为什么在hadoop中执行MapReduce代码时需要jar文件，而在hadoop中执行任何其他Non-MapReduce Java代码时却不需要

Question

I would like to know why is jar file needed not a .class file for executing MapReduce code in hadoop. Then if Jar file is used then why the same is not done while executing any other Non-MapReduce Java code in hadoop? Also while executing any other Non-MapReduce Java code in hadoop why a compiled class is directly mentioned along with hadoop keyword in the command line for eg. if I've a program to display file in hadoop using an url ie. for class FileSystemCat

    public class FileSystemCat {
    public static void main(String[] args) throws Exception {
    String uri = args[0];
    Configuration conf = new Configuration();
    FileSystem fs = FileSystem.get(URI.create(uri), conf);
    InputStream in = null;
    try {
    in = fs.open(new Path(uri));
    IOUtils.copyBytes(in, System.out, 4096, false);
    } finally {
    IOUtils.closeStream(in);

}
}
}

Command to execute the program after compiling the program is "hadoop FileSystemCat" not "hadoop Java FileSystemCat" . As in common environment the steps to execute the program would've been:

Javac FileSystemCat.java
Java FileSystemCat.class

Answer 1

 hadoop jar <jar> [mainClass] args... 

Runs a jar file. Users can bundle their Map Reduce code in a jar file and execute it using this command.

 hadoop CLASSNAME 

hadoop script can be used to invoke any class.

These above commands are doing two thinks.

1)Add all the jars in the Hadoop installation lib directory into the classpath of the running jar or class.

2)Add the configuration directory of hadoop installation to the classpath.

Thus the running JAR or CLASS will get all classes in the hadoop instalation classpath and all the configurtion files of the installation.

If you are running a JAR or CLASS using simple Java CLASSNAME command you have to separatly add the above two components to the java classpath.