Implement unary -

[yutannihilation]

Currently, this code

fn dsp() {
    2.0-2.0
}

cannot be parsed.

Filename: \\?\C:\Users\Yutani\Documents\GitHub\mimium-rs\tmp.mmm
, ror: unexpected token something else, expected ^, ;,
, (, |, -, /, ", :, *, <, !, =, 0, ., +, }, [, ), %, ,, &, {, ], >
   ╭─[\\?\C:\Users\Yutani\Documents\GitHub\mimium-rs\tmp.mmm:2:3]
   │
 2 │     2.0-2.0
   ·   ┬
   ·   ╰── Unexpected token
───╯
Error: unexpected token while parsing apply, expected ||, linebreak, }, |>, >, -, <=, +, !=, /, ;, >=, &&, %, (, ==, <, *, ^
   ╭─[\\?\C:\Users\Yutani\Documents\GitHub\mimium-rs\tmp.mmm:2:8]
   │
 2 │     2.0-2.0
   ·        ──┬─
   ·          ╰─── Unexpected token -2.0
───╯

This is because there's no parser for unary operations. Currently, the lexer parses - as part of a float literal.

mimium-rs/mimium-lang/src/compiler/parser/lexer.rs

Lines 24 to 30 in ee0ed1a

	let float = just('-')
	.or_not()
	.then(
	text::int(10)
	.then(just('.'))
	.then(text::digits(10).or_not()),
	)

It might be a choice to require spaces before and after - to distinguish the binary op and the unary op. But, I think users expect this syntax is valid, so mimium anyway needs a unary parser.

-x

[tomoyanonymous]

Thanks! Maybe should we add more test cases with nested operation to ensure the parse works correctly?

    run_simple_test("- -1.0", 1.0, 3);
    run_simple_test("-(-hoge)", 2.0, 3);

[yutannihilation]

Thanks. Sure, I'll add some more test cases.

[yutannihilation]

It seems a function call like this is not handled properly, so my implementation is not correct. Let me think.

-cos(0.0)

[tomoyanonymous]

Now parser emits ast for -cos(0.0) to (app (app neg (cos))((float 0.0)))). ((- cos) (0.0))
(You can emit raw ast as lisp-like syntax with --emit-ast option )

Now precedence of unary operator is bigger than apply. It should be just in between apply and other binary operators.

[yutannihilation]

Now precedence of unary operator is bigger than apply. It should be just in between apply and other binary operators.

You are right, thanks. Should be fixed now!