Yes, it’s frustrating—sometimes type and other programs
print gibberish, and sometimes they do not.
First of all, Unicode characters will only display if the
current console font contains the characters. So use
a TrueType font like Lucida Console instead of the default Raster Font.
But if the console font doesn’t contain the character you’re trying to display,
you’ll see question marks instead of gibberish. When you get gibberish,
there’s more going on than just font settings.
When programs use standard C-library I/O functions like printf, the
program’s output encoding must match the console’s output encoding, or
you will get gibberish. chcp shows and sets the current codepage. All
output using standard C-library I/O functions is treated as if it is in the
codepage displayed by chcp.
Matching the program’s output encoding with the console’s output encoding
can be accomplished in two different ways:
-
A program can get the console’s current codepage using
chcpor
GetConsoleOutputCP, and configure itself to output in that encoding, or -
You or a program can set the console’s current codepage using
chcpor
SetConsoleOutputCPto match the default output encoding of the program.
However, programs that use Win32 APIs can write UTF-16LE strings directly
to the console with
WriteConsoleW.
This is the only way to get correct output without setting codepages. And
even when using that function, if a string is not in the UTF-16LE encoding
to begin with, a Win32 program must pass the correct codepage to
MultiByteToWideChar.
Also, WriteConsoleW will not work if the program’s output is redirected;
more fiddling is needed in that case.
type works some of the time because it checks the start of each file for
a UTF-16LE Byte Order Mark
(BOM), i.e. the bytes 0xFF 0xFE.
If it finds such a
mark, it displays the Unicode characters in the file using WriteConsoleW
regardless of the current codepage. But when typeing any file without a
UTF-16LE BOM, or for using non-ASCII characters with any command
that doesn’t call WriteConsoleW—you will need to set the
console codepage and program output encoding to match each other.
How can we find this out?
Here’s a test file containing Unicode characters:
ASCII abcde xyz
German äöü ÄÖÜ ß
Polish ąęźżńł
Russian абвгдеж эюя
CJK 你好
Here’s a Java program to print out the test file in a bunch of different
Unicode encodings. It could be in any programming language; it only prints
ASCII characters or encoded bytes to stdout.
import java.io.*;
public class Foo {
private static final String BOM = "\ufeff";
private static final String TEST_STRING
= "ASCII abcde xyz\n"
+ "German äöü ÄÖÜ ß\n"
+ "Polish ąęźżńł\n"
+ "Russian абвгдеж эюя\n"
+ "CJK 你好\n";
public static void main(String[] args)
throws Exception
{
String[] encodings = new String[] {
"UTF-8", "UTF-16LE", "UTF-16BE", "UTF-32LE", "UTF-32BE" };
for (String encoding: encodings) {
System.out.println("== " + encoding);
for (boolean writeBom: new Boolean[] {false, true}) {
System.out.println(writeBom ? "= bom" : "= no bom");
String output = (writeBom ? BOM : "") + TEST_STRING;
byte[] bytes = output.getBytes(encoding);
System.out.write(bytes);
FileOutputStream out = new FileOutputStream("uc-test-"
+ encoding + (writeBom ? "-bom.txt" : "-nobom.txt"));
out.write(bytes);
out.close();
}
}
}
}
The output in the default codepage? Total garbage!
Z:\andrew\projects\sx\1259084>chcp
Active code page: 850
Z:\andrew\projects\sx\1259084>java Foo
== UTF-8
= no bom
ASCII abcde xyz
German ├ñ├Â├╝ ├ä├û├£ ├ƒ
Polish ąęźżńł
Russian ð░ð▒ð▓ð│ð┤ðÁð ÐìÐÄÐÅ
CJK õ¢áÕÑ¢
= bom
´╗┐ASCII abcde xyz
German ├ñ├Â├╝ ├ä├û├£ ├ƒ
Polish ąęźżńł
Russian ð░ð▒ð▓ð│ð┤ðÁð ÐìÐÄÐÅ
CJK õ¢áÕÑ¢
== UTF-16LE
= no bom
A S C I I a b c d e x y z
G e r m a n õ ÷ ³ ─ Í ▄ ▀
P o l i s h ♣☺↓☺z☺|☺D☺B☺
R u s s i a n 0♦1♦2♦3♦4♦5♦6♦ M♦N♦O♦
C J K `O}Y
= bom
■A S C I I a b c d e x y z
G e r m a n õ ÷ ³ ─ Í ▄ ▀
P o l i s h ♣☺↓☺z☺|☺D☺B☺
R u s s i a n 0♦1♦2♦3♦4♦5♦6♦ M♦N♦O♦
C J K `O}Y
== UTF-16BE
= no bom
A S C I I a b c d e x y z
G e r m a n õ ÷ ³ ─ Í ▄ ▀
P o l i s h ☺♣☺↓☺z☺|☺D☺B
R u s s i a n ♦0♦1♦2♦3♦4♦5♦6 ♦M♦N♦O
C J K O`Y}
= bom
■ A S C I I a b c d e x y z
G e r m a n õ ÷ ³ ─ Í ▄ ▀
P o l i s h ☺♣☺↓☺z☺|☺D☺B
R u s s i a n ♦0♦1♦2♦3♦4♦5♦6 ♦M♦N♦O
C J K O`Y}
== UTF-32LE
= no bom
A S C I I a b c d e x y z
G e r m a n õ ÷ ³ ─ Í ▄ ▀
P o l i s h ♣☺ ↓☺ z☺ |☺ D☺ B☺
R u s s i a n 0♦ 1♦ 2♦ 3♦ 4♦ 5♦ 6♦ M♦ N
♦ O♦
C J K `O }Y
= bom
■ A S C I I a b c d e x y z
G e r m a n õ ÷ ³ ─ Í ▄ ▀
P o l i s h ♣☺ ↓☺ z☺ |☺ D☺ B☺
R u s s i a n 0♦ 1♦ 2♦ 3♦ 4♦ 5♦ 6♦ M♦ N
♦ O♦
C J K `O }Y
== UTF-32BE
= no bom
A S C I I a b c d e x y z
G e r m a n õ ÷ ³ ─ Í ▄ ▀
P o l i s h ☺♣ ☺↓ ☺z ☺| ☺D ☺B
R u s s i a n ♦0 ♦1 ♦2 ♦3 ♦4 ♦5 ♦6 ♦M ♦N
♦O
C J K O` Y}
= bom
■ A S C I I a b c d e x y z
G e r m a n õ ÷ ³ ─ Í ▄ ▀
P o l i s h ☺♣ ☺↓ ☺z ☺| ☺D ☺B
R u s s i a n ♦0 ♦1 ♦2 ♦3 ♦4 ♦5 ♦6 ♦M ♦N
♦O
C J K O` Y}
However, what if we type the files that got saved? They contain the exact
same bytes that were printed to the console.
Z:\andrew\projects\sx\1259084>type *.txt
uc-test-UTF-16BE-bom.txt
■ A S C I I a b c d e x y z
G e r m a n õ ÷ ³ ─ Í ▄ ▀
P o l i s h ☺♣☺↓☺z☺|☺D☺B
R u s s i a n ♦0♦1♦2♦3♦4♦5♦6 ♦M♦N♦O
C J K O`Y}
uc-test-UTF-16BE-nobom.txt
A S C I I a b c d e x y z
G e r m a n õ ÷ ³ ─ Í ▄ ▀
P o l i s h ☺♣☺↓☺z☺|☺D☺B
R u s s i a n ♦0♦1♦2♦3♦4♦5♦6 ♦M♦N♦O
C J K O`Y}
uc-test-UTF-16LE-bom.txt
ASCII abcde xyz
German äöü ÄÖÜ ß
Polish ąęźżńł
Russian абвгдеж эюя
CJK 你好
uc-test-UTF-16LE-nobom.txt
A S C I I a b c d e x y z
G e r m a n õ ÷ ³ ─ Í ▄ ▀
P o l i s h ♣☺↓☺z☺|☺D☺B☺
R u s s i a n 0♦1♦2♦3♦4♦5♦6♦ M♦N♦O♦
C J K `O}Y
uc-test-UTF-32BE-bom.txt
■ A S C I I a b c d e x y z
G e r m a n õ ÷ ³ ─ Í ▄ ▀
P o l i s h ☺♣ ☺↓ ☺z ☺| ☺D ☺B
R u s s i a n ♦0 ♦1 ♦2 ♦3 ♦4 ♦5 ♦6 ♦M ♦N
♦O
C J K O` Y}
uc-test-UTF-32BE-nobom.txt
A S C I I a b c d e x y z
G e r m a n õ ÷ ³ ─ Í ▄ ▀
P o l i s h ☺♣ ☺↓ ☺z ☺| ☺D ☺B
R u s s i a n ♦0 ♦1 ♦2 ♦3 ♦4 ♦5 ♦6 ♦M ♦N
♦O
C J K O` Y}
uc-test-UTF-32LE-bom.txt
A S C I I a b c d e x y z
G e r m a n ä ö ü Ä Ö Ü ß
P o l i s h ą ę ź ż ń ł
R u s s i a n а б в г д е ж э ю я
C J K 你 好
uc-test-UTF-32LE-nobom.txt
A S C I I a b c d e x y z
G e r m a n õ ÷ ³ ─ Í ▄ ▀
P o l i s h ♣☺ ↓☺ z☺ |☺ D☺ B☺
R u s s i a n 0♦ 1♦ 2♦ 3♦ 4♦ 5♦ 6♦ M♦ N
♦ O♦
C J K `O }Y
uc-test-UTF-8-bom.txt
´╗┐ASCII abcde xyz
German ├ñ├Â├╝ ├ä├û├£ ├ƒ
Polish ąęźżńł
Russian ð░ð▒ð▓ð│ð┤ðÁð ÐìÐÄÐÅ
CJK õ¢áÕÑ¢
uc-test-UTF-8-nobom.txt
ASCII abcde xyz
German ├ñ├Â├╝ ├ä├û├£ ├ƒ
Polish ąęźżńł
Russian ð░ð▒ð▓ð│ð┤ðÁð ÐìÐÄÐÅ
CJK õ¢áÕÑ¢
The only thing that works is UTF-16LE file, with a BOM, printed to the
console via type.
If we use anything other than type to print the file, we get garbage:
Z:\andrew\projects\sx\1259084>copy uc-test-UTF-16LE-bom.txt CON
■A S C I I a b c d e x y z
G e r m a n õ ÷ ³ ─ Í ▄ ▀
P o l i s h ♣☺↓☺z☺|☺D☺B☺
R u s s i a n 0♦1♦2♦3♦4♦5♦6♦ M♦N♦O♦
C J K `O}Y
1 file(s) copied.
From the fact that copy CON does not display Unicode correctly, we can
conclude that the type command has logic to detect a UTF-16LE BOM at the
start of the file, and use special Windows APIs to print it.
We can see this by opening cmd.exe in a debugger when it goes to type
out a file:
After type opens a file, it checks for a BOM of 0xFEFF—i.e., the bytes
0xFF 0xFE in little-endian—and if there is such a BOM, type sets an
internal fOutputUnicode flag. This flag is checked later to decide
whether to call WriteConsoleW.
But that’s the only way to get type to output Unicode, and only for files
that have BOMs and are in UTF-16LE. For all other files, and for programs
that don’t have special code to handle console output, your files will be
interpreted according to the current codepage, and will likely show up as
gibberish.
You can emulate how type outputs Unicode to the console in your own programs like so:
#include <stdio.h>
#define UNICODE
#include <windows.h>
static LPCSTR lpcsTest =
"ASCII abcde xyz\n"
"German äöü ÄÖÜ ß\n"
"Polish ąęźżńł\n"
"Russian абвгдеж эюя\n"
"CJK 你好\n";
int main() {
int n;
wchar_t buf[1024];
HANDLE hConsole = GetStdHandle(STD_OUTPUT_HANDLE);
n = MultiByteToWideChar(CP_UTF8, 0,
lpcsTest, strlen(lpcsTest),
buf, sizeof(buf));
WriteConsole(hConsole, buf, n, &n, NULL);
return 0;
}
This program works for printing Unicode on the Windows console using the
default codepage.
For the sample Java program, we can get a little bit of correct output by
setting the codepage manually, though the output gets messed up in weird ways:
Z:\andrew\projects\sx\1259084>chcp 65001
Active code page: 65001
Z:\andrew\projects\sx\1259084>java Foo
== UTF-8
= no bom
ASCII abcde xyz
German äöü ÄÖÜ ß
Polish ąęźżńł
Russian абвгдеж эюя
CJK 你好
ж эюя
CJK 你好
你好
好
�
= bom
ASCII abcde xyz
German äöü ÄÖÜ ß
Polish ąęźżńł
Russian абвгдеж эюя
CJK 你好
еж эюя
CJK 你好
你好
好
�
== UTF-16LE
= no bom
A S C I I a b c d e x y z
…
However, a C program that sets a Unicode UTF-8 codepage:
#include <stdio.h>
#include <windows.h>
int main() {
int c, n;
UINT oldCodePage;
char buf[1024];
oldCodePage = GetConsoleOutputCP();
if (!SetConsoleOutputCP(65001)) {
printf("error\n");
}
freopen("uc-test-UTF-8-nobom.txt", "rb", stdin);
n = fread(buf, sizeof(buf[0]), sizeof(buf), stdin);
fwrite(buf, sizeof(buf[0]), n, stdout);
SetConsoleOutputCP(oldCodePage);
return 0;
}
does have correct output:
Z:\andrew\projects\sx\1259084>.\test
ASCII abcde xyz
German äöü ÄÖÜ ß
Polish ąęźżńł
Russian абвгдеж эюя
CJK 你好
The moral of the story?
typecan print UTF-16LE files with a BOM regardless of your current codepage- Win32 programs can be programmed to output Unicode to the console, using
WriteConsoleW. - Other programs which set the codepage and adjust their output encoding accordingly can print Unicode on the console regardless of what the codepage was when the program started
- For everything else you will have to mess around with
chcp, and will probably still get weird output.