Function name as text

[Don Clugston]

Craig Black wrote:
> "Don Clugston" <dac at nospam.com.au> wrote in message 
> news:fj8j1p$2per$1 at digitalmars.com...
>> Bill Baxter wrote:
>>> Don Clugston wrote:
>>>> Craig Black wrote:
>>>>> "Bill Baxter" <dnewsgroup at billbaxter.com> wrote in message 
>>>>> news:fj769h$7ou$1 at digitalmars.com...
>>>>>> Craig Black wrote:
>>>>>>> I have been considering porting some C++ code to D.  One of the 
>>>>>>> classes I would have to port is an event queue class where each event 
>>>>>>> on the queue has a delegate and a text string that points to the 
>>>>>>> function name that the delegate refers to.  The function name is used 
>>>>>>> to visualize the event queue for run-time debugging purposes.  It is 
>>>>>>> important to capture both the class name and the function name as 
>>>>>>> text.
>>>>>>>
>>>>>>> In C++ I had a macro called DISPATCH that used the stringize operator 
>>>>>>> # to capture the name of the function.  The good (and bad) thing 
>>>>>>> about C++ in this case is that when specifying a pointer to a member, 
>>>>>>> you must fully qualify the function name, so you would have something 
>>>>>>> like this.
>>>>>>>
>>>>>>> class Foo {
>>>>>>> public:
>>>>>>>   void bar() {}
>>>>>>> };
>>>>>>>
>>>>>>> Foo *foo = new Foo;
>>>>>>> Event event = DISPATCH(foo, &Foo::bar);
>>>>>>>
>>>>>>> Using the stringize operator, the DISPATCH macro could capture the 
>>>>>>> text string "Foo::bar" as well as the member function pointer.  Here 
>>>>>>> is the equivalent code in D..
>>>>>>>
>>>>>>> Foo foo = new Foo;
>>>>>>> Event event = dispatch(&foo.bar);
>>>>>>>
>>>>>>> Which is much more elegant, except that I can't figure out a way to 
>>>>>>> capture the name of the function and it's class.  I tried fiddling 
>>>>>>> with the stringof operator but that doesn't seem to work.
>>>>>>>
>>>>>>> Any ideas?
>>>>>> There probably isn't a way to do it right now without using a string 
>>>>>> mixin, which uglies things up on the calling side:
>>>>>>
>>>>>>      Event event = mixin(dispatch("&foo.bar"));
>>>>>>
>>>>>> Macros are supposed to give us a way to clean that up.  But for now 
>>>>>> you're probably better off just passing the name separately like 
>>>>>> dispatch(&foo.bar, "foo");
>>>>>>
>>>>>> --bb
>>>>> I guess that's not so bad.  It would be.
>>>>>
>>>>> Foo *foo = new Foo;
>>>>> Event event = dispatch(&foo.bar, "Foo.bar");
>>>>>
>>>>> That is probably easier on the eyes than the mixin syntax.  It stilll 
>>>>> would be cool if the compiler could somehow build the name 
>>>>> automatically.
>>>> Or you could use an alias template parameter, to give the syntax:
>>>>
>>>> Event event = dispatch!(foo.bar);
>>>>
>>> Really?  Can you take the stringof an alias parameter and get back 
>>> "foo.bar" ?  If so then nifty!
>>>
>>> --bb
>> Unfortunately it's not quite so simply, since there are so many bugs in 
>> .stringof. I'm unable to get the 'foo'.
>> ----
>> class Foo
>> {
>>     void bar(int d) {}
>> }
>>
>> void GetName(alias F)()
>> {
>>     pragma(msg, (&F).stringof); // prints "& bar"
>> }
>>
>> void main()
>> {
>>     GetName!(Foo.bar);
>> }
>> ----
>> I have however done this previously using .mangleof to retrieve the fully 
>> qualified name. The behaviour with alias template parameters was one of 
>> the parts of my NameOf module which didn't get into .stringof. Sure would 
>> be nice if
>> F.stringof worked.
> 
> I ran into this same thing with stringof.  Are you sure that this is a bug? 
It's hard to say. stringof was originally a 'hidden feature'/easter egg that 
didn't get mentioned in the changelog. It's in the docs, but it behaves quite 
differently from documented, except in simple cases.

> Has it been officially reported? 
Not that I know of. Even most of the examples in the docs don't work as 
advertised, so there hasn't been much motivation to report these secondary cases.