bearophile wrote:
Andrei Alexandrescu:
I just think it's the wrong problem to work on.<
Beside multiprocessing (that I am ignorant to comment on still), I
can see other purposes for having a way to tell the type system that
it exists only one reference/pointer to mutable data and ways to
safely change ownership of such pointer. It can also be used by the
compiler to optimize in various situations.
Correct. We've been trying valiantly to introduce unique in the type
system two times, first time in 2007. Our conclusion back then has been
that unique brings more problems than it solves. Last meeting the same
pattern ensued. Problems with unique cropped up faster than those solved.
Andrei
