BCS wrote:
> Regan Heath wrote:
>> On 2nd thought I think this only works by some sheer coincidence and fluke!
> 
> The whole idea is a hack. Hacks are only /suposed/ to work when the user does things exactly right

True.  There are many combinations which will break it, i.e.

const(char)
baz()

will get a function called "const", an extra check of p[start..end] using isKeyword would fix a few of these sorts of problems.

>> A better one could definately be written with the DMDFE.
> 
> yes but do you want to port DMDFE to run using CTFE?

Actually I think I have some code to lex C/C++ here somewhere and with very few changes it would lex D just as well.  That would be easier that using a full blown DMDFE port.  I might see what I can whip up tomorrow (it's 11pm here).

Regan

Amazing! I greatly appreciate your help!

> No.

The question is "Why not?". For a language that is supposed to have good compile time relfection, and wants to have good runtime relection, this basic inability illustrates an area in dire need of attention.

I don't think a solution is to introduce __FUNC__, that will only add on to the already big clutter pile. A more generalized solution is to introduce self-referencing (and then get rid of some of the clutter).

Eg

void function x()
{

   writeln( typeof(self).stringof );

}

Even better, using the uniform calling syntax convention

void function x()
{

   writeln( self.typeof.stringof );

}



--rt