用 Hadoop 统计词频并存入 HBase 中

发表于 更新于 Disqus：

统计一个 TXT 中的所有词语出现的平均频率（总出现次数／总共出现过的TXT文档数量），并写入 Hbase

一共用到 MapRecuce 的四个步骤：

Mapper 负责把把原来的任务分成很多Key-Value块。本题中，我们把任务分成这样的键值对：<Term#Doc, 1>

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
public class TokenizerMapper extends Mapper<LongWritable, Text, Text, IntWritable> {

    private static final IntWritable one = new IntWritable(1);
    private Text word = new Text();

    @Override
    public void map(LongWritable key, Text value, Context context) throws IOException, InterruptedException {
        StringTokenizer itr = new StringTokenizer(value.toString());
        while (itr.hasMoreTokens()) {
            word.set(itr.nextToken() + '#' + getSource((FileSplit) context.getInputSplit()));
            context.write(word, one);
        }
    }
    
    private static String getSource(FileSplit split) {
        String fileName = split.getPath().getName();
        return fileName.split("\\.", 2)[0];
    }
}

Combiner 是可选的，负责在 Mapper 后做个简单的合并，从而可以减少 Mapper 节点到 Reducer 节点之间的传输代价。比如对于上述的键值对，如果 Term#Doc 相同（换句话说，同一个词在同一篇文档中出现多次），可以直接替换成 <Term#Doc, n> , n为出现次数

1
2
3
4
5
6
7
8
9
10
11
12
13
public class LocalSumCombiner extends Reducer<Text, IntWritable, Text, IntWritable> {
    private IntWritable result = new IntWritable();

    @Override
    public void reduce(Text key, Iterable<IntWritable> values, Context context) throws IOException, InterruptedException {
        int sum = 0;
        for (IntWritable val : values) {
            sum += val.get();
        }
        result.set(sum);
        context.write(key, result);
    }
}

Partitioner 负责把 Key-Value 对 shuffle 到某个 Reducer 上。缺省的 Partitioner 就是对 Key 做哈希、再和 Reducer 数量取模。一方面 Partition 能实现 Reducer 的负载均衡，另一方面，我们常常希望这个分配的过程有意义，例如把同一个词（term）分到同一个 Reducer 上。

1
2
3
4
5
6
7
public class TermPartitioner extends Partitioner<Text, IntWritable> {
    @Override
    public int getPartition(Text key, IntWritable value, int numReduceTasks) {
        String term = key.toString().split("#")[0];
        return term.hashCode() % numReduceTasks;
    }
}

Reducer 负责汇总数据，产生结果。注意到 MapReduce 框架保证 shuffle 过来的数据是排序过的。

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
public class WordFreqReducer extends TableReducer<Text, IntWritable, ImmutableBytesWritable> {

    private String prevTerm = "";

    private List<Posting> postings = new ArrayList<>();

    @Override
    public void reduce(Text key, Iterable<IntWritable> values, WordFreqReducer.Context context)
            throws IOException, InterruptedException {
        String keySplits[] = key.toString().split("#");
        String term = keySplits[0];
        String document = keySplits[1];

        if (!term.equals(prevTerm)) {
            cleanup(context);
        }

        int count = 0;
        for (IntWritable val : values) {
            count += val.get();
        }

        postings.add(new Posting(document, count));
        prevTerm = term;
    }

    @Override
    public void cleanup(WordFreqReducer.Context context) throws IOException, InterruptedException {
        if (!"".equals(prevTerm)) {
            int total = postings.stream().mapToInt(Posting::getFrequency).sum();
            double average = (double) total / postings.size();

            Put put = new Put(Bytes.toBytes(prevTerm));
            put.addColumn(Bytes.toBytes(InvertedIndexer.COLUMN_FAMILY_NAME),
                    Bytes.toBytes(InvertedIndexer.COLUMN_NAME), Bytes.toBytes(String.format("%.2f", average)));
            context.write(null, put);
            postings.clear();
        }
    }

    static private class Posting {
        private final String document;
        private final int frequency;

        Posting(String document, int frequency) {
            this.document = document;
            this.frequency = frequency;
        }

        public String getDocument() {
            return document;
        }

        public int getFrequency() {
            return frequency;
        }

        @Override
        public String toString() {
            return String.format("%s:%d", document, frequency);
        }
    }
}

最后附上 main 函数的代码：

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
public class InvertedIndexer {
    private static Logger logger = Logger.getLogger(InvertedIndexer.class);
    
    private static final Configuration HBASE_CONFIG = HBaseConfiguration.create();
    
    public static final String TABLE_NAME = "wuxia";
    public static final String COLUMN_FAMILY_NAME = "cf";
    public static final String COLUMN_NAME = "freq";

    public static void main(String[] args) throws Exception {
        createTableIfNotExist(TABLE_NAME, COLUMN_FAMILY_NAME);

        Job job = Job.getInstance(new Configuration(), "word_freq");
        job.setJarByClass(InvertedIndexer.class);
        
        TableMapReduceUtil.initTableReducerJob(TABLE_NAME, WordFreqReducer.class, job, TermPartitioner.class);
        job.setMapperClass(TokenizerMapper.class);
        job.setPartitionerClass(TermPartitioner.class);
        job.setCombinerClass(LocalSumCombiner.class);
        job.setReducerClass(WordFreqReducer.class);

        job.setOutputKeyClass(Text.class);
        job.setOutputValueClass(IntWritable.class);
        FileInputFormat.addInputPath(job, new Path(args[0]));

        System.exit(job.waitForCompletion(true) ? 0 : 1);
    }

    private static void createTableIfNotExist(String tableName, String... families)
            throws Exception {
        Connection connection = ConnectionFactory.createConnection(HBASE_CONFIG);
        Admin admin = connection.getAdmin();
        
        HTableDescriptor desc = new HTableDescriptor(TableName.valueOf(tableName));
        for (String family : families) {
            desc.addFamily(new HColumnDescriptor(family));
        }
        
        if (admin.tableExists(TableName.valueOf(tableName))) {
            logger.info(String.format("Table '%s' already created", tableName));
        } else {
            admin.createTable(desc);
            logger.info(String.format("Created table '%s'", tableName));
        }
    }
}

注意，用到 HBase 的情况下，依赖只需要这两个就行了： 

1
2
3
4
5
6
7
8
9
10
<dependency>
    <groupId>org.apache.hbase</groupId>
    <artifactId>hbase-client</artifactId>
    <version>1.2.3</version>
</dependency>
<dependency>
    <groupId>org.apache.hbase</groupId>
    <artifactId>hbase-server</artifactId>
    <version>1.2.3</version>
</dependency>
以及 Apache 源 
1
2
3
4
<repository>
    <id>apache</id>
    <url>http://maven.apache.org</url>
</repository>