Specifying eponymous template "internal parameter"
monarch_dodra
monarchdodra at gmail.com
Tue Jan 8 02:16:31 PST 2013
On Tuesday, 8 January 2013 at 09:59:26 UTC, Simen Kjaeraas wrote:
> On 2013-55-08 09:01, monarch_dodra <monarchdodra at gmail.com>
> wrote:
>
>> Sometimes (especially in phobos), one defines a parametrized
>> template, that resolves to a templated function.
>>
>> This is a nifty trick, because it allows specifying a vararg
>> before the current type parameter, eg:
>>
>> //----
>> auto r = [1, 2, 3];
>> auto m = map!("++a", "--a")(r);
>> //----
>>
>> As you can see, the template guessed the type of r, even
>> though we used a vararg. This would not have worked with a
>> single template function.
>>
>> My question though: I have a similar use case, but I NEED to
>> be able to explicitly specify the type of r: as such:
>>
>> //----
>> auto m = fun!("++a", "--a", ubyte)(1);
>> auto m = fun!("++a", "--a")!(ubyte)(1);
>> auto m = fun!("++a", "--a").fun!ubyte(1);
>> //----
>> None of them work. In this specific case, I *need* to specify
>> that 1 is of type ubyte, but I really can't do it :/
>>
>> Simplified example: in my use case, it is a
>> "immutable(int[])": Failure to specify the type means the
>> compiler strips tail immutability...
>>
>> The only workaround I can find to make such a thing, is to
>> *not* use eponymous temples, and explicitly call an inner
>> function. This is ugly as sin, and it makes specifying the
>> internal parameter mandatory.
>>
>> Any thoughts?
>
> A non-eponymous template is currently the only way to do this.
> Strangely, this works:
>
> alias fun2 = fun!("++a", "--a");
> auto m = fun2!(ubyte)(1);
Nice!
And now, for the 1M$ question: Can I rely on this behavior, or is
this an accepts invalid...?
More information about the Digitalmars-d-learn
mailing list