自定义InputFormat

(http://youzitool.com 新博客，欢迎访问)

这几天准备好好看看MapReduce编程。要编程就肯定要涉及到输入、输出的问题。今天就先来谈谈自定义的InputFormat

我们先来看看系统默认的TextInputFormat.java

 public class TextInputFormat extends FileInputFormat<LongWritable, Text> { @Override public RecordReader<LongWritable, Text> createRecordReader(InputSplit split, TaskAttemptContext context) { return new LineRecordReader();//这里是系统实现的的RecordReader按行读取，等会我们就是要改写这个类。 } @Override protected boolean isSplitable(JobContext context, Path file) { CompressionCodec codec = new CompressionCodecFactory(context.getConfiguration()).getCodec(file); return codec == null;//而这里通过返回一个null，实际就是关闭了对当前读入文件的划分。 } }

这个类，没什么说的。接着我们来实现我们的读取类. MyRecordReader

//我实现的功能比较简单，只要明白了原理，剩下的就看自己发挥了。 //我们知道系统默认的TextInputFormat读取的key、value分别是偏移和行，而我就简单改下，改成key、value分别是行号和行 public class MyRecordReader extends RecordReader<LongWritable, Text>{ //这里继承RecordReader来实现我们自己的读取。 private static final Log LOG = LogFactory.getLog(MyRecordReader.class); private long pos; //记录行号 private boolean more; private LineReader in; private int maxLineLength; private LongWritable key = null; private Text value = null; public void initialize(InputSplit genericSplit, TaskAttemptContext context) throws IOException { pos = 1; more = true; FileSplit split = (FileSplit) genericSplit;//获取split Configuration job = context.getConfiguration(); this.maxLineLength = job.getInt("mapred.linerecordreader.maxlength", Integer.MAX_VALUE); final Path file = split.getPath();//得到文件路径 // open the file and seek to the start of the split FileSystem fs = file.getFileSystem(job); FSDataInputStream fileIn = fs.open(split.getPath()); //打开文件 in = new LineReader(fileIn, job); } public boolean nextKeyValue() throws IOException { //这个函数会被MapRunner循环读取<key、value> if (key == null) { key = new LongWritable(); } key.set(pos); //设置key if (value == null) { value = new Text(); } int newSize = 0; while (true) { newSize = in.readLine(value, maxLineLength,maxLineLength); //读取一行内容 pos++; //行号自加一 if (newSize == 0) { break; } if (newSize < maxLineLength) { break; } // line too long. try again LOG.info("Skipped line of size " + newSize + " at pos " + (pos - newSize)); } if (newSize == 0) { key = null; value = null; more = false; return false; } else { return true; } } //下面的函数都是必须实现的，跟系统默认的差不多，我就不多说了。 @Override public LongWritable getCurrentKey() { return key; } @Override public Text getCurrentValue() { return value; } /** * Get the progress within the split */ public float getProgress() { if (more) { return 0.0f; } else { return 100; } } public synchronized void close() throws IOException { if (in != null) { in.close(); } } }

有了，自己的MyInputFormat,就可以试试效果了。

Configuration conf = new Configuration(); Job job = new Job(conf, "helloword"); job.setJarByClass(helloHadoopV1.class); job.setInputFormatClass(MyInputFormat.class); //这里就是调用我们自己写的InputFormat job.setOutputKeyClass(LongWritable.class); job.setOutputValueClass(Text.class); job.setMapperClass(HelloMapper.class); //job.setReducerClass(HelloReducer.class); //这里不设置Reduce就会按Mapper的中间过程原样输出。 FileInputFormat.setInputPaths(job, new Path(args[1])); FileOutputFormat.setOutputPath(job, new Path(args[2])); checkAndDelete(args[2], conf); job.waitForCompletion(true);

我们上面都自定义的都比较简单，这样才容易懂。

最后再补充一句，上面RecordReader<LongWritable, Text>的<key、value>都是可以自己定义的。但key必须实现WritableComparable类，而value必须实现Writable类。