From 1687136e033f3f73494f228d83c80702933616a5 Mon Sep 17 00:00:00 2001
From: Matt <85322+mattmassicotte@users.noreply.github.com>
Date: Wed, 1 May 2024 13:18:50 -0400
Subject: [PATCH] Some actual parsing

---
 Package.swift                            |  10 ++
 Sources/Lowlight/Language.swift          |  61 ++++++++--
 Sources/Lowlight/Processor.swift         | 138 ++++++++++++++++++++---
 Tests/LowlightTests/LanguageTests.swift  |   2 +-
 Tests/LowlightTests/ProcessorTests.swift |  92 +++++++++++++++
 5 files changed, 274 insertions(+), 29 deletions(-)
 create mode 100644 Tests/LowlightTests/ProcessorTests.swift

diff --git a/Package.swift b/Package.swift
index 1581d3c..14920f4 100644
--- a/Package.swift
+++ b/Package.swift
@@ -27,3 +27,13 @@ let package = Package(
 		),
 	]
 )
+
+let swiftSettings: [SwiftSetting] = [
+	.enableExperimentalFeature("StrictConcurrency")
+]
+
+for target in package.targets {
+	var settings = target.swiftSettings ?? []
+	settings.append(contentsOf: swiftSettings)
+	target.swiftSettings = settings
+}
diff --git a/Sources/Lowlight/Language.swift b/Sources/Lowlight/Language.swift
index 70d35a0..5ac2933 100644
--- a/Sources/Lowlight/Language.swift
+++ b/Sources/Lowlight/Language.swift
@@ -1,20 +1,59 @@
 import Foundation
 
-/// The model of the Lowlight language.
-public struct Language: Codable, Hashable, Sendable {
-	public let keywords: Set<String>
-	public let symbols: Set<String>
+public struct Pattern: Hashable, Sendable {
+	public enum Behavior: Hashable, Sendable {
+		case exact(String)
+		case toEndOfLine(String)
+
+		public func matches(_ character: Character, at offset: Int) -> Bool {
+			switch self {
+			case let .exact(string):
+				let patternIdx = string.index(string.startIndex, offsetBy: offset)
 
-	public init(keywords: Set<String> = [], symbols: Set<String> = []) {
-		self.keywords = keywords
-		self.symbols = symbols
+				return string[patternIdx] == character
+			case let .toEndOfLine(string):
+				let patternIdx = string.index(string.startIndex, offsetBy: offset)
+
+				return string[patternIdx] == character
+			}
+		}
 	}
 
-	var keywordPattern: String {
-		"(" + keywords.joined(separator: "|") + ")"
+	public enum Element: Hashable, Sendable {
+		case keyword
+		case comment
+
+		public var treeSitterHighlightName: String {
+			switch self {
+			case .comment:
+				"comment"
+			case .keyword:
+				"keyword"
+			}
+		}
+	}
+
+	public let element: Element
+	public let match: Behavior
+
+	public init(element: Element, match: Behavior) {
+		self.element = element
+		self.match = match
 	}
+}
+
+/// The model of the Lowlight language.
+public struct Language: Hashable, Sendable {
+	public let patterns: Set<Pattern>
+
+	public init(patterns: Set<Pattern>) {
+		self.patterns = patterns
+	}
+
+	public init(keywords: Set<String>, lineComment: String) {
+		let keywordPatterns = keywords.map({ Pattern(element: .keyword, match: .exact($0)) })
+		let patterns = keywordPatterns + [.init(element: .comment, match: .toEndOfLine(lineComment))]
 
-	var symbolsPattern: String {
-		"(" + symbols.joined(separator: "|") + ")"
+		self.init(patterns: Set(patterns))
 	}
 }
diff --git a/Sources/Lowlight/Processor.swift b/Sources/Lowlight/Processor.swift
index 3f51e8e..ce63419 100644
--- a/Sources/Lowlight/Processor.swift
+++ b/Sources/Lowlight/Processor.swift
@@ -2,8 +2,20 @@ import Foundation
 
 /// Represents a range in the input with an associated identifier.
 public struct Token: Hashable, Sendable {
-	public let name: String
+	public let element: Pattern.Element
 	public let range: NSRange
+
+	public init(_ element: Pattern.Element, range: NSRange) {
+		self.element = element
+		self.range = range
+	}
+
+	init(_ pattern: Pattern, input: String, start: String.Index, offset: Int) {
+		let end = input.index(start, offsetBy: offset)
+
+		self.range = NSRange(start..<end, in: input)
+		self.element = pattern.element
+	}
 }
 
 /// Represents a textual scope.
@@ -11,6 +23,12 @@ public struct Scope: Hashable, Sendable {
 
 }
 
+extension CharacterSet {
+	func contains(_ character: Character) -> Bool {
+		character.unicodeScalars.allSatisfy { contains($0) }
+	}
+}
+
 /// Type that uses a language model to produce tokens and scopes.
 public struct Processor {
 	public struct Output: Hashable, Sendable {
@@ -18,6 +36,23 @@ public struct Processor {
 		public let scopes: [Scope]
 	}
 
+	private enum State {
+		case scanning
+		case matching(patterns: Set<Pattern>, start: String.Index, length: Int)
+		case matched(Pattern, start: String.Index, length: Int)
+		case advancingUntil(CharacterSet, pattern: Pattern, start: String.Index)
+		case skipping
+		case advanceToBoundary
+
+		mutating func advance() {
+			guard case let .matching(patterns: patterns, start: start, length: length) = self else {
+				fatalError()
+			}
+
+			self = .matching(patterns: patterns, start: start, length: length + 1)
+		}
+	}
+
 	public let language: Language
 
 	public init(language: Language) {
@@ -32,28 +67,97 @@ public struct Processor {
 	}
 
 	public func processTokens(for input: String) -> [Token] {
+		var state = State.scanning
 		var tokens = [Token]()
 
-		if let pattern = try? NSRegularExpression(pattern: language.keywordPattern), language.keywords.isEmpty == false {
-			let matches = pattern.matches(in: input, range: NSRange(0..<input.utf16.count))
-
-			for match in matches {
-				let range = match.range(at: 0)
-				guard range.length > 0 else { continue }
-
-				tokens.append(Token(name: "keyword", range: range))
+		var index = input.startIndex
+		let patterns = language.patterns
+		let anchors = CharacterSet.whitespacesAndNewlines
+
+		while index < input.endIndex {
+			let char = input[index]
+
+			switch state {
+			case .scanning:
+				let possible = patterns.filter { $0.match.matches(char, at: 0) }
+
+				if possible.isEmpty {
+					state = .skipping
+					continue
+				}
+
+				state = .matching(patterns: possible, start: index, length: 1)
+			case .skipping:
+				if anchors.contains(char) {
+					state = .advanceToBoundary
+				}
+			case .advanceToBoundary:
+				if anchors.contains(char) == false {
+					state = .scanning
+					continue
+				}
+			case let .matching(patterns: active, start: start, length: length):
+				let newLength = length + 1
+
+				let remaining = active.filter { $0.match.matches(char, at: length) }
+
+				// no matches, continue
+				guard remaining.isEmpty == false else {
+					state = .scanning
+					continue
+				}
+
+				// more than one match, keep going
+				guard let pattern = remaining.first, remaining.count == 1 else {
+					state = .matching(patterns: remaining, start: start, length: newLength)
+					break
+				}
+
+				switch pattern.match {
+				case let .exact(value):
+					if newLength < value.count {
+						state = .matching(patterns: remaining, start: start, length: newLength)
+						break
+					}
+
+					state = .matched(pattern, start: start, length: newLength)
+				case .toEndOfLine:
+					state = .advancingUntil(.newlines, pattern: pattern, start: start)
+				}
+			case let .matched(pattern, start: start, length: length):
+				if anchors.contains(char) == false {
+					state = .skipping
+					break
+				}
+
+				tokens.append(Token(pattern, input: input, start: start, offset: length))
+
+				state = .scanning
+			case let .advancingUntil(charSet, pattern: pattern, start: start):
+				guard charSet.contains(char) else {
+					break
+				}
+
+				let range = NSRange(start..<index, in: input)
+
+				tokens.append(Token(pattern.element, range: range))
+
+				state = .scanning
+				continue
 			}
-		}
 
-		if let pattern = try? NSRegularExpression(pattern: language.symbolsPattern), language.symbols.isEmpty == false {
-			let matches = pattern.matches(in: input, range: NSRange(0..<input.utf16.count))
+			index = input.index(after: index)
+		}
 
-			for match in matches {
-				let range = match.range(at: 0)
-				guard range.length > 0 else { continue }
+		switch state {
+		case let .advancingUntil(_, pattern: pattern, start: start):
+			let range = NSRange(start..<index, in: input)
 
-				tokens.append(Token(name: "keyword.operator.text", range: range))
-			}
+			tokens.append(Token(pattern.element, range: range))
+		case let .matched(pattern, start: start, length: length):
+			tokens.append(Token(pattern, input: input, start: start, offset: length))
+		default:
+			break
 		}
 
 		return tokens
diff --git a/Tests/LowlightTests/LanguageTests.swift b/Tests/LowlightTests/LanguageTests.swift
index 70a9e43..a28a5f6 100644
--- a/Tests/LowlightTests/LanguageTests.swift
+++ b/Tests/LowlightTests/LanguageTests.swift
@@ -3,7 +3,7 @@ import Lowlight
 
 final class LanguageTests: XCTestCase {
     func testLanguageWithOneKeyword() throws {
-        let language = Language()
+        let language = Language(patterns: [])
 
 		XCTAssertNotNil(language)
     }
diff --git a/Tests/LowlightTests/ProcessorTests.swift b/Tests/LowlightTests/ProcessorTests.swift
new file mode 100644
index 0000000..0a060ad
--- /dev/null
+++ b/Tests/LowlightTests/ProcessorTests.swift
@@ -0,0 +1,92 @@
+import XCTest
+import Lowlight
+
+final class ProcessorTests: XCTestCase {
+	func testLineComments() throws {
+		let language = Language(patterns: [Pattern(element: .comment, match: .toEndOfLine("//"))])
+		let processor = Processor(language: language)
+
+		let input = """
+// leading comment
+   // after whitespace
+abc def // trailing comment
+// comment inside // comment
+"""
+
+		let tokens = processor.processTokens(for: input)
+		let expected = [
+			Token(.comment, range: NSRange(0..<18)),
+			Token(.comment, range: NSRange(22..<41)),
+			Token(.comment, range: NSRange(50..<69)),
+			Token(.comment, range: NSRange(70..<98)),
+		]
+
+		XCTAssertEqual(tokens, expected)
+	}
+
+	func testSingleExactMatch() throws {
+		let language = Language(patterns: [Pattern(element: .keyword, match: .exact("abc"))])
+		let processor = Processor(language: language)
+
+		let tokens = processor.processTokens(for: "abc ")
+
+		let expected = [
+			Token(.keyword, range: NSRange(0..<3)),
+		]
+
+		XCTAssertEqual(tokens, expected)
+	}
+
+	func testSingleExactMatchEOF() throws {
+		let language = Language(patterns: [Pattern(element: .keyword, match: .exact("abc"))])
+		let processor = Processor(language: language)
+
+		let tokens = processor.processTokens(for: "abc")
+
+		let expected = [
+			Token(.keyword, range: NSRange(0..<3)),
+		]
+
+		XCTAssertEqual(tokens, expected)
+	}
+
+	func testExactMatches() throws {
+		let language = Language(patterns: [Pattern(element: .keyword, match: .exact("abc"))])
+		let processor = Processor(language: language)
+
+		let input = """
+abc
+dabc abc
+ abc
+abcabc
+"""
+
+		let tokens = processor.processTokens(for: input)
+		let expected = [
+			Token(.keyword, range: NSRange(0..<3)),
+			Token(.keyword, range: NSRange(9..<12)),
+			Token(.keyword, range: NSRange(14..<17)),
+		]
+
+		XCTAssertEqual(tokens, expected)
+	}
+
+	func testEndOfLineFollowedByNewlineThenExact() throws {
+		let language = Language(keywords: ["abc"], lineComment: "--")
+		let processor = Processor(language: language)
+
+		let input = """
+-- comment
+
+abc def
+"""
+
+		let tokens = processor.processTokens(for: input)
+		let expected = [
+			Token(.comment, range: NSRange(0..<10)),
+			Token(.keyword, range: NSRange(12..<15)),
+		]
+
+		XCTAssertEqual(tokens, expected)
+	}
+}