Would be nice if only prefix operator++ required for iterators passed to deserialize

[petebannister]

Thank you for this excellent library.

Didn't really seem to be a bug.. but iterators passed to deserialize require postfix operator++ to be defined.
The iterator requirements could be slightly relaxed to only require prefix version.

Postfix version may be slightly less efficient (and also more error-prone to implement), depending upon the iterator object (though compiler may be clever enough to optimize).. See https://people.csail.mit.edu/addy/effective_c++/MEC/MI6_FR.HTM

[petebannister]

Also it requires array subscript operator on the iterator

[fktn-k]

@petebannister
Thanks for your suggestion! I didn't notice those undocumented requirements.
Just as described above, deserialize() roughly requires random access iterators, which is surely unnecessary.
The requirement can be relaxed to LegacyInputIterator ((void)r++ isn't necessary though) so deserialize() covers more use cases, which should be achieved with small fixes.

Postfix version may be slightly less efficient (and also more error-prone to implement), depending upon the iterator object (though compiler may be clever enough to optimize)..

You're right. There's no need to use postfix version and we should avoid depending on such a fragile expectation if possible.

[fktn-k]

@petebannister
Iterators for deserialization are now accepted when they only have prefix increment operator and in/equality operators by merging the PR #478.
Could you confirm the changes in case I misunderstood something?

[petebannister]

Looks good! Thank you