4 ms·

Ah. So what isn't Haskell good for?

by pekk 13y ago

anon_d 13y ago
Haskell is not a great language for programs that need to fit within tight memory or latency limits. It's also slower in some cases than C/C++, but it's definitely competitive with other high-level languages. Other than that, it's pretty excellent for everything.
jerf 13y ago
Well, in all seriousness, a situation in which a mind-expanding language isn't called for, because the minds in question don't currently work that way, and I am being 100% serious. I really like Haskell. It's among my favorite languages. However, I don't advocate for it where I work and tend to sort of tamp down on anyone else who tries because despite the fairly polyglot nature of where I work, we don't have the background to use Haskell very effectively at scale. I do this even as I advocate to all and sundry that they learn it anyhow for the mind expansion. When one is fluent in Haskell I think the language naturally tends to drive you towards well-factored, reusable designs that can handle change well. I say "fluent" because I don't think one needs to be "expert" to get to this level, but certainly there's a level of experience required to be here. Beginners and pre-fluency intermediates in the language produce god-awful messes of code. (I know, because I have! Fortunately I had the foresight to do that to throwaway personal projects, rather than critical business projects.) I think a great deal of the Haskell propaganda is justified, but beware the Haskell enthusiast who recites the propaganda, but can't clearly explain in their own words the mechanisms that Haskell uses to accomplish it. Actually, scratch Haskell out of that. Beware any tech enthusiast who can't give a coherent explanation in their own words of why the tech stack has the claimed effects, but just repeats the propaganda again.
- AnimalMuppet 13y ago
  There's some sound wisdom in this post, jerf. A couple of questions, though: First, is it Haskell that drives you toward well-factored, reusable designs? Or is it functional programming? (I know, you can't do Haskell without functional programming. You can do functional programming without Haskell, though. Can you separate the two, and say which one is driving you here?) Second, have you ever seen Haskell deployed in a large, long-lived production code base? How large? How long-lived? How did it work? (I ask because I remember reading an article on Lisp, which said that X in Lisp is just like Y in an object-oriented language, "with the proviso that in Lisp, all the types exist in the programmer's head". I'm pretty sure that won't work very well on a million-line code base that lives for 30 years. Haskell has a very solid type system, so it won't run into that particular problem but... does it have a track record for large-scale work out in the real world?)
  - jerf 13y ago
    "First, is it Haskell that drives you toward well-factored, reusable designs? Or is it functional programming?" In my opinion, it's the purity, and the general simplicity of the individual pieces which forces you in certain directions. Since this conversation is dead-ish now, it's probably not a big deal to link to my longer musing on this: http://www.jerf.org/iri/post/2908 http://www.jerf.org/iri/post/2908 Even functional languages that make impurity easy and available end up making it too easy to jump down the escape hatch. "Second, have you ever seen Haskell deployed in a large, long-lived production code base?" I personally have not. As I said, where I work I tend to tamp down on any efforts to use Haskell, and there's not many! But I trust the people who have and do, and who knows, perhaps may join them someday.

Ah. So what isn't Haskell good for?

by pekk 13y ago

Ah. So what isn't Haskell good for?

Haskell is not a great language for programs that need to fit within tight memory or latency limits. It's also slower in some cases than C/C++, but it's definitely competitive with other high-level languages. Other than that, it's pretty excellent for everything.

jerf 13y ago

Well, in all seriousness, a situation in which a mind-expanding language isn't called for, because the minds in question don't currently work that way, and I am being 100% serious. I really like Haskell. It's among my favorite languages. However, I don't advocate for it where I work and tend to sort of tamp down on anyone else who tries because despite the fairly polyglot nature of where I work, we don't have the background to use Haskell very effectively at scale. I do this even as I advocate to all and sundry that they learn it anyhow for the mind expansion. When one is fluent in Haskell I think the language naturally tends to drive you towards well-factored, reusable designs that can handle change well. I say "fluent" because I don't think one needs to be "expert" to get to this level, but certainly there's a level of experience required to be here. Beginners and pre-fluency intermediates in the language produce god-awful messes of code. (I know, because I have! Fortunately I had the foresight to do that to throwaway personal projects, rather than critical business projects.) I think a great deal of the Haskell propaganda is justified, but beware the Haskell enthusiast who recites the propaganda, but can't clearly explain in their own words the mechanisms that Haskell uses to accomplish it. Actually, scratch Haskell out of that. Beware any tech enthusiast who can't give a coherent explanation in their own words of why the tech stack has the claimed effects, but just repeats the propaganda again.

AnimalMuppet 13y ago
There's some sound wisdom in this post, jerf. A couple of questions, though: First, is it Haskell that drives you toward well-factored, reusable designs? Or is it functional programming? (I know, you can't do Haskell without functional programming. You can do functional programming without Haskell, though. Can you separate the two, and say which one is driving you here?) Second, have you ever seen Haskell deployed in a large, long-lived production code base? How large? How long-lived? How did it work? (I ask because I remember reading an article on Lisp, which said that X in Lisp is just like Y in an object-oriented language, "with the proviso that in Lisp, all the types exist in the programmer's head". I'm pretty sure that won't work very well on a million-line code base that lives for 30 years. Haskell has a very solid type system, so it won't run into that particular problem but... does it have a track record for large-scale work out in the real world?)
- jerf 13y ago
  "First, is it Haskell that drives you toward well-factored, reusable designs? Or is it functional programming?" In my opinion, it's the purity, and the general simplicity of the individual pieces which forces you in certain directions. Since this conversation is dead-ish now, it's probably not a big deal to link to my longer musing on this: http://www.jerf.org/iri/post/2908 http://www.jerf.org/iri/post/2908 Even functional languages that make impurity easy and available end up making it too easy to jump down the escape hatch. "Second, have you ever seen Haskell deployed in a large, long-lived production code base?" I personally have not. As I said, where I work I tend to tamp down on any efforts to use Haskell, and there's not many! But I trust the people who have and do, and who knows, perhaps may join them someday.