I'm writing a Parser class. It takes 1 argument in its constructor which is a Lexer object.
The question is i dont know whether i should take it as a reference (leaving ownership with the user / caller) or not (taking ownership). It could be that the user would want to keep the lexer object alive for longer than the parser, but when not then it seems annoying to have to.
So i figured why not support both? It seems convenient that the user of the class can choose to instantiate a Parser with just a reference to a lexer, giving them control of when the lexer gets cleaned up, OR whether to pass on ownership to the Parser and then not have to worry about keeping the lexer object alive and cleaning it up when the parser is done with it.
The problem is im not entirely sure how this looks like. Should i have two constructors like this?
1 2
|
Parser::Parser(Lexer&)
Parser::Parser(Lexer&&)
|
or should it look like this?
1 2
|
Parser::Parser(std::unique_ptr<Lexer>)
Parser::Parser<std::shared_ptr<Lexer>)
|
i want to support both value semantics and pointer semantics
and how would you hook it all up with the member variables behind the scene?
also what are your throughts about this? is it silly? I think its convenient but i haven't ever seen anyone else do something like this