Why GitHub used Haskell for Semantic • Hacker News Board

Re: Why GitHub used Haskell for Semantic

#211

Post by spopejoy » Mon, Jun 10, 2019, 1:59 AM UTC

> it's worth mentioning that Semantic, as a rule, does not encounter runtime crashes: null pointer exceptions, missing-method exceptions, and invalid casts are entirely obviated, as Haskell makes it nigh-impossible to build programs that contain such bugs.

This I think is the real drug of Haskell.

There are lots of challenges involved with moving to Haskell for production software, and the article notes some of them. But once your code builds and you ship -- man, there is nothing remotely like it. v1.0 software running for weeks in prod without a bug.

Then, the super-sauce is: maintainers don't break your already-working code. Sure new code might have bugs, and deep semantic changes to code can break anything, but workaday fixes to one corner of the codebase simply can't break all the other corners. v2, v3, v4 all migrate to production with none of the working stuff falling over.

PS I don't buy that the "control flow" feature couldn't have been done in Java -- however I would bet huge dollars that you couldn't do it in a way that wouldn't require massive investment in maintainers understanding a complex and subtle data-driven control flow pattern that any single dev could easily subvert, intentionally or not. I was a huge fan of FP-style coding in Java, but it required buy-in from all maintainers; as soon as somebody got lazy and went back to mutability etc there was nothing to stop them.

Re: Why GitHub used Haskell for Semantic

#212

Post by emilypi » Mon, Jun 10, 2019, 3:54 AM UTC

spopejoy wrote: post #211

> it's worth mentioning that Semantic, as a rule, does not encounter runtime crashes: null pointer exceptions, missing-method exceptions, and invalid casts are entirely obviated, as Haskell makes it nigh-impossible to build programs that contain such bugs. This I think is the real drug of Haskell. There are lots of challenges involved with moving to Haskell for production software, and the article notes some of them.…

There's another "real drug" hidden in the subtext of your post and Semantic's post, right in between

"Sure new code might have bugs, and deep semantic changes to code can break anything..."

and

"Its language features allow concise, correct, and elegant expression of the data structures and algorithms we work with."

- the notion that fundamental abstractions (not just the ones someboy thought up like the Gang of Four) compose well. This is something that we've seen the industry slowly migrating towards in the form of functional JS flavors like typescript and purescript, Java's lambdas, Scala's Cats and so on.

The ability to refactor and as importantly be confident that your refactor is not messing with the semantics of anything upstream is a hidden feature of static types applied to expressive abstractions.

So far, I have only seen this achieved with Haskell's libraries, type system, and reliance on fundamental mathematical concepts while also being able to avoid any "there's something rotten in Denmark" moments. The content in the Hackage ecosystem is a wonder of the modern programming world, warts and all.

There are many rough edges to the development process, as always, but at least in Haskell you can limit the language's contributions to that edgeset. It actually makes you want to produce good code and be a better developer as a result!

Re: Why GitHub used Haskell for Semantic

#213

Post by yakshaving_jgt » Mon, Jun 10, 2019, 12:26 PM UTC

solidsnack9000 wrote: post #209

yakshaving_jgt wrote: post #195

Earlier quoted context omitted.

> No one is arguing that you need to be a genius to use Haskell. I’ve come across this sentiment so many times. Haskell definitely has a reputation for being an “ivory tower” language. > How does my argument come across that way? I’m not too familiar with Zelda, but it sounded like you were saying “some people are just able to do these things that most others can’t.” If that isn’t what you were saying, then I am sorr…

I think anybody can write Haskell and anybody can play Zelda or similar characters in fighting games. What the skill-cap / skill-floor thing is about, is how often do people bother. When the base level of skill required to play at all -- not necessarily play well -- is really high, often those characters don't get used as much. People find a character that demands less up front investment and play that character inst…

Ok, I think that's fair. I won't deny there was a significant time investment on my end.

Thanks for clarifying.

Re: Why GitHub used Haskell for Semantic

#214

Post by develop7 » Mon, Jun 17, 2019, 2:43 PM UTC

_bxg1 wrote: post #4

I'm really curious why Haskell has seen so little adoption in industry. Is it just the difficulty? Or a chicken-and-egg effect with tooling and libraries? One thing I've wondered is if it actually isn't ideal for a lot of cases. FP is beautiful for certain things. But in some domains (or pieces of domains), state and mutation aren't just unfortunate implementation details, but a core element of the problem space. For…

> Is it just the difficulty?

It's loads to unlearn for most of programmers and a perfectly natural knee-jerk reaction of a refusal to. I mean, the more one knows already, the more is the desire to reuse such knowledge. Well, at least I have been there.

> Haskell doesn't, really (from my cursory reading about it).

Of course it does, https://en.wikibooks.org/wiki/Haskell/Mutable_objects looks a good start.