Hadoop基于文件的数据结构及实例

基于文件的数据结构
两种文件格式：
1、SequenceFile
2、MapFile

SequenceFile

1、SequenceFile文件是Hadoop用来存储二进制形式的<key,value>对而设计的一种平面文件(Flat File)。

2、可以把SequenceFile当做一个容器，把所有文件打包到SequenceFile类中可以高效的对小文件进行存储和处理。

3、SequenceFile文件并不按照其存储的key进行排序存储，SequenceFile的内部类Writer**提供了append功能**。

4、SequenceFile中的key和value可以是任意类型Writable或者是自定义Writable类型。

SequenceFile压缩

1、SequenceFile的内部格式取决于是否启用压缩，如果是，要么是记录压缩，要么是块压缩。
2、三种类型：
A．无压缩类型：如果没有启用压缩（默认设置），那么每个记录就由它的记录长度（字节数）、键的长度，键和值组成。长度字段为四字节。

B．记录压缩类型：记录压缩格式与无压缩格式基本相同，不同的是值字节是用定义在头部的编码器来压缩。注意，键是不压缩的。

C．块压缩类型：块压缩一次压缩多个记录，因此它比记录压缩更紧凑，而且一般优先选择。当记录的字节数达到最小大小，才会添加到块。该最小值由io.seqfile.compress.blocksize中的属性定义。默认值是1000000字节。格式为记录数、键长度、键、值长度、值。

无压缩格式与记录压缩格式

块压缩格式

SequenceFile文件格式的好处：
A．支持基于记录（Record）或块（Block）的数据压缩。
B．支持splittable，能够作为MapReduce的输入分片。
C．修改简单：主要负责修改相应的业务逻辑，而不用考虑具体的存储格式。
SequenceFile文件格式的坏处：
坏处是需要一个合并文件的过程，且合并后的文件将不方便查看。因为它是二进制文件。

读写SequenceFile

写过程：
1）创建Configuration
2）获取FileSystem
3）创建文件输出路径Path
4）调用SequenceFile.createWriter得到SequenceFile.Writer对象
5）调用SequenceFile.Writer.append追加写入文件
6）关闭流
读过程：
1）创建Configuration
2）获取FileSystem
3）创建文件输出路径Path
4）new一个SequenceFile.Reader进行读取
5）得到keyClass和valueClass
6）关闭流

<code class="hljs livecodeserver has-numbering">org.apache.hadoop.io Class SequenceFileThere are <span class="hljs-constant">three</span> SequenceFile Writers based <span class="hljs-command"><span class="hljs-keyword">on</span> <span class="hljs-title">the</span> <span class="hljs-title">SequenceFile</span>.<span class="hljs-title">CompressionType</span> <span class="hljs-title">used</span> <span class="hljs-title">to</span> <span class="hljs-title">compress</span> <span class="hljs-title">key</span>/<span class="hljs-title">value</span> <span class="hljs-title">pairs</span>: </span><span class="hljs-number">1</span>、Writer : Uncompressed records. <span class="hljs-number">2</span>、RecordCompressWriter : Record-compressed <span class="hljs-built_in">files</span>, only <span class="hljs-built_in">compress</span> values. <span class="hljs-number">3</span>、BlockCompressWriter : Block-compressed <span class="hljs-built_in">files</span>, both <span class="hljs-built_in">keys</span> & values are collected <span class="hljs-operator">in</span> <span class="hljs-string">'blocks'</span> separately <span class="hljs-operator">and</span> compressed. The size <span class="hljs-operator">of</span> <span class="hljs-operator">the</span> <span class="hljs-string">'block'</span> is configurable</code>

无压缩方式、记录压缩、块压缩实例

<code class="hljs avrasm has-numbering">package SequenceFile<span class="hljs-comment">;</span>import java<span class="hljs-preprocessor">.io</span><span class="hljs-preprocessor">.IOException</span><span class="hljs-comment">;</span>import java<span class="hljs-preprocessor">.net</span><span class="hljs-preprocessor">.URI</span><span class="hljs-comment">;</span>import org<span class="hljs-preprocessor">.apache</span><span class="hljs-preprocessor">.hadoop</span><span class="hljs-preprocessor">.conf</span><span class="hljs-preprocessor">.Configuration</span><span class="hljs-comment">;</span>import org<span class="hljs-preprocessor">.apache</span><span class="hljs-preprocessor">.hadoop</span><span class="hljs-preprocessor">.fs</span><span class="hljs-preprocessor">.FileSystem</span><span class="hljs-comment">;</span>import org<span class="hljs-preprocessor">.apache</span><span class="hljs-preprocessor">.hadoop</span><span class="hljs-preprocessor">.fs</span><span class="hljs-preprocessor">.Path</span><span class="hljs-comment">;</span>import org<span class="hljs-preprocessor">.apache</span><span class="hljs-preprocessor">.hadoop</span><span class="hljs-preprocessor">.io</span><span class="hljs-preprocessor">.IOUtils</span><span class="hljs-comment">;</span>import org<span class="hljs-preprocessor">.apache</span><span class="hljs-preprocessor">.hadoop</span><span class="hljs-preprocessor">.io</span><span class="hljs-preprocessor">.IntWritable</span><span class="hljs-comment">;</span>import org<span class="hljs-preprocessor">.apache</span><span class="hljs-preprocessor">.hadoop</span><span class="hljs-preprocessor">.io</span><span class="hljs-preprocessor">.SequenceFile</span><span class="hljs-comment">;</span>import org<span class="hljs-preprocessor">.apache</span><span class="hljs-preprocessor">.hadoop</span><span class="hljs-preprocessor">.io</span><span class="hljs-preprocessor">.SequenceFile</span><span class="hljs-preprocessor">.CompressionType</span><span class="hljs-comment">;</span>import org<span class="hljs-preprocessor">.apache</span><span class="hljs-preprocessor">.hadoop</span><span class="hljs-preprocessor">.io</span><span class="hljs-preprocessor">.Text</span><span class="hljs-comment">;</span>import org<span class="hljs-preprocessor">.apache</span><span class="hljs-preprocessor">.hadoop</span><span class="hljs-preprocessor">.io</span><span class="hljs-preprocessor">.Writable</span><span class="hljs-comment">;</span>import org<span class="hljs-preprocessor">.apache</span><span class="hljs-preprocessor">.hadoop</span><span class="hljs-preprocessor">.io</span><span class="hljs-preprocessor">.compress</span><span class="hljs-preprocessor">.BZip</span>2Codec<span class="hljs-comment">;</span>import org<span class="hljs-preprocessor">.apache</span><span class="hljs-preprocessor">.hadoop</span><span class="hljs-preprocessor">.util</span><span class="hljs-preprocessor">.ReflectionUtils</span><span class="hljs-comment">;</span>public class Demo01 {    final static String uri= <span class="hljs-string">"hdfs://liguodong:8020/liguodong"</span><span class="hljs-comment">;</span>    final static String[] data = {        <span class="hljs-string">"apache,software"</span>,<span class="hljs-string">"chinese,good"</span>,<span class="hljs-string">"james,NBA"</span>,<span class="hljs-string">"index,pass"</span>    }<span class="hljs-comment">;</span>    public static void main(String[] args) throws IOException {        //<span class="hljs-number">1</span>        Configuration configuration = new Configuration()<span class="hljs-comment">;</span>        //<span class="hljs-number">2</span>        FileSystem fs = FileSystem<span class="hljs-preprocessor">.get</span>(URI<span class="hljs-preprocessor">.create</span>(uri),configuration)<span class="hljs-comment">;</span>        //<span class="hljs-number">3</span>        Path path = new Path(<span class="hljs-string">"/tmp.seq"</span>)<span class="hljs-comment">;</span>        write(fs,configuration,path)<span class="hljs-comment">;</span>        read(fs,configuration,path)<span class="hljs-comment">;</span>    }    public static void write(FileSystem fs,Configuration configuration,Path path) throws IOException{        //<span class="hljs-number">4</span>        IntWritable key = new IntWritable()<span class="hljs-comment">;</span>        Text value = new Text()<span class="hljs-comment">;</span>        //无压缩        <span class="hljs-comment">/*@SuppressWarnings("deprecation")        SequenceFile.Writer writer = SequenceFile.createWriter                (fs,configuration,path,key.getClass(),value.getClass());*/</span>        //记录压缩        @SuppressWarnings(<span class="hljs-string">"deprecation"</span>)        SequenceFile<span class="hljs-preprocessor">.Writer</span> writer = SequenceFile<span class="hljs-preprocessor">.createWriter</span>                (fs,configuration,path,key<span class="hljs-preprocessor">.getClass</span>(),                        value<span class="hljs-preprocessor">.getClass</span>(),CompressionType<span class="hljs-preprocessor">.RECORD</span>,new BZip2Codec())<span class="hljs-comment">;</span>        //块压缩        <span class="hljs-comment">/*@SuppressWarnings("deprecation")        SequenceFile.Writer writer = SequenceFile.createWriter                (fs,configuration,path,key.getClass(),                value.getClass(),CompressionType.BLOCK,new BZip2Codec());*/</span>        //<span class="hljs-number">5</span>        for (int i = <span class="hljs-number">0</span><span class="hljs-comment">; i < 30; i++) {</span>            key<span class="hljs-preprocessor">.set</span>(<span class="hljs-number">100</span>-i)<span class="hljs-comment">;</span>            value<span class="hljs-preprocessor">.set</span>(data[i%data<span class="hljs-preprocessor">.length</span>])<span class="hljs-comment">;</span>            writer<span class="hljs-preprocessor">.append</span>(key, value)<span class="hljs-comment">;</span>        }        //<span class="hljs-number">6</span>、关闭流        IOUtils<span class="hljs-preprocessor">.closeStream</span>(writer)<span class="hljs-comment">;        </span>    }    public static void read(FileSystem fs,Configuration configuration,Path path) throws IOException {        //<span class="hljs-number">4</span>        @SuppressWarnings(<span class="hljs-string">"deprecation"</span>)        SequenceFile<span class="hljs-preprocessor">.Reader</span> reader = new SequenceFile<span class="hljs-preprocessor">.Reader</span>(fs, path,configuration)<span class="hljs-comment">;</span>        //<span class="hljs-number">5</span>        Writable key = (Writable) ReflectionUtils<span class="hljs-preprocessor">.newInstance</span>                (reader<span class="hljs-preprocessor">.getKeyClass</span>(), configuration)<span class="hljs-comment">;</span>        Writable value = (Writable) ReflectionUtils<span class="hljs-preprocessor">.newInstance</span>                (reader<span class="hljs-preprocessor">.getValueClass</span>(), configuration)<span class="hljs-comment">;</span>        while(reader<span class="hljs-preprocessor">.next</span>(key,value)){            System<span class="hljs-preprocessor">.out</span><span class="hljs-preprocessor">.println</span>(<span class="hljs-string">"key = "</span> + key)<span class="hljs-comment">;</span>            System<span class="hljs-preprocessor">.out</span><span class="hljs-preprocessor">.println</span>(<span class="hljs-string">"value = "</span> + value)<span class="hljs-comment">;</span>            System<span class="hljs-preprocessor">.out</span><span class="hljs-preprocessor">.println</span>(<span class="hljs-string">"position = "</span>+ reader<span class="hljs-preprocessor">.getPosition</span>())<span class="hljs-comment">;</span>        }        IOUtils<span class="hljs-preprocessor">.closeStream</span>(reader)<span class="hljs-comment">;</span>    }}</code>

运行结果：

<code class="hljs makefile has-numbering"><span class="hljs-constant">key</span> = 100<span class="hljs-constant">value</span> = apache,software<span class="hljs-constant">position</span> = 164<span class="hljs-constant">key</span> = 99<span class="hljs-constant">value</span> = chinese,good<span class="hljs-constant">position</span> = 197<span class="hljs-constant">key</span> = 98<span class="hljs-constant">value</span> = james,NBA<span class="hljs-constant">position</span> = 227<span class="hljs-constant">key</span> = 97<span class="hljs-constant">value</span> = index,pass<span class="hljs-constant">position</span> = 258<span class="hljs-constant">key</span> = 96<span class="hljs-constant">value</span> = apache,software<span class="hljs-constant">position</span> = 294<span class="hljs-constant">key</span> = 95<span class="hljs-constant">value</span> = chinese,good<span class="hljs-constant">position</span> = 327......<span class="hljs-constant">key</span> = 72<span class="hljs-constant">value</span> = apache,software<span class="hljs-constant">position</span> = 1074<span class="hljs-constant">key</span> = 71<span class="hljs-constant">value</span> = chinese,good<span class="hljs-constant">position</span> = 1107</code>

MapFile

<code class="hljs java has-numbering"><span class="hljs-keyword">public</span> <span class="hljs-class"><span class="hljs-keyword">class</span> <span class="hljs-title">MapFile</span> {</span>  <span class="hljs-javadoc">/** The name of the index file. */</span>  <span class="hljs-keyword">public</span> <span class="hljs-keyword">static</span> <span class="hljs-keyword">final</span> String INDEX_FILE_NAME = <span class="hljs-string">"index"</span>;  <span class="hljs-javadoc">/** The name of the data file. */</span>  <span class="hljs-keyword">public</span> <span class="hljs-keyword">static</span> <span class="hljs-keyword">final</span> String DATA_FILE_NAME = <span class="hljs-string">"data"</span>;}</code>

MapFile是经过排序的索引的SequenceFile，可以根据key进行查找。

与SequenceFile不同的是， MapFile的Key一定要实现WritableComparable接口 ,即Key值是可比较的，而value是Writable类型的。
可以使用MapFile.fix()方法来重建索引，把SequenceFile转换成MapFile。
它有两个静态成员变量：

<code class="hljs java has-numbering"><span class="hljs-keyword">static</span> <span class="hljs-keyword">final</span> String INDEX_FILE_NAME<span class="hljs-keyword">static</span> <span class="hljs-keyword">final</span> String DATA_FILE_NAME</code>

通过观察其目录结构可以看到MapFile由两部分组成，分别是data和index。
index作为文件的数据索引，主要记录了每个Record的key值，以及该Record在文件中的偏移位置。

在MapFile被访问的时候，索引文件会被加载到内存，通过索引映射关系可迅速定位到指定Record所在文件位置。
因此，相对SequenceFile而言， MapFile的检索效率是高效的，缺点是会消耗一部分内存来存储index数据。
需注意的是， MapFile并不会把所有Record都记录到index中去，默认情况下每隔128条记录存储一个索引映射。当然，记录间隔可人为修改，通过MapFIle.Writer的setIndexInterval()方法，或修改io.map.index.interval属性；

读写MapFile

写过程：
1）创建Configuration
2）获取FileSystem
3）创建文件输出路径Path
4）new一个MapFile.Writer对象
5）调用MapFile.Writer.append追加写入文件
6）关闭流
读过程：
1）创建Configuration
2）获取FileSystem
3）创建文件输出路径Path
4）new一个MapFile.Reader进行读取
5）得到keyClass和valueClass
6）关闭流

具体操作与SequenceFile相似。

命令行查看二进制文件
hdfs dfs -text /liguodong/tmp.seq







原文转载至：http://blog.csdn.net/scgaliguodong123_/article/details/46391061