fix(es/parser): Re-lex << as two <-s if required

[kdy1]

Description:

Related issue:

• Closes Parser error for type parameters containing a generic function signature #7187.

[sandersn]

I looked at this and found the source of the bug, but I'm not sure how to fix it. It's in expr.rs:691, near the end of parse_member_expr_new_expr_inner. For the example, assertType<<K ..., << gets lexed as LShift not <,< because the context has should_not_lex_lt_or_gt_as_type set (and maybe !in_type, I'm not sure).

Problem is: this context is correct at the time. The lexing actually happens in parse_primary_expr, which should not be in a type context. But by the time parse_primary_expr returns, the lexer has already read LShift, so on line 691, let type_args = if self.syntax().typescript() && is!(self, '<') { will fail because it's LShift not <.

The Typescript parser relies on contexts less heavily than swc, and its lexer has a reScanLessThanToken method that it calls at this point; the lexer backs up 1 character and changes its token to < if the current token is LShift. However, I don't know if swc has this capability. It's possible that you can change the context in order to cause rescanning of the current token, but my initial reading of the code says No.

@kdy1 could you advise on whether rescanning is a possibility in the swc lexer?

A third option would be massive lookahead, in which parse_primary_expr is parsed in a context with should_not_lex_lt_or_gt_as_type set to false, but I expect this would be expensive.

[kdy1]

swc-bump:

• swc_ecma_parser

[kdy1]

@sandersn Thank you! Your comment was very helpful.

[swc-bot]

Automated review comment generated by auto-rebase script