On 2014-08-29 16:48:51 +0000, Andrei Alexandrescu said:
Worth a look:
http://stackoverflow.com/questions/25555329/d-finding-all-functions-with-certain-attribute
Andrei
That's some pretty neat code, I did something similar awhile ago for
deserializing classes. Here's a reduced case of it:
http://dpaste.dzfl.pl/4fffa339368f
The facility to be able to create such a factory through compile time
reflection is absolutely awesome. However, it "feels" as though the
ability to do this is making use of undefined behavior rather than
actual language features.
First:
I consider it sort of a language wart that alias hack(alias T) = T; is
necessary for aliasing to the particular symbol, and that the
__traits(getMember) expression can't be substituted in to the same
places as the alias to it without compiler errors. It's very
non-intuitive. I would have never figured that out on my own with out
some other D wizards helping me about a year ago. I still don't
understand why I can't simply do :
alias member = __traits(getMember, ...);
Also:
The way that the foreach works in the pasted example is very
surprising. I would be nice if this was a "static foreach" instead of
just a foreach which happens to run at compile time over a TypeTuple.
Normally doing a switch inside a foreach like that would not work at
all, but near as I can tell the compiler unrolls the whole thing since
it's inputs are known at compile time.
Finally:
If you actually want to have something that returns a TypeTuple!() of
only subclasses which you can foreach over in other places -- you have
to make a *very* ugly recursive template:
(Line 208 thru Line 235 returns a TypeTuple of all subclasses of the
Message class)
https://github.com/schancel/gameserver/blob/master/source/messages/core.d#L208
Making these awesome feature more intuitive and easy to use would go a
long way. I feel like this should be on your short list along with the
GC and C++ compatibility.
-S.
