stipulate sep parser must consume

[gdennie]

separated_list1 by multipspace0 sep failed #1691

the sep parser in the separated_list parser combinators cannot be "any" parser but must be a consuming parser, contrary to documentation. Ideally, we should have a trait to assert this constraint as well but better clarity in the documentation helps.

[epage]

many0 has a similar note that is separate from the Arguments list

Note: if the parser passed in accepts empty inputs (like alpha0 or digit0), many0 will return an error, to prevent going into an infinite loop

I'm assuming it'd be good to be consistent.

[gdennie]

Whether a parser consumes some input is clearly an important characteristic. As such, it would be good if that characteristic could be put into the type system.

It seems at least all of the multi combinators require that their subject parsers consume some input.

I wonder if this consuming requirement can be expressed in the type system through a ownership and return pattern and whether its worth the effort.

// Consuming Parsers...
fn(OldInput) -> NewInput {..}

// Optionally Consuming Parsers
fn(OldInput) -> Either<OldInput, NewInput> {..}

// Some machinery and ergonomics
impl From<NewInput> for OldInput {..}
impl From<Either<OldInput, NewInput>> for OldInput {..}

impl OldInput {
    fn next(&self) -> NewInput {..}
}
impl NewInput {
   fn next(&self) -> NewInput {..}
} 

[Geal]

thank you, it's a good idea to be explicit about it