[ruby-core:32236] Re: Ruby 2.0 Wiki/Wish-list?

On Thu, Sep 9, 2010 at 4:20 AM, "Martin J. D=C3=BCrst"
<[e-mail]> wrote:
> On 2010/09/08 23:44, Yukihiro Matsumoto wrote:
>> But just for curiosity, how do you define the semantics of x++ and
>> ++x, where you don't have explicit type for variables in Ruby.
>> For x +=3D 1, it's easily defined as x =3D x + 1, but not that easy for
>> x++ and ++x.
>
> I don't understand how this relates to explicit types for variables. ++x =
is
> just x =3D x + 1. x++ is something like (x'=3Dx, x =3D x+1, x'). Ruby doe=
sn't have
> the C comma operator, and this can't implemented as a method, because we
> can't pass a variable by reference (the equivalent of &x in C) in Ruby. B=
ut
> this isn't relevant because like +=3D and friends, ++ would just be synta=
ctic
> sugar implemented by the interpreter, not overridable operators.

A key component to this ++ discussion is the atomicity of updates. x
+=3D 1 is obviously not atomic, since it expands to x =3D x + 1 (one read,
one increment, one write). In C/C++ and in Java, the increment
operation is still not atomic, since it will usually compile to the
same three operations, albeit using memory locations and registers. As
it stands, +=3D, ||=3D, and friends are easier to accept as non-atomic. ++
is very often perceived to be atomic when it is not, and so
introducing it would make discussions about atomicity more
complicated.

FWIW, though, I still like the idea of having ++ as sugar for +=3D 1, if
only because it's two+ fewer characters to type.

In fact, if you slide the 1 over on top of the =3D, it almost looks like
++ (ok, it's more like +=C2=B1 or +=E2=89=A0, but hey, -> doesn't *really* =
look
like a =CE=BB).

- Charlie