ANTLR4: Using non-ASCII characters in token rules

Question

On page 74 of the ANTRL4 book it says that any Unicode character can be used in a grammar simply by specifying its codepoint in this manner:

'\uxxxx'

where xxxx is the hexadecimal value for the Unicode codepoint.

So I used that technique in a token rule for an ID token:

grammar ID;

id : ID EOF ;

ID : ('a' .. 'z' | 'A' .. 'Z' | '\u0100' .. '\u017E')+ ;
WS : [ \t\r\n]+ -> skip ;

When I tried to parse this input:

Gŭnter

ANTLR throws an error, saying that it does not recognize ŭ. (The ŭ character is hex 016D, so it is within the range specified)

What am I doing wrong please?

Answer 1

ANTLR is ready to accept 16-bit characters but, by default, many locales will read in characters as bytes (8 bits). You need to specify the appropriate encoding when you read from the file using the Java libraries. If you are using the TestRig, perhaps through alias/script grun, then use argument -encoding utf-8 or whatever. If you look at the source code of that class, you will see the following mechanism:

InputStream is = new FileInputStream(inputFile);
Reader r = new InputStreamReader(is, encoding); // e.g., euc-jp or utf-8
ANTLRInputStream input = new ANTLRInputStream(r);
XLexer lexer = new XLexer(input);
CommonTokenStream tokens = new CommonTokenStream(lexer);
...

Answer 2

For those having the same problem using antlr4 in java code, ANTLRInputStream beeing deprecated, here is a working way to pass multi-char unicode data from a String to a the MyLexer lexer :

    String myString = "\u2013";

    CharBuffer charBuffer = CharBuffer.wrap(myString.toCharArray());
    CodePointBuffer codePointBuffer = CodePointBuffer.withChars(charBuffer);
    CodePointCharStream cpcs = CodePointCharStream.fromBuffer(codePointBuffer);

    OneLexer lexer = new MyLexer(cpcs);       
    CommonTokenStream tokens = new CommonTokenStream(lexer);