CS SYD - Ad-hoc polymorphism erodes type-safety
cs-syd.eu
This blogpost explains and argues the claim that Ad-hoc polymorphism (Type-classes in Haskell/Scala/Purescript, Traits in Rust, Interfaces in Go/Java) makes code less type-safe.In other words: ad-hoc polymorphism makes it so that sometimes, after a refactor, code that is wrong and would not type-check without it, now still type-checks.
    • armchair_progamer@programming.devOPM
      3·
      10 months ago

      Plus the issue remains if the author replaced Vec with Vec of Vec instead of Option of Vec, or if iter().count() meant something completely different than “count # of iterated elements” (both pointed out in HN comments).

  • davawen@programming.devEnglish
    3·
    10 months ago

    ???
    I fail to see the point of this blog post.
    The example given makes no sense (maybe because it’s very simplistic, and a more complicated one would show the point better?), you would NEVER use .iter().count() if you had direct access to the Vec. The iterator is more general in this sense.

    You would use it if, say, your Settings struct was generic over an Iterator type, and in that case it’s the whole point, isn’t it?? Like what???

    Plus, I wouldn’t say this erodes type safety, it’s a lot more like a logic error.

    • armchair_progamer@programming.devOPM
      1·
      10 months ago

      Yeah this one’s a miss on my end. I saw “ad-hoc polymorphism is UNSAFE?” and well, it does a better job reinforcing that ad-hoc polymorphism is not unsafe.

      Author should have wrote a piece “how even type-safe programs can fail” and used his example to show that. Because what this really shows is that type-safety doesn’t prevent programs with the correct types but bad semantics. But that’s not ad-hoc polymorphism; it can happen anywhere (sans ultra-specific types) including even the author’s workaround if he used Vec<Vec instead