A small const problem

[Jose Armando Garcia]

On Thu, Mar 15, 2012 at 6:43 PM, bearophile <bearophileHUGS at lycos.com> wrote:
> In my code I have seen that in some situations I'd like to do
> something with structs that const doesn't allow me to do. There
> are many different cases of such usage, the following isn't my
> most common example, but it shows better what I mean:
>
>
> import std.random: randomShuffle;
> struct Foo {
>      const int x, y;
> }
> void main() {
>      auto foos = [Foo(1,2), Foo(3,4)];
>      randomShuffle(foos);
> }
>

Maybe in this particular case what you really want to do is shuffle the indices.

>
> Here the Foo struct has two fields, they are tagged with const
> because no method of Foo is ever allowed to change their value
> once a Foo is built, and no one outside of Foo is allowed to
> change them. So they are quite constant.
>
> On the other hand if I have two or more Foos I sometimes want to
> rebind an old Foo with a new one (a simple variable assign), or
> in this program I even want to shuffle them. Shuffling them
> doesn't change the logically constant nature of their x and y
> fields.
>
> So in a way here I am looking for a const that makes the struct
> fields not mutable, but allows me to rebind the whole structs.
>
> Currently std.typecons.rebindable works only on class instances,
> interfaces and arrays (std.typecons.rebindable seems to accept
> fixed-sized arrays too, but I don't know if this meaningful &
> good). So it's not a solution.
>
> I think the small problem shown here is not the logical const
> needed for storing things like lazily computed hash values for
> immutable structs. Here nothing inside Foo is allowed to change
> once Foo is created (here I have not added mutable fields to Foo).
>
> In such situations I sometimes don't use const, or sometimes I
> use class instances, or even pointers.
>
> Bye,
> bearophile