Merging with TLPLib?

[arturaz]

Hey there.

We have written our own functional programming library: https://github.com/tinylabproductions/tlplib

Then I discovered Smooth.

I was wondering - perhaps we want to join forces on this one and merge it to one project?

[pdo400]

I'm on break from Unity for a while, trying to quickly refresh my Scala skills and learn some of the current frameworks in order to go back to the real world with real pay, real productivity, and real value on API design, but I'd be down to merge at some point when I get the chance.

BTW, I noticed your Option is an interface, with value type implementations of Some and None.  This is nice because you can make T covariant and save some bytes and comparisons, but will lead to boxing as an Option typed var will need to be an Object.  It will in some cases anyway, and IMHO reducing allocations trumps minimizing stack usage / CPU operations in Unity land.

Patrick

On Tuesday, May 20, 2014 2:52 AM, Artūras Šlajus notifications@github.com wrote:

Hey there.
We have written our own functional programming library: https://github.com/tinylabproductions/tlplib
Then I discovered Smooth.
I was wondering - perhaps we want to join forces on this one and merge it to one project?
—
Reply to this email directly or view it on GitHub.

[arturaz]

Oh yeah, Scala, the land of elegant code :) I've moved from server app development with Scala to gaming for now, and all these horrible things make me think about Unreal Engine 4 and C++ :)

Yeah, this whole ordeal between reference types/value types is bad. Somehow I thought that making it an interface does not box it :/

Could I get some contacts of yours? An IM handle perhaps? Or a G+?

[pdo400]

While at my computer I'm generally on IRC.  I also have a crappy forums that is generally overrun with spam, gotta do something about that at some point.

My direct email is pdo400@yahoo.com, and the IRC and forum links can be found on the subreddit for my game:

Legacy: The Athletic Jetpack Shooter

Legacy: The Athletic Jetpack Shooter
reddit: the front page of the internet
View on www.reddit.com Preview by Yahoo

On Monday, May 26, 2014 7:01 AM, Artūras Šlajus notifications@github.com wrote:

Oh yeah, Scala, the land of elegant code :) I've moved from server app development with Scala to gaming for now, and all these horrible things make me think about Unreal Engine 4 and C++ :)
Yeah, this whole ordeal between reference types/value types is bad. Somehow I thought that making it an interface does not box it :/
Could I get some contacts of yours? An IM handle perhaps? Or a G+?
—
Reply to this email directly or view it on GitHub.

[arturaz]

Oh man. I've integrated Slinq into tlplib and boy, what you've done with it is awesome. Kudos! I should probably pitch this to be included into Unity as a replacement for LINQ :)

Can you share any docs on this pattern which you applied? Is it documented anywhere? How did you come up with it?

I was also wondering - is there a way to create an empty Slinq? One that doesn't contain any elements. Or should I create a specific context for that?

I'm trying to make my struct Range to return a Slinq without boxing to IEnumerable first. Basically it looks like this:

  [Serializable]
  // [from, to]
  public struct Range {
    // No it can't, Unity...
    // ReSharper disable once FieldCanBeMadeReadOnly.Local
    [SerializeField] private int _from, _to;
    public int from { get { return _from; } }
    public int to { get { return _to; } }

    public Range(int from, int to) {
      _from = from;
      _to = to;
    }

    public int random { get { return Random.Range(from, to + 1); } }

    public Slinq<int, FuncOptionContext<int, int>> Slinq { get {
      if (from > to) return F.throws<Slinq<int, FuncOptionContext<int, int>>>(new Exception("???"));

      return FuncOptionContext<int, int>.Sequence(
        from, 
        (i, toNum) => i == toNum ? F.none<int>() : F.some(i + 1),
        to
      );
    } }
  }

And I'm kind of lost on the from > to part :)

[arturaz]

        public static Slinq<T, FuncOptionContext<T>> Sequence(Option<T> seed, Fn<T, Option<T>> selector) {
            return new Slinq<T, FuncOptionContext<T>>(
                skip,
                remove,
                dispose,
                new FuncOptionContext<T>(seed, selector));
        }

        private FuncOptionContext(Option<T> seed, Fn<T, Option<T>> selector) {
            this.needsMove = false;
            this.acc = seed;
            this.selector = selector;

            this.bd = BacktrackDetector.Borrow();
        }

    private FuncOptionContext(T seed, Fn<T, Option<T>> selector) 
    : this(F.some(seed), selector) {}

    public Slinq<int, FuncOptionContext<int, int>> Slinq { get {
      return FuncOptionContext<int, int>.Sequence(
        from > to ? F.none<int>() : F.some(from), 
        (i, toNum) => i == toNum ? F.none<int>() : F.some(i + 1),
        to
      );
    } }

Seems to fix my issue :)

The remaining part of the question still stands.

[pdo400]

The default constructor for a Slinq<T, C> will result in a perfectly valid, empty Slinq, which works out quite nicely.  You can also do new Slinq<T, Unit>() if you don't care about the context type, though of course that is unlikely due to the strictness of value typing.  There is an "implicit" API that must be followed when writing Slinq operations or contexts that supports this behaviour.

Namely:

* Slinq operations should never call the delegates of an empty (ie: current is None) Slinq.
* When a Slinq delegate causes a Slinq to become empty, it must perform any required disposal.
* When disposing a Slinq, current must be set to None.

* (By these rules, whenever a delegate is called on a Slinq, the delegate can safely assume that the Slinq is non-empty and has not been disposed.)

Also, there is Slinqable.Range(int start, int count) that follows the Enumerable API and is close to what you want but with a bit of nasty fencepost math.  Or Slinqable.Sequence(int start, int step).TakeWhile(pred, param) which is a bit cleaner as you don't have to do any case logic (use the extra param version to pass in the max value so your predicate doesn't require a closure).  When writing the API I figured that ranges would be pretty common, but there are a lot of common things that are two operation chains, and adding all of them would be a slippery slope.  :)

But as you've discovered, Slinq<T, FuncOptionContext> is pretty much the "top type" of all Slinqs that can do anything. 

On Sunday, June 1, 2014 2:28 PM, Artūras Šlajus notifications@github.com wrote:

public static Slinq<T, FuncOptionContext> Sequence(Option seed, Fn<T, Option> selector) { return new Slinq<T, FuncOptionContext>( skip, remove, dispose, new FuncOptionContext(seed, selector)); } private FuncOptionContext(Option seed, Fn<T, Option> selector) { this.needsMove = false; this.acc = seed; this.selector = selector; this.bd = BacktrackDetector.Borrow(); } private FuncOptionContext(T seed, Fn<T, Option> selector) : this(F.some(seed), selector) {}
public Slinq<int, FuncOptionContext<int, int>> Slinq { get { return FuncOptionContext<int, int>.Sequence( from > to ? F.none() : F.some(from), (i, toNum) => i == toNum ? F.none() : F.some(i + 1), to ); } }
Seems to fix my issue :)
The remaining part of the question still stands.
—
Reply to this email directly or view it on GitHub.

[pdo400]

Oh, and yeah, I'm currently looking for Scala jobs and am interesting in hearing about any opportunities.  And I'm in the San Francisco Bay Area.

On , Patrick Whitesell pdo400@yahoo.com wrote:

The default constructor for a Slinq<T, C> will result in a perfectly valid, empty Slinq, which works out quite nicely.  You can also do new Slinq<T, Unit>() if you don't care about the context type, though of course that is unlikely due to the strictness of value typing.  There is an "implicit" API that must be followed when writing Slinq operations or contexts that supports this behaviour.

Namely:

* Slinq operations should never call the delegates of an empty (ie: current is None) Slinq.
* When a Slinq delegate causes a Slinq to become empty, it must perform any required disposal.
* When disposing a Slinq, current must be set to None.

* (By these rules, whenever a delegate is called on a Slinq, the delegate can safely assume that the Slinq is non-empty and has not been disposed.)

Also, there is Slinqable.Range(int start, int count) that follows the Enumerable API and is close to what you want but with a bit of nasty fencepost math.  Or Slinqable.Sequence(int start, int step).TakeWhile(pred, param) which is a bit cleaner as you don't have to do any case logic (use the extra param version to pass in the max value so your predicate doesn't require a closure).  When writing the API I figured that ranges would be pretty common, but there are a lot of common things that are two operation chains, and adding all of them would be a slippery slope.  :)

But as you've discovered, Slinq<T, FuncOptionContext> is pretty much the "top type" of all Slinqs that can do anything. 

On Sunday, June 1, 2014 2:28 PM, Artūras Šlajus notifications@github.com wrote:

public static Slinq<T, FuncOptionContext> Sequence(Option seed, Fn<T, Option> selector) { return new Slinq<T, FuncOptionContext>( skip, remove, dispose, new FuncOptionContext(seed, selector)); } private FuncOptionContext(Option seed, Fn<T, Option> selector) { this.needsMove = false; this.acc = seed; this.selector = selector; this.bd = BacktrackDetector.Borrow(); } private FuncOptionContext(T seed, Fn<T, Option> selector) : this(F.some(seed), selector) {}
public Slinq<int, FuncOptionContext<int, int>> Slinq { get { return FuncOptionContext<int, int>.Sequence( from > to ? F.none() : F.some(from), (i, toNum) => i == toNum ? F.none() : F.some(i + 1), to ); } }
Seems to fix my issue :)
The remaining part of the question still stands.
—
Reply to this email directly or view it on GitHub.