HTML to plain text conversion

Sources/HTMLStreamer/AttributedStringConverter.swift

@@ -17,7 +17,7 @@ private typealias PlatformFont = UIFont
 private typealias PlatformFont = NSFont
 #endif
 
-public struct AttributedStringConverter<Callbacks: AttributedStringCallbacks> {
+public struct AttributedStringConverter<Callbacks: HTMLConversionCallbacks> {
     private let configuration: AttributedStringConverterConfiguration
     private var fontCache: [FontTrait: PlatformFont] = [:]
     
@@ -58,10 +58,7 @@ public struct AttributedStringConverter<Callbacks: AttributedStringCallbacks> {
             case .startTag(let name, let selfClosing, let attributes):
                 let action = Callbacks.elementAction(name: name, attributes: attributes)
                 actionStack.append(action)
-                // self closing tags are ignored since they have no content
-                if !selfClosing {
-                    handleStartTag(name, attributes: attributes)
-                }
+                handleStartTag(name, selfClosing: selfClosing, attributes: attributes)
             case .endTag(let name):
                 handleEndTag(name)
                 // if we have a non-default action for the current element, the run finishes here
@@ -80,10 +77,17 @@ public struct AttributedStringConverter<Callbacks: AttributedStringCallbacks> {
         return str
     }
     
-    private mutating func handleStartTag(_ name: String, attributes: [HTMLStreamer.Attribute]) {
-        switch name {
-        case "br":
+    private mutating func handleStartTag(_ name: String, selfClosing: Bool, attributes: [Attribute]) {
+        if name == "br" {
             currentRun.append("\n")
+            return
+        }
+        // self closing tags are ignored since they have no content
+        guard !selfClosing else {
+            return
+        }
+        
+        switch name {
         case "a":
             // we need to always insert in attribute, because we need to always have one
             // to remove from the stack in handleEndTag
@@ -195,7 +199,6 @@ public struct AttributedStringConverter<Callbacks: AttributedStringCallbacks> {
     private mutating func finishBlockElement() {
         if str.length != 0 {
             previouslyFinishedBlockElement = true
-//            currentRun.append("\n\n")
         }
     }
     
@@ -310,37 +313,6 @@ public struct AttributedStringConverter<Callbacks: AttributedStringCallbacks> {
     }
 }
 
-public protocol AttributedStringCallbacks {
-    static func makeURL(string: String) -> URL?
-    static func elementAction(name: String, attributes: [Attribute]) -> ElementAction
-}
-
-public enum ElementAction: Equatable {
-    case `default`
-    case skip
-    case replace(String)
-    
-    var isReplace: Bool {
-        if case .replace(_) = self {
-            true
-        } else {
-            false
-        }
-    }
-}
-
-public extension AttributedStringCallbacks {
-    static func makeURL(string: String) -> URL? {
-        URL(string: string)
-    }
-    static func elementAction(name: String, attributes: [Attribute]) -> ElementAction {
-        .default
-    }
-}
-
-public struct DefaultCallbacks: AttributedStringCallbacks {
-}
-
 public struct AttributedStringConverterConfiguration {
     #if os(iOS)
     public var font: UIFont

Sources/HTMLStreamer/HTMLConversionCallbacks.swift

@@ -0,0 +1,39 @@
+//
+//  HTMLConversionCallbacks.swift
+//  HTMLStreamer
+//
+//  Created by Shadowfacts on 12/22/23.
+//
+
+import Foundation
+
+public protocol HTMLConversionCallbacks {
+    static func makeURL(string: String) -> URL?
+    static func elementAction(name: String, attributes: [Attribute]) -> ElementAction
+}
+
+public enum ElementAction: Equatable {
+    case `default`
+    case skip
+    case replace(String)
+    
+    var isReplace: Bool {
+        if case .replace(_) = self {
+            true
+        } else {
+            false
+        }
+    }
+}
+
+public extension HTMLConversionCallbacks {
+    static func makeURL(string: String) -> URL? {
+        URL(string: string)
+    }
+    static func elementAction(name: String, attributes: [Attribute]) -> ElementAction {
+        .default
+    }
+}
+
+public struct DefaultCallbacks: HTMLConversionCallbacks {
+}

Sources/HTMLStreamer/TextConverter.swift

@@ -0,0 +1,135 @@
+//
+//  TextConverter.swift
+//  HTMLStreamer
+//
+//  Created by Shadowfacts on 12/19/23.
+//
+
+import Foundation
+
+public struct TextConverter<Callbacks: HTMLConversionCallbacks> {
+    
+    private let configuration: TextConverterConfiguration
+    
+    private var tokenizer: Tokenizer<String.UnicodeScalarView.Iterator>!
+    private var str: String!
+    
+    private var actionStack: [ElementAction] = []
+    private var previouslyFinishedBlockElement = false
+    private var currentRun = ""
+    
+    public init(configuration: TextConverterConfiguration = .init()) where Callbacks == DefaultCallbacks {
+        self.init(configuration: configuration, callbacks: DefaultCallbacks.self)
+    }
+    
+    public init(configuration: TextConverterConfiguration = .init(), callbacks _: Callbacks.Type = Callbacks.self) {
+        self.configuration = configuration
+    }
+    
+    public mutating func convert(html: String) -> String {
+        tokenizer = Tokenizer(chars: html.unicodeScalars.makeIterator())
+        str = ""
+        
+        while let token = tokenizer.next() {
+            switch token {
+            case .character(let scalar):
+                currentRun.unicodeScalars.append(scalar)
+            case .characterSequence(let string):
+                currentRun.append(string)
+            case .startTag(let name, let selfClosing, let attributes):
+                let action = Callbacks.elementAction(name: name, attributes: attributes)
+                actionStack.append(action)
+                handleStartTag(name, selfClosing: selfClosing, attributes: attributes)
+            case .endTag(let name):
+                handleEndTag(name)
+                if actionStack.last != .default {
+                    finishRun()
+                }
+                actionStack.removeLast()
+            case .comment, .doctype:
+                break
+            }
+        }
+        
+        finishRun()
+        
+        return str
+    }
+    
+    private mutating func handleStartTag(_ name: String, selfClosing: Bool, attributes: [Attribute]) {
+        switch name {
+        case "br":
+            if configuration.insertNewlines {
+                currentRun.append("\n")
+            } else {
+                currentRun.append(" ")
+            }
+        case "pre", "blockquote", "p", "ol", "ul":
+            startBlockElement()
+            finishRun()
+        default:
+            break
+        }
+    }
+    
+    private mutating func startBlockElement() {
+        if !str.isEmpty {
+            previouslyFinishedBlockElement = false
+            if configuration.insertNewlines {
+                currentRun.append("\n\n")
+            } else {
+                currentRun.append(" ")
+            }
+        }
+    }
+    
+    private mutating func handleEndTag(_ name: String) {
+        switch name {
+        case "pre", "blockquote", "p", "ol", "ul":
+            finishRun()
+            finishBlockElement()
+        default:
+            break
+        }
+    }
+    
+    private mutating func finishBlockElement() {
+        if !str.isEmpty {
+            previouslyFinishedBlockElement = true
+        }
+    }
+    
+    private mutating func finishRun() {
+        guard !currentRun.isEmpty else {
+            return
+        }
+        
+        if actionStack.contains(.skip) {
+            currentRun = ""
+            return
+        } else if case .replace(let replacement) = actionStack.first(where: \.isReplace) {
+            currentRun = replacement
+        }
+        
+        if previouslyFinishedBlockElement {
+            previouslyFinishedBlockElement = false
+            if configuration.insertNewlines {
+                currentRun.insert(contentsOf: "\n\n", at: currentRun.startIndex)
+            } else {
+                currentRun.insert(" ", at: currentRun.startIndex)
+            }
+        }
+        
+        str.append(currentRun)
+        currentRun = ""
+    }
+    
+}
+
+public struct TextConverterConfiguration {
+    public var insertNewlines: Bool
+    
+    public init(insertWhitespace: Bool = true) {
+        self.insertNewlines = insertWhitespace
+    }
+}

Tests/HTMLStreamerTests/AttributedStringConverterTests.swift

@@ -41,7 +41,7 @@ final class AttributedStringConverterTests: XCTestCase {
         convert(html, callbacks: DefaultCallbacks.self)
     }
     
-    private func convert<Callbacks: AttributedStringCallbacks>(_ html: String, callbacks _: Callbacks.Type = Callbacks.self) -> NSAttributedString {
+    private func convert<Callbacks: HTMLConversionCallbacks>(_ html: String, callbacks _: Callbacks.Type = Callbacks.self) -> NSAttributedString {
         let config = AttributedStringConverterConfiguration(
             font: font,
             monospaceFont: monospaceFont,
@@ -57,6 +57,10 @@ final class AttributedStringConverterTests: XCTestCase {
             .font: font,
             .paragraphStyle: NSParagraphStyle.default,
         ]))
+        XCTAssertEqual(convert("a<br />b"), NSAttributedString(string: "a\nb", attributes: [
+            .font: font,
+            .paragraphStyle: NSParagraphStyle.default,
+        ]))
     }
     
     func testConvertA() {
@@ -208,7 +212,7 @@ final class AttributedStringConverterTests: XCTestCase {
     }
     
     func testMakeURLCallback() {
-        struct Callbacks: AttributedStringCallbacks {
+        struct Callbacks: HTMLConversionCallbacks {
             static func makeURL(string: String) -> URL? {
                 URL(string: "https://apple.com")
             }
@@ -222,7 +226,7 @@ final class AttributedStringConverterTests: XCTestCase {
     }
     
     func testElementActionCallback() {
-        struct Callbacks: AttributedStringCallbacks {
+        struct Callbacks: HTMLConversionCallbacks {
             static func elementAction(name: String, attributes: [Attribute]) -> ElementAction {
                 let clazz = attributes.attributeValue(for: "class")
                 if clazz == "invisible" {
@@ -236,8 +240,8 @@ final class AttributedStringConverterTests: XCTestCase {
         }
         let skipped = convert("<span class='invisible'>test</span>", callbacks: Callbacks.self)
         XCTAssertEqual(skipped, NSAttributedString())
-        let skipNestped = convert("<span class='invisible'><b>test</b></span>", callbacks: Callbacks.self)
-        XCTAssertEqual(skipNestped, NSAttributedString())
+        let skipNested = convert("<span class='invisible'><b>test</b></span>", callbacks: Callbacks.self)
+        XCTAssertEqual(skipNested, NSAttributedString())
         let skipNestped2 = convert("<b><span class='invisible'>test</span></b>", callbacks: Callbacks.self)
         XCTAssertEqual(skipNestped2, NSAttributedString())
         let replaced = convert("<span class='ellipsis'>test</span>", callbacks: Callbacks.self)

Tests/HTMLStreamerTests/TextConverterTests.swift

@@ -0,0 +1,66 @@
+//
+//  TextConverterTests.swift
+//  
+//
+//  Created by Shadowfacts on 12/22/23.
+//
+
+import XCTest
+@testable import HTMLStreamer
+
+final class TextConverterTests: XCTestCase {
+
+    private func convert(_ html: String, configuration: TextConverterConfiguration = .init()) -> String {
+        convert(html, configuration: configuration, callbacks: DefaultCallbacks.self)
+    }
+    
+    private func convert<Callbacks: HTMLConversionCallbacks>(_ html: String, configuration: TextConverterConfiguration = .init(), callbacks _: Callbacks.Type = Callbacks.self) -> String {
+        var converter = TextConverter<Callbacks>(configuration: configuration)
+        return converter.convert(html: html)
+    }
+    
+    func testConvertBR() {
+        XCTAssertEqual(convert("a<br>b"), "a\nb")
+        XCTAssertEqual(convert("a<br />b"), "a\nb")
+    }
+    
+    func testConvertA() {
+        XCTAssertEqual(convert("<a href='https://example.com'>link</a>"), "link")
+    }
+    
+    func testIncorrectNesting() {
+        XCTAssertEqual(convert("<strong>bold <em>both</strong> italic</em>"), "bold both italic")
+    }
+    
+    func testTextAfterBlockElement() {
+        XCTAssertEqual(convert("<blockquote>wee</blockquote>after"), "wee\n\nafter")
+        XCTAssertEqual(convert("<blockquote>wee</blockquote>after", configuration: .init(insertWhitespace: false)), "wee after")
+    }
+    
+    func testMultipleBlockElements() {
+        XCTAssertEqual(convert("<blockquote>a</blockquote><blockquote>b</blockquote>"), "a\n\nb")
+        XCTAssertEqual(convert("<blockquote>a</blockquote><blockquote>b</blockquote>", configuration: .init(insertWhitespace: false)), "a b")
+    }
+    
+    func testElementActionCallback() {
+        struct Callbacks: HTMLConversionCallbacks {
+            static func elementAction(name: String, attributes: [Attribute]) -> ElementAction {
+                let clazz = attributes.attributeValue(for: "class")
+                if clazz == "invisible" {
+                    return .skip
+                } else if clazz == "ellipsis" {
+                    return .replace("…")
+                } else {
+                    return .default
+                }
+            }
+        }
+        let skipped = convert("<span class='invisible'>test</span>", callbacks: Callbacks.self)
+        XCTAssertEqual(skipped, "")
+        let skipNested = convert("<span class='invisible'><b>test</b></span>", callbacks: Callbacks.self)
+        XCTAssertEqual(skipNested, "")
+        let replaced = convert("<span class='ellipsis'>test</span>", callbacks: Callbacks.self)
+        XCTAssertEqual(replaced, "…")
+    }
+    
+}