In parser generators like yacc, a simple parser for mathematical expressions is the most widely used example to become familiar with the syntax and usage, so we do this even here. The parser definition language itself is described below in a more detailed way.
Store the following grammar definition into a text file of your choice. For parser source files, the
.par file extension could be used, so the file is named calc.par in such case.
/~ Expression calculator written in JS/CC ~/
/~ Tokens to be ignored (e.g. whitespace, comments) ~/
! ' |\t'
;
/~ Grammar tokens ~/
'\+'
'\-'
'\*'
'/'
'\('
'\)'
'[0-9]+' INT [* %match = parseInt( %match ); *]
'[0-9]+\.[0-9]*|[0-9]*\.[0-9]+' FLOAT [* %match = parseFloat( %match ); *]
;
##
/~ The non-terminal "p" is the entry symbol, because it is the first one! ~/
program: expr [* print( %1 ); *]
;
/~ Don't confuse with the tokens: Here, we use the unescaped values because these
are not interpreted as regular expressions at this position! ~/
expr: expr '+' term [* %% = %1 + %3; *]
| expr '-' term [* %% = %1 - %3; *]
| term /~ Default semantic action fits here! ~/
;
term: term '*' factor [* %% = %1 * %3; *]
| term '/' factor [* %% = %1 / %3; *]
| factor /~ Default semantic action fits here! ~/
;
factor: '(' expr ')' [* %% = %2; *]
| INT /~ Default semantic action fits here! ~/
| FLOAT /~ Default semantic action fits here! ~/
;
/~ This is the parser entry point; Because this entry point could be very individual,
the compiler programmer has to decide which way he wants to read the source, parse
it and report the errors, if there are any. ~/
[*
var error_offsets = new Array();
var error_lookaheads = new Array();
var error_count = 0;
var str = new String( arguments[0] );
if( ( error_count = __parse( str,
error_offsets, error_lookaheads ) ) > 0 )
{
for( i = 0; i < error_count; i++ )
print( "Parse error near \""
+ str.substr( error_offsets[i] ) +
"\", expecting \"" +
error_lookaheads[i].join() +
"\"" );
}
*]