Wow ... I just saw type-safe macros / quasi-quoting.

It seems that Haskell is the language I've been searching for all my life :-) ... unfortunately code written in it looks really scary.

I mean ... when I look at good Python / Ruby code, I just get what's going on ... that's partly because I know these languages, but on the other hand they encourage programmers to do literary story-telling.

Maybe someday Haskell will get reinvented into some form meant for mere mortals :)

I really like the syntax of Haskell. You can get used to it fairly quickly.

> [...] but on the other hand they encourage programmers to do literary story-telling.

Try literate Haskell. (I say this only partially tongue-in-cheek.) Really, I think that Haskell is very geared towards being read by humans. This is mostly because it's so good at gluing, that you can break your problem apart in smaller pieces in the first place.

E.g. let's look at a program that computes sqrt(2). With Haskell you can talk first about how to produce infinite list of ever better approximations:

  next x = (2 / x + x) / 2
  approximations = iterate next 1

and then talk about when to stop and finally give an answer later on.

  goodEnough x = abs (2-x*x) <= 0.0001
  sqrtTwo = find goodEnough approximations

In a strict language, you usually intermingle producing new values with watching out for the ending condition:

  guess = 1
  while abs(2 - guess*guess) >= 0.0001:
    guess = (2/guess + guess) / 2

Python's generators help, but they are more restricted. (Which is intended, Python wants to stay simple.)

Just out of curiosity, what's the advantage of `(2/x + x)/2` over `x/2 + 1/x` (which, to me, is more readable)?

It is more readable. But I wanted to stick close to the general formula [1], so that you see what's going on. It would probably have been better to write it as:

  y = 2
  average x1 x2 = (x1 + x2) / 2
  next x = average (y/x + x)

It's just a co-incidence that I chose y as 2.

[1] Heron's Method (http://en.wikipedia.org/wiki/Herons_method#Babylonian_method)

My background is almost all C, C++, and some python. I have been fooling around with Haskell for a few months now. I can honestly say I haven't used anything on the list beyond the first two. Haskell falls very close to python in readability if you want it to. You can "go deep" and produce stuff that only an experienced, hard core functional programmer can understand as this guy does, but I tend to use it as a statically typed, fast python.

Flip through the first few chapters of Real World Haskell online and see that it is fairly readable.

> on the other hand they encourage programmers to do literary story-telling.

That is the part you must avoid. In Haskell, you don't tell a story. You depict a landscape. (There is a story, in the IO monad, but it is minimized and segregated.)

I'd say: just try it. After reading haskell-cafe for a day, I felt utterly incapable of ever learning Haskell. I started picking up the language using Real World Haskell (quite an acceptable book). I can read most programs now and crafted some short programs. I still find predicting memory/time use under a lazy-evaluation regime difficult. But it is certainly a nice experience ;)

The guys on Haskell Cafe delight in complex problems (and solutions). There's a list for beginners, too.

You will enjoy the Haskell Cafe later.

By the way, one small exercise (from a paper by Chris Okasaki): Write a function with the type

  Tree a -> Tree Int

where

  data Tree a = Node a (Tree a) (Tree a) | Empty

your function should all nodes in a tree with their position in level order traversal (also called breadth first search). E.g. (lot == lever order traversal)

  lot Empty = Empty
  lot (Node 'a' (Node 'b' Empty (Node 'c' Empty Empty))
                (Node 'd' Empty Empty)) =
      (Node '0' (Node '1' Empty (Node '3' Empty Empty))
                (Node '2' Empty Empty))

The thing with RWH I found was, chapters 1-4 great, a nice introduction to the language, and I did all the exercises without too much head-scratching. Then chapter 5 is utterly incomprehensible. I didn't get any further.

OCaml, that's a nice language, with much better introductary materials.

Dont' give up:

http://www.amazon.com/Programming-Haskell-Graham-Hutton/dp/0...

http://stackoverflow.com/questions/1012573/how-to-learn-hask...

http://learnyouahaskell.com/

What did you find so hard about chapter 5 ("Writing a library: working with JSON data")? Doesn't seem too be such a big step up to me in comparison to other chapters in the book?

I can actually read some of the syntax. I guess that a mathematical background helps when you're accustomed to seeing function definitions that map one real to another real or some of the other constructs. And I certainly recognize interfaces from Java (and why they can be a lot nicer than C++ style inheritance).

What still gets me are some of the basic functional programming concepts like monads and the y combinator, which obviously aren't unique to Haskell. Yeah, I've read Wikipedia on both of those (more than once). And several other things. I've even written some Lisp for my employer. But some of the concepts have yet to "click." I know monads encapsulate side effects or something (and the y combinator finds the fixed points of a function), but I feel like I just don't get the details of how it works.

After all, you don't need to know anything about monads to call the print function.

Monads are easy to understand: A monad is exactly anything that satisfies the three monad laws.

If you have worked with mathematical groups and have seen how mathematicians can look at very disparate things with their group-goggles and find similarities, you will appreciate monads.

The y combinator is a nice parlour trick (if you define it in lambda calculus without using any other recursion primitives), and research vehicle. But I did not apply it once, yet. And I've tried.

If you want to get combinatorial logic (including the y combinator), read "To Mock a Mockingbird". And do all the puzzles, of course. Actually you should read it, whether you want to learn anything about combinatorial logic or not. It is such a good book.

Thanks, I'll read that.

> I still find predicting memory/time use under a lazy-evaluation regime difficult.

It's not just you. Lazy evaluation makes predicting and understanding resource usage difficult.

Indeed. And that's the major shortcoming I see that could prevent me from using Haskell in a production environment.

The guys who wrote XMonad got around it, so it is possible. You just need to be careful, and use enough strictness annotations.

Laziness is one of those features I can never make my mind up about. It is sometimes very convenient, but most of the time I never think about the fact that the semantics mean evaluation is lazy. Sometimes this negligence has bad consequences, and then one must invoke a carefully honed skill to correct the issue. Definitely doable, but it certainly feels like fighting the language (to me, anyway).

One of the craziest things is that Haskell has decided that type families should also be lazily evaluated. So you can hop into the interactive interpreter to inspect some types and get not-so-useful information, like

    somevar :: SomeTypeFunction (SomeTypeFunction2 Int) (SomeTypeFunction3 (SomeTypeFunction 3 (SomeTypeFunction3 String)))

which might actually evaluate to something simple like (String, Int), though you wouldn't know it without either tracing through the functions (yuck!) or writing some silly helper functions whose sole job is to force evaluation of the type functions (yuck!). I've never personally been thankful for this behavior.

Interesting. I'll try to reproduce your example. Could you make something up that I can just copy-and-paste into ghci? Thanks!

Chris Okasaki argues in "Purely Functional Data Structures" for making lazy evaluation optional, but easy to use. Some data structures have optimal asymptotics only when evaluated strictly and some only when evalutated lazily.

With better strictness analysers most of the problems lazyness is causing at the moment, may go away. A Haskell compiler is free to make your program (or parts of it) strict, if it can prove that this preserves semantics.

Here's some code:

  {-# LANGUAGE
          TypeFamilies,
          MultiParamTypeClasses,
          FlexibleInstances #-}

  data True = T
  data False = F

  class LogicOr a b where
    type Or a b
    typeOr :: a -> b -> Or a b
  instance LogicOr True b where
    type Or True b = True
    typeOr T b = T
  instance LogicOr a True where
    type Or a True = True
    typeOr b T = T
  instance LogicOr False False where
    type Or False False = False
    typeOr F F = F

  class Decide tf a b where
    type If tf a b
    nonFunctionalIf :: tf -> a -> b -> If tf a b
  instance Decide True a b where
    type If True a b = a
    nonFunctionalIf T a b = a
  instance Decide False a b where
    type If False a b = b
    nonFunctionalIf F a b = b

  whatIsMyType tf1 tf2 a b = nonFunctionalIf (typeOr tf1 tf2) a b

If we load this into ghci, we have the following example:

  *Main> :t whatIsMyType F F 2 "foo"
  whatIsMyType F F 2 "foo" :: (Num t) => If (Or False False) t [Char]
  *Main> :t whatIsMyType F F 2 "foo" :: String
  whatIsMyType F F 2 "foo" :: String :: String

In the first :t, ghci lazily (though correctly) gives us the type of whatIsMyType. In the second example, where we provide an explicit type annotation, ghci is able to (correctly) conclude that our annotation was correct by evaluating the various type functions it had given us before.

Thanks. I can reproduce your example.

I am still learning Haskell but I thought things like foldl' and Iteratees allow you to use strict evaluation when it is appropriate.

For List-like structures getting stuff strict enough isn't too much of a problem. But not everything is a simple fold.

As a non-Haskell programmer, I didn't understand almost any of the examples. Some of them I understood, but didn't get what the feature is good for. Some of them I just plain didn't get (Haskell's syntax is to different from what I'm used to to easily understand what's going on).

I love the idea of this article, talking about specific features in your language you'd like to see gain wider adoption. It's too bad I had trouble understand what was going on.

(author here). This is good feedback. I definitely struggled with finding a good way to describe what each of the features does. I'm considering writing a sequence of followups, each focusing on a particular feature, trying to describe how (say) Java code could be refactored to take advantage of the feature.

I'd very much appreciate suggestions for things to address in such a series! Any time I write something like this, I always learn a lot in the ensuing discussions. Mostly I learn technical things, but it's good to learn more about the writing process as well.

Here's where I think the article goes wrong.

Firstly, I'm not sure what audience you had in mind when you wrote this. If it's Haskell programmer, then I think it's too detailed. If not, I think it's not detailed enough (but it does include all sorts of words/concepts a non-Haskell programmer won't understand).

I think a shorter article, focused on less features would have been better (or a series of articles). It feels to me like you could talk in length about each of these features, and the article is both too long because of so many features, but too short in each specific feature.

Finally and most importantly, what you're really missing is code examples in other languages. You mentioned doing this, and I agree 100%. That would make your article much easier for someone like me. Although I vote the examples be in Python :)

I'm looking forward to your next post!

Or make up some pseudo-code-y language. I.e. this is how it could look like in Java, if it included this awesome feature with some syntax I just made up for it.

(I like Python more in general, but I am not sure, if you can demonstrate some of the static goodness of Haskell-analogies in it. Polymorphism on return types is awesome, but not real imaginable in Python, but barely imaginable in Java.)

That's a shame. And I agree. I guess the author should provide translations in C or Java or whatever along with his examples as far as possible. I understand his Haskell examples, but that's preaching to the choir.

I read a blog post were somebody re-implemented some examples of Haskell concepts in C. I can't remember the URL, but searching for "Monads in C" or "Monads in Java" etc brings up some interesting posts. E.g. http://blog.tmorris.net/maybe-monad-in-java/

I've had a casual relationship with functional languages for a while now. I've decided to really get into one and chose F# as there is a possibility of using it on the job. For the functional gurus, should I ignore that benefit and go to what seems to be the holy grail of functional languages, Haskell?

If you work in a .Net environment and want to explore the possibility of using functional programming on the job, F# is definitely a natural way to go.

Depending on how much autonomy you have at work, Haskell is a reasonable language for getting "real work" done. The technology is ready to do the job: as with any language selection discussion, the issue is more about workplace culture (maintenance plans, hiring, etc).

In terms of advanced types, my understanding is that F# is playing catch-up to Haskell, but that's not really a slight against F#. I happen to prefer Haskell, but it's not obvious to me that a person who knows both languages would agree with me.

So the answer your question, or more precisely to not answer it, I think you'll find either language rewarding. If you think F# might be more useful at work, that's definitely worth considering. If you want to see the state of the art in (well supported) statically typed functional programming, Haskell is a strong contender. Regardless of which you decide to look at, the skills will translate fairly well if you later go to look at the other.

You work in a .NET environment, or have Windows lock-in? Otherwise, why not use Haskell on the job, like others do: http://haskell.org/haskellwiki/Haskell_in_industry

I work in a .NET environment. I am learning F# in my spare time with hopes of being able to write part of the app in F#. It may not be an entirely realistic hope, but at least that hope is there with F#.