Does functional programming mandate new naming conventions? - haskell

I recently started studying functional programming using Haskell and came upon this article on the official Haskell wiki: How to read Haskell.
The article claims that short variable names such as x, xs, and f are fitting for Haskell code, because of conciseness and abstraction. In essence, it claims that functional programming is such a distinct paradigm that the naming conventions from other paradigms don't apply.
What are your thoughts on this?

In a functional programming paradigm, people usually construct abstractions not only top-down, but also bottom-up. That means you basically enhance the host language. In this kind of situations I see terse naming as appropriate. The Haskell language is already terse and expressive, so you should be kind of used to it.
However, when trying to model a certain domain, I don't believe succinct names are good, even when the function bodies are small. Domain knowledge should reflect in naming.
Just my opinion.
In response to your comment
I'll take two code snippets from Real World Haskell, both from chapter 3.
In the section named "A more controlled approach", the authors present a function that returns the second element of a list. Their final version is this:
tidySecond :: [a] -> Maybe a
tidySecond (_:x:_) = Just x
tidySecond _ = Nothing
The function is generic enough, due to the type parameter a and the fact we're acting on a built in type, so that we don't really care what the second element actually is. I believe x is enough in this case. Just like in a little mathematical equation.
On the other hand, in the section named "Introducing local variables", they're writing an example function that tries to model a small piece of the banking domain:
lend amount balance = let reserve = 100
newBalance = balance - amount
in if balance < reserve
then Nothing
else Just newBalance
Using short variable name here is certainly not recommended. We actually do care what those amounts represent.

I think if the semantics of the arguments are clear within the context of the code then you can get away with short variable names. I often use these in C# lambdas for the same reason. However if it is ambiguous, you should be more explicit with naming.
map :: (a->b) -> [a] -> [b]
map f [] = []
map f (x:xs) = f x : map f xs
To someone who hasn't had any exposure to Haskell, that might seem like ugly, unmaintainable code. But most Haskell programmers will understand this right away. So it gets the job done.
var list = new int[] { 1, 2, 3, 4, 5 };
int countEven = list.Count(n => n % 2 == 0)
In that case, short variable name seems appropriate.
list.Aggregate(0, (total, value) => total += value);
But in this case it seems more appropriate to name the variables, because it isn't immediately apparent what the Aggregate is doing.
Basically, I believe not to worry too much about convention unless it's absolutely necessary to keep people from screwing up. If you have any choice in the matter, use what makes sense in the context (language, team, block of code) you are working, and will be understandable by someone else reading it hours, weeks or years later. Anything else is just time-wasting OCD.

I think scoping is the #1 reason for this. In imperative languages, dynamic variables, especially global ones need to be named properly, as they're used in several functions. With lexical scoping, it's clear what the symbol is bound to at compile time.
Immutability also contributes to this to some extent- in traditional languages like C/ C++/ Java, a variable can represent different data at different points in time. Therefore, it needs to be given a name to give the programmer an idea of its functionality.
Personally, I feel that features features like first-class functions make symbol names pretty redundant. In traditional languages, it's easier to relate to a symbol; based on its usage, we can tell if it's data or a function.

I'm studying Haskell now, but I don't feel that its naming conventions is so very different. Of course, in Java you're hardly to find a names like xs. But it is easy to find names like x in some mathematical functions, i, j for counters etc. I consider such names to be perfectly appropriate in right context. xs in Haskell is appropriate only generic functions over lists. There's a lot of them in Haskell, so this name is wide-spread. Java doesn't provide easy way to handle such a generic abstractions, that's why names for lists (and lists themselves) are usually much more specific, e.g. lists or users.

I just attended a number of talks on Haskell with lots of code samples. As longs as the code dealt with x, i and f the naming didn't bother me. However, as soon as we got into heavy duty list manipulation and the like I found the three letters or so names to be a lot less readable than I prefer.
To be fair a significant part of the naming followed a set of conventions, so I assume that once you get into the lingo it will be a little easier.
Fortunately, nothing prevents us from using meaningful names, but I don't agree that the language itself somehow makes three letter identifiers meaningful to the majority of people.

When in Rome, do as the Romans do
(Or as they say in my town: "Donde fueres, haz lo que vieres")

Anything that aids readability is a good thing - meaningful names are therefore a good thing in any language.
I use short variable names in many languages but they're reserved for things that aren't important in the overall meaning of the code or where the meaning is clear in the context.
I'd be careful how far I took the advice about Haskell names

My Haskell practice is only of mediocre level, thus, I dare to try to reply only the second, more general part of Your question:
"In essence, it claims that functional programming is such a distinct paradigm that the naming conventions from other paradigms don't apply."
I suspect, the answer is "yes", but my motivation behind this opinion is restricted only on experience in just one single functional language. Still, it may be interesting, because this is an extremely minimalistic one, thus, theoretically very "pure", and underlying a lot of practical functional languages.
I was curios how easy it is to write practical programs on such an "extremely" minimalistic functional programming language like combinatory logic.
Of course, functional programming languages lack mutable variables, but combinatory logic "goes further one step more" and it lacks even formal parameters. It lacks any syntactic sugar, it lacks any predefined datatypes, even booleans or numbers. Everything must be mimicked by combinators, and traced back to the applications of just two basic combinators.
Despite of such extreme minimalism, there are still practical methods for "programming" combinatory logic in a neat and pleasant way. I have written a quine in it in a modular and reusable way, and it would not be nasty even to bootstrap a self-interpreter on it.
For summary, I felt the following features in using this extremely minimalistic functional programming language:
There is a need to invent a lot of auxiliary functions. In Haskell, there is a lot of syntactic sugar (pattern matching, formal parameters). You can write quite complicated functions in few lines. But in combinatory logic, a task that could be expressed in Haskell by a single function, must be replaced with well-chosen auxiliary functions. The burden of replacing Haskell syntactic sugar is taken by cleverly chosen auxiliary functions in combinatory logic. As for replying Your original question: it is worth of inventing meaningful and catchy names for these legions of auxiliary functions, because they can be quite powerful and reusable in many further contexts, sometimes in an unexpected way.
Moreover, a programmer of combinatory logic is not only forced to find catchy names of a bunch of cleverly chosen auxiliary functions, but even more, he is forced to (re)invent whole new theories. For example, for mimicking lists, the programmer is forced to mimick them with their fold functions, basically, he has to (re)invent catamorphisms, deep algebraic and category theory concepts.
I conjecture, several differences can be traced back to the fact that functional languages have a powerful "glue".

In Haskell, meaning is conveyed less with variable names than with types. Being purely functional has the advantage of being able to ask for the type of any expression, regardless of context.

I agree with a lot of the points made here about argument naming but a quick 'find on page' shows that no one has mentioned Tacit programming (aka pointfree / pointless). Whether this is easier to read may be debatable so it's up to you & your team, but definitely worth a thorough consideration.
No named arguments = No argument naming conventions.

Related

Will I develop good/bad habits because of lazy evaluation?

I'm looking to learn functional programming with either Haskell or F#.
Are there any programming habits (good or bad) that could form as a result Haskell's lazy evaluation? I like the idea of Haskell's functional programming purity for the purposes of understanding functional programming. I'm just a bit worried about two things:
I may misinterpret lazy-evaluation-based features as being part of the "functional paradigm".
I may develop thought patterns that work in a lazy world but not in a normal order/eager evaluation world.
There are habits that you get into when programming in a lazy language that don't work in a strict language. Some of these seem so natural to Haskell programmers that they don't think of them as lazy evaluation. A couple of examples off the top of my head:
f x y = if x > y then .. a .. b .. else c
where
a = expensive
b = expensive
c = expensive
here we define a bunch of subexpressions in a where clause, with complete disregard for which of them will ever be evaluated. It doesn't matter: the compiler will ensure that no unnecessary work is performed at runtime. Non-strict semantics means that the compiler is able to do this. Whenever I write in a strict language I trip over this a lot.
Another example that springs to mind is "numbering things":
pairs = zip xs [1..]
here we just want to associate each element in a list with its index, and zipping with the infinite list [1..] is the natural way to do it in Haskell. How do you write this without an infinite list? Well, the fold isn't too readable
pairs = foldr (\x xs -> \n -> (x,n) : xs (n+1)) (const []) xs 1
or you could write it with explicit recursion (too verbose, doesn't fuse). There are several other ways to write it, none of which are as simple and clear as the zip.
I'm sure there are many more. Laziness is surprisingly useful, when you get used to it.
You'll certainly learn about evaluation strategies. Non-strict evaluation strategies can be very powerful for particular kinds of programming problems, and once you're exposed to them, you may be frustrated that you can't use them in some language setting.
I may develop thought patterns that work in a lazy world but not in a normal order/eager evaluation world.
Right. You'll be a more rounded programmer. Abstractions that provide "delaying" mechanisms are fairly common now, so you'd be a worse programmer not to know them.
I may misinterpret lazy-evaluation-based features as being part of the "functional paradigm".
Lazy evaluation is an important part of the functional paradigm. It's not a requirement - you can program functionally with eager evaluation - but it's a tool that naturally fits functional programming.
You see people explicitly implement/invoke it (notably in the form of lazy sequences) in languages that don't make it the default; and while mixing it with imperative code requires caution, pure functional code allows safe use of laziness. And since laziness makes many constructs cleaner and more natural, it's a great fit!
(Disclaimer: no Haskell or F# experience)
To expand on Beni's answer: if we ignore operational aspects in terms of efficiency (and stick with a purely functional world for the moment), every terminating expression under eager evaluation is also terminating under non-strict evaluation, and the values of both (their denotations) coincide.
This is to say that lazy evaluation is strictly more expressive than eager evaluation. By allowing you to write more correct and useful expressions, it expands your "vocabulary" and ability to think functionally.
Here's one example of why:
A language can be lazy-by-default but with optional eagerness, or eager by default with optional laziness, but in fact its been shown (c.f. Okasaki for example) that there are certain purely functional data structures which can only achieve certain orders of performance if implemented in a language that provides laziness either optionally or by default.
Now when you do want to worry about efficiency, then the difference does matter, and sometimes you will want to be strict and sometimes you won't.
But worrying about strictness is a good thing, because very often the cleanest thing to do (and not only in a lazy-by-default language) is to use a thoughtful mix of lazy and eager evaluation, and thinking along these lines will be a good thing no matter which language you wind up using in the future.
Edit: Inspired by Simon's post, one additional point: many problems are most naturally thought about as traversals of infinite structures rather than basically recursive or iterative. (Although such traversals themselves will generally involve some sort of recursive call.) Even for finite structures, very often you only want to explore a small portion of a potentially large tree. Generally speaking, non-strict evaluation allows you to stop mixing up the operational issue of what the processor actually bothers to figure out with the semantic issue of the most natural way to represent the actual structure you're using.
Recently, i found myself doing Haskell-style programming in Python. I took over a monolithic function that extracted/computed/generated values and put them in a file sink, in one step.
I thought this was bad for understanding, reuse and testing. My plan was to separate value generation and value processing. In Haskell i would have generated a (lazy) list of those computed values in a pure function and would have done the post-processing in another (side-effect bearing) function.
Knowing that non-lazy lists in Python can be expensive, if they tend to get big, i thought about the next close Python solution. To me that was to use a generator for the value generation step.
The Python code got much better thanks to my lazy (pun intended) mindset.
I'd expect bad habits.
I saw one of my coworkers try to use (hand-coded) lazy evaluation in our .NET project. Unfortunately the consequence of lazy evaluation hid the bug where it would try remote invocations before the start of main executed, and thus outside the try/catch to handle the "Hey I can't connect to the internet" case.
Basically, the manner of something was hiding the fact that something really expensive was hiding behind a property read and so made it look like a good idea to do inside the type initializer.
Contextual information missing.
Laziness (or more specifically, the assumption of the availabilty of the purity and equational reasoning) is sometimes quite useful for specific problem domains, but not necessarily better in general. If you're talking about general-purpose language settings, relying on the lazy evaluation rules by default is considered harmful.
Analysis
Any languages has functional combination (or the applicable terms combination; i.e. function call expression, function-like macro invocation, FEXPRs, etc.) enforces rules on evaluation, implying the order of different parts of subcomputation therein. For convenience and the simplicity of the specification of the language, a language usually specify the rules in a flavor paired to the reduction strategy:
The strict evaluation, or the applicative-order reduction, which evaluates all subexpression first, before the subcomputation of the remaining evaluation of the hole combination.
The non-strict evaluation, or the normal-order reduction, which does not necessarily evaluate every subexpression at first.
The remaining subcomputation finally determines the result of the whole evaluation of the expression. (For program-defined constructs, this usually implies the substitution of the evaluated argument into something like a function body, and the subsequent evaluation of the result.)
Lazy evaluation, or the call-by-need strategy, is a typical concrete instance of the non-strict evaluation kind. To make it practically usable, subexpression evaluations are required to be pure (side-effect-free), so the reductions implementing the strategy can have the Church-Rosser property whatever the order of subexpression evaluation is actually adopted.
One significant merit of such design is the availability of the equational resoning: users can encode the equality of expression evaluation in the program, and optimizing implementation of the language can perform the transformation depending directly on such constructs.
However, there are many serious problems behind such design.
Equational reasoning is not important as it in the first glance in practice.
The encoding is not a separate feature. It has some specific requirements on the other features to carry the encoding. For a pure language, it is even more difficult to encode them elsewhere, so there is certain pressure to make the type system more expressive, hence more complicated typing and typechecking.
Whether the compiler uses the equational reasoning directly encoded in the program or not is an implementation detail. It is more of a taste of style to promote the importance.
Syntatic equations are not powerful enough to encode semantic conditions like cases of "unspecified behavior" in ISO C. It still needs some additional primitives to express non-determinism of such semantic equivalence classes to make optimization techniques based on such equivalence possible.
It is computationally inefficient at the very basic level by default, and not amendable by the programmer easily.
There is no systemic way to reduce the cost on equations which are known not required by the programmer.
One of the significance comes from the clash between lazily evaluated combinations and proper tail recursion over the combinations.
The unpredictable abuse of thunks to memoize the lazily evaluated expressions also makes troubles on the utilization of the machine resources (e.g. registers and the cache memory).
Purely functional languages like Haskell may declare the referential transparency is a good thingTM. However, this is faulty in certain contexts.
There are semantic gaps over the terminology itself. The purity is not the only aspect for the referential transparency; moreover, there are other kinds of such property not readily provided by the evaluation strategy.
In general, referential transparency should not be a goal about programming. Instead, it is an optional manner to implement the composable components of programs. Composability is essentially about the expected invariance on the interface of the components. There are many ways to keep the composability without the aid of any kinds of referential transparency. Whether the guarantee should be enforced by the language rules? It depends. At least, it should not depend totally on the language designers' point.
The lack of impure evaluations requires more syntax noises to encode many constructs simply expressible by mutable state cells in the traditional impure languages. The workarounds of the practical problems do make the solution more difficult and hard to reason by humans.
For example, I/O operations are side-effectful, thus not directly expressible in Haskell expressions under the usual non-strict evaluation rules, otherwise the order of effects will be non-deterministic.
To overcoming the shortcoming, some indirect conventional constructs like the IO monad to simulate the traditional imperative style are proposed. Such monadic constructs are in essential "indirect" in the sense similar to the continuation-passing style, which is considerably low-level and difficult to read. Even though monads can be "powerful" than continuations in expresiveness, it does not naturally powerful than more high-level alternatives (like algebraic effect systems) when the lazy evaluation strategy is not enforced by default.
Besides the intuition problem above, the necessity of using monadic constructs are often difficult to prove formally (if ever possible). As the result, they are very easily abused (just like the design patterns for "OOP" languages derived from Simula). The related syntax sugar, notably, the famous do-notation, is abused for a few decades before well-known by the Haskell community.
Simulating strict language constructs in languages like Haskell usually needs monadic constructs, while simulating non-strict constructs in strict languages are considerably simpler and easier to implement efficiently. For instance, there is SRFI-45.
The lazy evaluation strategy does not deal with many other non-strict constructs well.
For example, seq has to be a compiler magic in GHC. This is not easily expressible by other Haskell constructs without massive changes in the core Haskell language rules.
Although traditional strict languages also do not allow user programs to simulate the enforcement of the order easily so such sequential constructs are therefore primitive (examples: C-like ; is primitive; the derivation of Scheme's begin is relying on the primitive lambda which in turn implying an implicit evaluation order on expressions), it can be implementable reusing the applicative order rules without additional ad-hoc primitives, like the derivation of the$sequence operator in the Kernel language.
Concerns about specific questions
Lazy evaluation is not a must for the "functional paradigm", though as mentioned above, purely functional languages are likely have the lazy evaluation strategy by default. The common properties are the usability of first-class functions. Impure languages like Lisp and ML family are considered "functional", which use eager evaluation by default. Also note the popularity of "functional paradigm" came after the introducing of function-level programming. The latter is quite different, but still somewhat similar to "functional programming" on the treatment of first-classness.
As mentioned above, the way to simulate laziness in eager languages are well-known. Additionally, for pure programs, there may be no non-trivially semantic difference between call-by-need and normal order reduction. To figure out something really only work in a lazy world is actually not easy. (Do you want to implement the language?) Just go ahead.
Conclusion
Be careful to the problem domain. Lazy evaluation may work well for specific scenarios. However, making it by default is likely to be a bad idea in general, because users (whoever to use the language to program, or to derive a new dialect based on the current language) will likely have few chances to ignore all of the problems it will cause.
Well, try to think of something that would work if lazily evaluated, that wouldn't if eagerly evaluated. The most common category of these would be lazy logical operator evaluation used to hide a "side effect". I'll use C#-ish language to explain, but functional languages would have similar analogs.
Take the simple C# lambda:
(a,b) => a==0 || ++b < 20
In a lazy-evaluated language, if a==0, the expression ++b < 20 is not evaluated (because the entire expression evaluates to true either way), which means that b is not incremented. In both imperative and functional languages, this behavior (and similar behavior of the AND operator) can be used to "hide" logic containing side effects that should not be executed:
(a,b) => a==0 && save(b)
"a" in this case may be the number of validation errors. If there were validation errors, the first half fails and the second half is not evaluated. If there were no validation errors, the second half is evaluated (which would include the side effect of trying to save b) and the result (apparently true or false) is returned to be evaluated. If either side evaluates to false, the lambda returns false indicating that b was not successfully saved. If this were evaluated "eagerly", we would try to save regardless of the value of "a", which would probably be bad if a nonzero "a" indicated that we shouldn't.
Side effects in functional languages are generally considered a no-no. However, there are few non-trivial programs that do not require at least one side effect; there's generally no other way to make a functional algorithm integrate with non-functional code, or with peripherals like a data store, display, network channel, etc.

Language features helpful for writing quines (self-printing programs)?

OK, for those who have never encountered the term, a quine is a "self-replicating" computer program. To be more specific, one which - upon execution - produces a copy of its own source code as its only output.
The quines can, of course, be developed in many programming languages (but not all); but some languages are obviously more suited to producing quines than others (to clearly understand the somewhat subjective-sounding "more suited", look at a Haskell example vs. C example in the Wiki page - and I provide my more-objective definition below).
The question I have is, from programming language perspective, what language features (either theoretical design ones or syntax sugar) make the language more suitable/helpful for writing quines?
My definition of "more suitable" is "quines are easier to write" and "are shorter/more readable/less obfuscated". But you're welcome to add more criteria that are at least somewhat objective.
Please note that this question explicitly excludes degenerate cases, like a language which is designed to contain "print_a_quine" primitive.
I am not entirely sure, so correct me if anyone of you knows better.
I agree with both other answers, going further by explaining, that a quine is this:
Y g
where Y is a Y fixed-point combinator (or any other fixed-point combinator), which means in lambda calculus:
Y g = g(Y g)
now, it is quite apparent, that we need the code to be data and g be a function which will print its arguments.
So to summarize we need for constructing such a quines functions, printing function, fixed-point combinator and call-by-name evaluation strategy.
The smallest language that satisfies this conditions is AFAIK Zot from the Iota and Jot family.
Languages like the Io Programming Language and others allow the treating of code as data. In tree walking systems, this typically allows the language implementer to expose the abstract syntax tree as a first class citizen. In the case of Io, this is what it does. Being object oriented, the AST is modelled around Message objects, and a special sentinel is created to represent the currently executing message; this sentinel is called thisMessage. thisMessage is a full Message like any other, and responds to the print message, which prints it to the screen. As a result, the shortest quine I've ever been able to produce in any language, has come from Io and looks like this:
thisMessage print
Anyway, I just couldn't help but sharing this with you on this subject. The above certainly makes writing quines easy, but not doing it this way certainly doesn't preclude easily creating a quine.
I'm not sure if this is useful answer from a practical point of view, but there is some useful theory in computability theory. In particular fixed points and Kleene's recursion theorem can be used for writing quines. Apparently, the theory can be used for writing quine in LISP (as the wikipedia page shows).

Experiences teaching or learning map/reduce/etc before recursion? [closed]

Closed. This question is opinion-based. It is not currently accepting answers.
Want to improve this question? Update the question so it can be answered with facts and citations by editing this post.
Closed 9 years ago.
Improve this question
As far as I can see, the usual (and best in my opinion) order for teaching iterting constructs in functional programming with Scheme is to first teach recursion and maybe later get into things like map, reduce and all SRFI-1 procedures. This is probably, I guess, because with recursion the student has everything that's necessary for iterating (and even re-write all of SRFI-1 if he/she wants to do so).
Now I was wondering if the opposite approach has ever been tried: use several procedures from SRFI-1 and only when they are not enough (for example, to approximate a function) use recursion. My guess is that the result would not be good, but I'd like to know about any past experiences with this approach.
Of course, this is not specific to Scheme; the question is also valid for any functional language.
One book that teaches "applicative programming" (the use of combinators) before recursion is Dave Touretsky's COMMON LISP: A Gentle Introduction to Symbolic Computation -- but then, it's a Common Lisp book, and he can teach imperative looping before that.
IMO starting with basic blocks of knowledge first is better, then derive the results. This is what they do in mathematics, i.e. they don't introduce exponentiation before multiplication, and multiplication before addition because the former in each case is derived from the latter. I have seen some instructors go the other way around, and I believe it is not as successful like when you go from the basics to the results. In addition, by delaying the more advanced topics, you give students a mental challenge to derive these results them selves using the knowledge they already have.
There is something fundamentally flawed in saying "with recursion the student has everything that's necessary for iterating". It's true that when you know how to write (recursive) functions you can do anything, but what is that better for a student? When you think about it, you also have everything you need if you know machine language, or to make a more extreme point, if I give you a cpu and a pile of wires.
Yes, that's an over-exaggeration, but it can relate to the way people teach. Take any language and remove any inessential constructs -- at the extreme of doing so in a functional language like Scheme, you'll be left with the lambda calculus (or something similar), which -- again -- is everything that you need. But obviously you shouldn't throw beginners into that pot before you cover more territory.
To put this in more concrete perspective, consider the difference between a list and a pair, as they are implemented in Scheme. You can do a lot with just lists even if you know nothing about how they're implemented as pairs. In fact, if you give students a limited form of cons that requires a proper list as its second argument then you'll be doing them a favor by introducing a consistent easier-to-grok concept before you proceed to the details of how lists are implemented. If you have experience teaching this stuff, then I'm sure that you've encountered many students that get hopelessly confused over using cons where the want append and vice versa. Problems that require both can be extremely difficult to newbies, and all the box+pointer diagrams in the world are not going to help them.
So, to give an actual suggestion, I recommend you have a look at HtDP: one of the things that this book is doing very carefully is to expose students gradually to programming, making sure that the mental picture at every step is consistent with what the student knows at that point.
I have never seen this order used in teaching, and I find it as backwards as you. There are quite a few questions on StackOverflow that show that at least some programmers think "functional programming" is exclusively the application of "magic" combinators and are at a loss when the combinator they need doesn't exist, even if what they would need is as simple as map3.
Considering this bias, I would make sure that students are able to write each combinator themselves before introducing it.
I also think introducing map/reduce before recursion is a good idea. (However, the classic SICP introduces recursion first, and implement map/reduce based on list and recursion. This is a building abstraction from bottom up approach. Its emphises is still abstraction.)
Here's the sum-of-squares example I can share with you using F#/ML:
let sumOfSqrs1 lst =
let rec sum lst acc =
match lst with
| x::xs -> sum xs (acc + x * x)
| [] -> acc
sum lst 0
let sumOfSqr2 lst =
let sqr x = x * x
lst |> List.map sqr |> List.sum
The second method is a more abstract way to do this sum-of-squares problem, while the first one expresses too much details. The strength of functional programming is better abstraction. The second program using the List library expresses the idea that the for loop can be abstracted out.
Once the student could play with List.*, they would be eager to know how these functions are implemented. At that time, you could go back to recursion. This is kind of top-down teaching approach.
I think this is a bad idea.
Recursion is one of the hardest basic subjects in programming to understand, and even harder to use. The only way to learn this is to do it, and a lot of it.
If the students will be handed the nicely abstracted higher order functions, they will use these over recursion, and will just use the higher order functions. Then when they will need to write a higher order function themselves, they will be clueless and will need you, the teacher, to practically write the code for them.
As someone mentioned, you've gotta learn a subject bottom-up if you want people to really understand a subject and how to customize it to their needs.
I find that a lot of people when programming functionally often 'imitate' an imperative style and try to imitate loops with recursion when they don't need anything resembling a loop and need map or fold/reduce instead.
Most functional programmers would agree that you shouldn't try to imitate an imperative style, I think recursion shouldn't be 'taught' at all, but should develop naturally and self-explanatory, in declarative programming it's often at various points evident that a function is defined in terms of itself. Recursion shouldn't be seen as 'looping' but as 'defining a function in terms of itself'.
Map however is a repetition of the same thing all over again, often when people use (tail) recursion to simulate loops, they should be using map in functional style.
The thing is that the "recursion" you really want to teach/learn is, ultimately, tail recursion, which is technically not recursion but a loop.
So I say go ahead and teach/learn the real recursion (the one nobody uses in real-life because it's impractical), then teach why they are useless, then teach tail-recursion, then teach why they are not recursions.
That seems to me to be the best way. If you're learning, do all this before using higher-order functions too much. If you're teaching, show them how they replace loops (and then they'll understand later when you teach tail-recursion how the looping is really just hidden but still there).

What technique in functional programming is difficult to learn but useful afterwards?

This question is of course inspired by Monads in Haskell.
wrapping my head around continuation passing style has helped my javascript coding a lot
I would say First-class functions.
In computer science, a programming
language is said to support
first-class functions (or function
literals) if it treats functions as
first-class objects. Specifically,
this means that the language supports
constructing new functions during the
execution of a program, storing them
in data structures, passing them as
arguments to other functions, and
returning them as the values of other
functions. This concept doesn't cover
any means external to the language and
program (metaprogramming), such as
invoking a compiler or an eval
function to create a new function.
Do you want to measure the usefulness in connection with functional-programming itself or programming in general?
In general, the positive experience of functional programming doesn't result from particular techniques but from the way it changes your thinking -
Holding immutable data
Formulating declaratively (recursion, pattern-matching)
Treating functions as data
So I'd say that functional programming is the answer to your question itself.
But to give a more specific answer too, I'd vote for functional abstraction mechanisms like
monads
arrows
continuation-passing-style
zippers
higher-order-functions
generics + typeclasses.
As already said, they are very abstract things on the first view, but once you have understood them, they are extremely cool and valueable techniques to write concise, error-safe and last but not least highly reusable code.
Compare the following (Pseudocode):
// Concrete
def sumList(Data : List[Int]) = ...
// Generic
def sumGeneric[C : Collection[T], T : Num](Data : C) = ...
The latter might be somewhat unintuitive compared with the first definition, but it allows you to work with any collection and numeric type in general!
All in all, many modern (mainstream) languages have discovered such benefits and introduced very functional features like lambda functios or Linq. Having understood these techniques will also improve writing code in this languages.
One from the "advanced" department: Programming with phantom types (sometimes also called indexed types). It's admittedly not a "standard" technique in functional programming but not entirely esoteric either, and it's something to keep your brain busy for awhile (you asked for something difficult, right? ;)).
In a nutshell, it is about parameterizing types to encode and statically enforce certain properties at compile time. One of the standard examples is the vector addition function that statically ensures that given two vectors of length N and M will return a vector of length N+M or otherwise you get a compile-time error. Yes, there are more interesting applications.
These techniques are not quite as useful in C++ as they are in a proper functional programming language, but so far I've managed to sneak some of this stuff in all of my recent projects at work to a varying degree, most recently in a C++ EDSL context where it worked out really well. You don't necessarily have to encode fancy stuff, learning this helped me catching the situations where a few type tags can reduce the verbosity of an EDSL or allowed a cleaner syntax, for example.
Admittedly, the usefulness is somewhat restricted by language support and what you're trying to achieve.
Some starters:
Generic and Indexed Type (slides with some brief applications overview)
Fun with Phantom Types
The Kennedy and Russo paper mentioned in the slides is Generalized Algebraic Data Types
and Object Oriented Programming and puts some of this stuff into the context of C#/Java.
Chapter 3 in Dave Abraham's book C++ Template Metaprogramming is available online as sample chapter and uses these techniques in C++ for dimensional analysis.
A practical FP project using phantom types is HaskellDB.
I would say that Structural typing in OCaml is particularly rewarding.
recursion. Difficult to wrap your head around it at times
The concept of higher-order functions, lambda functions and the power of generic algorithms that are easy to combine were very beneficial for me. I'm always excited when I see what I can do with a fold in haskell.
Likewise my programming in C# has changed a lot (to the better, I hope) since I got into functional programming (haskell specifically).

What functional language techniques can be used in imperative languages?

Which techniques or paradigms normally associated with functional languages can productively be used in imperative languages as well?
e.g.:
Recursion can be problematic in languages without tail-call optimization, limiting its use to a narrow set of cases, so that's of limited usefulness
Map and filter have found their way into non-functional languages, even though they have a functional sort of feel to them
I happen to really like not having to worry about state in functional languages. If I were particularly stubborn I might write C programs without modifying variables, only encapsulating my state in variables passed to functions and in values returned from functions.
Even though functions aren't first class values, I can wrap one in an object in Java say, and pass that into another method. Like Functional programming, just less fun.
So, for veterans of functional programming, when you program in imperative languages, what ideas from FP have you applied successfully?
Pretty nearly all of them?
If you understand functional languages, you can write imperative programs that are "informed" by a functional style. That will lead you away from side effects, and toward programs in which reading the program text at any particular point is sufficient to let you really know what the meaning of the program is at that point.
Back at the Dawn of Time we used to worry about "coupling" and "cohesion". Learning an FP will lead you to write systems with optimal (minimal) coupling, and high cohesion.
Here are things that get in the way of doing FP in a non-FP language:
If the language doesn't support lambda/closures, and doesn't have any syntactic sugar to easily mostly hack it, you are dead in the water. You don't call map/filter without closures.
If the language is statically-typed and doesn't support generics, you are dead in the water. All the good FP stuff uses genericity.
If the language doesn't support tail-recursion, you are hindered. You can write implementations of e.g. 'map' iteratively; also often your data may not be too large and recursion will be ok.
If the language does not support algebraic data types and pattern-matching, you will be mildly hindered. It's just annoying not to have them once you've tasted them.
If the language cannot express type classes, well, oh well... you'll get by, but darn if that's not just the awesomest feature ever, but Haskell is the only remotely popular language with good support.
Not having first-class functions really puts a damper on writing functional programs, but there are a few things that you can do that don't require them. The first is to eschew mutable state - try to have most or all of your classes return new objects that represent the modified state instead of making the change internally. As an example, if you were writing a linked list with an add operation, you would want to return the new linked list from add as opposed to modifying the object.
While this may make your programs less efficient (due to the increased number of objects being created and destroyed) you will gain the ability to more easily debug the program because the state and operation of the objects becomes more predictable, not to mention the ability to nest function calls more deeply because they have state inputs and outputs.
I've successfully used higher-order functions a lot, especially the kind that are passed in rather than the kind that are returned. The kind that are returned can be a bit tedious but can be simulated.
All sorts of applicative data structures and recursive functions work well in imperative languages.
The things I miss the most:
Almost no imperative languages guarantee to optimize every tail call.
I know of no imperative language that supports case analysis by pattern matching.

Resources