Warp wrote:

>   So-called functors are a common idiom in C++ template programming.
> For example many STL data containers and algorithms accept functors
> as parameters.
> 
>   A functor is something which acts like a function. In other words, it
> can be called like a function can be called, and it will return a value
> like a function returns a value.
> 
>   Obviously a regular function is a valid functor (and a common way to
> pass functors to the STL).

I see...

>   Another way of creating a functor is, as you point out, to create a
> struct or class which has the operator() overloaded. This way an instance
> of that class can be used exactly like it was a function.
> 
>   The advantage of the latter method is that the struct/class can have
> member variables (and member functions) which the operator() can use.
> This way you can have different values for those member variables for
> different instances of the functor. (If I'm not completely mistaken,
> this starts get closer to so-called closures.)

Yes, indeed. It looks like that means you could do something like write 
a struct that contains a number, and make it into a functor that adds 
that number when called. OTOH, if your function really had to be a 
*function*, you'd have to write a seperate function for every value you 
might want to add... which would be, uh, silly.

In Haskell, you can just say "map (*2)" and it'll work. It looks like 
what you're describing enables you to achieve something very similar in C++.

(Presumably the *other* thing about a C++ functor is that it could 
change some internal state each time you call it - say, to generate 
unique IDs or something...?)

>   In the next C++ standard there will be a third type of functor: Lambda
> functions.

Knowing what I now know about C++ (constructors, copy constructors, 
assignment constructors, pointers, references, etc.), I suspect this is 
going to be... uh... "interesting". :-}

I wonder what the scoping rules would be like?

>> I'm almost tempted to try just to see if I can implement curried 
>> functions with it in C++. You know, just to be totally perverse...
> 
>   C++ might not be the best language for that... ;)

Heh. It could probably work... but no, I'm sure it'll be pretty ugly. 
;-) [Not to mention being abnormal C++ style.]

-- 
http://blog.orphi.me.uk/
http://www.zazzle.com/MathematicalOrchid*

Post a reply to this message