Safe (bounds-checked) array lookup in Swift, through optional bindings? - xcode

If I have an array in Swift, and try to access an index that is out of bounds, there is an unsurprising runtime error:
var str = ["Apple", "Banana", "Coconut"]
str[0] // "Apple"
str[3] // EXC_BAD_INSTRUCTION
However, I would have thought with all the optional chaining and safety that Swift brings, it would be trivial to do something like:
let theIndex = 3
if let nonexistent = str[theIndex] { // Bounds check + Lookup
print(nonexistent)
...do other things with nonexistent...
}
Instead of:
let theIndex = 3
if (theIndex < str.count) { // Bounds check
let nonexistent = str[theIndex] // Lookup
print(nonexistent)
...do other things with nonexistent...
}
But this is not the case - I have to use the ol' if statement to check and ensure the index is less than str.count.
I tried adding my own subscript() implementation, but I'm not sure how to pass the call to the original implementation, or to access the items (index-based) without using subscript notation:
extension Array {
subscript(var index: Int) -> AnyObject? {
if index >= self.count {
NSLog("Womp!")
return nil
}
return ... // What?
}
}

Alex's answer has good advice and solution for the question, however, I've happened to stumble on a nicer way of implementing this functionality:
extension Collection {
/// Returns the element at the specified index if it is within bounds, otherwise nil.
subscript (safe index: Index) -> Element? {
return indices.contains(index) ? self[index] : nil
}
}
Example
let array = [1, 2, 3]
for index in -20...20 {
if let item = array[safe: index] {
print(item)
}
}

If you really want this behavior, it smells like you want a Dictionary instead of an Array. Dictionaries return nil when accessing missing keys, which makes sense because it's much harder to know if a key is present in a dictionary since those keys can be anything, where in an array the key must in a range of: 0 to count. And it's incredibly common to iterate over this range, where you can be absolutely sure have a real value on each iteration of a loop.
I think the reason it doesn't work this way is a design choice made by the Swift developers. Take your example:
var fruits: [String] = ["Apple", "Banana", "Coconut"]
var str: String = "I ate a \( fruits[0] )"
If you already know the index exists, as you do in most cases where you use an array, this code is great. However, if accessing a subscript could possibly return nil then you have changed the return type of Array's subscript method to be an optional. This changes your code to:
var fruits: [String] = ["Apple", "Banana", "Coconut"]
var str: String = "I ate a \( fruits[0]! )"
// ^ Added
Which means you would need to unwrap an optional every time you iterated through an array, or did anything else with a known index, just because rarely you might access an out of bounds index. The Swift designers opted for less unwrapping of optionals, at the expense of a runtime exception when accessing out of bounds indexes. And a crash is preferable to a logic error caused by a nil you didn't expect in your data somewhere.
And I agree with them. So you won't be changing the default Array implementation because you would break all the code that expects a non-optional values from arrays.
Instead, you could subclass Array, and override subscript to return an optional. Or, more practically, you could extend Array with a non-subscript method that does this.
extension Array {
// Safely lookup an index that might be out of bounds,
// returning nil if it does not exist
func get(index: Int) -> T? {
if 0 <= index && index < count {
return self[index]
} else {
return nil
}
}
}
var fruits: [String] = ["Apple", "Banana", "Coconut"]
if let fruit = fruits.get(1) {
print("I ate a \( fruit )")
// I ate a Banana
}
if let fruit = fruits.get(3) {
print("I ate a \( fruit )")
// never runs, get returned nil
}
Swift 3 Update
func get(index: Int) ->T? needs to be replaced by func get(index: Int) ->Element?

To build on Nikita Kukushkin's answer, sometimes you need to safely assign to array indexes as well as read from them, i.e.
myArray[safe: badIndex] = newValue
So here is an update to Nikita's answer (Swift 3.2) that also allows safely writing to mutable array indexes, by adding the safe: parameter name.
extension Collection {
/// Returns the element at the specified index if it is within bounds, otherwise nil.
subscript(safe index: Index) -> Element? {
return indices.contains(index) ? self[index] : nil
}
}
extension MutableCollection {
subscript(safe index: Index) -> Element? {
get {
return indices.contains(index) ? self[index] : nil
}
set(newValue) {
if let newValue = newValue, indices.contains(index) {
self[index] = newValue
}
}
}
}

extension Array {
subscript (safe index: Index) -> Element? {
0 <= index && index < count ? self[index] : nil
}
}
O(1) performance
type safe
correctly deals with Optionals for [MyType?] (returns MyType??, that can be unwrapped on both levels)
does not lead to problems for Sets
concise code
Here are some tests I ran for you:
let itms: [Int?] = [0, nil]
let a = itms[safe: 0] // 0 : Int??
a ?? 5 // 0 : Int?
let b = itms[safe: 1] // nil : Int??
b ?? 5 // nil : Int? (`b` contains a value and that value is `nil`)
let c = itms[safe: 2] // nil : Int??
c ?? 5 // 5 : Int?

Swift 4
An extension for those who prefer a more traditional syntax:
extension Array {
func item(at index: Int) -> Element? {
return indices.contains(index) ? self[index] : nil
}
}

Valid in Swift 2
Even though this has been answered plenty of times already, I'd like to present an answer more in line in where the fashion of Swift programming is going, which in Crusty's words¹ is: "Think protocols first"
• What do we want to do?
- Get an Element of an Array given an Index only when it's safe, and nil otherwise
• What should this functionality base it's implementation on?
- Array subscripting
• Where does it get this feature from?
- Its definition of struct Array in the Swift module has it
• Nothing more generic/abstract?
- It adopts protocol CollectionType which ensures it as well
• Nothing more generic/abstract?
- It adopts protocol Indexable as well...
• Yup, sounds like the best we can do. Can we then extend it to have this feature we want?
- But we have very limited types (no Int) and properties (no count) to work with now!
• It will be enough. Swift's stdlib is done pretty well ;)
extension Indexable {
public subscript(safe safeIndex: Index) -> _Element? {
return safeIndex.distanceTo(endIndex) > 0 ? self[safeIndex] : nil
}
}
¹: not true, but it gives the idea

Because arrays may store nil values, it does not make sense to return a nil if an array[index] call is out of bounds.
Because we do not know how a user would like to handle out of bounds problems, it does not make sense to use custom operators.
In contrast, use traditional control flow for unwrapping objects and ensure type safety.
if let index = array.checkIndexForSafety(index:Int)
let item = array[safeIndex: index]
if let index = array.checkIndexForSafety(index:Int)
array[safeIndex: safeIndex] = myObject
extension Array {
#warn_unused_result public func checkIndexForSafety(index: Int) -> SafeIndex? {
if indices.contains(index) {
// wrap index number in object, so can ensure type safety
return SafeIndex(indexNumber: index)
} else {
return nil
}
}
subscript(index:SafeIndex) -> Element {
get {
return self[index.indexNumber]
}
set {
self[index.indexNumber] = newValue
}
}
// second version of same subscript, but with different method signature, allowing user to highlight using safe index
subscript(safeIndex index:SafeIndex) -> Element {
get {
return self[index.indexNumber]
}
set {
self[index.indexNumber] = newValue
}
}
}
public class SafeIndex {
var indexNumber:Int
init(indexNumber:Int){
self.indexNumber = indexNumber
}
}

I realize this is an old question. I'm using Swift5.1 at this point, the OP was for Swift 1 or 2?
I needed something like this today, but I didn't want to add a full scale extension for just the one place and wanted something more functional (more thread safe?). I also didn't need to protect against negative indices, just those that might be past the end of an array:
let fruit = ["Apple", "Banana", "Coconut"]
let a = fruit.dropFirst(2).first // -> "Coconut"
let b = fruit.dropFirst(0).first // -> "Apple"
let c = fruit.dropFirst(10).first // -> nil
For those arguing about Sequences with nil's, what do you do about the first and last properties that return nil for empty collections?
I liked this because I could just grab at existing stuff and use it to get the result I wanted. I also know that dropFirst(n) is not a whole collection copy, just a slice. And then the already existent behavior of first takes over for me.

I found safe array get, set, insert, remove very useful. I prefer to log and ignore the errors as all else soon gets hard to manage. Full code bellow
/**
Safe array get, set, insert and delete.
All action that would cause an error are ignored.
*/
extension Array {
/**
Removes element at index.
Action that would cause an error are ignored.
*/
mutating func remove(safeAt index: Index) {
guard index >= 0 && index < count else {
print("Index out of bounds while deleting item at index \(index) in \(self). This action is ignored.")
return
}
remove(at: index)
}
/**
Inserts element at index.
Action that would cause an error are ignored.
*/
mutating func insert(_ element: Element, safeAt index: Index) {
guard index >= 0 && index <= count else {
print("Index out of bounds while inserting item at index \(index) in \(self). This action is ignored")
return
}
insert(element, at: index)
}
/**
Safe get set subscript.
Action that would cause an error are ignored.
*/
subscript (safe index: Index) -> Element? {
get {
return indices.contains(index) ? self[index] : nil
}
set {
remove(safeAt: index)
if let element = newValue {
insert(element, safeAt: index)
}
}
}
}
Tests
import XCTest
class SafeArrayTest: XCTestCase {
func testRemove_Successful() {
var array = [1, 2, 3]
array.remove(safeAt: 1)
XCTAssert(array == [1, 3])
}
func testRemove_Failure() {
var array = [1, 2, 3]
array.remove(safeAt: 3)
XCTAssert(array == [1, 2, 3])
}
func testInsert_Successful() {
var array = [1, 2, 3]
array.insert(4, safeAt: 1)
XCTAssert(array == [1, 4, 2, 3])
}
func testInsert_Successful_AtEnd() {
var array = [1, 2, 3]
array.insert(4, safeAt: 3)
XCTAssert(array == [1, 2, 3, 4])
}
func testInsert_Failure() {
var array = [1, 2, 3]
array.insert(4, safeAt: 5)
XCTAssert(array == [1, 2, 3])
}
func testGet_Successful() {
var array = [1, 2, 3]
let element = array[safe: 1]
XCTAssert(element == 2)
}
func testGet_Failure() {
var array = [1, 2, 3]
let element = array[safe: 4]
XCTAssert(element == nil)
}
func testSet_Successful() {
var array = [1, 2, 3]
array[safe: 1] = 4
XCTAssert(array == [1, 4, 3])
}
func testSet_Successful_AtEnd() {
var array = [1, 2, 3]
array[safe: 3] = 4
XCTAssert(array == [1, 2, 3, 4])
}
func testSet_Failure() {
var array = [1, 2, 3]
array[safe: 4] = 4
XCTAssert(array == [1, 2, 3])
}
}

Swift 5.x
An extension on RandomAccessCollection means that this can also work for ArraySlice from a single implementation. We use startIndex and endIndex as array slices use the indexes from the underlying parent Array.
public extension RandomAccessCollection {
/// Returns the element at the specified index if it is within bounds, otherwise nil.
/// - complexity: O(1)
subscript (safe index: Index) -> Element? {
guard index >= startIndex, index < endIndex else {
return nil
}
return self[index]
}
}

extension Array {
subscript (safe index: UInt) -> Element? {
return Int(index) < count ? self[Int(index)] : nil
}
}
Using Above mention extension return nil if anytime index goes out of bound.
let fruits = ["apple","banana"]
print("result-\(fruits[safe : 2])")
result - nil

I have padded the array with nils in my use case:
let components = [1, 2]
var nilComponents = components.map { $0 as Int? }
nilComponents += [nil, nil, nil]
switch (nilComponents[0], nilComponents[1], nilComponents[2]) {
case (_, _, .Some(5)):
// process last component with 5
default:
break
}
Also check the subscript extension with safe: label by Erica Sadun / Mike Ash: http://ericasadun.com/2015/06/01/swift-safe-array-indexing-my-favorite-thing-of-the-new-week/

The "Commonly Rejected Changes" for Swift list contains a mention of changing Array subscript access to return an optional rather than crashing:
Make Array<T> subscript access return T? or T! instead of T: The current array behavior is intentional, as it accurately reflects the fact that out-of-bounds array access is a logic error. Changing the current behavior would slow Array accesses to an unacceptable degree. This topic has come up multiple times before but is very unlikely to be accepted.
https://github.com/apple/swift-evolution/blob/master/commonly_proposed.md#strings-characters-and-collection-types
So the basic subscript access will not be changing to return an optional.
However, the Swift team/community does seem open to adding a new optional-returning access pattern to Arrays, either via a function or subscript.
This has been proposed and discussed on the Swift Evolution forum here:
https://forums.swift.org/t/add-accessor-with-bounds-check-to-array/16871
Notably, Chris Lattner gave the idea a "+1":
Agreed, the most frequently suggested spelling for this is: yourArray[safe: idx], which seems great to me. I am very +1 for adding this.
https://forums.swift.org/t/add-accessor-with-bounds-check-to-array/16871/13
So this may be possible out of the box in some future version of Swift. I'd encourage anyone who wants it to contribute to that Swift Evolution thread.

Not sure why no one, has put up an extension that also has a setter to automatically grow the array
extension Array where Element: ExpressibleByNilLiteral {
public subscript(safe index: Int) -> Element? {
get {
guard index >= 0, index < endIndex else {
return nil
}
return self[index]
}
set(newValue) {
if index >= endIndex {
self.append(contentsOf: Array(repeating: nil, count: index - endIndex + 1))
}
self[index] = newValue ?? nil
}
}
}
Usage is easy and works as of Swift 5.1
var arr:[String?] = ["A","B","C"]
print(arr) // Output: [Optional("A"), Optional("B"), Optional("C")]
arr[safe:10] = "Z"
print(arr) // [Optional("A"), Optional("B"), Optional("C"), nil, nil, nil, nil, nil, nil, nil, Optional("Z")]
Note: You should understand the performance cost (both in time/space) when growing an array in swift - but for small problems sometimes you just need to get Swift to stop Swifting itself in the foot

To propagate why operations fail, errors are better than optionals.
public extension Collection {
/// Ensure an index is valid before accessing an element of the collection.
/// - Returns: The same as the unlabeled subscript, if an error is not thrown.
/// - Throws: `AnyCollection<Element>.IndexingError`
/// if `indices` does not contain `index`.
subscript(validating index: Index) -> Element {
get throws {
guard indices.contains(index)
else { throw AnyCollection<Element>.IndexingError() }
return self[index]
}
}
}
public extension AnyCollection {
/// Thrown when `[validating:]` is called with an invalid index.
struct IndexingError: Error { }
}
XCTAssertThrowsError(try ["🐾", "🥝"][validating: 2])
let collection = Array(1...10)
XCTAssertEqual(try collection[validating: 0], 1)
XCTAssertThrowsError(try collection[validating: collection.endIndex]) {
XCTAssert($0 is AnyCollection<Int>.IndexingError)
}

I think this is not a good idea. It seems preferable to build solid code that does not result in trying to apply out-of-bounds indexes.
Please consider that having such an error fail silently (as suggested by your code above) by returning nil is prone to producing even more complex, more intractable errors.
You could do your override in a similar fashion you used and just write the subscripts in your own way. Only drawback is that existing code will not be compatible. I think to find a hook to override the generic x[i] (also without a text preprocessor as in C) will be challenging.
The closest I can think of is
// compile error:
if theIndex < str.count && let existing = str[theIndex]
EDIT: This actually works. One-liner!!
func ifInBounds(array: [AnyObject], idx: Int) -> AnyObject? {
return idx < array.count ? array[idx] : nil
}
if let x: AnyObject = ifInBounds(swiftarray, 3) {
println(x)
}
else {
println("Out of bounds")
}

I have made a simple extension for array
extension Array where Iterator.Element : AnyObject {
func iof (_ i : Int ) -> Iterator.Element? {
if self.count > i {
return self[i] as Iterator.Element
}
else {
return nil
}
}
}
it works perfectly as designed
Example
if let firstElemntToLoad = roots.iof(0)?.children?.iof(0)?.cNode,

You can try
if index >= 0 && index < array.count {
print(array[index])
}

To be honest I faced this issue too. And from performance point of view a Swift array should be able to throw.
let x = try a[y]
This would be nice and understandable.

When you only need to get values from an array and you don't mind a small performance penalty (i.e. if your collection isn't huge), there is a Dictionary-based alternative that doesn't involve (a too generic, for my taste) collection extension:
// Assuming you have a collection named array:
let safeArray = Dictionary(uniqueKeysWithValues: zip(0..., array))
let value = safeArray[index] ?? defaultValue;

2022
infinite index access and safe idx access(returns nil in case no such idex):
public extension Collection {
subscript (safe index: Index) -> Element? {
return indices.contains(index) ? self[index] : nil
}
subscript (infinityIdx idx: Index) -> Element where Index == Int {
return self[ abs(idx) % self.count ]
}
}
but be careful, it will throw an exception in case of array/collection is empty
usage
(0...10)[safe: 11] // nil
(0...10)[infinityIdx: 11] // 0
(0...10)[infinityIdx: 12] // 1
(0...10)[infinityIdx: 21] // 0
(0...10)[infinityIdx: 22] // 1

Swift 5 Usage
extension WKNavigationType {
var name : String {
get {
let names = ["linkAct","formSubm","backForw","reload","formRelo"]
return names.indices.contains(self.rawValue) ? names[self.rawValue] : "other"
}
}
}
ended up with but really wanted to do generally like
[<collection>][<index>] ?? <default>
but as the collection is contextual I guess it's proper.

Related

How to pop a value from cons list?

In the chapter 15.1 of The Book an example of Box<> usage for recursive type (cons list) implementation is shown. I tried to implement a method for this cons list to pop the outermost value out of the list, leaving the list with whatever left or Nil if nothing left. But it doesn't work, I can't figure out how to return the value while mutating the self after its deconstruction (and so borrowing?). Really none of the references in method make sense to me....
Is there no way to do this without creating a function that consumes the list and spits out both the value and new list?
Here is my code:
use crate::List::{Cons, Nil};
fn main() {
let mut list = Cons(1, Box::new(Cons(2, Box::new(Cons(3, Box::new(Nil))))));
println!("The first value is: {}.", list.pop().unwrap());
println!("The second value is: {}.", list.pop().unwrap());
}
#[derive(Debug)]
enum List {
Cons(i32, Box<List>),
Nil,
}
impl List {
// It seems to me I need to mutably borrow the list to change it
// but the way reference types behave later confuses me
fn pop(&mut self) -> Option<i32> {
if let Cons(value, list) = &self {
self = **list; // <- how to do this bit? self is borrowed...
Some(*value)
} else {
None
}
}
}
You can make your approach work by first moving the current list out of self, and replacing it with Nil. This way, you can match on the old list and still be able to assign to self:
fn pop(&mut self) -> Option<i32> {
let old_list = std::mem::replace(self, Nil);
match old_list {
Cons(value, tail) => {
*self = *tail;
Some(value)
}
Nil => None,
}
}
(Playground)

How to write a bidirectional mapping in Go?

I am writing a simple console game and would like to map a player to a symbol. With two players my approach looks like this:
func playerToString(p player) string {
if p == 0 {
return "X"
}
return "O"
}
func stringToPlayer(s string) player {
if s == "X" {
return 0
}
return 1
}
Of cause you could also write this as two maps mapping int to string and string to int. Both the above approach and the map approach seem error-prone. Is there a more idiomatic way to write this in go? Maybe some non-iota enum way?
[I assume your example is just minimal and that your actual mapping has more than two options. I also assume you meant bi-directonal mapping]
I would write one map:
var player2string = map[int]string{
0: "0",
1: "X",
// etc...
}
And then would create a function to populate a different map string2player programmatically. Something like this:
var player2string = map[int]string{
0: "0",
1: "X",
// etc...
}
var string2player map[string]int = convertMap(player2string)
func convertMap(m map[int]string) map[string]int {
inv := make(map[string]int)
for k, v := range m {
inv[v] = k
}
return inv
}
func main() {
fmt.Println(player2string)
fmt.Println(string2player)
}
Try it on the Go playground
In addition to Eli's answer, two other changes you could make. You could make the to-symbol function a method of the player type. And because the player values are integers (sequential starting from zero), you can use a slice instead of a map to store the int-to-symbol mapping -- it's a bit more efficient to store and for lookup.
type player int
var playerSymbols = []string{"X", "O", "A", "B", "C", "D", "E", "F", "G", "H"}
func (p player) Symbol() string {
if int(p) < 0 || int(p) >= len(playerSymbols) {
return "?" // or panic?
}
return playerSymbols[p]
}
This method signature could even be String() string so it's a fmt.Stringer, which can be useful for printing stuff out and debugging.
Assuming you don't need any specific mapping and the player integer values have the sequence 0,1,2,3,...,25, you can generate the players symbols directly without using the maps as shown in following snippet :-
type player int
func ToSymbol(p player) string {
return fmt.Sprintf("%c", 'A' + p)
}
func ToPlayer(symbol string) player {
return player([]rune(symbol)[0] - 'A')
}

Populating a priority queue based off boolean conditionals

I have a slice that I have to place 5 groupings of entries in.
I have to order everything relative to a one of the groups. So for example there is the base grouping and then each additional group placement is decided by a boolean, so true it needs to be appended or false prepended.
I can do this through a long and nasty looking block of if...else statements, but wondering if there is a better way I can do this semantically? Or an alternative method I should be using for the conditional that can help with this (each grouping can be of a couple or thousands of individual related entries that would be spread out). The final result needs to be a slice, as input into the configuration struct where it lives needs to be a slice.
func buildNewRoutesQueue(overrideGroup1 bool, overrideGroup2 bool, overrideGroup3 bool, overrideGroup4 bool, baseRoutes []*route.Route) []*route.Route {
var result []*route.Route
if overrideGroup1 {
result = append(routes.GetGroup1Routes(), baseRoutes...)
} else {
result = append(baseRoutes, routes.GetGroup1Routes()...)
}
if overrideGroup1 {
result = append(routes.GetGroup2Routes(), result...)
} else {
result = append(result, routes.GetGroup2Routes()...)
}
if overrideGroup3 {
result = append(routes.GetGroup3Routes(), result...)
} else {
result = append(result, routes.GetGroup3Routes()...)
}
if overrideGroup4 {
result = append(routes.GetGroup4Routes(), result...)
} else {
result = append(result, routes.GetGroup4Routes()...)
}
return result
}
Are there any performance issues here, and is there a better semantic way of doing this than what Im thinking? I can replace the boolean conditional with anything, its not set in stone.

Nil element during .sorted in Swift 3

So my problem is that I want to sort chatList from greatest to lesser TimeStamp. That works fine, the problem is when lastMessage is nil.
When is nil, I want to set that element of chatList to the first position and continue the sort.
Is there anyway to tell the sort to put that element in the first index to the first position and keep sorting. If no, then what you guys suggest me to do?
var sortedChatList: < Array AbstractChat>{
get{
return (chatList?.sorted(by:{
//here goes the timestamp comparison
if (($0.lastMessage != nil) && ($1.lastMessage != nil)) {
return Double($0.lastMessage!.timeStamp)! > Double($1.lastMessage!.timeStamp)!
}
else{
return true
}
}))!
}
}
You can use nil-coalescing operator to default nil's to distant future.
The following illustrates the idea:
import Foundation
let foo: [Date?] = [Date(), nil, Date(timeIntervalSinceNow: 100)]
let bar = foo.sorted(by: {
($0 ?? Date.distantFuture) > ($1 ?? Date.distantFuture)
})
bar.forEach({
print("\($0)")
})

Why does Linq know to increment

I am under the impression, code such as
var result = myList.Select(a => a == "Hello")
is the same as
public string Select(string a)
{
//logic
}
Therefore if I did
var result = myList.Select((a, v) => a == "Hello")
It would 'translate' to
public string Select(string a, object v)
{
//logic
}
I'm assuming my logic is wrong, because if you consider
var index = myList.Select((val, ind) => new { val, ind }).Single(a => a.v.Thing == "Condition").ind;
Then this returns the index. It's effectively done
public string Select(this List<string> s, string val, int ind)
{
//for loop, and keep track of the iteration
}
But how does Linq know my second value is an iteration? Surely all it knows is that I'm passing 2 values.
Select works more like a projection. When you break down your second example, it might get clearer:
var index = myList // a list of strings or, more general, an enumerable of string
.Select((val, ind) => new { val, ind }) // enumerable of anon. type with val and index
.Single(a => a.v.Thing == "Condition") // any element of enumerable matching condition
.ind; // the index of this element
The second optional parameter is just a counter. If you would do it 'traditionaly' with a for loop it would be your index. This comes in handy sometimes (as in your example) but I rarely use it. See here at MSDN:
selector Type: System.Func A transform
function to apply to each source element; the second parameter of the
function represents the index of the source element.

Resources