Default character encoding for java console output

Question

How does Java determine the encoding used for System.out?

Given the following class:

import java.io.File;
import java.io.PrintWriter;

public class Foo
{
    public static void main(String[] args) throws Exception
    {
        String s = "xxäñxx";
        System.out.println(s);
        PrintWriter out = new PrintWriter(new File("test.txt"), "UTF-8");
        out.println(s);
        out.close();
    }
}

It is saved as UTF-8 and compiled with javac -encoding UTF-8 Foo.java on a Windows system.

Afterwards on a git-bash console (using UTF-8 charset) I do:

$ java Foo
xxõ±xx
$ java -Dfile.encoding=UTF-8 Foo
xxäñxx
$ cat test.txt
xxäñxx
$ java Foo | cat
xxäñxx
$ java -Dfile.encoding=UTF-8 Foo | cat
xxäñxx

What is going on here?

Obviously java checks if it is connected to a terminal and is changing its encoding in that case. Is there a way to force Java to simply output plain UTF-8?

---

I tried the same with the cmd console, too. Redirecting STDOUT does not seem to make any difference there. Without the file.encoding parameter it outputs ansi encoding with the parameter it outputs utf8 encoding.

Solution

I'm assuming that your console still runs under cmd.exe. I doubt your console is really expecting UTF-8 - I expect it is really an OEM DOS encoding (e.g. 850 or 437.)

Java will encode bytes using the default encoding set during JVM initialization.

Reproducing on my PC:

java Foo

Java encodes as windows-1252; console decodes as IBM850. Result: Mojibake

java -Dfile.encoding=UTF-8 Foo

Java encodes as UTF-8; console decodes as IBM850. Result: Mojibake

cat test.txt

cat decodes file as UTF-8; cat encodes as IBM850; console decodes as IBM850.

java Foo | cat

Java encodes as windows-1252; cat decodes as windows-1252; cat encodes as IBM850; console decodes as IBM850

java -Dfile.encoding=UTF-8 Foo | cat

Java encodes as UTF-8; cat decodes as UTF-8; cat encodes as IBM850; console decodes as IBM850

This implementation of cat must use heuristics to determine if the character data is UTF-8 or not, then transcodes the data from either UTF-8 or ANSI (e.g. windows-1252) to the console encoding (e.g. IBM850.)

This can be confirmed with the following commands:

$ java HexDump utf8.txt
78 78 c3 a4 c3 b1 78 78

$ cat utf8.txt
xxäñxx

$ java HexDump ansi.txt
78 78 e4 f1 78 78

$ cat ansi.txt
xxäñxx

The cat command can make this determination because e4 f1 is not a valid UTF-8 sequence.

You can correct the Java output by:

• Setting the console encoding to the system ANSI value
• Using the Console type
• Using some shiv layer as you are doing with cat

HexDump is a trivial Java application:

import java.io.*;
class HexDump {
  public static void main(String[] args) throws IOException {
    try (InputStream in = new FileInputStream(args[0])) {
      int r;
      while((r = in.read()) != -1) {
        System.out.format("%02x ", 0xFF & r);
      }
      System.out.println();
    }
  }
}