This repository contains sample code used in the Flight School Guide to Swift Strings.
- Chapter 2: Working with Strings in Swift
- Chapter 3: Swift String Protocols and Supporting Types
- Chapter 4: Working with Foundation String APIs
- Chapter 5: Binary-to-Text Encoding
- Chapter 6: Parsing Data From Text
- Chapter 7: Natural Language Processing
You can construct string values in Swift using string literals. This Playground has examples of each variety, from the conventional, single-line to the raw, multi-line.
let multilineRawString = #"""
\-----------------------\
\ \
\ ___ \
\ (_ /'_ /_/ \ __
\ / (/(//)/ \ | \
> _/ >------| \ ______
/ __ / --- \_____/**|_|_\____ |
/ ( _ / / / \_______ --------- __>-}
/ __)( /)()()( / / \_____|_____/ |
/ / * |
/-----------------------/ {o}
"""#
Swift strings have opaque index types. One consequence of this is that you can't access character by integer position directly, as you might in other languages. This Playground shows various strategies for working with string indices and ranges.
let string = "Hello"
string[string.startIndex] // "H"
string[string.index(after: string.startIndex)] // "e"
string[string.index(string.startIndex, offsetBy: 4)] // "o"
In Swift,
two String
values are considered equal if they are
canonically equivalent,
even if they comprise different Unicode scalar values.
let precomposed = "expos\u{00E9}" // é LATIN SMALL LETTER E WITH ACUTE
let decomposed = "expose\u{0301}" // ´ COMBINING ACUTE ACCENT
precomposed == decomposed
precomposed.elementsEqual(decomposed) // true
precomposed.unicodeScalars.elementsEqual(decomposed.unicodeScalars) // false
Swift String
values
provide views to their UTF-8, UTF-16, and UTF-32 code units.
This Playground shows the correspondence between
the characters in a string and their various encoding forms.
let string = "東京 🇯🇵"
for unicodeScalar in character.unicodeScalars {
print(unicodeScalar.codePoint, terminator: "\t")
}
In Swift 5,
you can access several Unicode properties of Character
values,
which allow you to determine things like
Unicode general category membership,
whether a character has case mapping (lowercase / uppercase / titlecase),
and whether the character has an associated number value.
// U+2460 CIRCLED DIGIT ONE
("①" as Character).isNumber // true
("①" as Character).isWholeNumber // true
("①" as Character).wholeNumberValue // 1
For more direct access to the aforementioned character information,
you can do so through the properties
property on Unicode.Scalar
values.
For example,
the isEmoji
property does...
well, exactly what you'd expect it to do.
("👏" as Unicode.Scalar).properties.isEmoji // true
In Swift,
String
functionality is inherited from
a complex hierarchy of interrelated protocols,
including
Sequence
,
Collection
,
BidirectionalCollection
,
RangeReplaceableCollection
,
StringProtocol
,
and others.
Each of the protocols mentioned has their own Playground demonstrating the specific functionality they provide.
"Boeing 737-800".filter { $0.isCased }
.map { $0.uppercased() }
["B", "O", "E", "I", "N", "G"]
The print
function can direct its output
to a custom type conforming to the TextOutputStream
protocol.
This example implements a logger
that prints the Unicode code points of the provided string.
var logger = UnicodeLogger()
print("👨👩👧👧", to: &logger)
// 0: 👨 U+1F468 MAN
// 1: U+200D ZERO WIDTH JOINER
// 2: 👩 U+1F469 WOMAN
// 3: U+200D ZERO WIDTH JOINER
// 4: 👧 U+1F467 GIRL
// 5: U+200D ZERO WIDTH JOINER
// 6: 👧 U+1F467 GIRL
Text output streams can also be used to
direct print statements from the default stdout
destination.
In this example,
the print
function is directed to write to stderr
.
var standardError = StderrOutputStream()
print("Error!", to: &standardError)
Swift allows any type that conforms to ExpressibleByStringLiteral
to be initialized from a string literal.
This Playground provides a simple example through the BookingClass
type.
("J" as BookingClass) // Business Class
Types conforming to the LosslessStringConvertible
protocol
can be initialized directly from String
values.
This Playground shows a FlightCode
type that adopts both
the LosslessStringConvertible
and ExpressibleByStringLiteral
protocols.
let flight: FlightCode = "AA 1"
flight.airlineCode
flight.flightNumber
FlightCode(String(flight))
Swift 5 makes it possible to customize
the behavior of interpolation in string literals
by way of the ExpressibleByStringInterpolation
protocol.
To demonstrate this,
we implement a StyledString
type that
allows interpolation segments to specify a style,
such as bold, italic, and 𝔣𝔯𝔞𝔨𝔱𝔲𝔯.
let name = "Johnny"
let styled: StyledString = """
Hello, \(name, style: .fraktur(bold: true))!
"""
print(styled)
Objective-C APIs that take NSString
parameters
or have NSString
return values
are imported by Swift to use String
values instead.
However, some of these APIs still specify ranges using the NSRange
type
instead of Range<String.Index>
.
This Playground demonstrates how to convert back and forth
between the two range types.
import Foundation
let string = "Hello, world!"
let nsRange = NSRange(string.startIndex..<string.endIndex, in: string)
let range = Range(nsRange, in: string)
Foundation augments the Swift String
type
by providing localized string operations,
including
case mapping,
searching,
and comparison.
Be sure to use localized string operations
(ideally, the standard
variant, if applicable)
when working with text written or read by users.
import Foundation
"Éclair".contains("E") // false
"Éclair".localizedStandardContains("E") // true
Another consideration for localized string sorting is how to handle numbers.
By default, strings sort digits lexicographically;
7 follows 3, but 7 also follows 36.
This Playground demonstrates proper use of the
localizedStandardCompare
comparator,
which is what Finder uses to sort filenames.
import Foundation
let files: [String] = [
"File 3.txt",
"File 7.txt",
"File 36.txt"
]
let order: ComparisonResult = .orderedAscending
files.sorted { lhs, rhs in
lhs.localizedStandardCompare(rhs) == order
}
// ["File 3.txt", "File 7.txt", "File 36.txt"]
Foundation provides APIs for accessing normalization forms for strings, including NFC and NFD, as demonstrated in this example.
import Foundation
let string = "ümlaut"
let nfc = string.precomposedStringWithCanonicalMapping
nfc.unicodeScalars.first
let nfd = string.decomposedStringWithCanonicalMapping
nfd.unicodeScalars.first
Foundation offers support for many different legacy string encodings, as shown in this example.
import Foundation
"Hello, Macintosh!".data(using: .macOSRoman)
Foundation provides APIs to read and write String
values
from data values and files.
import Foundation
let url = Bundle.main.url(forResource: "file", withExtension: "txt")!
try String(contentsOf: url) // "Hello!"
let data = try Data(contentsOf: url)
String(data: data, encoding: .utf8) // "Hello!"
Another cool bit of functionality String
inherits from NSString
is the ability to apply
ICU string transforms,
as seen in this example.
import Foundation
"Avión".applyingTransform(.stripDiacritics, reverse: false)
// "Avion"
"©".applyingTransform(.toXMLHex, reverse: false)
// "©"
"🛂".applyingTransform(.toUnicodeName, reverse: false)
// "\\N{PASSPORT CONTROL}"
"マット".applyingTransform(.fullwidthToHalfwidth, reverse: false)
// "マット"
Foundation's CharacterSet
is used in various string APIs,
but it's perhaps most well-known for its role in the
trimmingCharacters(in:)
method,
as shown in this Playground.
import Foundation
"""
✈️
""".trimmingCharacters(in: .whitespacesAndNewlines) // "✈️"
Only certain characters are allowed in certain positions of a URLs.
By importing Foundation,
you can encode URL query parameters with confidence
with the addingPercentEncoding(withAllowedCharacters:)
method.
import Foundation
"q=lax to jfk".addingPercentEncoding(withAllowedCharacters: .urlQueryAllowed)
// q=lax%20to%20jfk
When you import the Foundation framework,
String
gets sprintf
-style initializers.
This Playground serves as an exhaustive reference
for all of the available
formatting specifiers, modifiers, flags, and arguments.
import Foundation
String(format: "%X", 127) // "7F"
These examples show you how to
use the String(_:radix:uppercase:)
initializer to
produce binary and hexadecimal representations of binary integer values.
let byte: UInt8 = 0xF0
String(byte, radix: 2) // "11110000"
String(byte, radix: 16, uppercase: true) // "F0"
Foundation provides APIs for base64 encoding and decoding data, which are demonstrated in this Playground.
import Foundation
let string = "Hello!"
let data = string.data(using: .utf8)!
let encodedString = data.base64EncodedString() // "SGVsbG8h"
Anticipating emoji's role in the forthcoming collapse of human communication, we present a novel binary-to-text encoding format that represents data using human face emoji combined with skin tone and hair style modifiers.
let data = "Fly".data(using: .utf8)!
let encodedString = data.base🧑EncodedString() // "👨🏽🦱👩🏻🦲👩🏽🦳👩🏿🦱"
In this example, we implement the 11-bit binary-to-text encoding described in RFC 1751: "A Convention for Human-Readable 128-bit Keys". "Why?" you ask? Why indeed!
import Foundation
let data = Data(bytes: [0xB2, 0x03, 0xE2, 0x8F,
0xA5, 0x25, 0xBE, 0x47])
data.humanReadableEncodedString()
// "LONG IVY JULY AJAR BOND LEE"
One of Foundation's many offerings is the Scanner
class:
a sort of lexer/parser combo deal with some convenient features.
This Playground demonstrates how to make it even more convenient in Swift,
and how to use it to parse information from an AFTN message.
import Foundation
let scanner = Scanner(string: string)
scanner.charactersToBeSkipped = .whitespacesAndNewlines
try scanner.scan("ZCZC")
let transmission = try scanner.scan(.alphanumerics)
let additionalServices = try scanner.scan(.decimalDigits)
let priority = try scanner.scan(.uppercaseLetters)
let destination = try scanner.scan(.uppercaseLetters)
let time = try scanner.scan(.decimalDigits)
let origin = try scanner.scan(.uppercaseLetters)
let text = try scanner.scan(upTo: "NNNN")
Foundation's NSRegularExpression
offers the closest thing to
built-in regex support in Swift.
Underneath the hood, it wraps the
ICU regular expression engine;
we take advantage of a bunch of its advanced features in this Playground
to parse the same message as before using a different approach.
import Foundation
let pattern = #"""
(?x-i)
\A
ZCZC \h
(?<transmission>[A-Z]{3}[0-9]{3}) \h (?<additionalService>[0-9]{0,8}) \n
(?<priority>[A-Z]{2}) \h (?<destination>[A-Z]{8}) \n
(?<time>[0-9]{6}) \h (?<origin>[A-Z]{8}) \n+
(?<text>[[A-Z][0-9]\h\n]+) \s*
NNNN
\Z
"""#
let regex = try NSRegularExpression(pattern: pattern,
options: [])
ANTLR is a parser generator with support for Swift code generation. This example provides a functional integration between ANTLR4 and the Swift Package Manager to demonstrate yet another approach to parsing the same AFTN message from the previous examples.
import AFTN
let message = try Message(string)!
message.priority
message.destination.location
message.destination.organization
message.destination.department
message.filingTime
message.text
The NaturalLanguage framework's NLTokenizer
class
can tokenize text by word, sentence, and paragraph,
as demonstrated in this example.
import NaturalLanguage
let string = "Welcome to New York, where the local time is 9:41 AM."
let tokenizer = NLTokenizer(unit: .word)
tokenizer.string = string
let stringRange = string.startIndex..<string.endIndex
tokenizer.enumerateTokens(in: stringRange) { (tokenRange, _) in
let token = string[tokenRange]
print(token, terminator: "\t")
return true // continue processing
}
// Prints: "Welcome to New York where the local time is 9 41 AM "
You can use the NLTagger
class to detect the language and script
for a piece of natural language text,
as seen in this Playground.
import NaturalLanguage
let string = """
Sehr geehrte Damen und Herren,
herzlich willkommen in Frankfurt.
"""
let tagSchemes: [NLTagScheme] = [.language, .script]
let tagger = NLTagger(tagSchemes: tagSchemes)
tagger.string = string
for scheme in tagSchemes {
if case let (tag?, _) = tagger.tag(at: string.startIndex,
unit: .word,
scheme: scheme) {
print(scheme.rawValue, tag.rawValue)
}
}
// Prints:
// "Language de"
// "Script Latn"
To tag part of speech for words (noun, verb, etc.)
use the NLTagger
class with the .lexicalClass
tag scheme.
import NaturalLanguage
let string = "The sleek white jet soars over the hazy fog."
let tagger = NLTagger(tagSchemes: [.lexicalClass])
tagger.string = string
let stringRange = string.startIndex..<string.endIndex
let options: NLTagger.Options = [.omitWhitespace, .omitPunctuation]
tagger.enumerateTags(in: stringRange,
unit: .word,
scheme: .lexicalClass,
options: options) { (tag, tagRange) in
if let partOfSpeech = tag?.rawValue {
print("\(string[tagRange]): \(partOfSpeech)")
}
return true // continue processing
}
// Prints:
// "The: Determiner"
// "sleek: Adjective"
// "white: Adjective"
// "jet: Noun"
// ...
NLTagger
can also be used to detect named entities,
including people, places, and organizations.
This example shows how to do just that.
import NaturalLanguage
let string = """
Fang Liu of China is the current Secretary General of ICAO.
"""
let tagger = NLTagger(tagSchemes: [.nameType])
tagger.string = string
let stringRange = string.startIndex..<string.endIndex
let options: NLTagger.Options = [.omitWhitespace, .omitPunctuation, .joinNames]
tagger.enumerateTags(in: stringRange,
unit: .word,
scheme: .nameType,
options: options) { (tag, tagRange) in
if let nameType = tag?.rawValue, tag != .otherWord {
print("\(string[tagRange]): \(nameType)")
}
return true // continue processing
}
// Prints:
// "Fang Liu: PersonalName"
// "China: PlaceName"
// "ICAO: OrganizationName"
Short of implementing a more complete natural language parser,
you can use NLTagger
to extract keywords by part of speech
as a first approximation for interpreting commands.
import NaturalLanguage
let string = "What's the current temperature in Tokyo?"
let tagger = NLTagger(tagSchemes: [.nameTypeOrLexicalClass])
tagger.string = string
var taggedKeywords: [(NLTag, String)] = []
let stringRange = string.startIndex..<string.endIndex
let options: NLTagger.Options = [.omitWhitespace,
.omitPunctuation,
.joinNames]
tagger.enumerateTags(in: stringRange,
unit: .word,
scheme: .nameTypeOrLexicalClass,
options: options) { (tag, tagRange) in
guard let tag = tag else { return true }
switch tag {
case .noun, .placeName:
print(tag.rawValue, String(string[tagRange]))
default:
break
}
return true // continue processing
}
// Prints:
// "Noun temperature"
// "PlaceName Tokyo"
This example demonstrates the .lemma
tag scheme
and how it resolves conjugations of various words.
import NaturalLanguage
let string = """
Flying flights fly flyers flown.
"""
let tagger = NLTagger(tagSchemes: [.lemma])
tagger.string = string
tagger.enumerateTags(in: string.startIndex..<string.endIndex,
unit: .word,
scheme: .lemma,
options: []) { (tag, tagRange) in
if let lemma = tag?.rawValue {
print("\(string[tagRange]): \(lemma)")
}
return true // continue processing
}
// Prints:
// "Flying: fly"
// "flights: flight"
// "fly: fly"
// "flyers: flyer"
// "flown: fly"
The NLLanguageRecognizer
provides a configurable classifier
for determining the language used in a piece of text.
Here, we demonstrate how to use the languageHints
property
to resolve a sentence that could be understood in either
Norwegian Bokmål (nb
) or Danish (da
).
import NaturalLanguage
let string = """
God morgen mine damer og herrer.
"""
let languageRecognizer = NLLanguageRecognizer()
languageRecognizer.processString(string)
languageRecognizer.dominantLanguage // da
languageRecognizer.languageHints = [.norwegian: 0.75,
.swedish: 0.25]
languageRecognizer.dominantLanguage // nb
This example provides a reference implementation for a Naive Bayes "bag of words" classifier in Swift.
enum Sentiment: String, Hashable {
case positive, negative
}
let classifier = NaiveBayesClassifier<Sentiment, String>()
classifier.trainText("great flight", for: .positive)
classifier.trainText("flight was late and turbulent", for: .negative)
classifier.classifyText("I had a great flight") // positive
Using Create ML, we can build a Core ML classifier model that can be used by the Natural Language framework to determine if a piece of natural language text expresses positive, negative, or neutral sentiment.
import NaturalLanguage
let url = Bundle.main.url(forResource: "SentimentClassifier",
withExtension: "mlmodelc")!
let model = try NLModel(contentsOf: url)
model.predictedLabel(for: "Nice, smooth flight") // positive
This Playground provides a Swift implementation of n-grams,
which, combined with NLTokenizer
,
can produce bigrams and trigrams of words in a piece of natural language text.
import NaturalLanguage
let string = """
Please direct your attention to flight attendants
as we review the safety features of this aircraft.
"""
let tokenizer = NLTokenizer(unit: .word)
tokenizer.string = string
let words = tokenizer.tokens(for: string.startIndex..<string.endIndex)
.map { String(string[$0]) }
bigrams(words)
// [("Please", "direct"), ("direct", "your"), ...]
Using n-grams to determine the conditional probability of transitions from one word to another, we can construct a model that randomly generates text that trivially resembles the provided source. In this example, we feed in a corpus of Air Traffic Control transcripts.
import Foundation
import NaturalLanguage
// https://catalog.ldc.upenn.edu/LDC94S14A
let url = Bundle.main.url(forResource: "LDC94S14A-sample",
withExtension: "txt")!
let text = try String(contentsOf: url)
var markovChain = MarkovChain(sentencesAndWords(for: text))
for word in markovChain {
print(word, terminator: " ")
}
// Prints: "CACTUS EIGHT OH EIGHT TURN LEFT HEADING ONE SEVENTY HEAVY"
Soundex is a classic phonetic coding system used to resolve ambiguity in the spelling of surnames. This example provides a Swift implementation of the standard algorithm.
let names: [String] = [
"Washington",
"Lee",
"Smith",
"Smyth"
]
for name in names {
print("\(name): \(soundex(name))")
}
// Prints:
// "Washington: W252"
// "Lee: L000"
// "Smith: S530"
// "Smyth: S530"
You can use a string metric like Levenshtein edit distance to quantify the similarity between two sequences.
/*
| | | S | a | t | u | r | d | a | y |
|-----|-----|-----|-----|-----|-----|-----|-----|-----|-----|
| | _0_ | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 |
| S | 1 | _0_ | _1_ | _2_ | 3 | 4 | 5 | 6 | 7 |
| u | 2 | 1 | 1 | 2 | _2_ | 3 | 4 | 5 | 6 |
| n | 3 | 2 | 2 | 2 | 3 | _3_ | 4 | 5 | 6 |
| d | 4 | 3 | 3 | 3 | 3 | 4 | _3_ | 4 | 5 |
| a | 5 | 4 | 3 | 4 | 4 | 4 | 4 | _3_ | 4 |
| y | 6 | 5 | 4 | 4 | 5 | 5 | 5 | 4 | _3_ |
*/
levenshteinDistance(from: "Saturday", to: "Sunday") // 3
Using the Levenshtein distance function from the previous example, and combining it with a corpus of frequently-used words, you can create a reasonably effective spell checker with very little additional code.
import Foundation
// https://catalog.ldc.upenn.edu/LDC2006T13
guard let url = Bundle.main.url(forResource: "LDC2006T13-sample",
withExtension: "txt")
else {
fatalError("Missing required resource")
}
let spellChecker = try SpellChecker(contentsOf: url)
spellChecker.suggestions(for: "speling")
// ["spelling", "spewing", "sperling"]
MIT
Flight School is a book series for advanced Swift developers that explores essential topics in iOS and macOS development through concise, focused guides.
If you'd like to get in touch, feel free to message us on Twitter or email us at [email protected].