Do something reasonable with string literals

[jlapeyre]

OQ3 supports neither a string type nor string literals. (It does support bit string literals, eg "1001")

Currently a string literal will cause a panic in syntax_to_semantics.rs here
here

openqasm3_parser/crates/oq3_semantics/src/syntax_to_semantics.rs

Line 702 in b9e9ae8

_ => todo!(), // error. can/should be caught at syntax level, obviously

The lexer tokenizes string literals (eg "abc") as token STRING. The parser then constructs the variant SyntaxKind::Literal

openqasm3_parser/crates/oq3_parser/src/grammar/expressions/atom.rs

Lines 11 to 30 in b9e9ae8

	pub(crate) const LITERAL_FIRST: TokenSet = TokenSet::new(&[
	T![true],
	T![false],
	INT_NUMBER,
	FLOAT_NUMBER,
	SIMPLE_FLOAT_NUMBER,
	BYTE,
	CHAR,
	STRING,
	BIT_STRING,
	]);
	pub(crate) fn literal(p: &mut Parser<'_>) -> Option<CompletedMarker> {
	if !p.at_ts(LITERAL_FIRST) {
	return None;
	}
	let m = p.start();
	p.bump_any();
	Some(m.complete(p, LITERAL))
	}

I'm pretty sure that everywhere a string literal is encountered in the code above it is a string literal that is not part of the language (ie bit string literals, pragmas, etc. have been filtered out). So we could record a parse error here. Since we currently do not proceed with semantic analysis if there are parse errors, the panic mentioned above would not occur. But we'd still want to do something reasonable in place of the todo!.

I should also note that you can get a string literal at the level of the AST (really CST) although this should not be part of the language:

openqasm3_parser/crates/oq3_syntax/src/ast/expr_ext.rs

Lines 288 to 297 in b9e9ae8

	pub enum LiteralKind {
	String(ast::String),
	BitString(ast::BitString),
	IntNumber(ast::IntNumber),
	FloatNumber(ast::FloatNumber),
	SimpleFloatNumber(ast::SimpleFloatNumber),
	TimingFloatNumber(ast::TimingFloatNumber),
	Char(ast::Char),
	Byte(ast::Byte),
	Bool(bool),

version/commit

b9e9ae8

[jlapeyre]

• With Record parse error on encountering string literal #123 a parse error is recorded.
  So the panic mentioned in the opening comment no longer occurs.