This is how they do it on the web, really. It works as expected and matches any flavor of newline (Win, Unix, Mac) despite the ambiguity that "\n\r" can be matched by two subrules, in the end it will be properly matched just by one rule. Also, grammar can be k=2 for lexer.

Several operators has been implemented (and fixed), but still is many "non deterministics" parser warnings.

Implemented command line test for testing values expression.