fix(scanner): correct implements keyword tokenization

[bastitva0-blip]

fixes #21

fix(scanner): correct implements keyword tokenization

What was broken

The implements keyword always tokenized as TOKEN_IDENTIFIER, making interface contracts silently non-functional. The parser had correct handling ready at parser.c:471 but TOKEN_IMPLEMENTS was never produced by the scanner.

Two bugs in the same line inside identifierType() in scanner.c:

// Before — broken
if (scanner->current - scanner->start > 2 && scanner->start[2] == 'l')
    return checkKeyword(scanner, 4, 7, "ements", TOKEN_IMPLEMENTS);

Bug 1: scanner->start[2] == 'l' — wrong character. implements is i-m-p-l-e-m-e-n-t-s, so index 2 is 'p', not 'l'. This condition was never true.

Bug 2: checkKeyword(scanner, 4, 7, "ements") — wrong length. "ements" is 6 characters, not 7. Even with the character check fixed this would always fail the memcmp.

The fix

// After — correct
if (scanner->current - scanner->start > 3 &&
    scanner->start[2] == 'p' && scanner->start[3] == 'l')
    return checkKeyword(scanner, 4, 6, "ements", TOKEN_IMPLEMENTS);

• Check start[2] == 'p' and start[3] == 'l' (the pl in im-pl-ements)
• Guard changed to > 3 since we now read up to index 3
• Length corrected to 6 to match "ements"

Diff

- if (scanner->current - scanner->start > 2 && scanner->start[2] == 'l')
-     return checkKeyword(scanner, 4, 7, "ements", TOKEN_IMPLEMENTS);
+ if (scanner->current - scanner->start > 3 &&
+     scanner->start[2] == 'p' && scanner->start[3] == 'l')
+     return checkKeyword(scanner, 4, 6, "ements", TOKEN_IMPLEMENTS);

File changed: src/compiler/lexer/scanner.c — 3 lines changed

Testing

// interface_test.prox
interface Greetable {
    func greet();
}

class Person implements Greetable {
    func greet() {
        print("Hello!");
    }
}

let p = new Person();
p.greet();

Before fix: implements parsed as identifier, parser errors or silently skips interface binding.
After fix: tokenizes correctly as TOKEN_IMPLEMENTS, parser enforces interface contract as expected.

[bastitva0-blip]

@ProgrammerKR
Hey! Just wanted to give a quick summary of why this is safe to merge — the change is 3 lines in a single file, no logic changes anywhere else, no new dependencies, and no risk of regressions since the broken branch was literally unreachable before. The implements keyword was silently doing nothing, so this only adds behaviour that was always intended. Parser side at line 471 already handles TOKEN_IMPLEMENTS correctly — this just finally lets the scanner produce it. Happy to answer any questions!

[ProgrammerKR]

Thanks for the detailed investigation and the well-explained fix.

I reviewed the changes, and the root cause makes sense:

• Corrected the character check from 'l' to 'p'/'l'
• Fixed the keyword length to match "ements"

The implementation is small, focused, and aligns with the intended scanner behavior.

I'll merge this after verifying the Windows CI failure is unrelated to this change.
Thanks for your contribution!

[ProgrammerKR]

Fixed by #22.

The scanner now correctly recognizes the implements keyword and produces TOKEN_IMPLEMENTS instead of TOKEN_IDENTIFIER.

Closing this issue. Thanks to @bastitva0-blip for the contribution.

[bastitva0-blip]

Hey, looks like the PR got closed instead of merged — was that intentional or did you mean to merge it? The CI failure should be unrelated to the scanner change since it's a pre-existing issue

[bastitva0-blip]

thanks for the merge @ProgrammerKR