Strange behaviour with reversed()

[Duncan Booth]

Steven D'Aprano <steve at REMOVE-THIS-cybersource.com.au> wrote:

>> Note that the starting index is determined at creation time, not when
>> the iteration begins. So, if you create a reversed object over a list
>> containing 3 elements, the first returned element will be seq[2], 
then
>> seq[1], then seq[0]. It doesn't matter if you modify the list after
>> creating the reversed object but before starting the iteration: it 
will
>> start at seq[2] even if it's not the last item, and will silently 
stop
>> if seq[2] is not a valid index anymore.
> 
> You know, that's probably the *least* intuitive behaviour possible.
> 
> For reversed() to iterate over the input as it was at creation time is 
a 
> reasonable design choice; for it to iterate over the input as it is at 
> call time is also a reasonable design choice; but for it to iterate 
over 
> some unholy mutant melding of the sequence as-it-was and as-it-is is 
> simply bizarre. I hope it just fell out of the implementation and 
wasn't 
> a deliberate design choice.

You mean that you don't find it intuitive that it iterates through the 
indices that existed at creation time and returns the values as they are 
now? I'd have said that was the *most* intuitive behaviour.

The only other behaviours I would regard as intuitive for iteration over 
a mutating sequence would be to throw an exception either for mutating 
the sequence while the iterator exists or for using the iterator after a 
mutation.