Syntax error when string union with escape

[ksss]

"\\" and 1 can create a Union type.
"a" and "\\" also can create a Union type,
whereas "\\" and "a" would result in a Syntax error.

Shouldn't it be possible to create a Union type in "\\" | "a" cases?

$ ruby -r rbs -e 'p RBS::Parser.parse_type(%q{"\\" | 1})'
#<RBS::Types::Union:0x00000001047b2118 @types=[#<RBS::Types::Literal:0x00000001047b22d0 @literal="\\", @location=#<RBS::Location:820 buffer=a.rbs, start=1:0, pos=0...3, children= source='"\"'>>, #<RBS::Types::Literal:0x00000001047b21b8 @literal=1, @location=#<RBS::Location:840 buffer=a.rbs, start=1:6, pos=6...7, children= source='1'>>], @location=#<RBS::Location:860 buffer=a.rbs, start=1:0, pos=0...7, children= source='"\" | 1'>>

$ ruby -r rbs -e 'p RBS::Parser.parse_type(%q{"a" | "\\"})'
#<RBS::Types::Union:0x0000000147162590 @types=[#<RBS::Types::Literal:0x0000000147162b58 @literal="a", @location=#<RBS::Location:820 buffer=a.rbs, start=1:0, pos=0...3, children= source='"a"'>>, #<RBS::Types::Literal:0x0000000147162720 @literal="\\", @location=#<RBS::Location:840 buffer=a.rbs, start=1:6, pos=6...9, children= source='"\"'>>], @location=#<RBS::Location:860 buffer=a.rbs, start=1:0, pos=0...9, children= source='"a" | "\"'>>

$ ruby -r rbs -e 'p RBS::Parser.parse_type(%q{"\\" | "a"})'
/Users/ksss/.rbenv/versions/3.1.1/lib/ruby/gems/3.1.0/gems/rbs-2.4.0/lib/rbs/parser_aux.rb:4:in `_parse_type': a.rbs:1:7...1:8: Syntax error: expected a token `pEOF`, token=`a` (tLIDENT) (RBS::ParsingError)
	from /Users/ksss/.rbenv/versions/3.1.1/lib/ruby/gems/3.1.0/gems/rbs-2.4.0/lib/rbs/parser_aux.rb:4:in `parse_type'
	from -e:1:in `<main>'

$ ruby -v
ruby 3.1.1p18 (2022-02-18 revision 53f5fc4236) [arm64-darwin21]
$ rbs -v
rbs 2.4.0

[soutaro]

%q{"\\" | "a"} is "\" | "a" in RBS. So, the syntax error is expected one.

The problem is when we give "\\" | "a" in RBS. This should be parse. But, it didn't because of the dqstring pattern in lexer.re

      dqstring = ["] ("\\"["] | [^"\x00])* ["];                     // Before
      dqstring = ["] ("\\"[abefnrstv"\\] | [^"\\\x00])* ["];        // After

With the Before pattern, the first \ in "\\" was eaten by the second alternation. And the next \" in "\\" was eaton by the first alternation. It cannot detect the end of string, and parsing fails.

The After pattern is more explicit on what can come just after \. The two \\ are eaten by the first alternation, and the closing " is processed correctly.