hadoop中带后缀的多路输出part文件

来源：互联网发布：税务总局网络培训学院编辑：程序博客网时间：2024/06/07 13:20

【导读】

hadoop源代码中org.apache.hadoop.mapred.lib.MultipleTextOutputFormat类实现了多路输出的框架。在此基础上，可以实现自定义的多路输出方案。

本篇给出了带后缀的多路输出方案。比如part-00000-[A-Z], part-00000-[a-z], part-00000-[0-9]。

【正文】
1. 约定reduce输出时的数据格式：

<key, value>#suffix_letter

suffix_letter表示后缀字母，目前支持[0-9A-Za-z]共62个字符。可以是常量，也可以是变量。

value可以为空。

通俗地说，只需要在输出的一行结尾加上“#suffix_letter” 就能支持多路

2. 实现

SuffixMultipleTextOutputFormat.java

package com.**.hadoop.mapred.lib;import org.apache.hadoop.mapred.lib.MultipleTextOutputFormat;import org.apache.hadoop.io.Text;public class SuffixMultipleTextOutputFormat extends MultipleTextOutputFormat<Text, Text> {private int tagPos = -1;private boolean has_value = true;public void test(Text key, Text value, String name) {System.out.println("file_name = " + this.generateFileNameForKeyValue(key, value, name));System.out.println("key = " + this.generateActualKey(key, value));System.out.println("value = " + this.generateActualValue(key, value));}@Overrideprotected Text generateActualKey(Text key, Text value) {if (!has_value && tagPos != -1) {return new Text(key.toString().substring(0, tagPos));}return key;}@Overrideprotected Text generateActualValue(Text key, Text value) {if (has_value && tagPos != -1) {return new Text(value.toString().substring(0, tagPos));}return value;}@Override  protected String generateFileNameForKeyValue(Text key, Text value, String name) {String val = value.toString();if (val.isEmpty()) {has_value = false;val = key.toString();}try {int pos = val.lastIndexOf('#');if (pos >= 0 && pos == val.length() - 2) {char suffix = val.charAt(pos+1);if (Character.isDigit(suffix) || Character.isLetter(suffix)) {tagPos = pos;return name + "-" + suffix;} else {throw new InvalidSuffixMultipleTextOutputFormatException("InvalidSuffixMultipleTextOutputFormatException : key = " + key.toString() + " , value = " + value.toString());}} else {throw new InvalidSuffixMultipleTextOutputFormatException("InvalidSuffixMultipleTextOutputFormatException : key = " + key.toString() + " , value = " + value.toString());}} catch (Exception e) {e.printStackTrace();}return name;}public static void main(String [] args) {new SuffixMultipleTextOutputFormat().test(new Text("abc"), new Text("#i"), "part-00000");new SuffixMultipleTextOutputFormat().test(new Text("abc"), new Text("#"), "part-00001");new SuffixMultipleTextOutputFormat().test(new Text("abc"), new Text("w#o"), "part-00001");new SuffixMultipleTextOutputFormat().test(new Text("abc#0"), new Text(""), "part-00001");new SuffixMultipleTextOutputFormat().test(new Text("abc#0"), new Text("a"), "part-00001");}}

InvalidSuffixMultipleTextOutputFormatException.java

package com.**.hadoop.mapred.lib;public class InvalidSuffixMultipleTextOutputFormatException extends Exception {private static final long serialVersionUID = -7900596082142417867L;public InvalidSuffixMultipleTextOutputFormatException(String error) {super(error);}}

0 0