Blog

Group and sort Swift collections like a pro
Apr 1, 2023 swift

Swift provides powerful ways to group and sort collections. Let’s take a look at some ways to do this and how to change the sort logic a bit.

Imagine having a type that can be grouped into named collections, for instance a Person that defines a city of residence:

struct Person {
  var name: String
  var city: String
}

Let’s now say that we have a couple of persons:

let persons: [Person] = [
  .init(name: "Johanna", city: "Stockholm"),
  .init(name: "Daniel", city: "Stockholm"),
  .init(name: "Joe", city: "Washington"),
  .init(name: "Kamala", city: "San Francisco")
]

Using Dictionary(grouping:by:) we can easily group persons by city:

extension Collection where Element == Person {
  func groupedByCity() -> Dictionary<String, [Element]> {
    Dictionary(grouping: self, by: { $0.city })
  }
}

and call it like this:

persons.groupedByCity()

This returns a dictionary, where the keys are city names and the dictionary values are an array of people:

[0] = {
  key = "Stockholm"
  value = 2 values {
    [0] = {
      name = "Johanna"
      city = "Stockholm"
    }
    [1] = {
      name = "Daniel"
      city = "Stockholm"
    }
  }
}
[1] = {
  key = "San Francisco"
  value = 1 value {
    [0] = {
      name = "Johanna"
      city = "Stockholm"
    }
  }
}
...

Dictionaries are great for key-value structured data, but a problem is that the keys are unordered, which means that the result above will be shuffled every time you create it.

We could get and sort all keys from the dictionary, then iterate over the sorted keys to fetch users, but let’s instead add this generic, named collection to get a bit more structure:

struct NamedCollection<Item> {

  var name: String
  var items: [Item]
}

We can now map the dictionary to a [NamedCollection] like this:

extension Collection where Element == Person {

    func groupedByCity() -> [NamedCollection<Element>] {
        Dictionary(grouping: persons, by: { $0.city })
            .map { NamedCollection(name: $0, items: $1) }
    }
}

Since the keys are unordered, we still get random order every time we do this:

[0] = {
  name = "Stockholm"
  items = 2 values {
    [0] = {
      name = "Johanna"
      city = "Stockholm"
    }
    [1] = {
      name = "Daniel"
      city = "Stockholm"
    }
  }
}
[1] = {
  name = "Washington"
  ...

But it’s now very easy to add sorting to the mix and sort the groups by city name:

extension Collection where Element == Person { 
    func groupedByCity() -> [NamedCollection<Element>] { 
        Dictionary(grouping: persons, by: { $0.city })
            .map { NamedCollection(name: $0, items: $1) }
            .sorted { $0.name < $1.name }   // <--
    } 
}

With this, we always get the cities in alphabetic ascending order:

[0] = {
  name = "San Francisco"
  items = 1 value {
    [0] = {
      name = "Kamala"
      city = "San Francisco"
    }
  }
}
[1] = {
  name = "Stockholm"
  ...

Now consider a situation where we want to sort this in alphabetical order, but also want to place one or several cities topmost. We may have a “city” of the week, a currently selected city etc.

This would have been easy to do with the dictionary, where we just would fetch and list a certain city key, then removed that key, then listed the remaining keys, although they would be listed in random order.

With the named collection, we instead have a sorted array, where we want to place a certain item first, then list the rest in alphabetical order. This could involve iterating the collection many times to find the city we’re after, remove it from the collection then append the remaining, sorted cities after it.

A more efficient approach is to create a custom sort function for the NamedCollection:

public extension Collection where Element == NamedCollection {

    func sorted(firstmost: [String]) -> [Element] {
        sorted {
            for name in firstmost {
                if $0.name == name { return true }
                if $1.name == name { return false }
            }
            return $0.name < $1.name
        }
    }
}

All this does is to look if either of the compared element names is in the list of names to place firstmost. Since we iterate over the firstmost collection, the provided order will be preserved.

We can now do this:

persons
  .groupedByCity()
  .sorted(firstmost: ["Washington"])

This will always place Washington first, then sort the rest in alphabetical ascending order. Replacing the return $0.name < $1.name with return $0.name > $1.name would apply a descending order while return Bool.random() would sort randomly.

You could also extend Collection where Element == Person further, to let you specify the topmost cities at the same time as grouping them. The possibilities are basically endless, so you are free to come up with a design that fits your domain and use-case.

That’s it, you’re now a Swift collection grouping and sorting pro! 🎉

Discussion

If you found this post interesting and would like to share your thoughts, ideas, feedback etc. please comment in the Disqus section below or reply to this tweet or this toot.