I have a generic enum where only one variant makes use of the generic type.
When “converting” a ParseResult<T>
to a ParseResult<U>
the compiler forces me to destruct the non-generic part, too. The non-generic part is immediately reassembled and returned in exactly the same way.
Is there a more elegant solution without the code duplication?
#[derive(PartialEq, Debug)]
enum ParseResult<'a, T> {
Success(T, &'a str),
Failure(&'static str),
}
fn char<'a>(c: char) -> impl Fn(&'a str) -> ParseResult<'a, char> {
move |input| match input.chars().next() {
Some(candidate) if c == candidate => ParseResult::Success(candidate, &input[1..]),
Some(_) => ParseResult::Failure("chars didn't match"),
None => ParseResult::Failure("unexpected end of stream"),
}
}
fn or<'a, T>(
mut lhs: impl FnMut(&'a str) -> ParseResult<T>,
mut rhs: impl FnMut(&'a str) -> ParseResult<T>,
) -> impl FnMut(&'a str) -> ParseResult<T> {
move |input| match lhs(input) {
ParseResult::Failure(_) => rhs(input),
ok => ok,
}
}
fn and<'a, T, U>(
mut lhs: impl FnMut(&'a str) -> ParseResult<T>,
mut rhs: impl FnMut(&'a str) -> ParseResult<U>,
) -> impl FnMut(&'a str) -> ParseResult<(T, U)> {
move |input| match lhs(input) {
ParseResult::Success(left, after) => match rhs(after) {
ParseResult::Success(right, after) => ParseResult::Success((left, right), after),
// Only explicit works:
ParseResult::Failure(why) => ParseResult::Failure(why),
},
// Same as line as above, same goes here.
ParseResult::Failure(why) => ParseResult::Failure(why),
}
}
Things that didn't work instead of the two ::Failure
lines towards the end:
bad => bad,
expected tuple, found type parameter U
bad => bad as ParseResult<(T, U)>,
bad @ ParseResult::Failure(why) => bad,
expected tuple, found type parameter U
But the non-generic part is immediately reassembled and returned in exactly the same way. Isn't there a smarter/more elegant solution without the code duplication?
It looks like it is reassembled the same way, but it isn't because it changed type from ParseResult<T>
to ParseResult<U>
(or ParseResult<(T, U)>
in your case). These have different sizes so ParseResult<T>::Failed(err)
is not the same as ParseResult<U>::Failed(err)
.
If we look at how the standard library handles this problem for Result::map
, you can see that the same pattern is used:
match self {
Ok(t) => Ok(op(t)),
Err(e) => Err(e),
}