Multiple inputs & results

[TheLastGimbus]

Hi there!

I was looking for something like @functools.lru_cache from Python in Dart, and came across this library. But it's currently remebering only one result - kinda useless in my opinion 😕

So I'm proposing a solution to write down results with a HashMap

Currently I added only memo1, but if you like it, I will extend it to all the others - probably nested HashMap<A, HashMap<...>> will be used for that

I also added an example code to demonstrate how it makes things way faster 🚀

Closes #10

[TheLastGimbus]

Hey @Aetet @brianegan, is this repo alive or something? I can make a fork/re-make of this myself...

[Aetet]

Hey.
Currently I'm not maintaining this repo. Some guys from Wrike company should.

@DisDis @Chudesnov @bunopus

[sinegovsky-ivan]

How about limit storage capacity? For example second argument (with default value) to memo1.

Otherwise, in some cases, we will waste a lot of memory.

[TheLastGimbus]

Oh, you mean to limit maximum cache? Yeah, that would be useful

Tho i'm not sure how to do this - if you just pop() one item from a HashMap in dart, does it do it in some order or just random?

[sinegovsky-ivan]

Better refer to the documentation https://api.dart.dev/stable/2.12.2/dart-collection/HashMap/HashMap.html :)

[TheLastGimbus]

Okay, so generally - you like/want my change in general sense? If so, I can implement rest of the features - just wasn't sure if I even should

[sinegovsky-ivan]

Yes, I like your changes:)

[missvalentinep]

@TheLastGimbus a gentle ping :)

[TheLastGimbus]

HIiiiiii!

I know it took some time, but here it is! I updated branch with master, and implemented rest of methods!

I also fixed (I hope that's the right word 😅) test - they were designed that function re-executes every time something changes - please take a look at 84c2854

@sinegovsky-ivan I don't have any good idea for limiting the capacity tho 😕 It's a HashMap, so we can't just pop the first one (to make some space) because it's un-ordered... maybe we can store not only the value but also some N counter which by we would decide?

Func1<A, R> memo1<A, R>(Func1<A, R> func, {int? maxSize}) {
  final argsToOutput = HashMap<List, R>(
    equals: (a, b) => _listEquals(a, b),
    hashCode: (l) => _listHashCode(l),
  );
  // External list to keep track of cache order
  final hashes = <int>[];

  return ((A aA) {
    final cached = argsToOutput[[aA]];
    if (cached != null) {
      return cached;
    } else {
      final res = func(aA);
      argsToOutput[[aA]] = res;
      // Skip the whole thing if user doesn't use it
      if (maxSize != null) {
        hashes.add(_listHashCode([aA]));
        if (argsToOutput.length > maxSize) {
          // Pop the first (oldest) one
          // I'm not sure how efficient this is :/
          argsToOutput.removeWhere((k, _) => _listHashCode(k) == hashes[0]);
          hashes.removeAt(0);
        }
      }
      return res;
    }
  });
}

Something like this?

Edit: I actually wrote simple test for the above:

test('should remove old cache when maxSize set', () {
  var count = 0;
  var func = memo1((int a) => ++count, maxSize: 2);
  expect(func(1), 1);
  expect(func(1), 1);
  expect(func(2), 2);
  expect(func(2), 2);
  expect(func(1), 1);
  expect(func(3), 3);
  // At this point, first cache element should be removed
  expect(func(1), 4);
});

and it passes!

...oh, and also - I didn't know what to do with imemoN functions - they compare by identical, and I'm not sure will this work in HashMap - but I can try...

[missvalentinep]

@TheLastGimbus hey! I come to you with many thoughts about this PR:

1. Creating a hash map for every memo is not very efficient (since we'll just increase memory usage for people who're already using these functions) and adding an "if" condition ("create hashmap only if maxSize was provided") would make the code hard to read. So I propose creating a separate set of functions (just like imemo), called memom1, memom2... (as in "memoMany") that will have this functionality you want to add.
2. As for _listEquals and _listHashCode, I'd propose you use quiver package and listEqual and hashObjects methods from there
3. For storing cache I believe you should use LinkedHashMap, as it remembers the order in which the keys were inserted. So it will allow you to do something like this:

if (argsToOutput.length == maxSize) {
  argsToOutput.remove(argsToOutput.keys.first);
}

4. Since we're creating a separate set of functions, the maxSize param should be mandatory.
5. [Optional] I'd personally rename argsToOutput to cachedResults/cache and maxSize to cacheLimit

[om-ha]

Great project and PR!

Thanks for your review @missvalentinep.

Thanks @TheLastGimbus for your PR, would love to see this merged.