Why is this allowed?

Why is this allowed?

[Jacques Carette]

type foo = Foo
let x = Foo

type foo2 = Foo | Bar
let y = Foo
let z = (x,y) ;;

I thought that re-using of algebraic labels was not allowed - but 
apparently it is?  Note that this means that it is impossible to "text" 
print such structures and hope to recover them uniquely.  This also 
causes very subtle issues when Marshal'ing, and grave issues for code 
generation [think metaocaml].

Jacques

Re: [Caml-list] Why is this allowed?

[Lukasz Stafiniak]

But what when someone includes or opens a module with Foo after "type
foo = Foo"? What when someone opens it locally in an expression?

Does a variant value have a unique type with a unique path so that it
would be possible to guarantee that within this unique path there are
no type *definitions* that override a variant name?

On Mon, May 31, 2010 at 4:46 PM, Jacques Carette <carette@mcmaster.ca> wrote:
> type foo = Foo
> let x = Foo
>
> type foo2 = Foo | Bar
> let y = Foo
> let z = (x,y) ;;
>
> I thought that re-using of algebraic labels was not allowed - but apparently
> it is?

Re: [Caml-list] Why is this allowed?

[Till Varoquaux]

AFAIK you are allowed to shadow just about anything in
implementations. If you wanted to keep the exact same interface but
retain the ability to avvoid the shadowing issue on the labels you
could do:

module Foo = struct
  type t = Foo
end
type foo = Foo.t = Foo

module Foo2 = struct
  type t = Foo | Bar
end

type foo2 = Foo2.t = Foo | Bar

HTH,
Till

On Mon, May 31, 2010 at 10:46 AM, Jacques Carette <carette@mcmaster.ca> wrote:
> type foo = Foo
> let x = Foo
>
> type foo2 = Foo | Bar
> let y = Foo
> let z = (x,y) ;;
>
> I thought that re-using of algebraic labels was not allowed - but apparently
> it is?  Note that this means that it is impossible to "text" print such
> structures and hope to recover them uniquely.  This also causes very subtle
> issues when Marshal'ing, and grave issues for code generation [think
> metaocaml].
>
> Jacques
>
> _______________________________________________
> Caml-list mailing list. Subscription management:
> http://yquem.inria.fr/cgi-bin/mailman/listinfo/caml-list
> Archives: http://caml.inria.fr
> Beginner's list: http://groups.yahoo.com/group/ocaml_beginners
> Bug reports: http://caml.inria.fr/bin/caml-bugs
>

Re: [Caml-list] Why is this allowed?

[Jacques Carette]

Lukasz Stafiniak wrote:
> But what when someone includes or opens a module with Foo after "type
> foo = Foo"? What when someone opens it locally in an expression?
>   
I would be fine if type-level shadowing was disallowed for all those cases.

> Does a variant value have a unique type with a unique path so that it
> would be possible to guarantee that within this unique path there are
> no type *definitions* that override a variant name?
>   
That would be nice.

Jacques

> On Mon, May 31, 2010 at 4:46 PM, Jacques Carette <carette@mcmaster.ca> wrote:
>   
>> type foo = Foo
>> let x = Foo
>>
>> type foo2 = Foo | Bar
>> let y = Foo
>> let z = (x,y) ;;
>>
>> I thought that re-using of algebraic labels was not allowed - but apparently
>> it is?
>>     

Re: [Caml-list] Why is this allowed?

[Jacques Carette]

Nice trick - but my 'problem' [1] regards arbitrary valid OCaml code and 
how to serialize code values.  The issue seems to be that one needs to 
attach an arbitrarily large typing environment to each part of a value, 
as there is no global typing environment which will do the job (because 
of shadowing).

Jacques

[1] Actually 'our' problem, in that this arose from some issues in 
metaocaml, and I am working with Oleg to figure out what is going on.  
Interestingly, this never arose in 3.09 and before.

Till Varoquaux wrote:
> AFAIK you are allowed to shadow just about anything in
> implementations. If you wanted to keep the exact same interface but
> retain the ability to avvoid the shadowing issue on the labels you
> could do:
>
> module Foo = struct
>   type t = Foo
> end
> type foo = Foo.t = Foo
>
> module Foo2 = struct
>   type t = Foo | Bar
> end
>
> type foo2 = Foo2.t = Foo | Bar
>
> HTH,
> Till
>
> On Mon, May 31, 2010 at 10:46 AM, Jacques Carette <carette@mcmaster.ca> wrote:
>   
>> type foo = Foo
>> let x = Foo
>>
>> type foo2 = Foo | Bar
>> let y = Foo
>> let z = (x,y) ;;
>>
>> I thought that re-using of algebraic labels was not allowed - but apparently
>> it is?  Note that this means that it is impossible to "text" print such
>> structures and hope to recover them uniquely.  This also causes very subtle
>> issues when Marshal'ing, and grave issues for code generation [think
>> metaocaml].
>>
>> Jacques
>>
>> _______________________________________________
>> Caml-list mailing list. Subscription management:
>> http://yquem.inria.fr/cgi-bin/mailman/listinfo/caml-list
>> Archives: http://caml.inria.fr
>> Beginner's list: http://groups.yahoo.com/group/ocaml_beginners
>> Bug reports: http://caml.inria.fr/bin/caml-bugs
>>
>>     
>
> _______________________________________________
> Caml-list mailing list. Subscription management:
> http://yquem.inria.fr/cgi-bin/mailman/listinfo/caml-list
> Archives: http://caml.inria.fr
> Beginner's list: http://groups.yahoo.com/group/ocaml_beginners
> Bug reports: http://caml.inria.fr/bin/caml-bugs
>   

Re: [Caml-list] Why is this allowed?

[Goswin von Brederlow]

Jacques Carette <carette@mcmaster.ca> writes:

> Nice trick - but my 'problem' [1] regards arbitrary valid OCaml code
> and how to serialize code values.  The issue seems to be that one
> needs to attach an arbitrarily large typing environment to each part
> of a value, as there is no global typing environment which will do the
> job (because of shadowing).
>
> Jacques

This is no different from

let x = 1
let x = 2

Shadowing is totally allowed in ocaml.

MfG
        Goswin

Re: [Caml-list] Why is this allowed?

[bluestorm]

[-- Attachment #1: Type: text/plain, Size: 237 bytes --]

It is actually not the case that type structure items can be shadowed :

# module A = struct type t = int;; type t = int end;;
Error: Multiple definition of the type name t.
       Names must be unique in a given structure or signature.

[-- Attachment #2: Type: text/html, Size: 312 bytes --]

Re: [Caml-list] Why is this allowed?

[Jacques Carette]

bluestorm wrote:
> It is actually not the case that type structure items can be shadowed :
>
> # module A = struct type t = int;; type t = int end;;
> Error: Multiple definition of the type name t.
>        Names must be unique in a given structure or signature.
>

Exactly.  My desired would be for this non-shadowing restriction to be 
extended to all types, not just type structure items.

Shadowing at the value level is an entirely different issue, and is 
perfectly fine as it is.

Jacques