Alias templates and anonymous delegates in pipe/map

[Adrien Chauve]

Thanks! Indeed it would make sense but I really agree it seems unintuitive.

Adrien

On Wed, May 18, 2011 at 21:44, Steven Schveighoffer <schveiguy at yahoo.com>wrote:

> On Wed, 18 May 2011 15:11:56 -0400, Adrien Chauve <adrien.chauve at gmail.com>
> wrote:
>
>  Hi,
>>
>> anyone on the alias difference?
>>
>
> I have a feeling that alias needs to be a symbol, and int is a keyword.
>
> a template alias can be any symbol, including a variable, function,
> delegate, etc.
>
> Though it does seem unintuitive that a template alias parameter can be a
> custom type, but not a keyword type.  I would say that's a bug, but I'm not
> sure.
>
>  On Sat, May 14, 2011 at 20:12, Adrien Chauve <adrien.chauve at gmail.com
>> >wrote:
>>
>>  Hi,
>>>
>>> I've been playing with templates and I get some (really dumb) questions.
>>> By the way, I'm using dmd 2.053 64 bits on linux 64 bits.
>>>
>>>
>>> First, I didn't really get the difference between an alias template and a
>>> type template when they are both instanciated with a type.
>>> For example:
>>>
>>> //////// File test1.d
>>> // alias
>>> template
>>> template Foo(alias
>>> X)
>>> {
>>>
>>>    alias X
>>> Y;
>>> }
>>>
>>>
>>>
>>> // type
>>> template
>>>
>>> template
>>> Foo2(X)
>>> {
>>>
>>>    alias X
>>> Y;
>>> }
>>>
>>>
>>>
>>> struct Bar
>>> {}
>>>
>>>
>>> void
>>> main()
>>>
>>> {
>>>
>>>    alias Foo!(Bar).Y BarAlias; //
>>> ok
>>>    alias Foo2!(int).Y IntAlias; //
>>> ok
>>>    alias Foo!(int).Y OtherIntAlias; //
>>> error
>>> }
>>> ////// end of file test1.d
>>>
>>> I get the following error from dmd:
>>> test1.d(19): Error: template instance Foo!(int) does not match template
>>> declaration Foo(alias X)
>>> test1.d(19): Error: identifier 'Y' of 'Foo!(int).Y' is not defined
>>>
>>> Why do the two templates behave differently? Why does the alias template
>>> yield an error? What did I miss about alias templates?
>>>
>>>
>>>
>>>
>>> My second question is about using anonymous delegates in map and pipe. I
>>> tried the following code:
>>>
>>> //////// File test2.d
>>> import
>>> std.functional;
>>>
>>> import
>>> std.algorithm;
>>>
>>>
>>>
>>> unittest
>>>
>>> {
>>>
>>>    immutable auto x = [1, 2, 3,
>>> 4];
>>>
>>>
>>>    alias map!((i){return 2*i;})
>>> mult2;
>>>    assert(equal(mult2(x),
>>> [2,4,6,8]));
>>>
>>>
>>>    alias pipe!(mult2)
>>> pipe_mult2;
>>>    assert(equal(pipe_mult2(x), [2,4,6,8])); // ok
>>>
>>>
>>>
>>>    alias pipe!(mult2, mult2) pipe_mult4;      // error
>>>
>>>    assert(equal(pipe_mult4(x),
>>> [4,8,12,16]));
>>> }
>>>
>>>
>>>
>>> void main(){}
>>> /////////// end of file test2.d
>>>
>>> and when compiling it with dmd -unittest test2.d I get:
>>> test2.d(368): Error: function
>>> test2.__unittest3.map!(__dgliteral1).map!(immutable(int[])).map is a
>>> nested
>>> function and cannot be accessed from doIt
>>> test2.d(368): Error: function
>>> test2.__unittest3.map!(__dgliteral1).map!(Result).map is a nested
>>> function
>>> and cannot be accessed from doIt
>>>
>>> So a single pipe!mult2 is ok but when a I put a second there's an error.
>>> What's also interesting is that all cases work fine when I use a named
>>> function instead of an anonymous delegate.
>>>
>>> I'm not sure to understand what's the error in compose (pipe uses
>>> compose).
>>> And the error returned by dmd is a bit strange because in no way test2.d
>>> has more than 20 lines (in fact line 368 is the first line of map in
>>> std/algorithm.d). Besides, without a template instanciation stack trace,
>>> I
>>> think it's a bit difficult to dive into template errors.
>>>
>>>
>>> Cheers,
>>> Adrien
>>>
>>>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.puremagic.com/pipermail/digitalmars-d/attachments/20110519/d9c59439/attachment-0001.html>