www.digitalmars.com         C & C++   DMDScript  

digitalmars.D - Required Reading: "Functional Programming in C++"

reply Walter Bright <newshound2 digitalmars.com> writes:
for core D devs. Of course, this is much easier in D than in C++ because of D's 
const and pure attributes.

"Functional Programming in C++" by John Carmack

https://www.gamasutra.com/view/news/169296/Indepth_Functional_programming_in_C.php

I'm as guilty as anyone for not understanding or following these guidelines. I 
expect we can do much better.
Oct 25 2017
next sibling parent codephantom <me noyb.com> writes:
On Wednesday, 25 October 2017 at 22:19:21 UTC, Walter Bright 
wrote:

 for core D devs. Of course, this is much easier in D than in 
 C++ because of D's const and pure attributes.


"But master," the apprentice started.. but the master 
interrupted, asking:

"Is it not true that the word for 'not functional' is 
'dysfunctional'?"

And then the apprentice understood.

https://medium.com/javascript-scene/the-dao-of-immutability-9f91a70c88cd
Oct 25 2017
prev sibling parent reply Ola Fosheim =?UTF-8?B?R3LDuHN0YWQ=?= writes:
On Wednesday, 25 October 2017 at 22:19:21 UTC, Walter Bright 
wrote:

 for core D devs. Of course, this is much easier in D than in 
 C++ because of D's const and pure attributes.


Nah, the type system isn't critical for doing functional style 
programming. People who want to do functional programming in C++ 
has plenty of opportunities, and libraries to support it…


 "Functional Programming in C++" by John Carmack


2012, and it doesn't really say anything, does it?
Oct 28 2017
parent reply Jonathan M Davis <newsgroup.d jmdavisprog.com> writes:
On Saturday, October 28, 2017 21:51:50 Ola Fosheim Grøstad via Digitalmars-d 
wrote:

 On Wednesday, 25 October 2017 at 22:19:21 UTC, Walter Bright

 wrote:

 for core D devs. Of course, this is much easier in D than in
 C++ because of D's const and pure attributes.


 Nah, the type system isn't critical for doing functional style
 programming. People who want to do functional programming in C++
 has plenty of opportunities, and libraries to support it…


The type system can help ensure the correctness of functional programming
(e.g. pure helps ensure that you didn't accidentally do something involving
global variables), but it's certainly not required to write in that style.

And actually, most D code that's functional doesn't do much with either
const or pure. Most of the functional code is range-based, which
fundamentally falls apart with const - at most, const or immutable is used
for the elements. And while pure is often involved, it's typically only by
inference. So, in those cases, the only places that pure actually gets
checked is in unittest blocks marked with pure or when you try to use a set
of range-based operations inside of a function that you've marked as pure.
And range-based operations that definitely aren't pure are done on a regular
basis (e.g. any lazy ranges grabbing their elements from I/O).

So, arguably, D is already showing how unnecessary const and pure are for
functional programming. They can help, to be sure, but they aren't
necessary.

- Jonathan M Davis
Oct 28 2017
parent Ola Fosheim =?UTF-8?B?R3LDuHN0YWQ=?= writes:
On Saturday, 28 October 2017 at 23:26:11 UTC, Jonathan M Davis 
wrote:

 The type system can help ensure the correctness of functional 
 programming (e.g. pure helps ensure that you didn't 
 accidentally do something involving global variables), but it's 
 certainly not required to write in that style.


Yes, the type-system would prevent some aspects of non-functional 
programming, but it doesn't enable functional programming…

Anyway, both C++ and D have abstraction mechanisms for building 
the tools you need to get functional-style programming. And I 
assume production backends also do tail-call optimization in D as 
well as C++.

What you don't get in either is a convenient applicative pattern 
matching syntax, and functional programming in the technical 
recursive sense is mostly not a very good idea (even with tail 
call optimization since you cannot enforce it, I think?).

Anyway, with the improvements to C++ lambdas it has become more 
attractive to do small-scale functional style programming in C++. 
So I think D and C++ are pretty much the same in that respect. (D 
has pure, but C++ has explicit capturing of references)


 And actually, most D code that's functional doesn't do much 
 with either const or pure. Most of the functional code is 
 range-based,


I guess it makes sense to call range-based programming functional 
in spirit (although not so much on a technical level, e.g. 
recursiveness). And you can do similar things with iterators, 
although the C++ stdlib does not provide much out-of-the-box.

In the context of the linked article I think maybe he is assuming 
that you should have SIMD access to the values, so maybe there is 
an underlying assumption there that libraries for 
ranges/iterators don't support SIMD well in their current 
incarnations.

Anyway, I think you can rely on RVO in  C++ in 2017, maybe not in 
2012. But the addition of constexpr to C++ has made some 
improvements to the C++ ecosystem where D used to be a little bit 
ahead.

The improvements to C++ in C++14 and C++17 are relative small, 
but as far as I am concerned, those improvements has enabled a 
more functional style of programming for initialization (small 
scale functional programming).

Ola.
Oct 29 2017