Variable no-shadowing problems

[Gregor Richards]

Chris Miller wrote:
> To those not sure what shadowing means, here is an example of variable  
> shadowing:
>    int i;
>    {
>       int i;
>    }
> The inner `i` variable shadows the outer one.
> 
> 
> D has made such shadowing illegal and flags it as an error. I'm against  
> this restriction, but I have a compromise.
> 
> I am proposing a `shadow` keyword be added to D. Actually, the general  
> idea is from someone else from #D IRC chat, but I have several reasons 
> why  explained below.
> 
> 1) Accidentally shadowing a variable. If this happens you will have to 
> go  back and manually fix up all uses of the shadowed name, which has a 
> very  high chance of error. Forget to fix one and it goes by silently; 
> bug. Or,  simply add `shadow` to its declaration, avoiding tedious, 
> error-prone  fixing, and make people aware of what's going on.

I was going to respond to this somehow, but I don't think I really have 
a response ... IMHO this doesn't actually cut down the debugging time at 
all, since in all likelihood coder-X will have inadvertantly 
mixed-and-matched in the inner scope. Maybe that's not true, I don't know :)

> 
> 2) Adding code. You can simply add a { } block and declare at will. 
> With  no shadowing, I am tempted to append random numbers to my variable 
> names  so I don't run into this, or moreso tempted to reuse variable 
> names like  is popular from C, as in declare once at the top of the 
> function and reuse  the variable for different things.

I think this scenario could be more cleanly filed under "code reuse." If 
you copy/paste code from one project to another, with shadowing you 
don't have to worry (too much) about variable collision.

> 
> 3) Code generators emitting D source code. They may not wish to keep 
> track  of such shadowing and could simply mark all declarations as 
> `shadow`.

This I agree with entirely.

> 
> 
> Notes:
> 
> `shadow` does not allow redeclarations in the exact same scope 
> (although  perhaps it could).

Redeclaration in the same scope ... hmmmmmm ... makes me gag a bit.

> 
> As for reason (1), if you are personally against shadowing, you are not  
> forced to use the keyword and can continue treating it how it is done now.

Yeah, if you never plan on reading anybody else's code ;)

> 
> - Chris

My opinion in short: I don't agree with 1, I somewhat agree with 2, I 
agree entirely with 3. You've covered the full gambit, bravo ;)

  - Gregor Richards