Remove <- and <-> forms

[nikomatsakis]

This involves:

• Converting a <- b to a = move b
• Converting a <-> b to util::swap(&mut a, &mut b) or util::replace() (I suspect most uses would be better written with the latter)
• Implementing util::swap() and util::replace() unsafely, probably by converting the arguments to unsafe pointers and then using the move_val_init intrinsic, which moves a value without dropping the previous value

[catamorphism]

I have a patch to remove <- that I'm just running tests on.

[catamorphism]

Just curious, why is replace better than swap?

[catamorphism]

There is a problem with removing <->: for example, in code that appears in pop_internal in send_map

bucket <-> self.buckets[idx];

Changing this to

swap(&mut bucket, &mut self.buckets[idx]);

doesn't work:

/Users/tchevalier/rust/src/libcore/send_map.rs:227:22: 227:38 error: illegal borrow unless pure: unique value in aliasable, mutable location
/Users/tchevalier/rust/src/libcore/send_map.rs:227             swap(&mut self.buckets[idx], &mut x);
                                                                         ^~~~~~~~~~~~~~~~
/Users/tchevalier/rust/src/libcore/send_map.rs:227:12: 227:16 note: impure due to access to impure function
/Users/tchevalier/rust/src/libcore/send_map.rs:227             swap(&mut self.buckets[idx], &mut x);

replace doesn't work either, for the same reason.

This may be a reason to keep <->. Or I guess we could cheat and make swap pure (it's going to use unsafe code anyway)? I'm not sure.

[brson]

If pure functions are allowed to mutate mutable in pointers (as was discussed at some point) then swap can be pure.

[nikomatsakis]

That ... is true. Interesting observation.

[catamorphism]

Left-arrow was removed already, but removing the swap operator is blocked on #5384. De-milestoning

[thestinger]

Fixed as by 9f106a6.