According to the docs for Split, there is a rev
method on the result of doing split
on a string:
fn main() {
let mut length = 0;
The problem is that rev()
is defined on the Split
iterator only if it implements DoubleEndedIterator
, but Split
only implements DoubleEndedIterator
if the searcher of the pattern you are splitting on satisfies DoubleEndedSearcher
:
impl<'a, P> DoubleEndedIterator for Split<'a, P>
where
P: Pattern<'a>,
<P as Pattern<'a>>::Searcher: DoubleEndedSearcher<'a>,
The documentation lists which types implement DoubleEndedSearcher. None of the types there correspond to &str
pattern, so you can't use rev()
when you split on string.
In your particular case, I guess, it is sufficient to change split(":")
to split(':')
(i.e. split on character instead of string) because the character pattern searcher does implement DoubleEndedSearcher
.
Such features of Rust (conditional trait implementation and trait bounds local to a method) allow writing really expressive code, but they can be sometimes hard to read through.
TLDR: StrSearcher
(the type that does searching for string patterns) doesn't implement DoubleEndedSearcher
and, as such, the split
iterator doesn't implement DoubleEndedIterator
and, as such, you can't call rev
on it.
If you look at the documentation on that page for rev
, you'll see where Self: DoubleEndedIterator
. This means that rev
is defined if and only if the type the Iterator
trait is being implemented for (which is Split
) also has an implementation of the DoubleEndedIterator
trait.
If you look further down, you'll see:
impl<'a, P> DoubleEndedIterator for Split<'a, P>
where P: Pattern<'a>, P::Searcher: DoubleEndedSearcher<'a>
Thus, DoubleEndedIterator
is implemented for Split
if and only if both of those conditions are satisfied: P
must be a Pattern
and the Searcher
type it defines must implement DoubleEndedSearcher
.
Now, you're using a string literal as the pattern, so if you check the documentation for the str type, you'll see:
impl<'a, 'b> Pattern<'a> for &'b str
Within that, the associated Searcher
type is defined as:
type Searcher = StrSearcher<'a, 'b>
Nearly there! Follow the link to the documentation for StrSearcher and...
...there's no implementation of DoubleEndedSearcher
. Thus, the required bounds are not satisfied and rev
can't be used on the Split
iterator.
The other answers are correct, but I want to point out rsplit. This is probably more obvious and more performant.
So, why can't you use rev
? As other answers state, it's not implemented for StrSearcher
. But why is it not implemented? From the DoubleEndedSearcher docs:
For this, the impl of
Searcher
andReverseSearcher
need to follow these conditions:
- All results of
next()
need to be identical to the results ofnext_back()
in reverse order.next()
andnext_back()
need to behave as the two ends of a range of values, that is they can not "walk past each other".
The problem with reversing the iterator using strings is this:
"baaab".split("aa") // -> ["b", "aa", "ab"];
However, if you were to start at the end of the string, you'd get something like:
"baaab".split("aa").rev() // -> ["b", "aa", "ba"]
Which is clearly not the same set of items in a different order!
Simply put, you can't reverse an iterator that is split on strings because there's no efficient way of knowing when the next result is. You'd have to split the entire string into a collection, then reverse the collection!
This is why rsplit
exists - it means start at the end of the string and split to the beginning, in an efficient manner.