Functional Programming In Cpp

"C++ = the "C" language, enhanced with ObjectOriented and FunctionalProgramming concepts."

Can anybody give examples how CeePlusPlus has support for FunctionalProgramming? The Y combinator in C++ would be nice... -- StephanHouben

This web site is one of several which describe libraries for function programming in C++: http://www.cc.gatech.edu/~yannis/fc++/. The authors give the library the name Functional Programming in C++, which makes it hard to invent a wiki name for it, as the obvious one is taken. FunctoidsInCpp is going to have to do. There is also the BoostPhoenixLibrary which is one of the BoostLibraries. -- JohnFletcher

November 2010 See CppNext for lively discussion of functional programming in C++.

It has support for currying (CurryingSchonfinkelling) using the binders binder1st and binder2nd. I.e., you can construct a unary_function by taking a binary_function and binding one of its args to a fixed value. More details can follow if you're interested... Is this the kind of thing the questioner meant?

Yes it is: BTW this sounds like partial application, not currying. In SchemeLanguage I can do:

  (define (apply-partial proc . arglist1) 
(lambda arglist2 
(apply proc (append arglist1 arglist2))))

How would one define something similar to apply-partial in C++? -- sh

C++ provides the following FunctionalProgramming tools and techniques:

FunctorObjects can be used to implement HigherOrderFunctions (FunctionPointers can too, but FunctorObjects are much better suited to the task), including lambdas. The StandardTemplateLibrary is the C++ user's first exposure to this, the BoostLibraries take it to the next level.
- However, the lack of GarbageCollection makes lambdas a bit of a pain--FunctorObjects used in this manner generally have usage patterns that make manual memory allocation/deallocation a pain in the butt.
  - Could you elaborate on that? I've used Phoenix and more, and don't have the slightest clue what you're talking about. -- ArneVogel
The template mechanism is a TuringComplete LazyFunctionalLanguage? it its own right (though not a very good one...)

BTW, there is InteLib which provides LispProgramming within C++ (actually, it is a kind of LispProgramming without Lisp). However, being the author of it, I still do not consider that CeePlusPlus is a FunctionalProgramming language. I'd rather consider it an algebraic language which allows almost any ProgrammingParadigm to be imported into. -- AndreyStolyarov

In the 2004 IcfpProgrammingContest (http://www.cis.upenn.edu/proj/plclub/contest//results.php#languages), there were more C++ entries (25) than entries in any other individual programming language. (In other words, C++ had a plurality--not a majority). In addition, Java was 4th (21 entries), Python 6th (16 entries), CeeLanguage 7th (15 entries), and PerlLanguage 9th (11 entries).

Among traditional function languages we find O'Caml 2nd (24 entries), Haskell 5th (20 entries), Lisp 8th (12 entries), and SchemeLanguage 10th (9 entries). The EmFour? preprocessor (which is TuringComplete) had more entries (2) than did ErlangLanguage, SmalltalkLanguage, ToolCommandLanguage, bash, and a bunch of other languages.

What's that all mean? Not much, other than a primitive indicator of geek mindshare. It's interesting how much TemplateMetaprogramming has rekindled interest in C++, especially when you consider what a royal pain C++ templates are to work with sometimes (especially when you are doing anything more complicated than simple abstraction over types).