Block comment ignores strings

[PavelVozenilek]

This failed:

/*
void main() {
    println("Hello /* world"); // <<< open block comment in a string
}
*/

void main() {
    println("Hello world");
}

with error:

main.delta:1:1: error: unterminated block comment
/*
^

I noticed, that in file parser/lex.cpp the function Lexer::readBlockComment doesn't handle strings.

[emillaine]

Good catch. But I don't this is feasible to implement, because comments may contain arbitrary text, making it nearly impossible to reliably detect if the /* is inside a "string" or not. For example:

/* Some explanation here.
var c = '"'; /* The ascii value of `"` is 34.
                More explanation. */
*/

Here the /* is enclosed between two " characters on the same line, but it's actually part of a nested block comment, and therefore should not be ignored.

I tested with Swift, which also has nested comments, and it doesn't seem to handle strings inside block comments in a special way either.

[PavelVozenilek]

I noticed it when reading lexer'code. One way would be to lex everything first (including comments start/stop), and then to filter comments away.

[emillaine]

That wouldn't work because comments can contain text that doesn't lex properly. For example the text "\" is valid in a comment, but lexing would treat is as an unterminated string literal (since the last quote is escaped).