Terminals with complex data type

[Lotes]

For Langium terminals it is possible to be defined as RegExp literal.
One feature of regular expressions are named groups.

The idea is to map named groups to properties. Here is one example for RGB values:

//before, parse later
ColorLiteral returns string: /#[a-f0-9]{6}/i;

//after
interface Color {
  red: string;
  green: string;
  blue: string;
}
ColorLiteral returns Color: /#(?<red>[a-f0-9]{2})(?<green>[a-f0-9]{2})(?<blue>[a-f0-9]{2})/i;

But maybe it is too much complexity for so little value. But now it is out of my head, more space for other stuff teehee

Values, I see, are:

• less duplicated code on TS side (the regex is duplicated if you need to parse the token again)
• complex types for terminals

[spoenemann]

Nice idea! Currently you'd have to do this with a parser rule:

ColorLiteral: '#' red=ColorToken green=ColorToken blue=ColorToken;

terminal ColorToken: /#[a-f0-9]{2}/;

But this would see the object as four distinct tokens and thus allow whitespace between them. The whitespace could be suppressed with a validation rule that looks at the CST to find "hidden tokens".