Hacker News Reader

Comment by peeters | Hacker News Reader

IMO this is the biggest thing available to modern languages that Java is missing. I would absolutely love to see this, particularly pattern decomposition. I wonder if you could do it without something analogous to Scala's sealed classes though--you really want your type checker to be able to assert every match has considered every branch (without having to specify a "default" everywhere). That means you need to be able to mark classes as not-dynamically-extendible, so the type checker has the full set of subtypes available.

Edit: Just got to the bottom of the article. Looks like sealed hierarchies is exactly what they explore.

View on HN

I mean, reading this article is basically a giant advertisement for Scala and its powerful pattern matching.

  functionThatReturnsAnyVal match {
    case i: Int if i > 2 => System.out.println("Greater than 2")
    case i: Int => System.out.println("Less than or equal to 2")
    case s: String => System.out.println("Hello, " + s)
    case _ =>
  }

This still seems a terrible example just to try and avoid naming an object. Isn't the comparison to:

    Object o = functionThatReturnsAnyVal();
    if (o instanceof Integer) {
        Integer i = (Integer) o;
        if (i > 2) {
           System.out.println("Greater than 2.");
        } else {
           System.out.println("Less than or equal to 2.");
        }
     } else if (o instanceof String) {
        String s = (String) s;
        System.out.println("Hello, " + s);
     }

I get that the case syntax is kinda nice, but this particular example just doesn't seem to get there for me. Roughly half the lines, which is good. None of them hard to reason about. Which makes it a wash.

Or is the comparison to something else?

> I get that the case syntax is kinda nice, but this particular example just doesn't seem to get there for me. Roughly half the lines, which is good. None of them hard to reason about. Which makes it a wash.

I suppose it's a matter of opinion, but to me the comparison of these two snippets is nowhere near "a wash". The readability of the former is leaps and bounds ahead of the java version, and it will only be more apparent as the example grows in complexity.

As either example grows in complexity, they both wash down to a check and a call to another method to do the heavy lifting. In that case, the length of the methods will not grow apart from each other heavily.

And let me be clear. My gripe is only with the example. Not the idea. I happen to like pattern matching, but have never used it like those. Usually I use it to tease apart data by parts. (though, to be fair, I have found I use it less than I thought I did. Not sure why...)

In reality my function would rarely exist in the Scala world since having a weakly-typed return like that is cringeworthy to say the least. But it's better to see this in action with Scala options.

  val myOptionalVar: Option[String] = functionThatReturnsOption()

  myOptionalVar match {
    case Some(str) if str.length > 10 => System.out.println("This is a long string.")
    case Some(str) => System.out.println("This is a short string.")
    case None => System.out.println("This code would be fifteen lines of null checking in Java.")
  }

With Java 8 having Optional and Lambdas, this is no longer the case:

  final Optional<String> myOptionalVar = functionThatReturnsOptional();
  
  final String output = myOptionalVar
    .map(str -> str.length() > 10
      ? "This is a long string."
      : "This is a short string.")
    .orElse("This code would be fifteen lines of null checking in Java.");
  
  System.out.println(output);

I think this is less-rare than you think. In my Scala code, I have a lot of matching on weakly-typed values which come from parsing JSON objects (where I also deal with a lot of optional values).

I think the prettier syntax actually really helps write code faster. A similar form of syntactic sugar appeared with lambda expressions. Instead of needing to explicitly subclass and write out the overriding method, you can just create an anonymous function.

It's way less code and it makes the resulting code easier to reason about. The code's intent is more elegantly conveyed.

(String) o;

Fair. :) I was much more worried about if I had the required spaces at the beginning of the line, than if I wrote this correctly. Still a static error, though, that the compiler would catch.

> reading this article is basically a giant advertisement for Scala and its powerful pattern matching.

Javas strength has always been to take the good parts of its competitors after they've checked out in production (and not just "wouldn't this be a great idea ..?") and implement them. It will probably never be ahead of the curve due to this, but at least it is remarkably free of "looks good in theory, useless in reality"-features.

In general, I agree with all but your last point. There are many things in Java that were good ideas at the time of introduction, but turned out horrible (serializable, finalize, etc.)

Fair enough. I think both of your examples exist since JDK 1.0 (or 1.1), though I'd have to look to be sure. I think they've wised up after that.

Type erasure for generics...

Wrong. Type erasure for generics is one of Java's luckiest breaks. At worst, it imposes a tiny inconvenience, but in exchange this is what allowed Java not to bake a variance model into the VM and libraries, and this is precisely what allows languages with different variance models -- like Java, Kotlin and Clojure -- to both share code and actual data objects with no runtime conversions. This is something that can't happen on .NET (see, e.g. how clumsily their Python implementation interacts with generics). This was a huge win for almost no cost.

I completely agree, what a blessing in disguise. Scala wouldn't be possible without it.

It was designed in part by Phil Wadler, so it's not surprising it's a success.

You mean it's a giant advertisement for ML and its powerful pattern matching?

> IMO this is the biggest thing available to modern languages that Java is missing.

It always amuses me to see features of ML, a language from 1973, described as "modern"; e.g. algebraic datatypes, pattern-matching, type inference, parametric polymorphism, etc.

C (from which many popular languages like C++, Java, C#, PHP, etc. are derived) came out in 1972.

I wouldn't say it's a case of being "modern", so much as paying attention to what else has been tried before, rather than sticking with what one already knows (when designing a language).

I didn't use "modern" to describe features, I used it to describe languages. Modern languages have a tendency toward multi-paradigm feature inclusion. So while the features themselves aren't new, combining well-received features from other languages and paradigms is largely a characteristic of modern languages. And newer languages have the upper hand here since it's easier to include a feature from the start than it is to retrofit it into a language.

I wonder what the programming world would look like now if functional programming concepts had broken through in the 80s and 90s, instead of OO. C would no doubt still be C, filling the same minimalistic imperative close-to-iron niche, but what about higher-level languages?

The move towards Interactive Applications has helped OO and scripting languages. There are still plenty of applications for functional programming in the most demanding of environments, for safety and high availability. However, most students are not good enough at math to learn functional programming concepts. Whether that's the teachers' fault or the students' fault or society's fault, I can't say. But if we paid programmers for safety critical systems much more than we pay for good front-end developers, you would see a much higher level of adoption of functional programming languages.

> However, most students are not good enough at math to learn functional programming concepts.

When does this misconception disappear?

You don't need to be good at math to use functional programming. It's nice that there is a correspondence between math and FP, but it's mostly irrelevant when coding. You could as well say you need a FP background when learning math. Both statements are nonsense and usually spread by people who mostly read complicated blogposts instead of writing actual code using FP.

FP just offers a nice bunch of intuitive and predictable ways of processing information.

Take Lisp as a perfect example, I doubt you'd find very many people who think that you need to be strong at maths to learn Lisp. I wonder if this misconception came about because of Haskell's (unfounded) reputation of being a "scary abstract maths language".

> However, most students are not good enough at math to learn functional programming concepts.

I don't agree with this. Many programming concepts outside functional programming can be very hard to grasp, yet we expect students to pick them up, and we expect front-end Web devs to use (some of) them every day.

Some examples off the top of my head:

- Pointers (e.g. the many incompatible meanings of "" in C).

- Classes and instances.

- Implicit state.

- Mutable global state!

- Concurrency (events, continuations, actors, etc.).

- Multithreading. Yes, it's a form of concurrency, but it also offers a unique combination of being a) the default go-to strategy for implementing concurrency, and b) so fundamentally hostile to any attempts at understanding.

- `x = x + 1`. This can be completely baffling when first encountered!

In this context, functional programming doesn't really require anything particularly difficult. Maybe recursion is hard to grasp, but that's not unique to FP; FP just uses it more often, making it harder to avoid. The only thing I can think of that's pretty much a requirement in FP, that we don't already require everyone to grasp, is first-class functions; and they're already pretty widespread in scripting languages.

What about monads? What about denotational semantics? What about cartesian closed categories? What about all that other complicated math stuff? None of it is needed to do FP*! Just grab a small Scheme interpreter, ignore "call/cc" or anything ending in "!", and start coding.

Let's look at, say, the accepted papers from the last OOPSLA (Object-Oriented Programming, Systems, Languages & Applications): http://2016.splashcon.org/track/splash-2016-oopsla#event-ove...

Looking through the titles, I don't know what an "enclave" is, or "conditional future synthesis", or "first-class effect reflection". I have no idea what a "non-linearizable concurrent object" is, or what constitutes a "dependent object type". Does that mean I'm not good enough at math to learn OOP? Not at all; I can just run `python` and start hacking.

Oops, I meant the incompatible meanings of the asterisk in C; HN swallowed it as markup, and emphasised the rest of the comment.

Half the world runs on Excel, a functional programming environment. If your average manager is mathy enough to understand functional programming, the average programmer should be as well.

I retrofitted F# style "discriminated unions" (which are basically sealed heirarchies) into C# by creating a series of generic types `OneOf<T0, ..., Tn>` which can hold exactly one value,

Each type has a `.Match` and `.Switch` methods, in to which you have to pass lambdas to handle each case `.Match(Func<T0, TResult, ..., Func<Tn, TResult>`.

I don't know if this would work in Java, given the generic type erasure, but it might...

1. https://github.com/mcintyre321/OneOf

Reply on news.ycombinator.com