Hi!

I need to iterate module members and find specific classes (and make tuple).

    class foo{};
    pragma (msg, __traits(allMembers,mixin(__MODULE__)));

gives me empty tuple. I found also this thread from 2011: http://forum.dlang.org/post/mailman.325.1293887146.4748.digitalmars-d-learn@puremagic.com
This is exactly what I need but this code also gives empty tuple.
Is this possible for now to iterate classes from module? Is this possible to iterate by classes, public imported to module? (in fact, I need to iterate classes imported using package.d file).

On Saturday, 7 January 2017 at 09:46:54 UTC, crimaniak wrote:
> Hi!
>
> I need to iterate module members and find specific classes (and make tuple).
>
>     class foo{};
>     pragma (msg, __traits(allMembers,mixin(__MODULE__)));
>
> gives me empty tuple.

works like a charm both in 2.071 and in git HEAD.

On Saturday, 7 January 2017 at 09:46:54 UTC, crimaniak wrote:
> Hi!
>
> I need to iterate module members and find specific classes (and make tuple).
>
>     class foo{};
>     pragma (msg, __traits(allMembers,mixin(__MODULE__)));
>
> gives me empty tuple. I found also this thread from 2011: http://forum.dlang.org/post/mailman.325.1293887146.4748.digitalmars-d-learn@puremagic.com
> This is exactly what I need but this code also gives empty tuple.
> Is this possible for now to iterate classes from module? Is this possible to iterate by classes, public imported to module? (in fact, I need to iterate classes imported using package.d file).

Do you require a module statement per chance?

On Saturday, 7 January 2017 at 10:27:51 UTC, Nicholas Wilson wrote:
> Do you require a module statement per chance?

it doesn't matter. if there is no explicit module declaration, compiler will insert implicit one. from the code PoV, there is no difference at all.

On Saturday, 7 January 2017 at 10:38:29 UTC, ketmar wrote:
> On Saturday, 7 January 2017 at 10:27:51 UTC, Nicholas Wilson wrote:
>> Do you require a module statement per chance?
>
> it doesn't matter. if there is no explicit module declaration, compiler will insert implicit one. from the code PoV, there is no difference at all.
 Yes, module statement is no reason. package.d is reason. The same code in two files:

class foo{};
struct bar{};
pragma (msg, "allMembers of "~__MODULE__~" in "~__FILE__);
pragma(msg, __traits(allMembers, mixin(__MODULE__)));

Result:

allMembers of vcm.bll.event in source/vcm/bll/event/package.d
tuple()
allMembers of vcm.bll.event.test in source/vcm/bll/event/test.d
tuple("object", "vcm", "foo", "bar")

So allMembers gives empty tuple for module in package.d file. Please check it. As I understand documentation, package.d is ordinary module and special here is only file name used when module name is directory.
DMD 2.071.1. Is this a bug or documentation problem?

On Saturday, 7 January 2017 at 12:11:20 UTC, crimaniak wrote:
> Is this a bug or documentation problem?

well, it's hard to say. package.d is one of the cases where you *have* to provide explicit module definition. ;-) 'cause implicit one creates module with name `package`, which is teh keyword. and then all the hell broke loose. this actually may be both bug and documentation problem.

p.s.: otherwise, it *should* work with explicit module declaration. but then, we have some well-known bugs with cross-module introspection, and some heisenbugs with package introspection.

On Saturday, 7 January 2017 at 12:29:34 UTC, ketmar wrote:
> On Saturday, 7 January 2017 at 12:11:20 UTC, crimaniak wrote:
>> Is this a bug or documentation problem?
>
> well, it's hard to say. package.d is one of the cases where you *have* to provide explicit module definition. ;-) 'cause implicit one creates module with name `package`, which is teh keyword. and then all the hell broke loose. this actually may be both bug and documentation problem.
  Yes, removing 'module' statement leads to error on __traits(allMembers, mixin(__MODULE__)) expression: Error: expression expected, not 'package'.
But it's not related, I provide module statement in package.d.

....
Ok, I found it: https://issues.dlang.org/show_bug.cgi?id=16044

On Saturday, 7 January 2017 at 13:45:53 UTC, crimaniak wrote:
> ....
> Ok, I found it: https://issues.dlang.org/show_bug.cgi?id=16044

yep. sorry, my search-fu is completely broken, so i didn't provide you with the number.