From what I understand, the purpose of size_type
and difference_type
is not merely the sign -- it was also meant to address e.g. segme
I don't think it is always safe. This historical issue existed since first specification of C language, where ptrdiff_t
is not guaranteed to cover the entire positive range of size_t
. For obvious reasons, this issue carries over to the specification of std::vector
.
With standard C++ containers, it is guaranteed that size_type
covers the non-negative range of difference_type
, but the reverse coverage is not guaranteed.
However, the relationship between size()
and end() - begin()
for a standard container can be guaranteed by other means. The implementation is free to impose its own limitations on the maximum container size, which are exposed through container::max_size()
function. It can artificially restrict the maximum size to make sure the subtraction never overflows.
P.S. I'd say that the reason for difference_type
's existence is just the sign and nothing else. To be completely "safe" difference_type
should be 1 bit longer than size_type
. This is often difficult to achieve in practice, which is why it is not required by the language specification.