www.digitalmars.com         C & C++   DMDScript  

digitalmars.D - [OT] Some neat ideas from the Kotlin language

reply Yuxuan Shui <yshuiv7 gmail.com> writes:
Just come across Kotlin today, and found some interesting ideas 
skimming through its tutorial:

1) Null check

Kotlin has Optional types, suffixed with a '?'. Like 'Int?', same 
as in Swift. But instead of explicitly unwrapping them (e.g. var! 
in Swift, or var.unwrap() in Rust), Kotlin let you do this:


     var: Int?
     if (var != null)
         //You can use var here


Skipping the null check will get you compile time error.

You can even do this:

     if (var == null) return;
     //You can use var from now on

2) Smart cast

This is a similar to previous one, instead of:

     var1: Object;
     var2 = cast(String)var1;

You do this:

     if (var1 is String)
         //You can use var1 as a String here


I think this two ideas are pretty neat, for more information, see:

http://kotlinlang.org/docs/reference/typecasts.html
Feb 18 2016
next sibling parent reply Jacob Carlborg <doob me.com> writes:
On 2016-02-19 00:33, Yuxuan Shui wrote:
 Just come across Kotlin today, and found some interesting ideas skimming
 through its tutorial:

 1) Null check

 Kotlin has Optional types, suffixed with a '?'. Like 'Int?', same as in
 Swift. But instead of explicitly unwrapping them (e.g. var! in Swift, or
 var.unwrap() in Rust), Kotlin let you do this:


      var: Int?
      if (var != null)
          //You can use var here


 Skipping the null check will get you compile time error.

 You can even do this:

      if (var == null) return;
      //You can use var from now on
I think the same applies to Swift. But I think you're supposed to use it like this: if let a = var { // a is no unwrapped } I like the way it's used in Scala: val a = Option(value) val b = a.map(e => e /* do something with the unrapped e).getOrElse(/* use a default value here */) The Option type in Scala acts like a zero or one element collection. The Scala way can be implemented in D as well without any language support.
 2) Smart cast

 This is a similar to previous one, instead of:

      var1: Object;
      var2 = cast(String)var1;

 You do this:

      if (var1 is String)
          //You can use var1 as a String here
It's similar how it works in D, for Objects: class Foo {} Object foo = new Foo; if (auto o = cast(Foo) foo) // use o as Foo here When casting to a subclass it will return null reference if the cast fails. -- /Jacob Carlborg
Feb 19 2016
next sibling parent reply Nemanja Boric <4burgos gmail.com> writes:
On Friday, 19 February 2016 at 12:16:49 UTC, Jacob Carlborg wrote:
 I like the way it's used in Scala:

 val a = Option(value)
 val b = a.map(e => e /* do something with the unrapped 
 e).getOrElse(/* use a default value here */)

 The Option type in Scala acts like a zero or one element 
 collection. The Scala way can be implemented in D as well 
 without any language support.
IMHO, without language support looks really bad - `map` really sticks out here. What Rust is doing: ``` let foo: Option<i32> = bar(); let new_stuff = match foo { Some(x) => x, None => 0 } ``` (or similar, I don't have compiler handy).
Feb 19 2016
next sibling parent Thiez <thiezz gmail.com> writes:
On Friday, 19 February 2016 at 12:28:14 UTC, Nemanja Boric wrote:
 What Rust is doing:

 ```
 let foo: Option<i32> = bar();

 let new_stuff = match foo {
    Some(x) => x,
    None => 0
 }
 ```

 (or similar, I don't have compiler handy).
I think most would write that as: let new_stuf = bar().unwrap_or(0); When you use the methods on `Option` you rarely need to match.
Feb 19 2016
prev sibling parent Jacob Carlborg <doob me.com> writes:
On 2016-02-19 13:28, Nemanja Boric wrote:

 IMHO, without language support looks really bad - `map` really sticks
 out here.

 What Rust is doing:

 ```
 let foo: Option<i32> = bar();

 let new_stuff = match foo {
     Some(x) => x,
     None => 0
 }
 ```
You can do the same in Scala, it has built-in pattern matching. Although I think the map approach is preferred, but I'm not a Scala expert. -- /Jacob Carlborg
Feb 21 2016
prev sibling next sibling parent Nick Sabalausky <SeeWebsiteToContactMe semitwist.com> writes:
On 02/19/2016 07:16 AM, Jacob Carlborg wrote:
 On 2016-02-19 00:33, Yuxuan Shui wrote:
 Just come across Kotlin today, and found some interesting ideas skimming
 through its tutorial:

 1) Null check

 Kotlin has Optional types, suffixed with a '?'. Like 'Int?', same as in
 Swift. But instead of explicitly unwrapping them (e.g. var! in Swift, or
 var.unwrap() in Rust), Kotlin let you do this:


      var: Int?
      if (var != null)
          //You can use var here
Me want.
 2) Smart cast

 This is a similar to previous one, instead of:

      var1: Object;
      var2 = cast(String)var1;

 You do this:

      if (var1 is String)
          //You can use var1 as a String here
It's similar how it works in D, for Objects: class Foo {} Object foo = new Foo; if (auto o = cast(Foo) foo) // use o as Foo here When casting to a subclass it will return null reference if the cast fails.
It'd be nice if it didn't require a new variable name though. When I do that, I usually find I wind up needing to resort to some hungarian-notation-inspired "start including the type in the variable's name" verbosity. It's a fantastic feature of D, but that Kotlin version seems much nicer.
Feb 19 2016
prev sibling parent reply Yuxuan Shui <yshuiv7 gmail.com> writes:
On Friday, 19 February 2016 at 12:16:49 UTC, Jacob Carlborg wrote:
 On 2016-02-19 00:33, Yuxuan Shui wrote:
 Just come across Kotlin today, and found some interesting 
 ideas skimming
 through its tutorial:

 1) Null check
 [snip]

 You can even do this:

      if (var == null) return;
      //You can use var from now on
I think the same applies to Swift. But I think you're supposed to use it like this: if let a = var { // a is no unwrapped }
But you would need to give the unwrapped value a new name.
 I like the way it's used in Scala:

 val a = Option(value)
 val b = a.map(e => e /* do something with the unrapped 
 e).getOrElse(/* use a default value here */)

 The Option type in Scala acts like a zero or one element 
 collection. The Scala way can be implemented in D as well 
 without any language support.

 2) Smart cast
 [snip]
It's similar how it works in D, for Objects: class Foo {} Object foo = new Foo; if (auto o = cast(Foo) foo) // use o as Foo here When casting to a subclass it will return null reference if the cast fails.
The point is, in other languages you would have to explicitly do an operation to the wrapped value, but in Kotlin the compiler is able to automatically unwrap it for you when it knows doing so is legal, based on static analysis results.
Feb 19 2016
parent reply Daniel N <ufo orbiting.us> writes:
On Friday, 19 February 2016 at 21:30:37 UTC, Yuxuan Shui wrote:
 I think the same applies to Swift. But I think you're supposed 
 to use it like this:

 if let a = var {
   // a is no unwrapped
 }
But you would need to give the unwrapped value a new name.
Swift also has guard, which solves this issue. func ex(a: Int?) { // unwrap (possible to add additional check if desired) guard let a = a where a > 0 else { return // fail } a.ok // a is now a plain validated Int }
Feb 20 2016
parent Steven Schveighoffer <schveiguy yahoo.com> writes:
On 2/20/16 5:04 AM, Daniel N wrote:
 On Friday, 19 February 2016 at 21:30:37 UTC, Yuxuan Shui wrote:
 I think the same applies to Swift. But I think you're supposed to use
 it like this:

 if let a = var {
   // a is no unwrapped
 }
But you would need to give the unwrapped value a new name.
Swift also has guard, which solves this issue. func ex(a: Int?) { // unwrap (possible to add additional check if desired) guard let a = a where a > 0 else { return // fail } a.ok // a is now a plain validated Int }
I'm glad you pointed this out, I did not know this. Having to rename variables all the time is quite annoying. -Steve
Feb 20 2016
prev sibling next sibling parent reply =?UTF-8?Q?Tobias=20M=C3=BCller?= <troplin bluewin.ch> writes:
Yuxuan Shui <yshuiv7 gmail.com> wrote:
 Just come across Kotlin today, and found some interesting ideas 
 skimming through its tutorial:
 
 1) Null check
 
 Kotlin has Optional types, suffixed with a '?'. Like 'Int?', same 
 as in Swift. But instead of explicitly unwrapping them (e.g. var! 
 in Swift, or var.unwrap() in Rust), Kotlin let you do this:
 
 
     var: Int?
     if (var != null)
         //You can use var here
In Rust that would be: let var : Option<i32> = ...; if let Some(var) = var { // You can use var here } It works for every enum (= tagged union), not just Option<T> Swift also has "if let". It's not much more verbose but more explicit. Changing the type of a variable based on static analysis is just advanced obfuscation. It hurts readability and the gain is questionable. At least it only works for nullable types. Tobi
Feb 20 2016
next sibling parent reply rsw0x <anonymous anonymous.com> writes:
On Saturday, 20 February 2016 at 09:40:40 UTC, Tobias Müller 
wrote:
 Yuxuan Shui <yshuiv7 gmail.com> wrote:
         [...]
In Rust that would be: let var : Option<i32> = ...; if let Some(var) = var { // You can use var here } It works for every enum (= tagged union), not just Option<T> Swift also has "if let". It's not much more verbose but more explicit. Changing the type of a variable based on static analysis is just advanced obfuscation. It hurts readability and the gain is questionable. At least it only works for nullable types. Tobi
D has this too, but only for nullable types afaik. if(byte* ptr = someFunc()){ //... }
Feb 20 2016
parent reply =?UTF-8?Q?Tobias=20M=C3=BCller?= <troplin bluewin.ch> writes:
rsw0x <anonymous anonymous.com> wrote:
 On Saturday, 20 February 2016 at 09:40:40 UTC, Tobias Müller 
 wrote:
 Yuxuan Shui <yshuiv7 gmail.com> wrote:
 [...]
In Rust that would be: let var : Option<i32> = ...; if let Some(var) = var { // You can use var here } It works for every enum (= tagged union), not just Option<T> Swift also has "if let". It's not much more verbose but more explicit. Changing the type of a variable based on static analysis is just advanced obfuscation. It hurts readability and the gain is questionable. At least it only works for nullable types. Tobi
D has this too, but only for nullable types afaik. if(byte* ptr = someFunc()){ //... }
That's not quite the same as there are no non-nullable pointers in D. There's no guarantee from the type system that the byte* is not null and there are no compiler checks involved. It's a simple runtime check. OTOH in the examples in Kotlin/Rust the variable 'var' changes its type from 'int?' to plain 'int'. In Kotlin this is done with static analysis, in Rust with rebinding of the name. Tobi
Feb 22 2016
parent reply rsw0x <anonymous anonymous.com> writes:
On Tuesday, 23 February 2016 at 06:49:46 UTC, Tobias Müller wrote:
 rsw0x <anonymous anonymous.com> wrote:
 On Saturday, 20 February 2016 at 09:40:40 UTC, Tobias Müller 
 wrote:
 [...]
D has this too, but only for nullable types afaik. if(byte* ptr = someFunc()){ //... }
That's not quite the same as there are no non-nullable pointers in D. There's no guarantee from the type system that the byte* is not null and there are no compiler checks involved. It's a simple runtime check. OTOH in the examples in Kotlin/Rust the variable 'var' changes its type from 'int?' to plain 'int'. In Kotlin this is done with static analysis, in Rust with rebinding of the name. Tobi
Rust's Option<T> checks are not done at compile-time in most cases unless something changed drastically in the past ~18 months.
Feb 22 2016
parent reply Xinok <xinok live.com> writes:
On Tuesday, 23 February 2016 at 07:18:09 UTC, rsw0x wrote:
 On Tuesday, 23 February 2016 at 06:49:46 UTC, Tobias Müller 
 wrote:
 OTOH in the examples in Kotlin/Rust the variable 'var' changes 
 its type
 from 'int?' to plain 'int'.
 In Kotlin this is done with static analysis, in Rust with 
 rebinding of the
 name.

 Tobi
Rust's Option<T> checks are not done at compile-time in most cases unless something changed drastically in the past ~18 months.
Option<T> is an enum type in Rust (i.e. algebraic data type). The Rust compiler forces you to check all possible cases so you can't use it improperly. So you would have to use something like pattern matching or "if let" to check the state. https://doc.rust-lang.org/book/match.html#matching-on-enums
Feb 23 2016
next sibling parent reply rsw0x <anonymous anonymous.com> writes:
On Tuesday, 23 February 2016 at 14:58:21 UTC, Xinok wrote:
 On Tuesday, 23 February 2016 at 07:18:09 UTC, rsw0x wrote:
 On Tuesday, 23 February 2016 at 06:49:46 UTC, Tobias Müller 
 wrote:
 OTOH in the examples in Kotlin/Rust the variable 'var' 
 changes its type
 from 'int?' to plain 'int'.
 In Kotlin this is done with static analysis, in Rust with 
 rebinding of the
 name.

 Tobi
Rust's Option<T> checks are not done at compile-time in most cases unless something changed drastically in the past ~18 months.
Option<T> is an enum type in Rust (i.e. algebraic data type). The Rust compiler forces you to check all possible cases so you can't use it improperly. So you would have to use something like pattern matching or "if let" to check the state. https://doc.rust-lang.org/book/match.html#matching-on-enums
How does this differ from the example I gave where the branch is only taken if the pointer is non-null?
Feb 23 2016
parent reply Xinok <xinok live.com> writes:
On Tuesday, 23 February 2016 at 19:43:43 UTC, rsw0x wrote:
 ...
 How does this differ from the example I gave where the branch 
 is only taken if the pointer is non-null?
D doesn't prevent you from dereferencing a null pointer whereas these scenarios should be impossible in Kotlin as well as Rust. Case and point, this code compiles without issue but crashes at runtime: int* foo() { return null; } void main() { if(int* ptr = new int) { ptr = foo(); // Whoops... *ptr = 35; // Crash } } In D, pointers and reference types can spontaneously become null under almost any context. That's the difference.
Feb 23 2016
parent Kagamin <spam here.lot> writes:
On Tuesday, 23 February 2016 at 22:03:32 UTC, Xinok wrote:
 On Tuesday, 23 February 2016 at 19:43:43 UTC, rsw0x wrote:
 ...
 How does this differ from the example I gave where the branch 
 is only taken if the pointer is non-null?
D doesn't prevent you from dereferencing a null pointer whereas these scenarios should be impossible in Kotlin as well as Rust. Case and point, this code compiles without issue but crashes at runtime: int* foo() { return null; } void main() { if(int* ptr = new int) { ptr = foo(); // Whoops... *ptr = 35; // Crash } } In D, pointers and reference types can spontaneously become null under almost any context. That's the difference.
The idea is to match non-null pointers, this is how you do it: int* foo() { return null; } void main() { if(int* ptr = foo()) { *ptr = 35; // no crash } }
Feb 24 2016
prev sibling parent reply Andrei Alexandrescu <SeeWebsiteForEmail erdani.org> writes:
On 02/23/2016 09:58 AM, Xinok wrote:
 Option<T> is an enum type in Rust (i.e. algebraic data type). The Rust
 compiler forces you to check all possible cases so you can't use it
 improperly. So you would have to use something like pattern matching or
 "if let" to check the state.

 https://doc.rust-lang.org/book/match.html#matching-on-enums
Probably we can do the same in a library: never allow the value out of an Option, but provide a visitor with two aliases (ham/spam). There was a related PR in Phobos, wasn't there? -- Andrei
Feb 24 2016
next sibling parent Kagamin <spam here.lot> writes:
On Wednesday, 24 February 2016 at 14:50:37 UTC, Andrei 
Alexandrescu wrote:
 Probably we can do the same in a library: never allow the value 
 out of an Option, but provide a visitor with two aliases 
 (ham/spam).
Rust provides value accessor for Option: https://doc.rust-lang.org/stable/std/option/enum.Option.html#method.unwrap
Feb 25 2016
prev sibling next sibling parent Kagamin <spam here.lot> writes:
On Wednesday, 24 February 2016 at 14:50:37 UTC, Andrei 
Alexandrescu wrote:
 There was a related PR in Phobos, wasn't there?
https://forum.dlang.org/post/n0j9nj$10e9$1 digitalmars.com - probably this.
Feb 25 2016
prev sibling parent crimaniak <crimaniak gmail.com> writes:
On my experience programming using Optional's functional 
interfaces is more reliable then explicit logic, so in case when 
we have value semantic and Optional then Kotlin's approach is not 
very useful.
My old experiments about this (obviously java-inspired): 
optional.d

import std.typecons;
import std.stdio;

Nullable!T Optional(T)(){ return Nullable!T(); }
Nullable!T Optional(T)(T value){ return Nullable!T(value); }
Nullable!OUT map(IN, OUT)(Nullable!IN n, OUT delegate(IN) fn)
{
     return n.isNull() ? Nullable!OUT() : 
Nullable!OUT(fn(n.get()));
}
T orElse(T)(Nullable!T n, T value){ return n.isNull() ? value : 
n.get(); }
void ifPresent(T)(Nullable!T n, void delegate(T) fn){ 
if(!n.isNull()) fn(n.get()); }
void ifPresent(T)(Nullable!T n, void function(T) fn){ 
if(!n.isNull()) fn(n.get()); }
bool isPresent(T)(Nullable!T n){ return !n.isNull(); }
Nullable!T coalesce(T)(Nullable!T n, Nullable!T other){ return 
n.isNull() ? other : n; }



unittest
{
  assert(Optional!string().isNull());
  assert(Optional!string("asd").isPresent());
  assert(Optional!string("asd").map((string v) => v~"$") == 
"asd$");
  assert(Optional!string().map((string v) => v~"$").isNull());
  assert(Optional!string("true").map((string v) => v=="1" || 
v=="true").orElse(false) == true);
  assert(Optional!string().coalesce(Optional!string("test")) == 
"test");
  
assert(Optional!string("test1").coalesce(Optional!string("test2")) == "test1");

  string a="-";
  Optional!string("A").ifPresent((string b){a=b;});
  Optional!string().ifPresent((string b){a=b;});
  assert(a == "A");

}
Feb 26 2016
prev sibling parent Xinok <xinok live.com> writes:
On Saturday, 20 February 2016 at 09:40:40 UTC, Tobias Müller 
wrote:
 ...
 It's not much more verbose but more explicit.
 Changing the type of a variable based on static analysis is 
 just advanced
 obfuscation. It hurts readability and the gain is questionable. 
 At least it
 only works for nullable types.
I consider this feature similar to requiring all control paths to initialize a variable or return a value. Imagine requiring some explicit or verbose syntax to enforce this behavior. I don't see this being an issue as long as the behavior is consistent between compilers (if code works in one compiler, it works in all compilers).
Feb 20 2016
prev sibling parent reply Ice Create Man <ice cream.com> writes:
On Thursday, 18 February 2016 at 23:33:45 UTC, Yuxuan Shui wrote:
 Just come across Kotlin today, and found some interesting ideas 
 skimming through its tutorial:

 [...]
Both those issues predate both Kotlin and Swift. C# has had both of them for as long as I've been coding in the language. Nothing new here.
Feb 22 2016
next sibling parent Yuxuan Shui <yshuiv7 gmail.com> writes:
On Monday, 22 February 2016 at 19:33:12 UTC, Ice Create Man wrote:
 On Thursday, 18 February 2016 at 23:33:45 UTC, Yuxuan Shui 
 wrote:
 Just come across Kotlin today, and found some interesting 
 ideas skimming through its tutorial:

 [...]
Both those issues predate both Kotlin and Swift. C# has had both of them for as long as I've been coding in the language. Nothing new here.
Sure that Optional types and dynamic casts have been there for ages. But this is the first time I see a compiler does the unwrapping & casting for you based on static analyses. And Xinok pointed out we have been using the same technique for detecting uninitialized variables, so what Kotlin does can be seen as a small, but clever, extension of this.
Feb 22 2016
prev sibling parent Russel Winder via Digitalmars-d <digitalmars-d puremagic.com> writes:
On Mon, 2016-02-22 at 19:33 +0000, Ice Create Man via Digitalmars-d
wrote:
 On Thursday, 18 February 2016 at 23:33:45 UTC, Yuxuan Shui wrote:
 Just come across Kotlin today, and found some interesting ideas=C2=A0
 skimming through its tutorial:
=20
 [...]
=20 Both those issues predate both Kotlin and Swift. C# has had both=C2=A0 of them for as long as I've been coding in the language. Nothing=C2=A0 new here.
But lots of what Kotlin, and Ceylon, are doing are new=E2=80=A6 on the JVM.= I like new, I like progress and evolution. Just because an idea is not objectively new, doesn't mean it isn't new and useful. Context and application can be as refreshing as having a brand new idea. Just think in programming languages today actors, dataflow, CSP are new, despite being 50-ish, and 40-ish years old. =C2=A0 --=20 Russel. =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D Dr Russel Winder t: +44 20 7585 2200 voip: sip:russel.winder ekiga.n= et 41 Buckmaster Road m: +44 7770 465 077 xmpp: russel winder.org.uk London SW11 1EN, UK w: www.russel.org.uk skype: russel_winder
Feb 22 2016