Trying to parse a soft keyword

What is your question?

I’m currently making a Python transpiler, using lark/grammars/python.lark for parsing. The grammar is almost fine, but it lacks match…case support.

I’ve adapted grammar from official docs, but now I encounter a new problem

If you’re having trouble with your code or grammar Here’s a minimal example, with semicolons and braces added to drop indentation checks

from lark import Lark

grammar = """
%ignore /[\t \f\n]+/x
NAME: /[a-zA-Z_]\w*/
MATCH: "match"
CASE: "case"
NUMBER: ("+" | "-")? "0".."9"

expr: NAME | NUMBER

agmt: NAME "=" expr
match: MATCH expr "{" case+ "}"
case: CASE expr block

block: "{" code "}"
code: stmt+

stmt: (expr | match | agmt) ";"
"""

parser = Lark(grammar, start="code", parser="lalr")

code = """match = 1;
case = 2;
match N {
    case Z {
        5;
    } 
};"""

parser.parse(code)

What happens is “match” in first line treated as the beginning of match rule and fails to parse. Is there a way to resolve this while using LALR(1), not Earley?

[MegaIng]

Not in general. There is a reason python3.10 does no longer use a lalr compatible parser/grammar. You can maybe get around this with a postlex that looks at the next token/symbol/end of the line for a colon l and change it according to that, but this is simply not context free.

[MegaIng]

@erezsh Yeah, you are correct it works.

@evtn #1016 should contain full support for parsing match statements.