Hello, World! Is it still a fact that there is no...

# stdlib

[bod]

Hello, World! Is it still a fact that there is no MultiMap in the stdlib?

[gildor]

There is no MultiMap in stdlib

[bod]

too bad 🙂 Thanks!

[gildor]

do you need one for MPP?

[bod]

I'm on Android actually

[gildor]

You can use com.google.common.collect.Multimap

[bod]

yeahhh it's just too bad it's not directly in Kotlin imho - don't really want to add a dependency just for this little bit of code I'm working on right now

[bod]

in the end, I'll just code it myself, but I feel thousands of devs are re-implementing the wheel on all their projects 🙂

[gildor]

so you don’t want to add library to you project because it’s unnecessary, but want to add a new class to stdlib, so every user in world will have it?

[gildor]

re-implementing the wheel on all their projects

Or use many existing ones, Guava probably the most common dependency

[gildor]

One thing with stdlib, that it has very limited own collections, because it uses Java implementations on runtime, so it allows to keep stdlib quite small

[bod]

Well of course everything is a compromise, but given that IMO the vast majority of projects will at some point need a Multimap, and its implementation is probably quite modest, yes this addition is worth it. Also as you noted Java libraries are OK, but only for JVM or Android. A stdlib solution would work everywhere. And also Java solutions are probably not very good with nullability.

[gildor]

Yes, it always trade off

[Ruckus]

given that IMO the vast majority of projects will at some point need a Multimap

IMO this is an extreme over estimation, but that may just be selection bias.

[bod]

😄 yeah I mean, if you wait long enough, the probability for a project to need it will tend to 1

[Cody Engel]

I don't know that I've had a reason to use a MultiMap. Honestly this seems like it could make for an advanced collections library that someone could in theory open source and give back to the community.

[bod]

Another thing to consider is that since it's so easy to implement, most people will prefer to just do it themselves than add a dependency (especially on mobile where adding a dependency is a big deal). But then the situation is not ideal (the everybody re invents the wheel part).

[Cody Engel]

Honestly whenever I've needed a Map that requires multiple values for a key I just do Map<K, List<V>> which isn't that much work. It looks like iterating over the maps is the same amount of effort

[bod]

exactly, that's a multimap 🙂 You also have to deal with the empty list vs no element for the key. Essentially you're implementing it yourself, when such a prevalent and simple utility could be in the stdlib

[bod]

... I mean for instance, shuffle has just been added to arrays. Don't get me wrong, that's cool! But is it really used that often? I don't know if there's an easy standard to decide what should be in the stdlib vs what should be kept in external libs. Probably a very hard question.

[Derek Peirce]

You could use a typealias and a few extension methods to effectively turn any Map<K, MutableCollection<V>> into a MultiMap<K, V>, just slightly less performant than Guava's implementation. Same deal for Map<E, Int> into a MultiSet<E>.

[Cody Engel]

I use shuffle a ton for unit testing

[gildor]

I mean for instance, shuffle has just been added to arrays

Shuffle is just a small extension function, it was added recently because stdlib Random is introduced

[bod]

@Derek Peirce I mean yes, part of my point is that it's easy/small to implement indeed 🙂 More reason for it to be there already once and for all. I mean let is also easy to implement but we wouldn't like to do it ourselves in every project.

[gildor]

Special collection type is not so simple as let

[gildor]

if you have a few extension function to simplify work with Map<K, List<V>>, be free to suggest it for stdlib

[bod]

I guess. The chances of it being accepted seem low however judging by this thread 🙂 But I may try, it could be a fun exercise in any case.

[gildor]

because it depends on what will be suggested

[gildor]

because idea to add a special collection for it (and actually multiple collections) doesn’t look reasonsable