UTF8 filenames with Java

Using java.io.*

Functions from java.io solely depend on the locale settings of your operating system. If you want to create the directory /tmp/foö/bär, save this snippet to a file called javaapplication1/JavaApplication1.java:

package javaapplication1;

public class JavaApplication1 {

	static {
		System.setProperty("file.encoding",    "UTF-8");
		System.setProperty("sun.jnu.encoding", "UTF-8");
	}
	
	public static void main(String[] args) {
		new java.io.File("/tmp/foö/bär").mkdirs();
	}
}

Since the source file contains UTF-8 characters, the Java compiler needs to be aware of it:

$ javac -encoding UTF-8 javaapplication/JavaApplication1.java

Now run that program:

$ java -Dfile.encoding=UTF-8 -Dsun.jnu.encoding=UTF-8 javaapplication.JavaApplication1

If your environment is set to ASCII or not set at all, you will get /tmp/fo?/b?r. If your environment is set to ISO-8859-1 instead, you will get /tmp/fo\366/b\344r. System properties given on the commandline and in the static initialization block have no effect. You have to set your environment to an *.utf-8 locale:

$ LC_ALL=de_DE.utf-8 java javaapplication.JavaApplication1

Now you have a directory tree named /tmp/f\303\266/b\303\244r, which is the correct byte level encoding for UTF-8.

Using java.nio.*

In contrast to java.io, java.nio honors the sun.jnu.encoding system property set in the static initialization block:

package javaapplication1;

public class JavaApplication1 {

	static {
		System.setProperty("sun.jnu.encoding", "UTF-8");
	}

	public static void main(String[] args) throws java.io.IOException {
		java.nio.file.Files.createDirectories(java.nio.file.Paths.get("/tmp/foö/bär"));
	}
}

Compilation is not different:

$ javac -encoding UTF-8 javaapplication/JavaApplication1.java

But, you can omit any locale settings and system properties on the command line, and will still get directories with UTF-8 names:

$ java javaapplication.JavaApplication1

If you remove the static initialization block, e.g.

package javaapplication1;

public class JavaApplication1 {

	public static void main(String[] args) throws java.io.IOException {
		java.nio.file.Files.createDirectories(java.nio.file.Paths.get("/tmp/foö/bär"));
	}
}

and if your environment is not set to a *.utf-8 locale, then you have to set it manually again:

$ LC_ALL=de_DE.utf-8 java javaapplication.JavaApplication1

Btw, java.nio just defines static functions plus several interfaces. For an object oriented language I find this highly unclean.