Since my phobos PR for better static assertions was clearly never getting merged (https://github.com/dlang/phobos/pull/3677), I moved the code to dub instead:

http://code.dlang.org/packages/concepts

Basically, as long as you pair up your template constraints (e.g. isFoo) with a suitably named function that should compile (e.g. checkFoo, not a lambda), then you can write this:

@models!(Foo, isFoo)
struct Foo {
    //...
}

And in the case that isFoo is false, you'll get to know _why_ it didn't compile instead of an AssertError that leaves you confused.

At least it used to and the unit tests pass so I'm assuming it still does :P

Atila

On Friday, 16 December 2016 at 20:51:19 UTC, Atila Neves wrote:
> Atila

Surprisingly little code. Thx.

On Friday, 16 December 2016 at 20:51:19 UTC, Atila Neves wrote:
> @models!(Foo, isFoo)


What happened to the __UDA_ATTACHMENT__ proposal?

On Friday, 16 December 2016 at 23:49:18 UTC, Nicholas Wilson wrote:
> On Friday, 16 December 2016 at 20:51:19 UTC, Atila Neves wrote:
>> @models!(Foo, isFoo)
>
>
> What happened to the __UDA_ATTACHMENT__ proposal?

Also are UDAs legal after the symbol? i.e. with __UDA_ATTACHMENT__ could you do struct Foo @models!isFoo
{
    void foo();
}

 ?

On 2016-12-16 21:51, Atila Neves wrote:
> Since my phobos PR for better static assertions was clearly never
> getting merged (https://github.com/dlang/phobos/pull/3677), I moved the
> code to dub instead:
>
> http://code.dlang.org/packages/concepts
>
> Basically, as long as you pair up your template constraints (e.g. isFoo)
> with a suitably named function that should compile (e.g. checkFoo, not a
> lambda), then you can write this:
>
> @models!(Foo, isFoo)
> struct Foo {
>     //...
> }
>
> And in the case that isFoo is false, you'll get to know _why_ it didn't
> compile instead of an AssertError that leaves you confused.
>
> At least it used to and the unit tests pass so I'm assuming it still
> does :P

It's not very well tested ;). When a check fails I get this error:

Error: can only import from a module, not from package std.algorithm.searching.countUntil

After fixing that to a proper selective import I get:

Error: template instance moduleName!P template 'moduleName' is not defined

Adding an import for std.traits will fix that. Then after fixing all that it seems to work properly :)

Asserting that something does not compile is very risky because it does not tell you why it didn't compile. One should always first compile without __traits(compiles) to verify that the correct error is reported.

-- 
/Jacob Carlborg

On Saturday, 17 December 2016 at 14:58:24 UTC, Jacob Carlborg wrote:
> On 2016-12-16 21:51, Atila Neves wrote:
>> [...]
>
> It's not very well tested ;). When a check fails I get this error:

Oops!

Fixed it now. Look, now it's version 0.0.2! :P

Atila

On Saturday, 17 December 2016 at 04:00:32 UTC, Nicholas Wilson wrote:
> On Friday, 16 December 2016 at 23:49:18 UTC, Nicholas Wilson wrote:
>> On Friday, 16 December 2016 at 20:51:19 UTC, Atila Neves wrote:
>>> @models!(Foo, isFoo)
>>
>>
>> What happened to the __UDA_ATTACHMENT__ proposal?

Nothing AFAIK.

>
> Also are UDAs legal after the symbol? i.e. with __UDA_ATTACHMENT__ could you do struct Foo @models!isFoo
> {
>     void foo();
> }
>
>  ?

Nope.