

Writable接口，是根据DataInput和DataOutput实现的简单、有效的序列化对象。

MR的任意key和value必须实现Writable接口。

MR的任意key必须实现WritableComparable接口。

自定义Writable类（上面代码中有）

实现Writable：

1、write是把每个对象序列化到输出流。

2、readFields是把输入流字节反序列化。

实现WritableComparable:

Java值对象的比较：一般需要重写toString(),hashCode(),equals()方法。