MapReduce中的自定义多目录/文件名输出<转>

最近考虑到这样一个需求:

需要把原始的日志文件用hadoop做清洗后,按业务线输出到不同的目录下去,以供不同的部门业务线使用。

这个需求需要用到MultipleOutputFormat和MultipleOutputs来实现自定义多目录、文件的输出。

需要注意的是,在hadoop 0.21.x之前和之后的使用方式是不一样的:

hadoop 0.21 之前的API 中有 org.apache.hadoop.mapred.lib.MultipleOutputFormat 和 org.apache.hadoop.mapred.lib.MultipleOutputs,而到了 0.21 之后 的API为 org.apache.hadoop.mapreduce.lib.output.MultipleOutputs ,

新版的API 整合了上面旧API两个的功能,没有了MultipleOutputFormat。

本文将给出新旧两个版本的API code。

1、旧版0.21.x之前的版本:

01 import java.io.IOException;
02  
03 import org.apache.hadoop.conf.Configuration;
04 import org.apache.hadoop.conf.Configured;
05 import org.apache.hadoop.fs.Path;
06 import org.apache.hadoop.io.LongWritable;
07 import org.apache.hadoop.io.NullWritable;
08 import org.apache.hadoop.io.Text;
09 import org.apache.hadoop.mapred.FileInputFormat;
10 import org.apache.hadoop.mapred.FileOutputFormat;
11 import org.apache.hadoop.mapred.JobClient;
12 import org.apache.hadoop.mapred.JobConf;
13 import org.apache.hadoop.mapred.MapReduceBase;
14 import org.apache.hadoop.mapred.Mapper;
15 import org.apache.hadoop.mapred.OutputCollector;
16 import org.apache.hadoop.mapred.Reporter;
17 import org.apache.hadoop.mapred.TextInputFormat;
18 import org.apache.hadoop.mapred.lib.MultipleTextOutputFormat;
19 import org.apache.hadoop.util.Tool;
20 import org.apache.hadoop.util.ToolRunner;
21  
22 public class MultiFile extends Configured implements Tool {
23  
24     public static class MapClass extends MapReduceBase implements
25             Mapper<LongWritable, Text, NullWritable, Text> {
26  
27         @Override
28         public void map(LongWritable key, Text value,
29                 OutputCollector<NullWritable, Text> output, Reporter reporter)
30                 throws IOException {
31             output.collect(NullWritable.get(), value);
32         }
33  
34     }
35  
36     // MultipleTextOutputFormat 继承自MultipleOutputFormat,实现输出文件的分类
37  
38     public static class PartitionByCountryMTOF extends
39             MultipleTextOutputFormat<NullWritable, Text> { // key is
40                                                             // NullWritable,
41                                                             // value is Text
42         protected String generateFileNameForKeyValue(NullWritable key,
43                 Text value, String filename) {
44             String[] arr = value.toString().split(",", -1);
45             String country = arr[4].substring(13); // 获取country的名称
46             return country + "/" + filename;
47         }
48     }
49  
50     // 此处不使用reducer
51     /*
52      * public static class Reducer extends MapReduceBase implements
53      * org.apache.hadoop.mapred.Reducer<LongWritable, Text, NullWritable, Text>
54      * {
55      *
56      * @Override public void reduce(LongWritable key, Iterator<Text> values,
57      * OutputCollector<NullWritable, Text> output, Reporter reporter) throws
58      * IOException { // TODO Auto-generated method stub
59      *
60      * }
61      *
62      * }
63      */
64     @Override
65     public int run(String[] args) throws Exception {
66         Configuration conf = getConf();
67         JobConf job = new JobConf(conf, MultiFile.class);
68  
69         Path in = new Path(args[0]);
70         Path out = new Path(args[1]);
71  
72         FileInputFormat.setInputPaths(job, in);
73         FileOutputFormat.setOutputPath(job, out);
74  
75         job.setJobName("MultiFile");
76         job.setMapperClass(MapClass.class);
77         job.setInputFormat(TextInputFormat.class);
78         job.setOutputFormat(PartitionByCountryMTOF.class);
79         job.setOutputKeyClass(NullWritable.class);
80         job.setOutputValueClass(Text.class);
81  
82         job.setNumReduceTasks(0);
83         JobClient.runJob(job);
84         return 0;
85     }
86  
87     public static void main(String[] args) throws Exception {
88         int res = ToolRunner.run(new Configuration(), new MultiFile(), args);
89         System.exit(res);
90     }
91  
92 }
测试数据及结果:
1 hadoop fs -cat /tmp/multiTest.txt
2 5765303,1998,14046,1996,"AD","",,1,12,42,5,59,11,1,0.4545,0,0,1,67.3636,,,,
3 5785566,1998,14088,1996,"AD","",,1,9,441,6,69,3,0,1,,0.6667,,4.3333,,,,
4 5894770,1999,14354,1997,"AD","",,1,,82,5,51,4,0,1,,0.625,,7.5,,,,
5 5765303,1998,14046,1996,"CN","",,1,12,42,5,59,11,1,0.4545,0,0,1,67.3636,,,,
6 5785566,1998,14088,1996,"CN","",,1,9,441,6,69,3,0,1,,0.6667,,4.3333,,,,
7 5894770,1999,14354,1997,"CN","",,1,,82,5,51,4,0,1,,0.625,,7.5,,,,

MapReduce中的自定义多目录/文件名输出<转>_第1张图片

from:

MultipleOutputFormat Example

http://mazd1002.blog.163.com/blog/static/665749652011102553947492/


2、新版0.21.x及之后的版本:

01 public class TestwithMultipleOutputs extends Configured implements Tool {
02  
03   public static class MapClass extends Mapper<LongWritable,Text,Text,IntWritable> {
04  
05     private MultipleOutputs<Text,IntWritable> mos;
06  
07     protected void setup(Context context) throws IOException,InterruptedException {
08       mos = new MultipleOutputs<Text,IntWritable>(context);
09     }
10  
11     public void map(LongWritable key, Text value, Context context) throws IOException, InterruptedException{
12       String line = value.toString();
13       String[] tokens = line.split("-");
14  
15       mos.write("MOSInt",new Text(tokens[0]), new IntWritable(Integer.parseInt(tokens[1])));  //(第一处)
16       mos.write("MOSText"new Text(tokens[0]),tokens[2]);     //(第二处)
17       mos.write("MOSText"new Text(tokens[0]),line,tokens[0]+"/");  //(第三处)同时也可写到指定的文件或文件夹中
18     }
19  
20     protected void cleanup(Context context) throws IOException,InterruptedException {
21       mos.close();
22     }
23  
24   }
25   public int run(String[] args) throws Exception {
26  
27     Configuration conf = getConf();
28  
29     Job job = new Job(conf,"word count with MultipleOutputs");
30  
31     job.setJarByClass(TestwithMultipleOutputs.class);
32  
33     Path in = new Path(args[0]);
34     Path out = new Path(args[1]);
35  
36     FileInputFormat.setInputPaths(job, in);
37     FileOutputFormat.setOutputPath(job, out);
38  
39     job.setMapperClass(MapClass.class);
40     job.setNumReduceTasks(0);  
41  
42     MultipleOutputs.addNamedOutput(job,"MOSInt",TextOutputFormat.class,Text.class,IntWritable.class);
43     MultipleOutputs.addNamedOutput(job,"MOSText",TextOutputFormat.class,Text.class,Text.class);
44  
45     System.exit(job.waitForCompletion(true)?0:1);
46     return 0;
47   }
48  
49   public static void main(String[] args) throws Exception {
50  
51     int res = ToolRunner.run(new Configuration(), new TestwithMultipleOutputs(), args);
52     System.exit(res);
53   }
54  
55 }

测试的数据:

abc-1232-hdf
abc-123-rtd
ioj-234-grjth
ntg-653-sdgfvd
kju-876-btyun
bhm-530-bhyt
hfter-45642-bhgf
bgrfg-8956-fmgh
jnhdf-8734-adfbgf
ntg-68763-nfhsdf
ntg-98634-dehuy
hfter-84567-drhuk

结果截图:(结果输出到/test/testMOSout)

MapReduce中的自定义多目录/文件名输出<转>_第2张图片

你可能感兴趣的:(MapReduce中的自定义多目录/文件名输出<转>)