Why is rev() defined on Iterator?

community
1

I just read the docs of Rev for unrelated reasons.
I wonder why rev() is defined on Iterator?
It requires Self: DoubleEndedIterator, so why isn't it defined there? After all, DoubleEndedIterator is a subtrait of Iterator.

2

I believe that's just historical baggage, which we can't change since 1.0 because it would be a breaking change. The rposition method is similarly weird, and this was debated a bit when rfind was added, ultimately on DoubleEndedIterator in spite of the PR title:

3

I suspected as much, but both were introduced in 1.0 according to the docs, so I thought that this was a deliberate design decision that I just didn't understand.

4

A couple of minor advantages (even if unintentional) that come to mind:

  • Much better discoverability in the docs
  • No need for an extra import to bring the method to scope.
5

DoubleEndedIterator is in the prelude.

6

Don't forget that “introduced in 1.0” means “introduced in nine years between 2006 and 2015”. It's not like “introduced in 1.x” for any other x

7

Thanks, should've double-checked.

8

I wonder. Would Evolving trait hierarhies allow to "fix" this. That is move rev method definition from Iterator to DobuleEndedIterator?

9

I think not, the supertrait relationship is backwards for that proposal.