|
| 1 | +# Chunked |
| 2 | + |
| 3 | +[[Source](https://github.com/apple/swift-algorithms/blob/main/Sources/Algorithms/Chunked.swift) | |
| 4 | + [Tests](https://github.com/apple/swift-algorithms/blob/main/Tests/AlgorithmsTests/ChunkedTests.swift)] |
| 5 | + |
| 6 | +Break a collection into subsequences where consecutive elements pass a binary |
| 7 | +predicate, or where all elements in each chunk project to the same value. |
| 8 | + |
| 9 | +There are two variations of the `chunked` method: `chunked(by:)` and |
| 10 | +`chunked(on:)`. `chunked(by:)` uses a binary predicate to test consecutive |
| 11 | +elements, separating chunks where the predicate returns `false`. For example, |
| 12 | +you can chunk a collection into ascending sequences using this method: |
| 13 | + |
| 14 | +```swift |
| 15 | +let numbers = [10, 20, 30, 10, 40, 40, 10, 20] |
| 16 | +let chunks = numbers.chunked(by: { $0 <= $1 }) |
| 17 | +// [[10, 20, 30], [10, 40, 40], [10, 20]] |
| 18 | +``` |
| 19 | + |
| 20 | +The `chunk(on:)` method, by contrast, takes a projection of each element and |
| 21 | +separates chunks where the projection of two consecutive elements is not equal. |
| 22 | + |
| 23 | +```swift |
| 24 | +let names = ["David", "Kyle", "Karoy", "Nate"] |
| 25 | +let chunks = names.chunked(on: \.first!) |
| 26 | +// [["David"], ["Kyle", "Karoy"], ["Nate"]] |
| 27 | +``` |
| 28 | + |
| 29 | +These methods are related to the [existing SE proposal][proposal] for chunking a |
| 30 | +collection into subsequences of a particular size, potentially named something |
| 31 | +like `chunked(length:)`. Unlike the `split` family of methods, the entire |
| 32 | +collection is included in the chunked result — joining the resulting chunks |
| 33 | +recreates the original collection. |
| 34 | + |
| 35 | +```swift |
| 36 | +c.elementsEqual(c.chunked(...).joined()) |
| 37 | +// true |
| 38 | +``` |
| 39 | + |
| 40 | +[proposal]: https://github.com/apple/swift-evolution/pull/935 |
| 41 | + |
| 42 | +## Detailed Design |
| 43 | + |
| 44 | +The two methods are added as extension to `Collection`, with two matching |
| 45 | +versions that return a lazy wrapper added to `LazyCollectionProtocol`. |
| 46 | + |
| 47 | +```swift |
| 48 | +extension Collection { |
| 49 | + public func chunked( |
| 50 | + by belongInSameGroup: (Element, Element) -> Bool |
| 51 | + ) -> [SubSequence] |
| 52 | + |
| 53 | + public func chunked<Subject: Equatable>( |
| 54 | + on projection: (Element) -> Subject |
| 55 | + ) -> [SubSequence] |
| 56 | + } |
| 57 | + |
| 58 | + extension LazyCollectionProtocol { |
| 59 | + public func chunked( |
| 60 | + by belongInSameGroup: @escaping (Element, Element) -> Bool |
| 61 | + ) -> LazyChunked<Elements> |
| 62 | + |
| 63 | + public func chunked<Subject: Equatable>( |
| 64 | + on projection: @escaping (Element) -> Subject |
| 65 | + ) -> LazyChunked<Elements> |
| 66 | +} |
| 67 | +``` |
| 68 | + |
| 69 | +The `LazyChunked` type is bidirectional when the wrapped collection is |
| 70 | +bidirectional. |
| 71 | + |
| 72 | +### Complexity |
| 73 | + |
| 74 | +The eager methods are O(_n_), the lazy methods are O(_1_). |
| 75 | + |
| 76 | +### Naming |
| 77 | + |
| 78 | +The operation performed by these methods is similar to other ways of breaking a collection up into subsequences. In particular, the predicate-based `split(where:)` method looks similar to `chunked(on:)`. You can draw a distinction between these different operations based on the resulting subsequences: |
| 79 | + |
| 80 | +- `split`: *In the standard library.* Breaks a collection into subsequences, removing any elements that are considered "separators". The original collection cannot be recovered from the result of splitting. |
| 81 | +- `chunked`: *In this package.* Breaks a collection into subsequences, preserving each element in its initial ordering. Joining the resulting subsequences re-forms the original collection. |
| 82 | +- `sliced`: *Not included in this package or the stdlib.* Breaks a collection into potentially overlapping subsequences. |
| 83 | + |
| 84 | + |
| 85 | +### Comparison with other langauges |
| 86 | + |
| 87 | +**Ruby:** Ruby’s `Enumerable` class defines `chunk_while` and `chunk`, which map |
| 88 | +to the proposed `chunked(by:)` and `chunked(on:)` methods. |
| 89 | + |
| 90 | +**Rust:** Rust defines a variety of size-based `chunks` methods, but doesn’t |
| 91 | +include any with the functionality described here. |
0 commit comments