Mastering Time-to-Market with Haskell
fpcomplete.com
fpcomplete.com
But once the core abstractions are settled, you start to reap its power. The type system catches tons of potential errors. Combinators allow for enormous expressiveness. Here you're rolling: Haskell is in its zone!
But then you hit a wall. Laziness makes for brutal debugging. Singly linked lists actually suck. Performance optimization is a black art. You find yourself longing for a language with simple semantics and mechanical sympathy. Now Haskell is bumping up against the real world.
Haskell has its sweet spot somewhere between "bang this out by 5pm" and "ship this to a million users". (No surprise it's popular in academia.)
This is so true. Once you get the types settled for your project, then it becomes a breeze to quickly add things and refactor over and over again, because maybe the spec or idea changed.
I've recently experienced this in both a Haskell project (using the Yesod framework) and I'm currently working on an Elm project, where I very much feel this strength also.
I'll have to agree with you on the debugging though - it is somewhat of an art, and not one I've completely mastered yet, since I haven't had that much need for it for now (haven't done that many production haskell projects).
To be fair, these issue are more-or-less solved by newer functional languages, such as OCaml or Rust.
Well, even then, Rust does not only give you control over execution order, but also tight control over memory management, while keeping it safe.
Haskell allows for good reasoning about correctness. In OCaml, you can additionally reason about execution order, but memory management and garbage collector are still somewhat "unpredictable" (i.e. hard to reason about). On top of that, in Rust, you can reason clearly about memory usage and aliasing.
These new, modern languages follow a clear path about keeping quality and correctness (hence safety) while having more and more control over performance, without going back to low-level where you would loose clarity about correctness and safety.
(Note, however, that the distinction between correctness and performance is over-simplified here, because performance can also be part of correctness, e.g. for real-time systems, or in systems where scalability is an important requirement.)
Pure is not quite the opposite of imperative; declarative is the opposite of imperative, although in practice it is true that pure functional languages tend to be more declarative that imperative functional ones.
You can have pure (AKA referentially transparent) imperative languages if they are synchronous[1]. The synchronous style is especially well suited to reactive/interactive applications, and quite popular in hard realtime applications.
Also, we don't really have exact definitions to any of these terms (imperative, pure, functional, declarative). Here is my attempt for approximate, problematic definitions:
* Functional -- a language that models most/all computations as (possibly partial) mathematical functions, and constructs a program by assembling those functions.
* Imperative -- a language that models computation as state transitions, specifying what state should the computation have at each step.
* Declarative -- a language that describes what result should the program give (as a function, a relation or a behavior) rather than at each step.
* Pure -- a language where the semantic value of the composition of syntactic terms can be determined by the semantic value of each of the component terms and no others.
[1]: https://en.wikipedia.org/wiki/Synchronous_programming_langua...
---
@pron
Nowadays I prefer the terms “effect-free” and “effectful”, rather than “pure” and “impure”, since I don't want to suggest that effects are somehow “wrong” or “evil”. By definition, values are effect-free. An effect is anything that invalidates some equational law that holds for values.
For example, integer equality is decidable, and `x == x` evaluates to `true` when `x` is any integer value. However, if we substitute `x` with the expression `foo()`, where:
int counter = 0; // or any other initial value
int foo() { return counter++; }
Of course, nontermination is an effect too. If we had defined `foo` as:
int foo() { while(true); return 0; }
---
> An effect is anything that invalidates some equational law that holds for values.
Ah, I can certainly accept that as a definition, except that I'm not sure how different it is from mine, because the equational law you refer to applies to syntactic terms (`foo` in your example) and the language's equality operator (`==` in your example). It's easy to come with an equivalence relation that would hold for `foo`, but not at the syntax level (e.g., equality that takes `foo` to mean `foo`'s definition and the content of the heap, and the value of `foo` would not be its return value but its behavior -- i.e., the behavior of foo for equal heap contents is always the same).
An effectful language can be given a denotational semantics. At least from what I've seen, when most people say “pure”, they mean “effect-free”.
> It's easy to come with an equivalence relation that would hold for `foo`, but not at the syntax level [emphasis mine]
Yep, indeed, that's the whole point.
But saying "effect free" doesn't mean much unless you define what an effect is, and what constitutes an effect depends on the language. I meant to define the same thing without referring to another vague definition.
Not the runtime equality testing operator. Rather, the language's static notion of contextual equivalence.
It just happens to be the case that most high-level languages provide a built-in equality testing operator that works on primitive types, returning `true` iff its operands are contextually equivalent values.
> I think assembly language and BASIC would still qualify as effect-free
I don't even know what a good notion of contextual equivalence for any assembly language would be.
As for BASIC, I haven't used the old ones, so I can't comment on them. I've used Visual Basic, which most certainly has effectful procedures.
How is that different from denotational semantics?
> The ones that can be effectful or effect-free are specific computations, not whole languages.
Well, we could define an effect-free language as one where all programs are effect-free. But anyway, what is the "static notion of contextual equivalence" in assembly (or BASIC), and how can you write an effectful program in such a language?
On the other hand, neither Haskell nor the MLs tackle the problem of safely managing ephemeral resources. Rust does this, but it sacrifices the ability to conveniently share immutable data structures (Rc and Arc exist, but they aren't very convenient). The good news is that this doesn't have to be a dichotomy. You can have GC for in-memory data structures and deterministic destruction for everything else: https://news.ycombinator.com/item?id=12766781
That's very interesting. Please correct me if I'm wrong, but I think this is not the final solution, either, because you still have the GC running, which prevents strict proofs about performance e.g. for real-time systems. Here, the Rust approach has benefits as it enables you to get very far without a GC.
Also, OCaml is still two decades old. I find it very bizarre that you attempted to group it in with Rust as a "new" language.
The opposite is true: here Haskell helps to find the right data structures/abstractions incredibly fast. The ability to quickly construct and reconstruct a high-level skeleton of your program without actually having to implement anything makes prototyping much easier and faster than in comparable dynamic languages.
> But then you hit a wall. Laziness makes for brutal debugging.
Laziness can sometimes surprise you in bad ways, but it's not like it's impossible to debug.
> Singly linked lists actually suck.
Just replace them with the right data structure then? Even for big codebases and fundamental program abstractions that's a 15min job, because of types.
> Performance optimization is a black art.
I agree. But then again - you don't have to write everything in Haskell. Writing the critical 1% in C, testing the shit out of it and calling it from Haskell is absolutely fine.
> Laziness can sometimes surprise you in bad ways, but it's not like it's impossible to debug.
The last message I get was something like "Array.Array: Index out of bounds". Program crashed. No idea where the error originated. The answer I got was "yeah, in GHC 8 there will be some support for stack traces".
This is much faster than writing a large chunk of code, playing around in the REPL thinking everything is ok, only to realize one week later that you didn't think through some core aspect of your design and then having to refactor a sizeable piece of your codebase.
> The last message I get was something like "Array.Array: Index out of bounds". Program crashed. No idea where the error originated.
It is an unfortunate historical artefact that there are still partial functions (i.e. functions that may crash) in the Haskell base libraries. It is generally advised to avoid such functions, but this is not very clearly communicated to newcomers to Haskell.
For example, Elm, a language largely inspired by Haskell, doesn't have partial functions. This is why it prides itself in "not having any runtime errors in production" (excluding things like compiler bugs or unsafe JavaScript libraries, of course).
Such errors can't be avoided. I need arrays for performance.
If performance is no concern I can use Maps, but I still need indexing ("partial functions") for architectural reasons: separation of concerns - can't make "safe" graphs where everything holds a reference to everything "related".
I just don't buy the idea that partial functions can be avoided.
[0] https://pursuit.purescript.org/packages/purescript-arrays/3....
Or in other words, an "AssertionMonad". No thanks, I'm perfectly fine with assertion errors (i.e. "Index out of bounds"). I just need help with finding my own errors. Stacktraces are a good thing.
For the time being, make a helper function:
ixWithString :: String -> Int -> Array a -> a
check the bounds inside and error out with a custom message when the index is outside the array bounds.
I may be wrong, but I don't think this is technically true. I don't think Elm has only total functions, but it does avoid a lot of common, error-prone partial functions (e.g. accessing elements from a collection).
It's even impossible to upload a package to the official repository containing JavaScript code without prior approval, because the native code may crash!
THAT is where Haskell shines, the ability to refactor and expand fearlessly and aggressively.
I think for large or complex programs Haskell is much faster for constructing a skeleton because it's usually not the logical operations that trip us programmers up, it's the data flow and correct alignment of types, which are implicit in a dynamically typed language and explicit in Haskell.
So, it's kinda-sorta like writing your ideas down in a structured language which enables you to "see" the abstractions in a rigorous light, then the ability to refactor aggressively and safely kicks in so once you can see your abstractions you can quickly adapt for better ones.
So I think the initial work is made quicker in that it forces you to excogitate then because it's in some concrete form you can see it clearly and take advantage of quick turnaround time in refactoring.
“Simple semantics” as measured how? The only good metric I can think of is the size of a formal specification.
There's some evidence Haskell is pretty good at rapid prototyping.
Are you familiar with this '94 paper on rapid prototyping [ http://www.cs.yale.edu/publications/techreports/tr1049.pdf ] by Paul Hudak (RIP), sponsored by the Office of Naval Research & ARPA? The paper is a report on an experimental comparison of programmming languages for prototyping, and it's pretty cool because they were asked to implement a portion of a (heavily simplified) AEGIS weapons system! Unfortunately the paper and experiment are dated and have a number of methodological shortcomings (results were self reported; the examining panel didn't look at running code; requirements were very vague and mostly determined by each team, which makes the comparison difficult), but the results are pretty interesting regardless:
- Of all the teams, only the two Haskell teams delivered working code (though the examiners didn't look at working programs). This must be stressed: no other team delivered running code at all, much less on time!
- One of the Haskell prototypes delivered more features than the initial requirements asked for.
- The programmer of one of the Haskell teams had NO experience with Haskell before the experiment.
- Both Haskell prototypes were also the shortest in line count. Way shorter than the C++ prototype, which wasn't even finished or working in time!
I'd like to see the same experiment repeated under more rigorous conditions and with more modern languages, but the results certainly hint at Haskell being more than adequate at prototyping.
Of course, today it would have to go toe to toe against modern dynamic languages. In this case it'd be essential to demonstrate running code, which is arguably a huge advantage of static typing over dynamic typing; if you're only comparing source code, dynamic languages can get away with being terse and claiming "trust me, it works".
I imagine Haskell would still leave modern C++ coughing dust at the start line :)
Not so ago there was some discussion about "batteries" included with Haskell. [1] It compared situation of Haskell with that of Python/Java etc, worth reading if you are about to go the Haskell route.
It seems, the priorities (academic, commercial, library support and so on) of the members of Haskell community are at crossroads and they cannot seem to resolve those very good, IMHO.
My take: Haskell is good for learning some really deep concepts, but may be not so good when it comes to commercial projects, unless you are a Haskell veteran and also have an army of Haskell veterans with you.
[1] http://osdir.com/ml/haskell-cafe@haskell.org/2016-10/msg0001...
https://github.com/Gabriel439/post-rfc/blob/master/sotu.md
Which backs up your statements; many of the sections under "Common Programming Needs" are rated immature: Databases, Debugging, IDEs...
I'm not saying everything is great with Haskell, but so far, I haven't run into major holes for what I'm doing, library-wise.
Is it rigorous in theory or in practice?
An academic veteran might want to better express his ideas in code while a professional veteran might want more readable code with better test coverage. And that's just 1 axis where the two viewpoints might diverge or even come in conflict.
Industrial rigor regarding code means:
* consistence in using a coding style
* having adequately named modules, functions and variables
* having adequate comments
* having an adequate level of code coverage through automated tests
* having performance and regression tests
* having good release notes
* etc.
Many of those things, required for high quality libraries, are often skipped for the academic projects Haskell is known for.
So that's why Python or even PHP or Javascript, as less "rigorous" theoretical languages, can have more "rigorous" libraries in practice.
Have we actually seen that or have you just asserted that? Is this really true, and if it is, by how much? Haskell has been around for a couple of decades now, and has had least two hype cycles (I remember that when I was in university in the late '90s, Haskell was the next big thing). It does not seem to expand significantly even within organizations that have tried it (and that's a very negative signal), with at least one notable case where the language has been abandoned by a company that was among the flagship adopters.
In general, we know that often linguistic abstractions that seem like a good idea in theory -- or even seem to work nicely in small programs -- don't end up having a significant effect on the bottom line when larger software is concerned. People say that scientific evidence of actual contribution is hard to collect, but we don't even have well-researched anecdotes. Not only do we not have strong evidence in favor of this hypothesis, but there aren't even promising hints. All we do have is people who really like Haskell based on its aesthetics and really wish that the the nice theoretical arguments translated to significant bottom-line gains.
This blog post by Dan Ghica, a PL researcher, really addresses this point: there is nothing to suggest that aesthetically nice theory translates to actual software development gains, and wishful thinking (or personal affinity) simply cannot replace gathering of data: http://danghica.blogspot.com/2016/09/what-else-are-we-gettin...
How can you be sure? Have you seen Haskell shops churn out software significantly faster than others?
Are other banks really spending orders of magnitude more on that kind of software? I've heard that SC adopted Haskell when Lennart Augustsson -- who wrote the first Haskell compiler -- moved to SC from Credit Suisse. When he left CS, they stopped using Haskell. If Haskell gives an order-of-magnitude difference in productivity as you claim, why did CS drop it after Augustsson had left?
I never claimed "orders of magnitude" productivity from a language alone. Haskell has every weakly-typed mainstream imperative language as a valid subset. You need good people, teams and processes to get the best out of the language. This means it is difficult to do case studies without getting dismissed as anecdotal.
Most other banks have settled on Python for similar use cases. I hear on the grapevine that they have significant maintenance problems (unsurprisingly). Haskell is spreading to larger software in the bank and we are currently hiring.
So you are building in hours and days applications that other banks are spending weeks and months on.
I'd love to see a few case studies!!
Second, that Haskell "has every weakly typed mainstream imperative language as a valid subset" (which I don't think is accurate, because projects are not built from scratch, and libraries match a certain programming style) does not mean that Haskell is worth it. For the sake of argument, suppose Haskell has a 5% improvement in maintenance cost over Python but a 10% higher cost of adoption, then it's not worth it.
> CS did end up adopting functional programming and F# became their strategic choice.
That's great, but doesn't have much to do with Haskell. The investment required for a pure-FP language like Haskell seems to be much greater than the investment in an imperative functional language like F# (or Clojure). It seems like CS, after having evaluated Haskell, decided that it may not be worth it.
Not true. If enough FUD is spread about Haskell, for example it being a failure at CS, then it becomes difficult to convince management.
> If I don't have information, choosing Haskell would be far less rational than not choosing it
No one is suggesting choices be made blindly. Managers should hire good people and then delegate those decisions to those people. If they want to use Haskell, don't stop them due to lack of case studies.
> For the sake of argument, suppose Haskell has a 5% improvement in maintenance cost over Python but a 10% higher cost of adoption, then it's not worth it.
I don't subscribe to nineteenth-century management science. My point was that software development is more about people than technologies. You will never be able to compare Haskell and Python like that, without comparing the people behind the code. What we can say for sure, is that many good developers are asking to develop in Haskell. The best way for other developers to understand why, is to try it.
> That's great, but doesn't have much to do with Haskell.
You were implying that Haskell had failed somehow because SC had stopped using it. Decisions within large organisations are rarely made on purely technical grounds. I can assure you there are many Haskell advocates still there.
Haskell is not yet at an adoption level where anything said about it would qualify as FUD. If I told you I have a great cold remedy and all of my friends really like it, you saying that you may want some more carefully researched information first doesn't qualify as FUD. I don't see why anyone can be expected to make a rational decision based on nothing but marketing from enthusiasts. Haskell is not even at the point where management is the part of the organization that needs convincing.
> If they want to use Haskell, don't stop them due to lack of case studies.
That is precisely the job of responsible management: to stop technical teams from using expensive technologies, that are not only unproven, but hardly have any actual anecdotal information at all. The precise difference between management and developers is that management sees the bigger picture. As a manager I wouldn't mind technical people experimenting with Haskell or any other unproven technology on side projects, though.
> The best way for other developers to understand why, is to try it.
Yes, but when you come out with an experimental, very expensive drug, you don't say, "the best way for people to see if it works is to try it". Instead, some people try it, and then they research and report on their experience. After all, you're not asking me to try a new sorting routine, but a whole new language ecosystem. I understand why some early adopters and PL enthusiasts would be happy to try, but most organizations don't and shouldn't act in this way. There are tons of promising technologies. Is every company supposed to just try each one? Do we all need to try Elixir, Kotlin, Clojure, Scala, Haskell, F#, and Go? Now, it may not be fair, but the fact is that the more expensive the experiment, the more data we'd want before we'd allocate resources to conducting it. A language like Kotlin is nearly free for Java shops to try. Haskell? Not so much. The best way for other developers to understand is, therefore, for the few early adopters to research and report on their experience, with some actual data (cost, duration, size etc.).
> You were implying that Haskell had failed somehow because SC had stopped using it.
Yes, that is a very negative signal. I don't see how choosing F# changes things.
> Decisions within large organisations are rarely made on purely technical grounds.
True, but in general, organizations shouldn't make decisions on "technical grounds", but "bottom line grounds". Even that doesn't always happen, but it generally does, and it's reasonable to guess that if Haskell had had a huge advantage, it wouldn't have been abandoned. Also, remember that the converse is also true. If abandoning a language isn't conclusive evidence that it isn't cost-effective, adopting it isn't conclusive evidence that it is. That's why we need some data. It can be anecdotal data, but it must be actual data, and not a marketing lecture.
> I can assure you there are many Haskell advocates still there.
I'm sure there are. Unfortunately, the number or enthusiasm of advocates is never a substitute for evidence. It's perfectly fine that some early adopters are swayed by the advocates, but it's also perfectly rational not to be swayed by it. I absolutely believe that some people truly enjoy Haskell and feel that it makes them more productive, but without some data, it would simply be irresponsible to pay so much based on marketing alone.
In most cases management isn't qualified to make technical decisions. Managers rarely see the bigger picture.
I've worked on a Haskell project that worked, never crashed, was lightning fast and was meeting deadlines for new features without a problem. Then higher management decided to throw away probably around 7-10 man-years and rewrite the whole thing in Java, quite literally "just because Java". We even had a consultant from FP complete come over to take a look at our codebase and talk in front of management about the advantages of a mature codebase in Haskell. Management didn't even come.
Abandoning Haskell, in my case, had absolutely nothing to do with its technical merits and everything to do with incompetent management. When I left one year after they started rewriting the project in Java (I was staying to help maintain the Haskell code while the transition was happening) everything was buggy as hell, slow and unmaintainable. Not to mention that half the Java guys they hired had already left again.
All I could do is collect metrics like lines of code and intentionally vague experience reports from former colleagues who are still there, but I guess that's hardly gonna convince anybody.
Maybe "FUD" isn't precisely the right term, but Haskell does seem to attract a special level of attacks that isn't applied to other languages at a similar level of maturity.
> Is every company supposed to just try each one? Do we all need to try Elixir, Kotlin, Clojure, Scala, Haskell, F#, and Go?
Well, companies have to make the choice one way or another, and they don't seem willing to pay what it would cost to study these questions to a scientific level. And the decision to stick with "industry best practice" (probably currently Java) is not a safe option either - anyone who does that runs the risk of being outcompeted by those who pick a better lanugage.
> A language like Kotlin is nearly free for Java shops to try.
Oh? To my mind a language that young, backed by a relatively small company, and whose popularity is heavily based on marketing (including a history of repeatedly announcing features before they were actually implemented) would come with a high risk factor.
> The best way for other developers to understand is, therefore, for the few early adopters to research and report on their experience, with some actual data (cost, duration, size etc.).
Sure - but the business incentives don't seem to be there for those early-adopter businesses to publish that kind of report (whether the outcome is positive or negative). How can we make that happen?
It's a level of attack reserved for a special level of unsupported hype. I have absolutely no problem with those who say they really like Haskell, it fits well with how they think, and they feel more productive with it. I even have no problem with those who say that they believe the language has objective benefits in some domains and circumstances and that it deserves a closer look. But if you go around claiming that we can significantly reduce the cost of software development using your language (or any tool), I think an intsy-bitsy amount of even anecdotal evidence is required, especially after two decades.
You may not believe me, but I truly admire Haskell's elegance (even though I believe pure-FP is not the right path for most kinds of software; I have my own pet hypotheses). I just don't like the religious ecstasy that surrounds it.
> and they don't seem willing to pay what it would cost to study these questions to a scientific level.
Because in order for a business to invest in a costly experiment, there should first be some evidence that there's a good chance for the experiment to pay off significantly. Businesses aren't research institution, and they can't afford to invest in researching every new technology that somebody really likes.
> anyone who does that runs the risk of being outcompeted by those who pick a better lanugage.
As soon as the competition pulls ahead, people will be quick to flock to the new language. That's what happened with late Java adopters.
> would come with a high risk factor.
You're right that the risk isn't nonexistent, but it's relatively low (and Kotlin is backed by a larger company than the ones backing most new languages, with the exceptions of Go, Swift and maybe Rust). There's automatic translation and seamless interop, and gradual adoption is very easy. You can pick a class at random and translate it to Kotlin to test if build/debug/profile/deploy etc. works well, and then translate more classes. I'm not sure such a tool exists, but it should be fairly trivial to translate Kotlin back to very readable Java if the need arises to revert for whatever reason.
> but the business incentives don't seem to be there for those early-adopter businesses to publish that kind of report
I think that the business interests are there. If you invest heavily in a language, you'd want it to gain wide adoption so you get more maintenance, more libraries, easier recruiting, more tools and lower risk. The business interests would not be there if Haskell provided an incredible boost in productivity, that is so large that the competitive advantage offsets the downsides of low adoption. I don't think anyone is even making that claim.
How carefully researched was your very public assessment of what happened at CS?
> Do we all need to try Elixir, Kotlin, Clojure, Scala, Haskell, F#, and Go?
Of course not. Those interested in learning more about pure functional programming should try Haskell, the others have different niches they are trying to fill.
> If Haskell had had a huge advantage, it wouldn't have been abandoned
No one is claiming Haskell is some sort of magic bullet for all use cases. If you require ease-of-use, are writing disposable software, or high performance software, then it may not be the best choice.
Nobody is arguing against further studies, but I suspect that most of our industry isn't really interested. As I have said elsewhere, Haskell does not optimise for ease-of-use or aesthetics. It has a growing niche and will continue to be successful in that niche, with or without further productivity studies.
It wasn't. I simply stated that it's a negative signal; that's not FUD. It is a negative signal whether or not I mention it. Mentioning it gives Haskell advocates a chance to address it. You don't have to, but whatever you do, the signal is there. But it's just one signal. I don't think anyone should make a decision based on just one signal.
> Those interested in learning more about pure functional programming should try Haskell, the others have different niches they are trying to fill.
What should those interested in increasing productivity learn? This entire discussion isn't about learning a new programming style, but about achieving a goal of reducing development costs.
> No one is claiming Haskell is some sort of magic bullet for all use cases.
So what is the claim, and is there actual evidence (even anecdotal) to support it?
> I suspect that most of our industry isn't really interested.
The industry is extremely interested in anything that can significantly reduce the cost of developing software. We don't care if it's called Haskell or an omniscient debugger, or a static analysis tool.
> and will continue to be successful in that niche, with or without further productivity studies.
That's terrific, but don't blame the industry for being short-sighted and investing heavily in something that has little evidence of a great promise. You can only blame the industry for ignoring evidence, not for ignoring hype.
Productivity for whom? And over what timescales? As I have been trying to say, it is not that simple, you are oversimplifying.
> So what is the claim, and is there actual evidence (even anecdotal) to support it?
It sounds like you are completely satisfied with existing approaches, in which case there is little point in me trying to convince you.
> The industry is extremely interested in anything that can significantly reduce the cost of developing software.
Industry as a whole, does not have a good track record of adopting even good ideas in a timely manner. c.f. Garbage collection. So again, I think you are oversimplifying.
I'm only simplifying as much as the claims in the article. It's just pretty hard to understand what the claims about Haskell are given all of the vague hype. The article we're discussing says, "Haskell decreases development time by narrowing the scope of development to your domain." Obviously, I understand there are caveats, and it's unclear how big the savings are. I just want reports with some meat.
> It sounds like you are completely satisfied with existing approaches, in which case there is little point in me trying to convince you.
The question of being "completely satisfied" is irrelevant. I'm not a researcher and my goal isn't to dream up better approaches, nor do I have the resources to invest in experimenting with every new tool. I am, however, very interested if someone says they have a tool that can provide significant savings. If there were only some more actionable information, convincing me would be quite easy. I want to be convinced, but throwing general promises around serves to confuse more than enlighten.
Although the study in the paper isn't very practical, it's still an interesting experiment.
I'll settle for well-researched case studies.
> there is a related paper regarding that
That's a step in the right direction, but the paper doesn't discuss software development, but prototyping. We know that "theoretically aesthetic" languages do well in specification and prototyping.
Don't know if it helps, but they have some case studies here https://www.fpcomplete.com/case-studies.
I'd say they have probably seen it. The author of the post himself works at FPComplete, and they use Haskell for their consulting business. That and there are some case studies on their site (I'm fairly sure they all have experience in other languages too).
The same can be said about nearly every language. Haskell is first and foremost a research language motivated by answering a research questions, which is what it is like to program using a certain mathematical formalism that is aesthetically appealing to some. Its designers didn't set out to find the best abstractions etc., but to test the validity of a specific theoretical approach.
> but to that I would say don't knock it until you have tried it.
I'm not knocking it. I'm saying that given how expensive it is to try, perhaps some actual information rather than marketing would do a better job in persuading people to use it. Saying that I'm not persuaded by virtually zero information is no knocking anything, except perhaps the advocacy effort.
> The same can be said about nearly every language.
This clearly isn't true. For example many languages optimise for ease of use, shallow learning curve, programming in the small etc. Some optimise for efficiency and runtime determinism. Features like type-systems are great for software development (programming in the large), but do negatively impact ease-of-use.
> Haskell is first and foremost a research language motivated by answering a research questions
This isn't true. The Haskell committee contains many people from industry. Haskell just happens to have come from academia and is still heavily used in academia.
> Saying that I'm not persuaded by virtually zero information is no knocking anything
That's a huge exaggeration to suggest there is zero information. There are many experience reports published all the time at conferences and in academic journals, going back decades. None however will give you the quantitative metrics you seem to be asking for, because it is almost impossible to do such a comparison.
> Haskell just happens to have come from academia and is still heavily used in academia.
Haskell isn't heavily used in academia at all. My guess is that most CS professors -- like most developers -- have hardly even heard of it, let alone used it. It is heavily used among researchers who's job it is to study the design of languages like Haskell. Languages that are heavily used in academia are C, Python, Java and maybe a few more.
> There are many experience reports published all the time at conferences and in academic journals,
Where? I've looked. If you want to see what a good report looks like, take a look at this one[1]. It's got some background, and then lot's of analysis of pros and cons, with plenty of numbers. I found it so convincing that I tried the language it advocated, and ended up adopting it. I'd seen similar (even more comprehensive) reports for Java back in the early '00s before deciding our team should adopt it. I haven't seen anything similar for Haskell. And again -- the more costly the adoption, the more information required before trying.
> going back decades. None however will give you the quantitative metrics you seem to be asking for, because it is almost impossible to do such a comparison.
I don't want a comparison, and what I'm asking for has been done at some point for nearly every language that has gained widespread adoption. I just want to know how much time and effort has gone into a single project, a breakdown of costs, and a subjective analysis of pros and cons. You see such reports being written from time to time about other languages (although usually with far more adoption than Haskell).
[1]: Slides: http://tla2012.loria.fr/contributed/newcombe-slides.pdf
Report 1: http://glat.info/pdf/formal-methods-amazon-2014-11.pdf
Report 2: http://link.springer.com/chapter/10.1007/978-3-662-43652-3_3
We can certainly say that uncontrolled side-effects, manual memory-management and dynamic types is bad for software development in the large.
I'll stop here because I don't which to argue the semantics of what does are does not constitute significant use or significant research.
We most certainly cannot say that. There's no indication or even a hint that uncontrolled side-effects are a cause of expensive bugs. As a Haskell developer, however, I assume that you consider mutation to be a side-effect (because in pure-FP mutation is a side effect), and it is true that uncontrolled, non-transactional global-state mutation is not such a great idea, but most developers know not to do that in any language, and if you want language-level enforcement on mutation, you can get that for a fraction of the price of Haskell. There's also nothing to suggest that Haskell's approach is a good one (as opposed to, say, Clojure's or Erlang's) or even that Haskell's cure isn't worse than the disease.
> manual memory-management
Absolutely, and you can get automatic memory management for a fraction of the cost.
> and dynamic types
You get static types for a fraction of the cost, too.
> argue the semantics of what does are does not constitute significant use
It's fine for a language to not have wide adoption. That doesn't mean it's not good or even not great. It's also fine -- and I do that, too -- for people to say, I really like that language, I find it elegant, and I feel it helps me develop. What isn't fine is endless hype with claims of significant bottom line benefit but little effort to collect even anecdotal evidence (and I don't mean evidence that Haskell programs can run, or that you really like the language; I mean evidence of the extent of the claimed benefit).
Uncontrolled side-effects destroy the ability to reason about software and when one can no longer reason effectively, bugs will happen. This is obvious to anyone who has worked on large systems. Whether it matters or not, depends on the domain. In finance, it is absolutely critical that we can reproduce a calculation and that for example, the current date is not read from the system deep inside the code somewhere. Having the compiler guarantee this for us is a huge win.
> Absolutely, and you can get automatic memory management for a fraction of the cost. > You get static types for a fraction of the cost, too.
What do you actually know about the cost of Haskell adoption? If you want to accuse me of making unsubstantiated claims, your position would be more tenable without making them yourself!
I have worked on systems comprising 10MLOCs over a career of ~20 years, both as a developer and as a manager, and this isn't obvious to me at all. But I separate mutation from side-effects, because mutation is better defined, and what side-effects mean varies by language. Global mutation is important to control, but there are lots of ways to do that. Other side-effects (and especially IO)? Not so much.
> Having the compiler guarantee this for us is a huge win.
It is almost as trivial to guarantee that in Java, although maybe not during compilation.
> What do you actually know about the cost of Haskell adoption? If you want to accuse me of making unsubstantiated claims, your position would be more tenable without making them yourself!
Oh, while I haven't written much more than Hello, World programs in Haskell (but I have programmed in SML a bit years ago, and a fair bit in Scheme and Clojure), I am no stranger to the theory of pure-FP (simply typed LC w/ polymorphism, functional semantics, monadic composition etc; I'm less familiar with call-by-value semantics, though). It is pretty radically different from most other styles of programming, including imperative-functional, both in terms of coding as well as tools such as debuggers and profilers.
While I could probably stomach giving up purity in a small project, you'd only be able to take away sum types and pattern matching from my cold dead hands.
[1]: This isn't even about a particular language (although different languages do have different cost of adoption). Any language that doesn't have excellent interop with existing libraries and legacy code has a very high cost of adoption (and yeah, we've tried Scala years ago, and the experience was far from good).
I would have liked to see a comparison to other functional languages (say Elixir or OCaml) and not just Java and C#. I'd also argue that picking Java instead of a more agile environment (there are some cool lightweight Java frameworks but most people will associate it with the rather heavy enterprise stack) when comparing regarding time to market is a bit odd. Granted I'm mostly thinking about webapps (but the article mentions Yesode).
Still a nice article (since my post sounds overly negative upon rereading).
Well, the best language for an early-stage business is the one the founder knows, but that's always going to be true. There's no reason that language can't be Haskell.
Having said that, Time-to-Market is only partially influenced by my-code, the biggest part is the code that I don't have to write, i.e. third-party libraries.
In my Haskell adventures I am having trouble finding third-party libraries for even the most popular things, e.g. Cassandra. As far as I can tell there are two libraries, the `cassandra-cql` and the `cql-io`, the first hasn't been updated for a year now, and the second has only 3 stars, which makes me uneasy.
So, although I can see where the author is coming from, I don't think you can beat Java, Ruby, JS or Python in that sense. Unless of course your code/project doesn't have a lot of dependencies.
It's some incredibly productive yak shaving, it's even fun, but it's still yak shaving.
But I'll point that Haskell has also many wonderful libraries, for stuff you won't see in other communities. The libraries that are available are of an unmatched quality. Odds are a library marked as experimental on Hackage is more reliable than a mature one on the Python Index.
F# looks like a more pragmatic choice, and I've been looking hard for a good place to bite off to use it going forward. But again, it'll be bad for the bus factor.
So the best option I've found is to just start using more functional concepts and patterns in Java/C#. C# in particular seems to be leaning in this direction with the features added in the latest language versions.
https://m.reddit.com/r/rust/comments/5dtfp2/haskell_more_pro...
(Tried to ask on HN first, which once again proves the value of subreddits, as small focused communities: https://news.ycombinator.com/item?id=12989041)
This is not going to change.
I love Haskell, and I appreciate the effort many are making to evangelize the language, but I am experienced (ie. cynical) enough to believe that it's never going to become truly mainstream.
Clojure is quite quick to write and time to market can be very fast, as with most lisps, but you pay the price at debug time. We recently shipped a production app that had a small typo that the compiler would have easily caught, but instead it crashed the site when this one particular task was run.
What was the typo?
> it crashed the site when this one particular task was run.
Might that also be because that was the first time the task was run at all, since the code was written or altered?
Code that compiles (such that we are confident it is probably free of silly typos, other than those somewhat rare typos which play out in some way that still compiles), might not be correct; it still requires testing.
Not even a problem in a Lisp-2 like ANSI CL!
(let ((list '(1 2)))
(list list list)) -> ((1 2) (1 2))
Even without a compiler, we can implement a warning for this. The code walker which expands macros is aware of lexical environments and can issue diagnostics when suspicious-looking shadowing is going on, or unbound variables are referenced and such.
Not only is this not the fault of the language being dynamic, but the problem could exist in a static language, like, oh, C:
#define DECLARE_MY_PRINTF int (*printf)(const char *, ...) = my_printf;
{
DECLARE_MY_PRINTF:
printf("hello, %s\n", "world"); // goes to my_printf via shadowing local var
}
Of course, the above depends on the local printf pointer actually having the right type so that the call is well-formed. If we just have "int printf" or whatever, the type system will catch it.
Maybe with F# author would see decrease of development time compared to Haskell?
If you're looking for enlightenment, pick up Haskell, maybe OCaml and skip F#.
However, Haskell and OCaml are the more classic functional, static-typed languages, and while Haskell has more job opportunities than OCaml, it is probably the steepest learning curve and highest time commitment of all these languages.
[1]: https://www.coursera.org/specializations/scalaShort answer: whichever you can get a job in. That's more important than the design differences between them, which are relatively small.
Long answer: I wouldn't use F# because nice as it is you're eventually going to want HKT. Between Scala and Haskell I'd pick based on your learning style - Scala lets you write familiar comfortable code and incrementally adopt functional features, Haskell forces you to dive straight in and do things the Right Way from the start.
Having gone down the Haskell rathole, it's clear to me that the advertising is better than the product. It's complex and difficult to learn and the "less errors because of typing" mantra turns into 'how in the world am I supposed to find my error'. In practice, I very rarely run into errors caused by type problems.
I know nothing about F#, but that's actually my criticism of it. The F# community is small. I occasionally hear about people playing with F#, but even that is rare. (although it could be that I just don't know enough about it and F# and it's community is awesome)
Scala has a reasonably sized community and is being used in production plenty. I personally don't see any benefit beyond the fact that some people like the syntax better than Java and can do less boilerplate (which, if not paying attention could lead to higher resource consumption).
If I were looking for something new, I'd look into Go and Rust. Both have strong followings, a lot of active development to learn from, and strong job markets moving forward (IMO). Or alternatively reverse-expand back into C and start playing with embedded RTOSs - which has the benefit of both being very interesting and an area where senior people could make a significant impact.
... Now, if you really wanted one of the three - I would choose Haskell for learning, Scala for job market, F# if you are already a .net developer.
There is more professional work being done in Scala, and it makes it more palatable to managers that it runs on the JVM and allows them to leverage some of the existing knowledge of their Java programmers.
In the financial world, however, F# use is on the rise.
I currently work in F# but not in finance, and the lack of mature libraries for common things (like JSON serialization!) makes it a major pain in the behind.
Were I told to select a functional language, "quick!", I would select Clojure when caffeinated, and Javascript (ES7) when not.
FWIW: I subscribe to the notion that dynamic typing with unit testing will get you a correct and working program faster than static typing. In the case of Clojure, you now have clojure.spec, which means you can write conformance criteria as functions and generate tests and test data.
http://games.greggman.com/game/dynamic-typing-static-typing/