Group elements of an array by some property - arrays

Question

I have an array of objects with property date.
What I want is to create array of arrays where each array will contain objects with the same date.
I understand, that I need something like .filter to filter objects, and then .map to add every thing to array.
But how to tell .map that I want separate array for each group from filtered objects and that this array must be added to "global" array and how to tell .filter that I want objects with the same date ?

Answer 1

It might be late but new Xcode 9 sdk dictionary has new init method
init<S>(grouping values: S, by keyForValue: (S.Element) throws -> Key) rethrows where Value == [S.Element], S : Sequence
Documentation has simple example what this method does.
I just post this example below:
let students = ["Kofi", "Abena", "Efua", "Kweku", "Akosua"]
let studentsByLetter = Dictionary(grouping: students, by: { $0.first! })
Result will be:
["E": ["Efua"], "K": ["Kofi", "Kweku"], "A": ["Abena", "Akosua"]]

Answer 2

improving on oriyentel solution to allow ordered grouping on anything:
extension Sequence {
func group<GroupingType: Hashable>(by key: (Iterator.Element) -> GroupingType) -> [[Iterator.Element]] {
var groups: [GroupingType: [Iterator.Element]] = [:]
var groupsOrder: [GroupingType] = []
forEach { element in
let key = key(element)
if case nil = groups[key]?.append(element) {
groups[key] = [element]
groupsOrder.append(key)
}
}
return groupsOrder.map { groups[$0]! }
}
}
Then it will work on any tuple, struct or class and for any property:
let a = [(grouping: 10, content: "a"),
(grouping: 20, content: "b"),
(grouping: 10, content: "c")]
print(a.group { $0.grouping })
struct GroupInt {
var grouping: Int
var content: String
}
let b = [GroupInt(grouping: 10, content: "a"),
GroupInt(grouping: 20, content: "b"),
GroupInt(grouping: 10, content: "c")]
print(b.group { $0.grouping })

Answer 3

With Swift 5, you can group the elements of an array by one of their properties into a dictionary using Dictionary's init(grouping:by:) initializer. Once done, you can create an array of arrays from the dictionary using Dictionary's values property and Array init(_:) initializer.
The following Playground sample code shows how to group the elements of an array by one property into a new array of arrays:
import Foundation
struct Purchase: CustomStringConvertible {
let id: Int
let date: Date
var description: String {
return "Purchase #\(id) (\(date))"
}
}
let date1 = Calendar.current.date(from: DateComponents(year: 2010, month: 11, day: 22))!
let date2 = Calendar.current.date(from: DateComponents(year: 2015, month: 5, day: 1))!
let date3 = Calendar.current.date(from: DateComponents(year: 2012, month: 8, day: 15))!
let purchases = [
Purchase(id: 1, date: date1),
Purchase(id: 2, date: date1),
Purchase(id: 3, date: date2),
Purchase(id: 4, date: date3),
Purchase(id: 5, date: date3)
]
let groupingDictionary = Dictionary(grouping: purchases, by: { $0.date })
print(groupingDictionary)
/*
[
2012-08-14 22:00:00 +0000: [Purchase #4 (2012-08-14 22:00:00 +0000), Purchase #5 (2012-08-14 22:00:00 +0000)],
2010-11-21 23:00:00 +0000: [Purchase #1 (2010-11-21 23:00:00 +0000), Purchase #2 (2010-11-21 23:00:00 +0000)],
2015-04-30 22:00:00 +0000: [Purchase #3 (2015-04-30 22:00:00 +0000)]
]
*/
let groupingArray = Array(groupingDictionary.values)
print(groupingArray)
/*
[
[Purchase #3 (2015-04-30 22:00:00 +0000)],
[Purchase #4 (2012-08-14 22:00:00 +0000), Purchase #5 (2012-08-14 22:00:00 +0000)],
[Purchase #1 (2010-11-21 23:00:00 +0000), Purchase #2 (2010-11-21 23:00:00 +0000)]
]
*/

Answer 4

Abstracting one step, what you want is to group elements of an array by a certain property. You can let a map do the grouping for you like so:
protocol Groupable {
associatedtype GroupingType: Hashable
var grouping: GroupingType { get set }
}
extension Array where Element: Groupable {
typealias GroupingType = Element.GroupingType
func grouped() -> [[Element]] {
var groups = [GroupingType: [Element]]()
for element in self {
if let _ = groups[element.grouping] {
groups[element.grouping]!.append(element)
} else {
groups[element.grouping] = [element]
}
}
return Array<[Element]>(groups.values)
}
}
Note that this grouping is stable, that is groups appear in order of appearance, and inside the groups the individual elements appear in the same order as in the original array.
Usage Example
I'll give an example using integers; it should be clear how to use any (hashable) type for T, including Date.
struct GroupInt: Groupable {
typealias GroupingType = Int
var grouping: Int
var content: String
}
var a = [GroupInt(grouping: 1, content: "a"),
GroupInt(grouping: 2, content: "b") ,
GroupInt(grouping: 1, content: "c")]
print(a.grouped())
// > [[GroupInt(grouping: 2, content: "b")], [GroupInt(grouping: 1, content: "a"), GroupInt(grouping: 1, content: "c")]]

Answer 5

Rapheal's solution does work. However, I would propose altering the solution to support the claim that the grouping is in fact stable.
As it stands now, calling grouped() will return a grouped array but subsequent calls could return an array with groups in a different order, albeit the elements of each group will be in the expected order.
internal protocol Groupable {
associatedtype GroupingType : Hashable
var groupingKey : GroupingType? { get }
}
extension Array where Element : Groupable {
typealias GroupingType = Element.GroupingType
func grouped(nilsAsSingleGroup: Bool = false) -> [[Element]] {
var groups = [Int : [Element]]()
var groupsOrder = [Int]()
let nilGroupingKey = UUID().uuidString.hashValue
var nilGroup = [Element]()
for element in self {
// If it has a grouping key then use it. Otherwise, conditionally make one based on if nils get put in the same bucket or not
var groupingKey = element.groupingKey?.hashValue ?? UUID().uuidString.hashValue
if nilsAsSingleGroup, element.groupingKey == nil { groupingKey = nilGroupingKey }
// Group nils together
if nilsAsSingleGroup, element.groupingKey == nil {
nilGroup.append(element)
continue
}
// Place the element in the right bucket
if let _ = groups[groupingKey] {
groups[groupingKey]!.append(element)
} else {
// New key, track it
groups[groupingKey] = [element]
groupsOrder.append(groupingKey)
}
}
// Build our array of arrays from the dictionary of buckets
var grouped = groupsOrder.flatMap{ groups[$0] }
if nilsAsSingleGroup, !nilGroup.isEmpty { grouped.append(nilGroup) }
return grouped
}
}
Now that we track the order that we discover new groupings, we can return a grouped array more consistently than just relying on a Dictionary's unordered values property.
struct GroupableInt: Groupable {
typealias GroupingType = Int
var grouping: Int?
var content: String
}
var a = [GroupableInt(groupingKey: 1, value: "test1"),
GroupableInt(groupingKey: 2, value: "test2"),
GroupableInt(groupingKey: 2, value: "test3"),
GroupableInt(groupingKey: nil, value: "test4"),
GroupableInt(groupingKey: 3, value: "test5"),
GroupableInt(groupingKey: 3, value: "test6"),
GroupableInt(groupingKey: nil, value: "test7")]
print(a.grouped())
// > [[GroupableInt(groupingKey: 1, value: "test1")], [GroupableInt(groupingKey: 2, value: "test2"),GroupableInt(groupingKey: 2, value: "test3")], [GroupableInt(groupingKey: nil, value: "test4")],[GroupableInt(groupingKey: 3, value: "test5"),GroupableInt(groupingKey: 3, value: "test6")],[GroupableInt(groupingKey: nil, value: "test7")]]
print(a.grouped(nilsAsSingleGroup: true))
// > [[GroupableInt(groupingKey: 1, value: "test1")], [GroupableInt(groupingKey: 2, value: "test2"),GroupableInt(groupingKey: 2, value: "test3")], [GroupableInt(groupingKey: nil, value: "test4"),GroupableInt(groupingKey: nil, value: "test7")],[GroupableInt(groupingKey: 3, value: "test5"),GroupableInt(groupingKey: 3, value: "test6")]]

Answer 6

+1 to GolenKovkosty answer.
init<S>(grouping values: S, by keyForValue: (S.Element) throws -> Key) rethrows where Value == [S.Element], S : Sequence
More Examples:
enum Parity {
case even, odd
init(_ value: Int) {
self = value % 2 == 0 ? .even : .odd
}
}
let parity = Dictionary(grouping: 0 ..< 10 , by: Parity.init )
Equilvalent to
let parity2 = Dictionary(grouping: 0 ..< 10) { $0 % 2 }
In your case:
struct Person : CustomStringConvertible {
let dateOfBirth : Date
let name :String
var description: String {
return "\(name)"
}
}
extension Date {
init(dateString:String) {
let formatter = DateFormatter()
formatter.timeZone = NSTimeZone.default
formatter.dateFormat = "MM/dd/yyyy"
self = formatter.date(from: dateString)!
}
}
let people = [Person(dateOfBirth:Date(dateString:"01/01/2017"),name:"Foo"),
Person(dateOfBirth:Date(dateString:"01/01/2017"),name:"Bar"),
Person(dateOfBirth:Date(dateString:"02/01/2017"),name:"FooBar")]
let parityFields = Dictionary(grouping: people) {$0.dateOfBirth}
Output:
[2017-01-01: [Foo, Bar], 2017-02-01: [FooBar] ]

Answer 7

This is a clean way to perform group by:
let grouped = allRows.group(by: {$0.groupId}) // Dictionary with the key groupId
Assuming you have array of contacts like :
class ContactPerson {
var groupId:String?
var name:String?
var contactRecords:[PhoneBookEntry] = []
}
To achieve this, add this extension:
class Box<A> {
var value: A
init(_ val: A) {
self.value = val
}
}
public extension Sequence {
func group<U: Hashable>(by key: (Iterator.Element) -> U) -> [U: [Iterator.Element]] {
var categories: [U: Box<[Iterator.Element]>] = [:]
for element in self {
let key = key(element)
if case nil = categories[key]?.value.append(element) {
categories[key] = Box([element])
}
}
var result: [U: [Iterator.Element]] = Dictionary(minimumCapacity: categories.count)
for (key, val) in categories {
result[key] = val.value
}
return result
}
}