1.2.9 // Structural optimizations. (#72)

* Compositor // Update things related to walkedAnchors(). * Node // Add spanLength property. * NodeAnchor // Structural optimization. * Fix malfunctioned fixNodeWithCandidate() since v1.2.8. * Sync fixes from MegrezNT v1.2.9 update. * NodeAnchor // Remove KeyLength.
vChewing · Aug 1, 2022 · 0af1710 · 0af1710
1 parent 3c152fb
commit 0af1710
Show file tree

Hide file tree

Showing 7 changed files with 77 additions and 70 deletions.
diff --git a/Sources/Megrez/1_Compositor.swift b/Sources/Megrez/1_Compositor.swift
@@ -26,7 +26,7 @@ CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 extension Megrez {
   /// 組字器。
   public class Compositor: Grid {
-    /// 文字輸入方向
+    /// 就文字輸入方向而言的方向。
     public enum TypingDirection { case front, rear }
     /// 給被丟掉的節點路徑施加的負權重。
     private let kDroppedPathScore: Double = -999
@@ -38,7 +38,14 @@ extension Megrez {
     private var langModel: LangModelProtocol
     /// 允許查詢當前游標位置屬於第幾個幅位座標（從 0 開始算）。
     private(set) var cursorRegionMap: [Int: Int] = .init()
-    private(set) var walkedAnchors: [Megrez.NodeAnchor] = []  // 用以記錄爬過的節錨的陣列
+    /// 用以記錄爬過的節錨的陣列。
+    private(set) var walkedAnchors: [NodeAnchor] = []
+
+    /// 該函式用以更新爬過的節錨的陣列。
+    /// - Parameter nodes: 傳入的節點陣列。
+    public func updateWalkedAnchors(with nodes: [Node]) {
+      walkedAnchors = nodes.map { Megrez.NodeAnchor(node: $0) }
+    }
 
     /// 公開：多字讀音鍵當中用以分割漢字讀音的記號，預設為空。
     public var joinSeparator: String = "-"
@@ -47,7 +54,7 @@ extension Megrez {
     public var length: Int { readings.count }
 
     /// 按幅位來前後移動游標。
-    /// - Parameter direction: 移動方向
+    /// - Parameter direction: 移動方向。
     /// - Returns: 該操作是否順利完成。
     @discardableResult public func jumpCursorBySpan(to direction: TypingDirection) -> Bool {
       switch direction {
@@ -64,8 +71,9 @@ extension Megrez {
         case currentRegionBorderRear:
           switch direction {
             case .front:
-              if currentRegion > walkedAnchors.count { cursor = readings.count }
-              else { cursor = walkedAnchors[0...currentRegion].map(\.spanLength).reduce(0, +) }
+              cursor =
+                (currentRegion > walkedAnchors.count)
+                ? readings.count : walkedAnchors[0...currentRegion].map(\.spanLength).reduce(0, +)
             case .rear:
               cursor = walkedAnchors[0..<aRegionForward].map(\.spanLength).reduce(0, +)
           }
@@ -87,7 +95,7 @@ extension Megrez {
     ///   - separator: 多字讀音鍵當中用以分割漢字讀音的記號，預設為空。
     public init(lm: LangModelProtocol, length: Int = 10, separator: String = "-") {
       langModel = lm
-      super.init(spanLength: abs(length))  // 防呆
+      super.init(spanLengthLimit: abs(length))  // 防呆
       joinSeparator = separator
     }
 
@@ -180,7 +188,7 @@ extension Megrez {
 
       var paths = [[NodeAnchor]]()
       let nodes = nodesEndingAt(location: location).stableSorted {
-        $0.scoreForSort > $1.scoreForSort
+        $0.node.score > $1.node.score
       }
 
       guard !nodes.isEmpty else { return .init() }  // 防止下文出現範圍外索引的錯誤
@@ -269,7 +277,7 @@ extension Megrez {
           if hasMatchedNode(location: p, spanLength: q, key: combinedReading) { continue }
           let unigrams: [Unigram] = langModel.unigramsFor(key: combinedReading)
           if unigrams.isEmpty { continue }
-          let n = Node(key: combinedReading, unigrams: unigrams)
+          let n: Node = .init(key: combinedReading, spanLength: q, unigrams: unigrams)
           insertNode(node: n, location: p, spanLength: q)
         }
       }
@@ -281,6 +289,7 @@ extension Megrez {
 
     internal func updateCursorJumpingTables(_ anchors: [NodeAnchor]) {
       var cursorRegionMapDict = [Int: Int]()
+      cursorRegionMapDict[-1] = 0  // 防呆
       var counter = 0
       for (i, anchor) in anchors.enumerated() {
         for _ in 0..<anchor.spanLength {
@@ -289,7 +298,6 @@ extension Megrez {
         }
       }
       cursorRegionMapDict[counter] = anchors.count
-      cursorRegionMapDict[-1] = 0  // 防呆
       cursorRegionMap = cursorRegionMapDict
     }
   }

diff --git a/Sources/Megrez/2_Grid.swift b/Sources/Megrez/2_Grid.swift
@@ -41,8 +41,8 @@ extension Megrez {
     public var isEmpty: Bool { spans.isEmpty }
 
     /// 初期化轨格。
-    public init(spanLength: Int = 10) {
-      maxBuildSpanLength = spanLength
+    public init(spanLengthLimit: Int = 10) {
+      maxBuildSpanLength = spanLengthLimit
       spans = [Megrez.SpanUnit]()
     }
 
@@ -98,7 +98,7 @@ extension Megrez {
           spans.remove(at: location)
       }
       for i in 0..<location {
-        // zaps overlapping spans
+        // 處理掉被損毀的或者重複的幅位。
         spans[i].dropNodesBeyond(length: location - i)
       }
     }
@@ -114,13 +114,7 @@ extension Megrez {
       let span = spans[location]
       for i in 1...maxBuildSpanLength {
         if let np = span.nodeOf(length: i) {
-          results.append(
-            .init(
-              node: np,
-              location: location,
-              spanLength: i
-            )
-          )
+          results.append(.init(node: np))
         }
       }
       return results  // 已證實不會有空節點產生。
@@ -137,13 +131,7 @@ extension Megrez {
         let span = spans[i]
         if i + span.maxLength < location { continue }
         if let np = span.nodeOf(length: location - i) {
-          results.append(
-            .init(
-              node: np,
-              location: i,
-              spanLength: location - i
-            )
-          )
+          results.append(.init(node: np))
         }
       }
       return results  // 已證實不會有空節點產生。
@@ -162,13 +150,7 @@ extension Megrez {
         for j in 1...span.maxLength {
           if i + j < location { continue }
           if let np = span.nodeOf(length: j) {
-            results.append(
-              .init(
-                node: np,
-                location: i,
-                spanLength: location - i
-              )
-            )
+            results.append(.init(node: np))
           }
         }
       }
@@ -193,7 +175,7 @@ extension Megrez {
     @discardableResult public func fixNodeWithCandidateLiteral(_ value: String, at location: Int) -> NodeAnchor {
       let location = abs(location)  // 防呆
       var node = NodeAnchor()
-      for theAnchor in nodesOverlappedAt(location: location) {
+      for theAnchor in nodesCrossingOrEndingAt(location: location) {
         let candidates = theAnchor.node.candidates
         // 將該位置的所有節點的候選字詞鎖定狀態全部重設。
         theAnchor.node.resetCandidate()
@@ -217,7 +199,7 @@ extension Megrez {
     @discardableResult public func fixNodeWithCandidate(_ pair: KeyValuePaired, at location: Int) -> NodeAnchor {
       let location = abs(location)  // 防呆
       var node = NodeAnchor()
-      for theAnchor in nodesOverlappedAt(location: location) {
+      for theAnchor in nodesCrossingOrEndingAt(location: location) {
         let candidates = theAnchor.node.candidates
         // 將該位置的所有節點的候選字詞鎖定狀態全部重設。
         theAnchor.node.resetCandidate()

diff --git a/Sources/Megrez/3_NodeAnchor.swift b/Sources/Megrez/3_NodeAnchor.swift
@@ -30,28 +30,35 @@ extension Megrez {
     public var isEmpty: Bool { node.key.isEmpty }
     /// 節點。一個節锚內不一定有節點。
     public var node: Node = .init()
-    /// 節锚所在的位置。
-    public var location: Int = 0
     /// 指定的幅位長度。
-    public var spanLength: Int = 0
+    public var spanLength: Int { node.spanLength }
+    /// 獲取用來比較的權重。
+    public var scoreForSort: Double { node.score }
     /// 累計權重。
     public var mass: Double = 0.0
-    /// 索引鍵的長度。
-    public var keyLength: Int {
-      isEmpty ? node.key.count : 0
+    /// 單元圖陣列。
+    public var unigrams: [Unigram] { node.unigrams }
+    /// 雙元圖陣列。
+    public var bigrams: [Bigram] { node.bigrams }
+    /// 鍵。
+    public var key: String { node.key }
+
+    /// 初期化一個節錨。
+    public init(node: Node = .init(), mass: Double? = nil) {
+      self.node = node
+      self.mass = mass ?? self.node.score
     }
 
+    /// 將該節錨雜湊化。
     public func hash(into hasher: inout Hasher) {
       hasher.combine(node)
-      hasher.combine(location)
-      hasher.combine(spanLength)
       hasher.combine(mass)
     }
 
     /// 將當前節锚列印成一個字串。
     public var description: String {
       var stream = ""
-      stream += "{@(" + String(location) + "," + String(spanLength) + "),"
+      stream += "{@(" + String(spanLength) + "),"
       if node.key.isEmpty {
         stream += node.description
       } else {
@@ -60,11 +67,6 @@ extension Megrez {
       stream += "}"
       return stream
     }
-
-    /// 獲取用來比較的權重。
-    public var scoreForSort: Double {
-      isEmpty ? node.score : 0
-    }
   }
 }
 

diff --git a/Sources/Megrez/4_Node.swift b/Sources/Megrez/4_Node.swift
@@ -30,14 +30,15 @@ extension Megrez {
       lhs.key == rhs.key && lhs.score == rhs.score && lhs.unigrams == rhs.unigrams && lhs.bigrams == rhs.bigrams
         && lhs.candidates == rhs.candidates && lhs.valueUnigramIndexMap == rhs.valueUnigramIndexMap
         && lhs.precedingBigramMap == rhs.precedingBigramMap && lhs.isCandidateFixed == rhs.isCandidateFixed
-        && lhs.selectedUnigramIndex == rhs.selectedUnigramIndex
+        && lhs.selectedUnigramIndex == rhs.selectedUnigramIndex && lhs.spanLength == rhs.spanLength
     }
 
     public func hash(into hasher: inout Hasher) {
       hasher.combine(key)
       hasher.combine(score)
       hasher.combine(unigrams)
       hasher.combine(bigrams)
+      hasher.combine(spanLength)
       hasher.combine(candidates)
       hasher.combine(valueUnigramIndexMap)
       hasher.combine(precedingBigramMap)
@@ -50,9 +51,11 @@ extension Megrez {
     /// 當前節點的當前被選中的候選字詞「在該節點內的」目前的權重。
     private(set) var score: Double = 0
     /// 單元圖陣列。
-    private var unigrams: [Unigram]
+    private(set) var unigrams: [Unigram]
     /// 雙元圖陣列。
-    private var bigrams: [Bigram]
+    private(set) var bigrams: [Bigram]
+    /// 指定的幅位長度。
+    public var spanLength: Int = 0
     /// 候選字詞陣列，以鍵值陣列的形式存在。
     private(set) var candidates: [KeyValuePaired] = []
     /// 專門「用單元圖資料值來調查索引值」的辭典。
@@ -83,10 +86,11 @@ extension Megrez {
     ///   - key: 索引鍵。
     ///   - unigrams: 單元圖陣列。
     ///   - bigrams: 雙元圖陣列（非必填）。
-    public init(key: String = "", unigrams: [Megrez.Unigram] = [], bigrams: [Megrez.Bigram] = []) {
+    public init(key: String = "", spanLength: Int = 0, unigrams: [Megrez.Unigram] = [], bigrams: [Megrez.Bigram] = []) {
       self.key = key
       self.unigrams = unigrams
       self.bigrams = bigrams
+      self.spanLength = spanLength
 
       self.unigrams.sort {
         $0.score > $1.score

diff --git a/Sources/Megrez/5_LanguageModel.swift b/Sources/Megrez/5_LanguageModel.swift
@@ -28,7 +28,7 @@ public protocol LangModelProtocol {
   func unigramsFor(key: String) -> [Megrez.Unigram]
 
   /// 給定當前鍵與前述鍵，讓語言模型找給一組雙元圖陣列。
-  func bigramsForKeys(precedingKey: String, key: String) -> [Megrez.Bigram]
+  func bigramsFor(precedingKey: String, key: String) -> [Megrez.Bigram]
 
   /// 給定鍵，確認是否有單元圖記錄在庫。
   func hasUnigramsFor(key: String) -> Bool
@@ -47,7 +47,7 @@ extension Megrez {
     }
 
     /// 給定當前鍵與前述鍵，讓語言模型找給一組雙元圖陣列。
-    open func bigramsForKeys(precedingKey: String, key: String) -> [Megrez.Bigram] {
+    open func bigramsFor(precedingKey: String, key: String) -> [Megrez.Bigram] {
       precedingKey == key ? [Megrez.Bigram]() : [Megrez.Bigram]()
     }
 

diff --git a/Tests/MegrezTests/LMDataForTests.swift b/Tests/MegrezTests/LMDataForTests.swift
@@ -28,7 +28,7 @@ import Megrez
 // MARK: - 用以測試的語言模型（簡單範本型）
 
 class SimpleLM: LangModelProtocol {
-  func bigramsForKeys(precedingKey _: String, key _: String) -> [Megrez.Bigram] {
+  func bigramsFor(precedingKey _: String, key _: String) -> [Megrez.Bigram] {
     .init()
   }
 
@@ -65,7 +65,7 @@ class SimpleLM: LangModelProtocol {
 }
 
 class MockLM: LangModelProtocol {
-  func bigramsForKeys(precedingKey _: String, key _: String) -> [Megrez.Bigram] {
+  func bigramsFor(precedingKey _: String, key _: String) -> [Megrez.Bigram] {
     .init()
   }