Hacker News Reader

Comment by msie | Hacker News Reader

I always thought that introducing keys in function parameters was a mistake.

Not having keys in function parameters is a mistake.

Swift does have an unconventional approach to function keys though, wherein you can specify an "external" label and an "internal" label:

  func greeting(for person: String) -> String {
      "Hello, " + person + "!"  // the argument is "person" inside the function body...
  }
  print(greeting(for: "Dave"))  // ... but it's "for" when calling the function

I'm not sure if I like it or hate it, but it is cool.

I love the internal/external labeling. With a little thought it can make for code that reads more nicely than is possible otherwise.

JavaScript has that with:

    const hello = ({ for: person }) => `Hello ${person}`

    hello({ for: "Dorian" })

    'Hello Dorian'

Javascript's one is a little bit different in that you are renaming the destructured variable - it's not a function-specific thing - whereas Swift's version is a property of the function.

e.g.

  const user = {id: "123"};
  const { id } = user;          // This extracts the `id` field
  const { id: userId } = user;  // This also extracts the `id` field, but renames it to userId
  console.log(userId)

It does indeed, but the cost is an object allocation for every function call.

it's one of those small things that are also brilliant in hindsight

Objective-C sort of had it unofficially through how methods tended to be named.

I’m glad they found a nice way to do it in Swift.

crystal has it too, and i agree, it's a very nice feature

Named parameters and nothing new and I like them. To this day I don’t understand why swift/objc have two names. I get that one is external, one internal so changing external breaks APIs. So is the only advantage that you can rename the parameter name for function implementation? So what? You can do that by a variable at the top of the function. Those two names make function signatures hard to read (which one is which etc)

It’s one of the things I miss most in Kotlin.

Even C/C++ have the internal/external mismatch:

// in greeting.h

void greeting(char* for);

// in greeting.c

void greeting(char* person) { ... }

For me, I felt the same way initially, but I found it depends on the case. The case you gave is probably now one I would do, I would probably have the external be 'for' if I expected the caller to pass in a list that I was gonna iterate over inside of the method.

An example I like to use to show when and how I used it. I have been playing with Metal. If I write a function called translate, I will have it's external parameter as 'by' with the internal parameter being 'vector'. As a simple example:

    func translate(_ origin: simd_float3, by vector: simd_float3) -> simd_float3 {
      // do some stuff
    }

    let newVector = translate(originalVector, by: translatingVector)

I disagree, I think the more self documenting the code is, the better. It's also nice with code completion because when I start typing a function, I'm given a list of all the overloads for that function, and when I pick one I get all the arguments sitting there ready for me to fill in.

You still get that with traditional languages that don’t use keywords like java. With Swift you have to read a lot more to get the right function since many vary in the number and type of parameters.

I disagree. It makes the code readable.

It's still a tradeoff: in a previous team switching languages it openned the door to function with 5+ parameters, when that was highly throwned upon otherwise.

It's a tool to help better manage complexity, and that naturally invites more complexity as well.

is that an issue of the language or of the team?

adding additional names does not add additional complexity. complexity arises in interactions between components.

I see it as mostly influenced by the language and tooling. From there, some people will be more attracted by one side or the other, but when moving stack you tend to follow the stack's natural incline.

For instance ruby helps to have shorter, less verbose code, but with more hidden components. Rails rose from there, tooling tends to support that style. I've seen java devs moving to ruby and naturally writing less verbose code.

In contract Java tooling makes it a lot more manageable to write 40 character long classes, deeper dependency trees and injections and potentially auto-generate half of the needed methods. The cost is low, and there's less incentive to fight that trend when tools help you abstract part of it.

i agree but you discuss verbosity and not complexity

Fair point.

You're right that total complexity across the system fundamentally doesn't change (it's up to the devs and the problems at hand). Local complexity in each part of the system can vary, but the whole will probably always be least as complex as it needs to be.

The best function calling syntax I've ever used is the video scripting language AVISynth, which is mostly key based. It's basically Swift but more flexible and informal.

http://avisynth.nl/index.php/Grammar#Functions.2C_Filters_an...

Reply on news.ycombinator.com