Alan Kay (yes, that Alan Kay[1] -- the guy that's won a Turing award) formalized spreadsheets as a limited form of first-order functional programming.[0]
In that case any languages supporting function is "first-order functional programming." It's kind of meaningless when we are talking about real functional languages. PHP can define a function that takes a value and returns a value, and thus it can do "first-order functional programming." But no one claims PHP is a functional language. (Not to degrade PHP, it's a good language doing what it does best.)
All these twisting the definition of functional programming just produce more confusion.
On the other hand, a "formula" in excel is inherently a "functional" function or "first order function". Why? two things: No side effects, and statelessness- that is, the order of the computation is theoretically implicit: a value can be computed in a number of different arbitrary orders without effecting the final result.-- not explicit- like in php, where you are specifying an exact order of operations by the order of statements in a procedure ("function").
Another thing about statelessness is that given the same input, a "first order function" must always return the same output.
The fact that order is not explicit in excel, that there is no user controlled "state" means that there's certain associative, commutative and compositional properties available to "first order" functions that are not possible with a procedure that is not guaranteed to be a first order function.
it's the difference between deciding to do functional programming in an imperative language, and having functional programming structurally enforced by the language.
Likewise, Excel can do stateful update with side effect. The cell formula function happens to be a subset of all it can do. Sure, we can claim it's "first-order functional programming." But do we want to call it a functional language?
My point is that "first-order functional programming" is a really weak qualifier to call a language functional.
In this case, I believe the "Excel is a functional language" meme is using "Excel" not to include ALL the things that excel does, which includes things like VB and Javascript scripting, but as shorthand for the spreadsheet model of application, and more specifically aimed at Excel's formula language, which enforces the creation of first order functions by excluding the possibility of state-fulness and side-effects from the language's design. It is the fact that this is inherent in its design that merits calling it a functional language, while in php there are no such design choices which enforce or even encourage a functional way of writing code.
The most important aspect of being functional, high order function, is completely missing.
Common declarative languages include those of database query languages (e.g., SQL, XQuery), regular expressions, logic programming, and functional programming.
I guess you're both right then. :)
One FP attribute that spreadsheets do have is referential transparency: the same inputs given to the same formula will always produce the same result.
let x = 3, y = 5 in
x + y
No (mutable) variables there, but if I change the y = 3 bit to y = 4, the value of the expression changes. That's what it's like in excel, as well.You're right, of course. But it depends on how you draw the boundaries of the system. If you think of the numbers in your example as code, then the program is stateless, and each time you change a number you get a new program. But if you think of them as parameters that live outside the code, then the edit-recompute cycle is a state change. In a similar if trivial way, if you take (say) a Java program running over a database and decide that the data in the database is "code", that "program" (consisting of Java code plus database) is now "stateless" too, and anyone who updates a database record is changing the "program".
Given that "stateful" vs. "stateless" depends on how you draw the boundaries of the system, the question is how best to draw the boundaries of a spreadsheet. I'm not arguing there's a single correct way to do that, but in my view the user's mental model of a spreadsheet is closer to "a calculating machine with state that I can update" than it is to "a stateless calculating machine with a lot of hard-coded literals".
Psychologically, updating the numbers in a spreadsheet doesn't feel like editing the source code of a program and re-running it. It feels stateful, like mutating something that triggers a cascade of side effects (recalculation). For this reason I think that spreadsheets are closer to the Smalltalk vision of a world of objects that respond to user interaction (as well as to each other) than they are to the functional programming vision of pure code. Put differently, the spreadsheet's I/O, its grid UI, is part of its essence. You can't abstract away from that without losing the heart of the thing. So the analogy with functional programming, though tempting, leads in the wrong direction. Every individual formula that lives in a cell is certainly a functional program when taken in isolation—but the memory model by which the cells reference one another and get updated is stateful.
Kay's interest in spreadsheets wasn't about functional programming, it was about interactive and dynamic computation. I have a pdf of the 1984 Scientific American article that Wikipedia is quoting from. It does include the phrase "value rule"—by which he simply meant what we would call a spreadsheet formula—but I'm pretty sure it makes no argument about functional programming (it's all images so I can't search to be sure). If you'd like a copy, email me. It's a pretty neat article, ahead of its time as one would expect from Alan Kay.
I guess we'll agree to disagree, I don't think someone needs to use the phrase "first-order functional programming" when they give the very definition of it, which is what Wikipedia does: summarize Kay's argument.
I do agree the article itself was quite interesting, and certainly ahead of its time.
I was making a textual point about the Wikipedia article. Its use of the phrase "first-order functional programming" is hyperlinked to http://journals.cambridge.org/action/displayAbstract?aid=727.... It's not citing Alan Kay.
I have no idea if Alan Kay ever used the latter phrase, but it was easier to use that phrase here on HN than Alan Kay's made up phrase, which would have be difficult to understand without the content of the article explaining it.
Totally agree. My startup is using hierarchical grids as our core datatypes for just that reason (we also support Function cells that are used as values).
"Naming" is one of the core problems in computer science, and grids/spreadsheets elegantly solve that problem for many ad hoc use cases, where functional programming (in all its forms) does not.
Whereas in functional languages, the functions declare relationships between values and the language uses the evaluation model to propagate the results between function evaluation. Where's the difference?
For example, you can't define the abstraction "square" in Excel. You can compute 2 * 2, A1 * A1 and so on. But there's no way to define a construct which given x produces x * x, then reuse that construct everywhere you want to square things. Everywhere you want to square something, you must inline the computation. That's the absence of functional abstraction.
Second, OO is an orthogonal attribute to functional.
Third, your argument commits a formal fallacy of this form:
All cats have whiskers
Cats can have stripes
Tony has stripes
therefore tony is a cat.
---The possibility of an attribute in X, and Y containing that attribute does not imply that Y is an X.
the point is that whether something is declarative has no bearing on whether it is functional or not. It's an irrelevant point to bring up. Whether something is OO is equally irrelevant. That is what "orthogonal" means. I would go on to define for you "functional" "declarative", "formal fallacy" "logic", but this seems like a bottomless rabbit hole. I can only hope you'll try and find out what these words actually mean yourself.
However, my original point was addressing the claim in the blog that functional and declarative are equivalent in that functional has aspect of declarative. And since Excel is declarative, it's functional. I was basically saying that's not the case.
While declarative and functional can be orthogonal, and can both live in a language is an interesting but separate topic that is not really related to the original comment.
A mathematical (ie pure) function is declarative in nature as it defines a relation between sets. However, in practice you often use it imperatively, ie as a means to get an output from some input.
An object is imperative in nature as it encapsulates mutable state. However, the set of messages defines an abstract interface which is arguably declarative.
In principle, a purely functional language cannot be object-oriented, and a purely object-oriented language cannot be functional. In practice, this doesn't matter as pure languages are rare.
It's not functional at all. This reaffirms my belief that blog posts are a terrible place to learn. People who know the least shout the loudest.
Functional programming is declarative. Especially when it's lazy and the program's instantaneous state is abstracted out. One of the motivating goals in functional programming is to be able to define a computation once, in terms of other computations, and have that relationship be maintained with minimal regard to the state of the program or its order of execution. Which is what it seems like (this is the first time I've specifically encountered them) Promises are a powerful tool for accomplishing.
In contrast, threading explicit callbacks/continuations through a program, which is explicitly managing the order of execution, is relatively more imperative, which I think is the point of the article -- you don't need the wider control-flow flexibility of explicitly and manually threading callbacks to do the type of computations that most async web stuff does. You can abstract the common callback pattern out into something like Promises, and make all your shit more consistent and concise.
I'm saying his characterization of promises as functional is so close to being right that it's the worst kind of wrongness.
It's worse to say, "The capital of Kenya is Nairobi Central" than to say "The capital of Kenya is London." It is worse because errors that are only slightly wrong and maybe even partially correct are more deceptive, durable, and confusing.
A promise is not a functional concept. Functions don't make promises. You CAN, as a thinking human, re-conceptualize the immutable return value of a function as being "pretty much the same" as a promise, but when you re-conceptualize functional programming you are just deconstructing what is already understood to be an arbitrary conceptual construct. Any conceptual construct--especially one based on simple metaphors--can be deconstructed into some other form. This doesn't invalidate the construct and it doesn't equate the construct as such with the deconstructed components.
You might as well say that an airplane is not really an airplane. It's just two wings, a cockpit, some jet engines and a fuselage assembled together in a certain way so that it can fly. It's not really an airplane; it's just those things in that way. And those things are really just all made out of aluminium so really it's not an airplane it's just aluminium. Anyone who says an airplane is more than just aluminium is making a tyrannical dichotomy.
So according to you promises are functional because under the hood promises have some conceptual similarity to the immutable return values provided in functional programming. But under the hood the Boeing 737 is just scrap metal. Yet it's NOT just scrap metal. The difference is that "the promise" is a precise concept based on some metaphors that is specifically used in declarative styles and not in functional styles. Functions return immutable values. Declarations make promises. Under the hood there may be similarities but no one made any claims to the contrary.
If people don't think these high level conceptual distinctions are important, why are they even reading this article that is entirely about splitting hairs between these conceptual distinctions? AND the article is even getting it wrong because the author is a self-admitted amateur. Why bother learning from the village loudmouth when there are geniuses who publish books and give lectures just down the hall?
The stuff that reaches the front page of HN makes no sense. Most of it is written by terribly ignorant people. Programmers have no respect for expertise.
If you insist on setting up semantic games: if Oleg hacking Haskell isn't functional programming then nothing meaningfully is.
----
> Functions return immutable values. Declarations make promises. Under the hood there may be similarities but no one made any claims to the contrary.
What does `foldr (+)` return? What about `const 1`?
Semantic games are meaningful or this blog post wouldn't be on HN. The title of the post makes it clear that we're playing a semantic game.
Kiselyov is deconstructing the constructs. That's fine. He's playing the semantic game of showing how different concepts share ontology. This doesn't make the concepts identical; it makes their deconstructions identical. These paradigms are not physical things they are conceptual abstractions. Anyway I'm obviously not getting through to you.
const is not a promise, it's a constant. Different concepts, similar ontology.
A "promise" is a delayed computation. It's a stand-in for a value, and the computation referencing it will suspend its execution until a value is available for it to consume. Similarly with Iteratees. They encapsulate specific patterns for building computations out of other computations.
What makes the "promise" pattern relatively less imperative than threading callbacks is that it abstracts out a specific pattern for ordering a computation, so that it doesn't need to be restated every time.
What makes the "promise" pattern relatively less declarative than "purely declarative" is that the control of the computation can be specified in the language itself. It's just abstracted out to the pattern in common. Similarly to re-writing certain stateful `for` loops with `map`.
Promises are a pattern in the functional programming paradigm. The "functional programming community" is where the construct comes from, and using them lets you write code that is closer to the declarative ideals of functional programming than the article's specified alternative, threading callbacks, which has relatively more in common with the explicit control of the order of computation that is implied by imperative programming.
I don't see anything in the article that remotely justifies "It's not functional at all. This reaffirms my belief that blog posts are a terrible place to learn. People who know the least shout the loudest.". In fact, I don't see anything wrong with it at all, or anything that substantially disagrees with what I've said or what I've seen any of the type of (yet-unspecified) authorities of certified expertise that you've appealed to have said. Yes, Promises are relatively more declarative, but they're also relatively more functional, relatively less imperative, and I don't see any basis for calling somebody an idiot for not using those terms with total mutual exclusion.
So please, raise, any specific objections you have which I have not addressed, or else have the honesty to retract the undeserved insults you've thrown at the author of this article.
You define promise:
>A "promise" is a delayed computation. It's a stand-in for a value, and the computation referencing it will suspend its execution until a value is available for it to consume.
I define promise:
>A "promise" is a declaration or assurance that one will do a particular thing or that guarantees that a particular thing will happen.
My definition of promise is conceptually and formally more accurate than yours, because your definition raises issues like "what is a computation" and "what is a value" and "what is a stand-in" and "what is execution" and "what does consume mean"
As programmers, we understand what these words mean concretely in terms of lower-level abstractions or in some cases actual hardware operation.
My point is that the declarative ideal that you reference is not defined in terms of lower level programmer abstractions, but rather in terms of natural conceptual metaphors.
Since functional programming is generally seen as a subset of declarative programming, it's natural that functional programmers use declarative concepts and it's even understandable that they call those concepts functional. But they're mixing metaphors and creating conceptual confusion. They're losing track of the original metaphors that draw hair-splitting analytic distinctions between concepts in order to create clarity. It's semantics.
Like I already said, I'm not criticizing his design sense and I'm not criticizing yours. I'm criticizing his and your use of language and I'm asserting that your language is degraded and impure.
To me, aside from the design advice he gives, his blog post is about drawing semantic distinctions. I just find the way he does this to be horribly flawed in a way is "so close to being right that it's the worst kind of wrong."
Similarly, I would describe your definition of promise in this way. So close to being right that it's the worst kind of wrong. Your definition of promise is not a definition; it's a deconstruction. It's actually an imperative deconstruction--you're describing the promise using imperative language--concepts like execution and computation.
HN is full of very "practical" people who have a view that goes something like, "I'm correct enough for this to WORK, I'm correct enough to get the correct answer, therefore I'm correct." It's a reductionist view of truth; that which I can deconstruct, I understand. I don't share this view.
I would say, you ARE correct, but you're not as correct as you could be. Likewise with the author.
When I want to learn about the distinction between declarative paradigms and functional paradigms I talk to people who specialize in drawing that distinction. Since the blog post is prima facie drawing a distinction between functional and imperative, the value I'm looking for is an analytically rigorous distinction between abstract concepts.
That's not what I found, so I'm critical. Sorry. This just isn't the author's area of expertise and it shows. Perhaps most of his readers don't care. More power to them. But if you want to engage in intellectual life you can't throw down the gauntlet every time someone criticizes you.
Who specializes in making that distinction? Point me in the right direction. I am absolutely throwing down the gauntlet, I don't believe for a second you know what you're talking about, and I suspect that your understanding of functional programming does not include the possibility of functions returning functions, and the term "thunk" would be lost on you. If you're going to put so much weight on expertise, please point me in the direction of the researcher who you suspect would take the author of this article to task for his terminology. Or make a positive argument yourself.
"Promise" here refers to a specific computation strategy, it has a well-documented intellectual heritage and has been defined as a lambda calculus. It has its basis in the structuring of computations. With how you've constructed this ontology, I don't see any room for "functional programming" to mean anything at all.
I suppose that does leave a lot of room for calling the OP "the village loudmouth" for having any actual content behind his words.
I'm just inline with my own discipline where promise has a different meaning, a super ordinate meaning that includes your meaning. As a consequence, anyone from your discipline just thinks I'm clueless. I'm definitely not going to convince you that the typical definition of promise, the one you're used to, is actually just an operational definition--an instrumental label assigned to a particular instance of a promise-like thing.
I'd cite the tiny field of cranks who think like me, but that would just bring shame on them by association. I'm not doing research in functional programming. I'm working on hair-splitting tyrannical distinctions-without-a-difference. According to your paradigm, I'm an outright fraud spouting bullshit. So enjoy your victory I guess.