LMAssembly // Pack LMUserOverride inside LMInstantiator, etc.

Packages/vChewing_LangModelAssembly/README.md

@@ -1,17 +1,17 @@
 # LangModelAssembly
 
-威注音輸入法的語言模組總成套裝。
+威注音輸入法的語言模組總成套裝，以 LMAssembly 命名空間承載下述唯二對外物件：
 
-- vChewingLM：總命名空間，也承載一些在套裝內共用的工具函式。
 - LMConsolidator：自動格式整理模組。
-- LMInstantiator：語言模組副本化模組。另有其日期時間擴充模組可用（對 CIN 磁帶模式無效）。
+- LMInstantiator：語言模組副本化模組，亦集成一些自身功能擴展。
+
+LMAssembly 總命名空間也承載一些在套裝內共用的工具函式。
 
 以下是子模組：
 
-- lmCassette：專門用來處理 CIN 磁帶檔案的模組，命名為「遠野」引擎。
 - LMAssociates：關聯詞語模組。
+- lmCassette：專門用來處理 CIN 磁帶檔案的模組，命名為「遠野」引擎。
 - LMCoreEX：可以直接讀取 TXT 格式的帶有權重資料的語彙檔案的模組。
-- LMCoreJSON：專門用來讀取原廠 JSON 檔案的模組。
 - lmPlainBopomofo：專門用來讀取使用者自訂ㄅ半候選字順序覆蓋定義檔案（plist）的模組。
 - lmReplacements：專門用來讀取使用者語彙置換模式的辭典資料的模組。
 - lmUserOverride：半衰記憶模組。

Packages/vChewing_LangModelAssembly/Sources/LangModelAssembly/InputToken.swift

@@ -11,7 +11,8 @@ import Foundation
 /// 工作原理：先用 InputToken.parse 分析原始字串，給出準確的 Token。
 /// 然後再讓這個 Token 用 .translated() 自我表述出轉換結果。
 
-public enum InputToken {
+extension LMAssembly {
+  enum InputToken {
     case timeZone(shortened: Bool)
     case timeNow(shortened: Bool)
     case date(dayDelta: Int = 0, yearDelta: Int = 0, shortened: Bool = true, luna: Bool = false)
@@ -20,20 +21,21 @@ public enum InputToken {
     case yearGanzhi(yearDelta: Int = 0)
     case yearZodiac(yearDelta: Int = 0)
   }
+}
 
 // MARK: - 正式對外投入使用的 API。
 
 public extension String {
   func parseAsInputToken(isCHS: Bool) -> [String] {
-    InputToken.parse(from: self).map { $0.translated(isCHS: isCHS) }.flatMap { $0 }.deduplicated
+    LMAssembly.InputToken.parse(from: self).map { $0.translated(isCHS: isCHS) }.flatMap { $0 }.deduplicated
   }
 }
 
 // MARK: - Parser parsing raw token value to construct token.
 
-public extension InputToken {
+extension LMAssembly.InputToken {
-  static func parse(from rawToken: String) -> [InputToken] {
+  static func parse(from rawToken: String) -> [LMAssembly.InputToken] {
-    var result: [InputToken] = []
+    var result: [LMAssembly.InputToken] = []
     guard rawToken.prefix(6) == "MACRO@" else { return result }
     var mapParams: [String: Int] = [:]
     let tokenComponents = rawToken.dropFirst(6).split(separator: "_").map { param in
@@ -69,7 +71,7 @@ public extension InputToken {
 
 // MARK: - Parser parsing token itself.
 
-public extension InputToken {
+extension LMAssembly.InputToken {
   func translated(isCHS: Bool) -> [String] {
     let locale = Locale(identifier: isCHS ? "zh-Hans" : "zh-Hant-TW")
     let formatter = DateFormatter()

Packages/vChewing_LangModelAssembly/Sources/LangModelAssembly/LMConsolidator.swift

@@ -10,7 +10,7 @@ import Foundation
 import LineReader
 import Shared
 
-public extension vChewingLM {
+public extension LMAssembly {
   enum LMConsolidator {
     public static let kPragmaHeader = "# 𝙵𝙾𝚁𝙼𝙰𝚃 𝚘𝚛𝚐.𝚊𝚝𝚎𝚕𝚒𝚎𝚛𝙸𝚗𝚖𝚞.𝚟𝚌𝚑𝚎𝚠𝚒𝚗𝚐.𝚞𝚜𝚎𝚛𝙻𝚊𝚗𝚐𝚞𝚊𝚐𝚎𝙼𝚘𝚍𝚎𝚕𝙳𝚊𝚝𝚊.𝚏𝚘𝚛𝚖𝚊𝚝𝚝𝚎𝚍"
 

Packages/vChewing_LangModelAssembly/Sources/LangModelAssembly/LMInstantiator.swift

@@ -11,7 +11,7 @@ import Megrez
 import Shared
 import SQLite3
 
-public extension vChewingLM {
+public extension LMAssembly {
   /// 語言模組副本化模組（LMInstantiator，下稱「LMI」）自身為符合天權星組字引擎內
   /// 的 LangModelProtocol 協定的模組、統籌且整理來自其它子模組的資料（包括使
   /// 用者語彙、繪文字模組、語彙濾除表、原廠語言模組等）。
@@ -56,8 +56,12 @@ public extension vChewingLM {
     public var config = Config()
 
     // 這句需要留著，不然無法被 package 外界存取。
-    public init(isCHS: Bool = false) {
+    public init(
+      isCHS: Bool = false,
+      uomDataURL: URL? = nil
+    ) {
       self.isCHS = isCHS
+      lmUserOverride = .init(dataURL: uomDataURL)
     }
 
     public func setOptions(handler: (inout Config) -> Void) {
@@ -109,6 +113,9 @@ public extension vChewingLM {
     var lmAssociates = LMAssociates()
     var lmPlainBopomofo = LMPlainBopomofo()
 
+    // 半衰记忆模组
+    var lmUserOverride: LMUserOverride
+
     // MARK: - 工具函式
 
     public func resetFactoryJSONModels() {}

Packages/vChewing_LangModelAssembly/Sources/LangModelAssembly/LMInstantiator_CassetteExtension.swift

@@ -10,7 +10,7 @@ import Foundation
 import Megrez
 import Shared
 
-public extension vChewingLM.LMInstantiator {
+public extension LMAssembly.LMInstantiator {
   /// 磁帶模式專用：當前磁帶所規定的花牌鍵。
   var cassetteWildcardKey: String { Self.lmCassette.wildcardKey }
   /// 磁帶模式專用：當前磁帶規定的最大碼長。

Packages/vChewing_LangModelAssembly/Sources/LangModelAssembly/LMInstantiator_DateTimeExtension.swift

@@ -11,7 +11,7 @@ import Megrez
 
 // MARK: - 日期時間便捷輸入功能
 
-extension vChewingLM.LMInstantiator {
+extension LMAssembly.LMInstantiator {
   func queryDateTimeUnigrams(with key: String = "") -> [Megrez.Unigram] {
     guard let tokenTrigger = TokenTrigger(rawValue: key) else { return [] }
     var results = [Megrez.Unigram]()

Packages/vChewing_LangModelAssembly/Sources/LangModelAssembly/LMInstantiator_NumPadExtension.swift

@@ -9,7 +9,7 @@
 import Foundation
 import Megrez
 
-public extension vChewingLM.LMInstantiator {
+public extension LMAssembly.LMInstantiator {
   func supplyNumPadUnigrams(key: String) -> [Megrez.Unigram] {
     guard let status = config.numPadFWHWStatus else { return [] }
     let initials = "_NumPad_"

Packages/vChewing_LangModelAssembly/Sources/LangModelAssembly/LMInstantiator_SQLExtension.swift

@@ -31,6 +31,7 @@ import SQLite3
    ) WITHOUT ROWID;
  */
 
+extension LMAssembly.LMInstantiator {
   enum CoreColumn: Int32 {
     case theDataCHS = 1 // 簡體中文
     case theDataCHT = 2 // 繁體中文
@@ -53,8 +54,9 @@ enum CoreColumn: Int32 {
       }
     }
   }
+}
 
-extension vChewingLM.LMInstantiator {
+extension LMAssembly.LMInstantiator {
   fileprivate static func querySQL(strStmt sqlQuery: String, coreColumn column: CoreColumn, handler: (String) -> Void) {
     guard Self.ptrSQL != nil else { return }
     performStatementSansResult { ptrStatement in
@@ -134,7 +136,9 @@ extension vChewingLM.LMInstantiator {
   /// - parameters:
   ///   - key: 讀音索引鍵。
   ///   - column: 資料欄位。
-  func factoryUnigramsFor(key: String, column: CoreColumn) -> [Megrez.Unigram] {
+  func factoryUnigramsFor(
+    key: String, column: LMAssembly.LMInstantiator.CoreColumn
+  ) -> [Megrez.Unigram] {
     if key == "_punctuation_list" { return [] }
     var grams: [Megrez.Unigram] = []
     var gramsHW: [Megrez.Unigram] = []
@@ -210,7 +214,7 @@ extension vChewingLM.LMInstantiator {
   }
 }
 
-private extension vChewingLM.LMInstantiator {
+private extension LMAssembly.LMInstantiator {
   /// 內部函式，用以將注音讀音索引鍵進行加密。
   ///
   /// 使用這種加密字串作為索引鍵，可以增加對 json 資料庫的存取速度。
@@ -258,7 +262,7 @@ private extension vChewingLM.LMInstantiator {
   ]
 }
 
-public extension vChewingLM.LMInstantiator {
+public extension LMAssembly.LMInstantiator {
   @discardableResult static func connectToTestSQLDB() -> Bool {
     Self.connectSQLDB(dbPath: #":memory:"#) && sqlTestCoreLMData.runAsSQLExec(dbPointer: &ptrSQL)
   }

Packages/vChewing_LangModelAssembly/Sources/LangModelAssembly/LMInstantiator_UOMRepresentable.swift

@@ -0,0 +1,60 @@
+// (c) 2021 and onwards The vChewing Project (MIT-NTL License).
+// ====================
+// This code is released under the MIT license (SPDX-License-Identifier: MIT)
+// ... with NTL restriction stating that:
+// No trademark license is granted to use the trade names, trademarks, service
+// marks, or product names of Contributor, except as required to fulfill notice
+// requirements defined in MIT License.
+
+import Foundation
+import Megrez
+
+public extension LMAssembly.LMInstantiator {
+  func performUOMObservation(
+    walkedBefore: [Megrez.Node],
+    walkedAfter: [Megrez.Node],
+    cursor: Int,
+    timestamp: Double,
+    saveCallback: (() -> Void)? = nil
+  ) {
+    lmUserOverride.performObservation(
+      walkedBefore: walkedBefore,
+      walkedAfter: walkedAfter,
+      cursor: cursor,
+      timestamp: timestamp,
+      saveCallback: saveCallback
+    )
+  }
+
+  func fetchUOMSuggestion(
+    currentWalk: [Megrez.Node],
+    cursor: Int,
+    timestamp: Double
+  ) -> LMAssembly.OverrideSuggestion {
+    lmUserOverride.fetchSuggestion(
+      currentWalk: currentWalk,
+      cursor: cursor,
+      timestamp: timestamp
+    )
+  }
+
+  func loadUOMData(fromURL fileURL: URL? = nil) {
+    lmUserOverride.loadData(fromURL: fileURL)
+  }
+
+  func saveUOMData(toURL fileURL: URL? = nil) {
+    lmUserOverride.saveData(toURL: fileURL)
+  }
+
+  func clearUOMData(withURL fileURL: URL? = nil) {
+    lmUserOverride.clearData(withURL: fileURL)
+  }
+
+  func bleachSpecifiedUOMSuggestions(targets: [String], saveCallback: (() -> Void)? = nil) {
+    lmUserOverride.bleachSpecifiedSuggestions(targets: targets, saveCallback: saveCallback)
+  }
+
+  func bleachUOMUnigrams(saveCallback: (() -> Void)? = nil) {
+    lmUserOverride.bleachUnigrams(saveCallback: saveCallback)
+  }
+}

Packages/vChewing_LangModelAssembly/Sources/LangModelAssembly/SubLMs/lmAssociates.swift

@@ -10,8 +10,8 @@ import Megrez
 import PinyinPhonaConverter
 import Shared
 
-public extension vChewingLM {
+extension LMAssembly {
-  @frozen struct LMAssociates {
+  struct LMAssociates {
     public private(set) var filePath: String?
     var rangeMap: [String: [(Range<String.Index>, Int)]] = [:]
     var strData: String = ""

Packages/vChewing_LangModelAssembly/Sources/LangModelAssembly/SubLMs/lmCassette.swift

@@ -12,9 +12,9 @@ import LineReader
 import Megrez
 import Shared
 
-public extension vChewingLM {
+extension LMAssembly {
   /// 磁帶模組，用來方便使用者自行擴充字根輸入法。
-  @frozen struct LMCassette {
+  struct LMCassette {
     public private(set) var filePath: String?
     public private(set) var nameShort: String = ""
     public private(set) var nameENG: String = ""
@@ -45,7 +45,7 @@ public extension vChewingLM {
   }
 }
 
-public extension vChewingLM.LMCassette {
+extension LMAssembly.LMCassette {
   /// 計算頻率時要用到的東西 - fscale
   private static let fscale = 2.7
   /// 萬用花牌字符，哪怕花牌鍵仍不可用。
@@ -86,7 +86,7 @@ public extension vChewingLM.LMCassette {
     if FileManager.default.fileExists(atPath: path) {
       do {
         guard let fileHandle = FileHandle(forReadingAtPath: path) else {
-          throw vChewingLM.FileErrors.fileHandleError("")
+          throw LMAssembly.FileErrors.fileHandleError("")
         }
         let lineReader = try LineReader(file: fileHandle)
         var theMaxKeyLength = 1

Packages/vChewing_LangModelAssembly/Sources/LangModelAssembly/SubLMs/lmCoreEX.swift

@@ -10,12 +10,12 @@ import Megrez
 import PinyinPhonaConverter
 import Shared
 
-public extension vChewingLM {
+extension LMAssembly {
   /// 與之前的 LMCore 不同，LMCoreEX 不在辭典內記錄實體，而是記錄 range 範圍。
   /// 需要資料的時候，直接拿 range 去 strData 取資料。
   /// 資料記錄原理與上游 C++ 的 ParselessLM 差不多，但用的是 Swift 原生手段。
   /// 主要時間消耗仍在 For 迴圈，但這個算法可以顯著減少記憶體佔用。
-  @frozen struct LMCoreEX {
+  struct LMCoreEX {
     public private(set) var filePath: String?
     /// 資料庫辭典。索引內容為注音字串，資料內容則為字串首尾範圍、方便自 strData 取資料。
     var rangeMap: [String: [Range<String.Index>]] = [:]

Packages/vChewing_LangModelAssembly/Sources/LangModelAssembly/SubLMs/lmPlainBopomofo.swift

@@ -9,7 +9,7 @@
 import Foundation
 import Shared
 
-public extension vChewingLM {
+public extension LMAssembly {
   @frozen struct LMPlainBopomofo {
     public private(set) var filePath: String?
     var dataMap: [String: String] = [:]
@@ -29,13 +29,8 @@ public extension vChewingLM {
 
       do {
         let rawData = try Data(contentsOf: URL(fileURLWithPath: path))
-        if let rawJSON = try? JSONSerialization.jsonObject(with: rawData) as? [String: String] {
+        let rawJSON = try JSONDecoder().decode([String: String].self, from: rawData)
         dataMap = rawJSON
-        } else {
-          filePath = oldPath
-          vCLog("↑ Exception happened when reading JSON file at: \(path).")
-          return false
-        }
       } catch {
         filePath = oldPath
         vCLog("\(error)")

Packages/vChewing_LangModelAssembly/Sources/LangModelAssembly/SubLMs/lmReplacements.swift

@@ -8,8 +8,8 @@
 
 import Shared
 
-public extension vChewingLM {
+extension LMAssembly {
-  @frozen struct LMReplacements {
+  struct LMReplacements {
     public private(set) var filePath: String?
     var rangeMap: [String: Range<String.Index>] = [:]
     var strData: String = ""

Packages/vChewing_LangModelAssembly/Sources/LangModelAssembly/SubLMs/lmRevLookup.swift

@@ -1,76 +0,0 @@
-// (c) 2021 and onwards The vChewing Project (MIT-NTL License).
-// ====================
-// This code is released under the MIT license (SPDX-License-Identifier: MIT)
-// ... with NTL restriction stating that:
-// No trademark license is granted to use the trade names, trademarks, service
-// marks, or product names of Contributor, except as required to fulfill notice
-// requirements defined in MIT License.
-
-import Foundation
-import Shared
-
-public extension vChewingLM {
-  @frozen struct LMRevLookup {
-    public private(set) var dataMap: [String: [String]] = [:]
-    public private(set) var filePath: String = ""
-
-    public init(data dictData: (dict: [String: [String]]?, path: String)) {
-      guard let theDict = dictData.dict else {
-        vCLog("↑ Exception happened when reading JSON file at: \(dictData.path).")
-        return
-      }
-      filePath = dictData.path
-      dataMap = theDict
-    }
-
-    public init(path: String) {
-      if path.isEmpty { return }
-      do {
-        let rawData = try Data(contentsOf: URL(fileURLWithPath: path))
-        if let rawJSON = try? JSONSerialization.jsonObject(with: rawData) as? [String: [String]] {
-          dataMap = rawJSON
-        } else {
-          vCLog("↑ Exception happened when reading JSON file at: \(path).")
-          return
-        }
-      } catch {
-        vCLog("↑ Exception happened when reading JSON file at: \(path).")
-        return
-      }
-      filePath = path
-    }
-
-    public func query(with kanji: String) -> [String]? {
-      guard let resultData = dataMap[kanji] else { return nil }
-      let resultArray = resultData.compactMap {
-        let result = restorePhonabetFromASCII($0)
-        return result.isEmpty ? nil : result
-      }
-      return resultArray.isEmpty ? nil : resultArray
-    }
-
-    /// 內部函式，用以將被加密的注音讀音索引鍵進行解密。
-    ///
-    /// 如果傳入的字串當中包含 ASCII 下畫線符號的話，則表明該字串並非注音讀音字串，會被忽略處理。
-    /// - parameters:
-    ///   - incoming: 傳入的已加密注音讀音字串。
-    func restorePhonabetFromASCII(_ incoming: String) -> String {
-      var strOutput = incoming
-      if !strOutput.contains("_") {
-        for entry in Self.dicPhonabet4ASCII {
-          strOutput = strOutput.replacingOccurrences(of: entry.key, with: entry.value)
-        }
-      }
-      return strOutput
-    }
-
-    // MARK: - Constants
-
-    static let dicPhonabet4ASCII: [String: String] = [
-      "b": "ㄅ", "p": "ㄆ", "m": "ㄇ", "f": "ㄈ", "d": "ㄉ", "t": "ㄊ", "n": "ㄋ", "l": "ㄌ", "g": "ㄍ", "k": "ㄎ", "h": "ㄏ",
-      "j": "ㄐ", "q": "ㄑ", "x": "ㄒ", "Z": "ㄓ", "C": "ㄔ", "S": "ㄕ", "r": "ㄖ", "z": "ㄗ", "c": "ㄘ", "s": "ㄙ", "i": "ㄧ",
-      "u": "ㄨ", "v": "ㄩ", "a": "ㄚ", "o": "ㄛ", "e": "ㄜ", "E": "ㄝ", "B": "ㄞ", "P": "ㄟ", "M": "ㄠ", "F": "ㄡ", "D": "ㄢ",
-      "T": "ㄣ", "N": "ㄤ", "L": "ㄥ", "R": "ㄦ", "2": "ˊ", "3": "ˇ", "4": "ˋ", "5": "˙",
-    ]
-  }
-}

Packages/vChewing_LangModelAssembly/Sources/LangModelAssembly/SubLMs/lmUserOverride.swift

@@ -11,72 +11,40 @@ import Foundation
 import Megrez
 import Shared
 
-public extension vChewingLM {
+// MARK: - Public Types.
-  class LMUserOverride {
-    // MARK: - Main
 
+public extension LMAssembly {
+  struct OverrideSuggestion {
+    public var candidates = [(String, Megrez.Unigram)]()
+    public var forceHighScoreOverride = false
+    public var isEmpty: Bool { candidates.isEmpty }
+  }
+}
+
+// MARK: - LMUserOverride Class Definition.
+
+extension LMAssembly {
+  class LMUserOverride {
     var mutCapacity: Int
     var mutDecayExponent: Double
     var mutLRUList: [KeyObservationPair] = []
     var mutLRUMap: [String: KeyObservationPair] = [:]
     let kDecayThreshold: Double = 1.0 / 1_048_576.0 // 衰減二十次之後差不多就失效了。
-    var fileSaveLocationURL: URL
+    var fileSaveLocationURL: URL?
 
     public static let kObservedOverrideHalfLife: Double = 3600.0 * 6 // 6 小時半衰一次，能持續不到六天的記憶。
 
-    public init(capacity: Int = 500, decayConstant: Double = LMUserOverride.kObservedOverrideHalfLife, dataURL: URL) {
+    public init(capacity: Int = 500, decayConstant: Double = LMUserOverride.kObservedOverrideHalfLife, dataURL: URL? = nil) {
       mutCapacity = max(capacity, 1) // Ensures that this integer value is always > 0.
       mutDecayExponent = log(0.5) / decayConstant
       fileSaveLocationURL = dataURL
     }
-
-    public func performObservation(
-      walkedBefore: [Megrez.Node], walkedAfter: [Megrez.Node],
-      cursor: Int, timestamp: Double, saveCallback: @escaping () -> Void
-    ) {
-      // 參數合規性檢查。
-      guard !walkedAfter.isEmpty, !walkedBefore.isEmpty else { return }
-      guard walkedBefore.totalKeyCount == walkedAfter.totalKeyCount else { return }
-      // 先判斷用哪種覆寫方法。
-      var actualCursor = 0
-      guard let currentNode = walkedAfter.findNode(at: cursor, target: &actualCursor) else { return }
-      // 當前節點超過三個字的話，就不記憶了。在這種情形下，使用者可以考慮新增自訂語彙。
-      guard currentNode.spanLength <= 3 else { return }
-      // 前一個節點得從前一次爬軌結果當中來找。
-      guard actualCursor > 0 else { return } // 該情況應該不會出現。
-      let currentNodeIndex = actualCursor
-      actualCursor -= 1
-      var prevNodeIndex = 0
-      guard let prevNode = walkedBefore.findNode(at: actualCursor, target: &prevNodeIndex) else { return }
-
-      let forceHighScoreOverride: Bool = currentNode.spanLength > prevNode.spanLength
-      let breakingUp = currentNode.spanLength == 1 && prevNode.spanLength > 1
-
-      let targetNodeIndex = breakingUp ? currentNodeIndex : prevNodeIndex
-      let key: String = vChewingLM.LMUserOverride.formObservationKey(
-        walkedNodes: walkedAfter, headIndex: targetNodeIndex
-      )
-      guard !key.isEmpty else { return }
-      doObservation(
-        key: key, candidate: currentNode.currentUnigram.value, timestamp: timestamp,
-        forceHighScoreOverride: forceHighScoreOverride, saveCallback: { saveCallback() }
-      )
-    }
-
-    public func fetchSuggestion(
-      currentWalk: [Megrez.Node], cursor: Int, timestamp: Double
-    ) -> Suggestion {
-      var headIndex = 0
-      guard let nodeIter = currentWalk.findNode(at: cursor, target: &headIndex) else { return .init() }
-      let key = vChewingLM.LMUserOverride.formObservationKey(walkedNodes: currentWalk, headIndex: headIndex)
-      return getSuggestion(key: key, timestamp: timestamp, headReading: nodeIter.joinedKey())
-    }
   }
 }
 
 // MARK: - Private Structures
 
-extension vChewingLM.LMUserOverride {
+extension LMAssembly.LMUserOverride {
   enum OverrideUnit: CodingKey { case count, timestamp, forceHighScoreOverride }
   enum ObservationUnit: CodingKey { case count, overrides }
   enum KeyObservationPairUnit: CodingKey { case key, observation }
@@ -153,10 +121,52 @@ extension vChewingLM.LMUserOverride {
   }
 }
 
-// MARK: - Hash and Dehash the entire UOM data, etc.
+// MARK: - Internal Methods in LMAssembly.
 
-public extension vChewingLM.LMUserOverride {
+extension LMAssembly.LMUserOverride {
-  func bleachSpecifiedSuggestions(targets: [String], saveCallback: @escaping () -> Void) {
+  func performObservation(
+    walkedBefore: [Megrez.Node], walkedAfter: [Megrez.Node],
+    cursor: Int, timestamp: Double, saveCallback: (() -> Void)? = nil
+  ) {
+    // 參數合規性檢查。
+    guard !walkedAfter.isEmpty, !walkedBefore.isEmpty else { return }
+    guard walkedBefore.totalKeyCount == walkedAfter.totalKeyCount else { return }
+    // 先判斷用哪種覆寫方法。
+    var actualCursor = 0
+    guard let currentNode = walkedAfter.findNode(at: cursor, target: &actualCursor) else { return }
+    // 當前節點超過三個字的話，就不記憶了。在這種情形下，使用者可以考慮新增自訂語彙。
+    guard currentNode.spanLength <= 3 else { return }
+    // 前一個節點得從前一次爬軌結果當中來找。
+    guard actualCursor > 0 else { return } // 該情況應該不會出現。
+    let currentNodeIndex = actualCursor
+    actualCursor -= 1
+    var prevNodeIndex = 0
+    guard let prevNode = walkedBefore.findNode(at: actualCursor, target: &prevNodeIndex) else { return }
+
+    let forceHighScoreOverride: Bool = currentNode.spanLength > prevNode.spanLength
+    let breakingUp = currentNode.spanLength == 1 && prevNode.spanLength > 1
+
+    let targetNodeIndex = breakingUp ? currentNodeIndex : prevNodeIndex
+    let key: String = LMAssembly.LMUserOverride.formObservationKey(
+      walkedNodes: walkedAfter, headIndex: targetNodeIndex
+    )
+    guard !key.isEmpty else { return }
+    doObservation(
+      key: key, candidate: currentNode.currentUnigram.value, timestamp: timestamp,
+      forceHighScoreOverride: forceHighScoreOverride, saveCallback: saveCallback
+    )
+  }
+
+  func fetchSuggestion(
+    currentWalk: [Megrez.Node], cursor: Int, timestamp: Double
+  ) -> LMAssembly.OverrideSuggestion {
+    var headIndex = 0
+    guard let nodeIter = currentWalk.findNode(at: cursor, target: &headIndex) else { return .init() }
+    let key = LMAssembly.LMUserOverride.formObservationKey(walkedNodes: currentWalk, headIndex: headIndex)
+    return getSuggestion(key: key, timestamp: timestamp, headReading: nodeIter.joinedKey())
+  }
+
+  func bleachSpecifiedSuggestions(targets: [String], saveCallback: (() -> Void)? = nil) {
     if targets.isEmpty { return }
     for neta in mutLRUMap {
       for target in targets {
@@ -166,44 +176,50 @@ public extension vChewingLM.LMUserOverride {
       }
     }
     resetMRUList()
-    saveCallback()
+    saveCallback?() ?? saveData()
   }
 
   /// 自 LRU 辭典內移除所有的單元圖。
-  func bleachUnigrams(saveCallback: @escaping () -> Void) {
+  func bleachUnigrams(saveCallback: (() -> Void)? = nil) {
     for key in mutLRUMap.keys {
       if !key.contains("(),()") { continue }
       mutLRUMap.removeValue(forKey: key)
     }
     resetMRUList()
-    saveCallback()
+    saveCallback?() ?? saveData()
   }
 
-  internal func resetMRUList() {
+  func resetMRUList() {
     mutLRUList.removeAll()
     for neta in mutLRUMap.reversed() {
       mutLRUList.append(neta.value)
     }
   }
 
-  func clearData(withURL fileURL: URL) {
+  func clearData(withURL fileURL: URL? = nil) {
     mutLRUMap = .init()
     mutLRUList = .init()
     do {
       let nullData = "{}"
+      guard let fileURL = fileURL ?? fileSaveLocationURL else {
+        throw "given fileURL is invalid or nil."
+      }
       try nullData.write(to: fileURL, atomically: false, encoding: .utf8)
     } catch {
-      vCLog("UOM Error: Unable to clear data. Details: \(error)")
+      vCLog("UOM Error: Unable to clear the data in the UOM file. Details: \(error)")
       return
     }
   }
 
   func saveData(toURL fileURL: URL? = nil) {
+    guard let fileURL: URL = fileURL ?? fileSaveLocationURL else {
+      vCLog("UOM saveData() failed. At least the file Save URL is not set for the current UOM.")
+      return
+    }
     // 此處不要使用 JSONSerialization，不然執行緒會炸掉。
     let encoder = JSONEncoder()
     do {
       guard let jsonData = try? encoder.encode(mutLRUMap) else { return }
-      let fileURL: URL = fileURL ?? fileSaveLocationURL
       try jsonData.write(to: fileURL, options: .atomic)
     } catch {
       vCLog("UOM Error: Unable to save data, abort saving. Details: \(error)")
@@ -211,7 +227,11 @@ public extension vChewingLM.LMUserOverride {
     }
   }
 
-  func loadData(fromURL fileURL: URL) {
+  func loadData(fromURL fileURL: URL? = nil) {
+    guard let fileURL: URL = fileURL ?? fileSaveLocationURL else {
+      vCLog("UOM loadData() failed. At least the file Load URL is not set for the current UOM.")
+      return
+    }
     // 此處不要使用 JSONSerialization，不然執行緒會炸掉。
     let decoder = JSONDecoder()
     do {
@@ -228,20 +248,14 @@ public extension vChewingLM.LMUserOverride {
       return
     }
   }
-
-  struct Suggestion {
-    public var candidates = [(String, Megrez.Unigram)]()
-    public var forceHighScoreOverride = false
-    public var isEmpty: Bool { candidates.isEmpty }
-  }
 }
 
-// MARK: - Private Methods
+// MARK: - Other Non-Public Internal Methods
 
-extension vChewingLM.LMUserOverride {
+extension LMAssembly.LMUserOverride {
   func doObservation(
     key: String, candidate: String, timestamp: Double, forceHighScoreOverride: Bool,
-    saveCallback: @escaping () -> Void
+    saveCallback: (() -> Void)?
   ) {
     guard mutLRUMap[key] != nil else {
       var observation: Observation = .init()
@@ -258,7 +272,7 @@ extension vChewingLM.LMUserOverride {
         mutLRUList.removeLast()
       }
       vCLog("UOM: Observation finished with new observation: \(key)")
-      saveCallback()
+      saveCallback?() ?? saveData()
       return
     }
     // 這裡還是不要做 decayCallback 判定「是否不急著更新觀察」了，不然會在嘗試覆寫掉錯誤的記憶時失敗。
@@ -269,11 +283,11 @@ extension vChewingLM.LMUserOverride {
       mutLRUList.insert(theNeta, at: 0)
       mutLRUMap[key] = theNeta
       vCLog("UOM: Observation finished with existing observation: \(key)")
-      saveCallback()
+      saveCallback?() ?? saveData()
     }
   }
 
-  func getSuggestion(key: String, timestamp: Double, headReading: String) -> Suggestion {
+  func getSuggestion(key: String, timestamp: Double, headReading: String) -> LMAssembly.OverrideSuggestion {
     guard !key.isEmpty, let kvPair = mutLRUMap[key] else { return .init() }
     let observation: Observation = kvPair.observation
     var candidates: [(String, Megrez.Unigram)] = .init()

Packages/vChewing_LangModelAssembly/Sources/LangModelAssembly/TestCoreLMSQLData.swift

@@ -1,9 +1,5 @@
-//
+// 下述詞頻資料取自 libTaBE 資料庫 (http://sourceforge.net/projects/libtabe/)
-//  File.swift
+// (2002 最終版). 該專案於 1999 年由 Pai-Hsiang Hsiao 發起、以 BSD 授權發行。
-//
-//
-//  Created by ShikiSuen on 2023/11/26.
-//
 
 import Foundation
 

Packages/vChewing_LangModelAssembly/Sources/LangModelAssembly/vChewingLMAssembly_Common.swift

@@ -10,7 +10,7 @@ import Foundation
 import Shared
 import SQLite3
 
-public enum vChewingLM {
+public enum LMAssembly {
   enum FileErrors: Error {
     case fileHandleError(String)
   }

Packages/vChewing_LangModelAssembly/Tests/LangModelAssemblyTests/InputTokenTests.swift

@@ -57,8 +57,8 @@ final class InputTokenTests: XCTestCase {
   }
 
   func testGeneratedResultsFromLMInstantiator() throws {
-    let instance = vChewingLM.LMInstantiator(isCHS: true)
+    let instance = LMAssembly.LMInstantiator(isCHS: true)
-    XCTAssertTrue(vChewingLM.LMInstantiator.connectToTestSQLDB())
+    XCTAssertTrue(LMAssembly.LMInstantiator.connectToTestSQLDB())
     instance.setOptions { config in
       config.isCNSEnabled = false
       config.isSymbolEnabled = false
@@ -70,6 +70,6 @@ final class InputTokenTests: XCTestCase {
     )
     let x = instance.unigramsFor(keyArray: ["ㄐㄧㄣ", "ㄊㄧㄢ", "ㄖˋ", "ㄑㄧˊ"]).description
     print(x)
-    vChewingLM.LMInstantiator.disconnectSQLDB()
+    LMAssembly.LMInstantiator.disconnectSQLDB()
   }
 }

Packages/vChewing_LangModelAssembly/Tests/LangModelAssemblyTests/LMCassetteTests.swift

@@ -20,7 +20,7 @@ private let testDataPath: String = packageRootPath + "/Tests/TestCINData/"
 final class LMCassetteTests: XCTestCase {
   func testCassetteLoadWubi86() throws {
     let pathCINFile = testDataPath + "wubi.cin"
-    var lmCassette = vChewingLM.LMCassette()
+    var lmCassette = LMAssembly.LMCassette()
     NSLog("LMCassette: Start loading CIN.")
     lmCassette.open(pathCINFile)
     NSLog("LMCassette: Finished loading CIN. Entries: \(lmCassette.count)")
@@ -41,7 +41,7 @@ final class LMCassetteTests: XCTestCase {
 
   func testCassetteLoadArray30() throws {
     let pathCINFile = testDataPath + "array30.cin2"
-    var lmCassette = vChewingLM.LMCassette()
+    var lmCassette = LMAssembly.LMCassette()
     NSLog("LMCassette: Start loading CIN.")
     lmCassette.open(pathCINFile)
     NSLog("LMCassette: Finished loading CIN. Entries: \(lmCassette.count)")

Packages/vChewing_LangModelAssembly/Tests/LangModelAssemblyTests/LMCoreEXTests.swift

@@ -38,7 +38,7 @@ private let sampleData: String = #"""
 
 final class LMCoreEXTests: XCTestCase {
   func testLMCoreEXAsFactoryCoreDict() throws {
-    var lmTest = vChewingLM.LMCoreEX(
+    var lmTest = LMAssembly.LMCoreEX(
       reverse: false, consolidate: false, defaultScore: 0, forceDefaultScore: false
     )
     lmTest.replaceData(textData: sampleData)

Packages/vChewing_LangModelAssembly/Tests/LangModelAssemblyTests/LMInstantiatorSQLTests.swift

@@ -22,8 +22,8 @@ private let expectedReverseLookupResults: [String] = [
 
 final class LMInstantiatorSQLTests: XCTestCase {
   func testSQL() throws {
-    let instance = vChewingLM.LMInstantiator(isCHS: true)
+    let instance = LMAssembly.LMInstantiator(isCHS: true)
-    XCTAssertTrue(vChewingLM.LMInstantiator.connectToTestSQLDB())
+    XCTAssertTrue(LMAssembly.LMInstantiator.connectToTestSQLDB())
     instance.setOptions { config in
       config.isCNSEnabled = false
       config.isSymbolEnabled = false
@@ -41,13 +41,13 @@ final class LMInstantiatorSQLTests: XCTestCase {
     XCTAssertEqual(instance.unigramsFor(keyArray: strRefutationKey).count, 10)
     XCTAssertEqual(instance.unigramsFor(keyArray: strBoobsKey).last?.description, "(☉☉,-13.0)")
     // 再測試反查。
-    XCTAssertEqual(vChewingLM.LMInstantiator.getFactoryReverseLookupData(with: "和"), expectedReverseLookupResults)
+    XCTAssertEqual(LMAssembly.LMInstantiator.getFactoryReverseLookupData(with: "和"), expectedReverseLookupResults)
-    vChewingLM.LMInstantiator.disconnectSQLDB()
+    LMAssembly.LMInstantiator.disconnectSQLDB()
   }
 
   func testCNSMask() throws {
-    let instance = vChewingLM.LMInstantiator(isCHS: false)
+    let instance = LMAssembly.LMInstantiator(isCHS: false)
-    XCTAssertTrue(vChewingLM.LMInstantiator.connectToTestSQLDB())
+    XCTAssertTrue(LMAssembly.LMInstantiator.connectToTestSQLDB())
     instance.setOptions { config in
       config.isCNSEnabled = false
       config.isSymbolEnabled = false

Packages/vChewing_LangModelAssembly/Tests/LangModelAssemblyTests/LMUserOverrideTests.swift

@@ -17,12 +17,12 @@ private let halfLife: Double = 5400
 private let nullURL = URL(fileURLWithPath: "/dev/null")
 
 final class LMUserOverrideTests: XCTestCase {
-  private func observe(who uom: vChewingLM.LMUserOverride, key: String, candidate: String, timestamp stamp: Double) {
+  private func observe(who uom: LMAssembly.LMUserOverride, key: String, candidate: String, timestamp stamp: Double) {
     uom.doObservation(key: key, candidate: candidate, timestamp: stamp, forceHighScoreOverride: false, saveCallback: {})
   }
 
   func testUOM_1_BasicOps() throws {
-    let uom = vChewingLM.LMUserOverride(capacity: capacity, decayConstant: Double(halfLife), dataURL: nullURL)
+    let uom = LMAssembly.LMUserOverride(capacity: capacity, decayConstant: Double(halfLife), dataURL: nullURL)
     let key = "((ㄕㄣˊ-ㄌㄧˇ-ㄌㄧㄥˊ-ㄏㄨㄚˊ,神里綾華),(ㄉㄜ˙,的),ㄍㄡˇ)"
     let headReading = "ㄍㄡˇ"
     let expectedSuggestion = "狗"
@@ -45,7 +45,7 @@ final class LMUserOverrideTests: XCTestCase {
   }
 
   func testUOM_2_NewestAgainstRepeatedlyUsed() throws {
-    let uom = vChewingLM.LMUserOverride(capacity: capacity, decayConstant: Double(halfLife), dataURL: nullURL)
+    let uom = LMAssembly.LMUserOverride(capacity: capacity, decayConstant: Double(halfLife), dataURL: nullURL)
     let key = "((ㄕㄣˊ-ㄌㄧˇ-ㄌㄧㄥˊ-ㄏㄨㄚˊ,神里綾華),(ㄉㄜ˙,的),ㄍㄡˇ)"
     let headReading = "ㄍㄡˇ"
     let valRepeatedlyUsed = "狗" // 更常用
@@ -74,7 +74,7 @@ final class LMUserOverrideTests: XCTestCase {
     let b = (key: "((ㄆㄞˋ-ㄇㄥˊ,派蒙),(ㄉㄜ˙,的),ㄐㄧㄤˇ-ㄐㄧㄣ)", value: "伙食費", head: "ㄏㄨㄛˇ-ㄕˊ-ㄈㄟˋ")
     let c = (key: "((ㄍㄨㄛˊ-ㄅㄥ,國崩),(ㄉㄜ˙,的),ㄇㄠˋ-ㄗ˙)", value: "帽子", head: "ㄇㄠˋ-ㄗ˙")
     let d = (key: "((ㄌㄟˊ-ㄉㄧㄢˋ-ㄐㄧㄤ-ㄐㄩㄣ,雷電將軍),(ㄉㄜ˙,的),ㄐㄧㄠˇ-ㄔㄡˋ)", value: "腳臭", head: "ㄐㄧㄠˇ-ㄔㄡˋ")
-    let uom = vChewingLM.LMUserOverride(capacity: 2, decayConstant: Double(halfLife), dataURL: nullURL)
+    let uom = LMAssembly.LMUserOverride(capacity: 2, decayConstant: Double(halfLife), dataURL: nullURL)
     observe(who: uom, key: a.key, candidate: a.value, timestamp: nowTimeStamp)
     observe(who: uom, key: b.key, candidate: b.value, timestamp: nowTimeStamp + halfLife * 1)
     observe(who: uom, key: c.key, candidate: c.value, timestamp: nowTimeStamp + halfLife * 2)

Packages/vChewing_LangModelAssembly/Tests/LangModelAssemblyTests/NumPadDataTests.swift

@@ -13,7 +13,7 @@ import XCTest
 
 final class LMInstantiatorNumericPadTests: XCTestCase {
   func testSQL() throws {
-    let instance = vChewingLM.LMInstantiator(isCHS: true)
+    let instance = LMAssembly.LMInstantiator(isCHS: true)
     instance.setOptions { config in
       config.numPadFWHWStatus = nil
     }

Packages/vChewing_MainAssembly/Sources/MainAssembly/InputHandler/InputHandler_Core.swift

@@ -19,8 +19,7 @@ import Tekkon
 // MARK: - InputHandler 自身協定 (Protocol).
 
 public protocol InputHandlerProtocol {
-  var currentLM: vChewingLM.LMInstantiator { get set }
+  var currentLM: LMAssembly.LMInstantiator { get set }
-  var currentUOM: vChewingLM.LMUserOverride { get set }
   var delegate: InputHandlerDelegate? { get set }
   var keySeparator: String { get }
   static var keySeparator: String { get }
@@ -99,8 +98,7 @@ public class InputHandler: InputHandlerProtocol {
   var composer: Tekkon.Composer = .init() // 注拼槽
   var compositor: Megrez.Compositor // 組字器
 
-  public var currentUOM: vChewingLM.LMUserOverride
+  public var currentLM: LMAssembly.LMInstantiator {
-  public var currentLM: vChewingLM.LMInstantiator {
     didSet {
       compositor.langModel = .init(withLM: currentLM)
       clear()
@@ -108,10 +106,9 @@ public class InputHandler: InputHandlerProtocol {
   }
 
   /// 初期化。
-  public init(lm: vChewingLM.LMInstantiator, uom: vChewingLM.LMUserOverride, pref: PrefMgrProtocol) {
+  public init(lm: LMAssembly.LMInstantiator, pref: PrefMgrProtocol) {
     prefs = pref
     currentLM = lm
-    currentUOM = uom
     /// 同步組字器單個詞的幅位長度上限。
     Megrez.Compositor.maxSpanLength = prefs.maxCandidateLength
     /// 組字器初期化。因為是首次初期化變數，所以這裡不能用 ensureCompositor() 代勞。
@@ -369,8 +366,8 @@ public class InputHandler: InputHandlerProtocol {
     let currentNode = currentWalk.findNode(at: actualNodeCursorPosition, target: &accumulatedCursor)
     guard let currentNode = currentNode else { return }
 
-    uom: if currentNode.currentUnigram.score > -12, prefs.fetchSuggestionsFromUserOverrideModel {
+    uomProcessing: if currentNode.currentUnigram.score > -12, prefs.fetchSuggestionsFromUserOverrideModel {
-      if skipObservation { break uom }
+      if skipObservation { break uomProcessing }
       vCLog("UOM: Start Observation.")
       // 這個過程可能會因為使用者半衰記憶模組內部資料錯亂、而導致輸入法在選字時崩潰。
       // 於是在這裡引入災後狀況察覺專用變數，且先開啟該開關。順利執行完觀察後會關閉。
@@ -378,9 +375,9 @@ public class InputHandler: InputHandlerProtocol {
       prefs.failureFlagForUOMObservation = true
       // 令半衰記憶模組觀測給定的三元圖。
       // 這個過程會讓半衰引擎根據當前上下文生成三元圖索引鍵。
-      currentUOM.performObservation(
+      currentLM.performUOMObservation(
         walkedBefore: previousWalk, walkedAfter: currentWalk, cursor: actualNodeCursorPosition,
-        timestamp: Date().timeIntervalSince1970, saveCallback: { self.currentUOM.saveData() }
+        timestamp: Date().timeIntervalSince1970
       )
       // 如果沒有出現崩框的話，那就將這個開關復位。
       prefs.failureFlagForUOMObservation = false
@@ -432,7 +429,7 @@ public class InputHandler: InputHandlerProtocol {
     /// 如果這個開關沒打開的話，直接放棄執行這個函式。
     if !prefs.fetchSuggestionsFromUserOverrideModel { return arrResult }
     /// 獲取來自半衰記憶模組的建議結果
-    let suggestion = currentUOM.fetchSuggestion(
+    let suggestion = currentLM.fetchUOMSuggestion(
       currentWalk: compositor.walkedNodes, cursor: actualNodeCursorPosition, timestamp: Date().timeIntervalSince1970
     )
     arrResult.append(contentsOf: suggestion.candidates)

Packages/vChewing_MainAssembly/Sources/MainAssembly/LangModelManager/LMMgr_Core.swift

@@ -15,26 +15,20 @@ import SwiftExtension
 // MARK: - Input Mode Extension for Language Models
 
 public extension Shared.InputMode {
-  private static let lmCHS = vChewingLM.LMInstantiator(isCHS: true)
+  private static let lmCHS = LMAssembly.LMInstantiator(
-  private static let lmCHT = vChewingLM.LMInstantiator(isCHS: false)
+    isCHS: true, uomDataURL: LMMgr.userOverrideModelDataURL(.imeModeCHS)
-  private static let uomCHS = vChewingLM.LMUserOverride(dataURL: LMMgr.userOverrideModelDataURL(.imeModeCHS))
+  )
-  private static let uomCHT = vChewingLM.LMUserOverride(dataURL: LMMgr.userOverrideModelDataURL(.imeModeCHT))
+  private static let lmCHT = LMAssembly.LMInstantiator(
+    isCHS: false, uomDataURL: LMMgr.userOverrideModelDataURL(.imeModeCHT)
+  )
 
-  var langModel: vChewingLM.LMInstantiator {
+  var langModel: LMAssembly.LMInstantiator {
     switch self {
     case .imeModeCHS: return Self.lmCHS
     case .imeModeCHT: return Self.lmCHT
     case .imeModeNULL: return .init()
     }
   }
-
-  var uom: vChewingLM.LMUserOverride {
-    switch self {
-    case .imeModeCHS: return Self.uomCHS
-    case .imeModeCHT: return Self.uomCHT
-    case .imeModeNULL: return .init(dataURL: LMMgr.userOverrideModelDataURL(IMEApp.currentInputMode))
-    }
-  }
 }
 
 // MARK: - Language Model Manager.
@@ -54,14 +48,14 @@ public class LMMgr {
     Self.loadUserPhrasesData()
   }
 
-  public static var isCoreDBConnected: Bool { vChewingLM.LMInstantiator.isSQLDBConnected }
+  public static var isCoreDBConnected: Bool { LMAssembly.LMInstantiator.isSQLDBConnected }
 
   public static func connectCoreDB(dbPath: String? = nil) {
     guard let path: String = dbPath ?? Self.getCoreDictionaryDBPath() else {
       assertionFailure("vChewing factory SQLite data not found.")
       return
     }
-    let result = vChewingLM.LMInstantiator.connectSQLDB(dbPath: path)
+    let result = LMAssembly.LMInstantiator.connectSQLDB(dbPath: path)
     assert(result, "vChewing factory SQLite connection failed.")
     Notifier.notify(
       message: NSLocalizedString("Core Dict loading complete.", comment: "")
@@ -71,10 +65,10 @@ public class LMMgr {
   /// 載入磁帶資料。
   /// - Remark: cassettePath() 會在輸入法停用磁帶時直接返回
   public static func loadCassetteData() {
-    vChewingLM.LMInstantiator.loadCassetteData(path: cassettePath())
+    LMAssembly.LMInstantiator.loadCassetteData(path: cassettePath())
   }
 
-  public static func loadUserPhrasesData(type: vChewingLM.ReplacableUserDataType? = nil) {
+  public static func loadUserPhrasesData(type: LMAssembly.ReplacableUserDataType? = nil) {
     guard let type = type else {
       Shared.InputMode.validCases.forEach { mode in
         mode.langModel.loadUserPhrasesData(
@@ -82,7 +76,7 @@ public class LMMgr {
           filterPath: userDictDataURL(mode: mode, type: .theFilter).path
         )
         mode.langModel.loadUserSymbolData(path: userDictDataURL(mode: mode, type: .theSymbols).path)
-        mode.uom.loadData(fromURL: userOverrideModelDataURL(mode))
+        mode.langModel.loadUOMData()
       }
 
       if PrefMgr.shared.associatedPhrasesEnabled { Self.loadUserAssociatesData() }
@@ -187,12 +181,12 @@ public class LMMgr {
   // MARK: UOM
 
   public static func saveUserOverrideModelData() {
-    let globalQueue = DispatchQueue(label: "vChewingLM_UOM", qos: .unspecified, attributes: .concurrent)
+    let globalQueue = DispatchQueue(label: "LMAssembly_UOM", qos: .unspecified, attributes: .concurrent)
     let group = DispatchGroup()
     Shared.InputMode.validCases.forEach { mode in
       group.enter()
       globalQueue.async {
-        mode.uom.saveData(toURL: userOverrideModelDataURL(mode))
+        mode.langModel.saveUOMData()
         group.leave()
       }
     }
@@ -201,11 +195,11 @@ public class LMMgr {
   }
 
   public static func bleachSpecifiedSuggestions(targets: [String], mode: Shared.InputMode) {
-    mode.uom.bleachSpecifiedSuggestions(targets: targets, saveCallback: { mode.uom.saveData() })
+    mode.langModel.bleachSpecifiedUOMSuggestions(targets: targets)
   }
 
   public static func removeUnigramsFromUserOverrideModel(_ mode: Shared.InputMode) {
-    mode.uom.bleachUnigrams(saveCallback: { mode.uom.saveData() })
+    mode.langModel.bleachUOMUnigrams()
   }
 
   public static func relocateWreckedUOMData() {
@@ -227,6 +221,6 @@ public class LMMgr {
   }
 
   public static func clearUserOverrideModelData(_ mode: Shared.InputMode = .imeModeNULL) {
-    mode.uom.clearData(withURL: userOverrideModelDataURL(mode))
+    mode.langModel.clearUOMData()
   }
 }

Packages/vChewing_MainAssembly/Sources/MainAssembly/LangModelManager/LMMgr_PhraseEditorDelegate.swift

@@ -17,23 +17,23 @@ import Shared
 extension LMMgr: PhraseEditorDelegate {
   public var currentInputMode: Shared.InputMode { IMEApp.currentInputMode }
 
-  public func openPhraseFile(mode: Shared.InputMode, type: vChewingLM.ReplacableUserDataType, using app: FileOpenMethod) {
+  public func openPhraseFile(mode: Shared.InputMode, type: LMAssembly.ReplacableUserDataType, using app: FileOpenMethod) {
     Self.openPhraseFile(fromURL: Self.userDictDataURL(mode: mode, type: type), using: app)
   }
 
   public func consolidate(text strProcessed: inout String, pragma shouldCheckPragma: Bool) {
-    vChewingLM.LMConsolidator.consolidate(text: &strProcessed, pragma: shouldCheckPragma)
+    LMAssembly.LMConsolidator.consolidate(text: &strProcessed, pragma: shouldCheckPragma)
   }
 
   public func checkIfPhrasePairExists(userPhrase: String, mode: Shared.InputMode, key unigramKey: String) -> Bool {
     Self.checkIfPhrasePairExists(userPhrase: userPhrase, mode: mode, keyArray: [unigramKey])
   }
 
-  public func retrieveData(mode: Shared.InputMode, type: vChewingLM.ReplacableUserDataType) -> String {
+  public func retrieveData(mode: Shared.InputMode, type: LMAssembly.ReplacableUserDataType) -> String {
     Self.retrieveData(mode: mode, type: type)
   }
 
-  public static func retrieveData(mode: Shared.InputMode, type: vChewingLM.ReplacableUserDataType) -> String {
+  public static func retrieveData(mode: Shared.InputMode, type: LMAssembly.ReplacableUserDataType) -> String {
     vCLog("Retrieving data. Mode: \(mode.localizedDescription), type: \(type.localizedDescription)")
     let theURL = Self.userDictDataURL(mode: mode, type: type)
     do {
@@ -44,12 +44,12 @@ extension LMMgr: PhraseEditorDelegate {
     }
   }
 
-  public func saveData(mode: Shared.InputMode, type: vChewingLM.ReplacableUserDataType, data: String) -> String {
+  public func saveData(mode: Shared.InputMode, type: LMAssembly.ReplacableUserDataType, data: String) -> String {
     Self.saveData(mode: mode, type: type, data: data)
   }
 
   @discardableResult public static func saveData(
-    mode: Shared.InputMode, type: vChewingLM.ReplacableUserDataType, data: String
+    mode: Shared.InputMode, type: LMAssembly.ReplacableUserDataType, data: String
   ) -> String {
     DispatchQueue.main.async {
       let theURL = Self.userDictDataURL(mode: mode, type: type)

Packages/vChewing_MainAssembly/Sources/MainAssembly/LangModelManager/LMMgr_UserPhraseStructure.swift

@@ -88,7 +88,7 @@ public extension LMMgr {
       /// 有些使用者的語彙檔案已經過於龐大了（超過一千行），
       /// 每次寫入時都全文整理格式的話，會引發嚴重的效能問題。
       /// 所以這裡不再強制要求整理格式。
-      let theType: vChewingLM.ReplacableUserDataType = toFilter ? .theFilter : .thePhrases
+      let theType: LMAssembly.ReplacableUserDataType = toFilter ? .theFilter : .thePhrases
       let theURL = LMMgr.userDictDataURL(mode: inputMode, type: theType)
       var fileSize: UInt64?
       do {
@@ -143,7 +143,7 @@ public extension LMMgr {
         }
       }
       let theURL = LMMgr.userDictDataURL(mode: inputMode, type: .theFilter)
-      if forceConsolidate, !vChewingLM.LMConsolidator.consolidate(path: theURL.path, pragma: false) { return false }
+      if forceConsolidate, !LMAssembly.LMConsolidator.consolidate(path: theURL.path, pragma: false) { return false }
       // Get FileSize.
       var fileSize: UInt64?
       do {

Packages/vChewing_MainAssembly/Sources/MainAssembly/LangModelManager/LMMgr_Utilities.swift

@@ -61,7 +61,7 @@ public extension LMMgr {
   ///   - mode: 繁簡模式。
   ///   - type: 辭典資料類型
   /// - Returns: 資料路徑（URL）。
-  static func userDictDataURL(mode: Shared.InputMode, type: vChewingLM.ReplacableUserDataType) -> URL {
+  static func userDictDataURL(mode: Shared.InputMode, type: LMAssembly.ReplacableUserDataType) -> URL {
     var fileName: String = {
       switch type {
       case .thePhrases: return "userdata"
@@ -271,7 +271,7 @@ public extension LMMgr {
     return true
   }
 
-  static func openUserDictFile(type: vChewingLM.ReplacableUserDataType, dual: Bool = false, alt: Bool) {
+  static func openUserDictFile(type: LMAssembly.ReplacableUserDataType, dual: Bool = false, alt: Bool) {
     let app: FileOpenMethod = alt ? .textEdit : .finder
     openPhraseFile(fromURL: userDictDataURL(mode: IMEApp.currentInputMode, type: type), using: app)
     guard dual else { return }
@@ -324,7 +324,7 @@ public extension LMMgr {
     /// 前者的話，需要該檔案存在的人自己會建立。
     /// 後者的話，你在敲字時自己就會建立。
     var failed = false
-    caseCheck: for type in vChewingLM.ReplacableUserDataType.allCases {
+    caseCheck: for type in LMAssembly.ReplacableUserDataType.allCases {
       let templateName = Self.templateName(for: type, mode: mode)
       if !ensureFileExists(userDictDataURL(mode: mode, type: type), deployTemplate: templateName) {
         failed = true
@@ -334,7 +334,7 @@ public extension LMMgr {
     return !failed
   }
 
-  internal static func templateName(for type: vChewingLM.ReplacableUserDataType, mode: Shared.InputMode) -> String {
+  internal static func templateName(for type: LMAssembly.ReplacableUserDataType, mode: Shared.InputMode) -> String {
     switch type {
     case .thePhrases: return kTemplateNameUserPhrases
     case .theFilter: return kTemplateNameUserFilterList

Packages/vChewing_MainAssembly/Sources/MainAssembly/RevLookupUI/CtlRevLookupWindow.swift

@@ -150,7 +150,7 @@ class FrmRevLookupWindow: NSWindow {
         strBuilder.append("Maximum 15 results returnable.".localized + "\n")
         break theLoop
       }
-      let arrResult = vChewingLM.LMInstantiator.getFactoryReverseLookupData(with: char)?.deduplicated ?? []
+      let arrResult = LMAssembly.LMInstantiator.getFactoryReverseLookupData(with: char)?.deduplicated ?? []
       if !arrResult.isEmpty {
         strBuilder.append(char + "\t")
         strBuilder.append(arrResult.joined(separator: ", "))

Packages/vChewing_MainAssembly/Sources/MainAssembly/SessionController/SessionCtl_Core.swift

@@ -178,7 +178,6 @@ public class SessionCtl: IMKInputController {
         // ----------------------------
         /// 重設所有語言模組。這裡不需要做按需重設，因為對運算量沒有影響。
         inputHandler?.currentLM = inputMode.langModel // 會自動更新組字引擎內的模組。
-        inputHandler?.currentUOM = inputMode.uom
         /// 清空注拼槽＋同步最新的注拼槽排列設定。
         inputHandler?.ensureKeyboardParser()
         /// 將輸入法偏好設定同步至語言模組內。
@@ -214,9 +213,7 @@ public class SessionCtl: IMKInputController {
       // 關掉所有之前的副本的視窗。
       Self.current?.hidePalettes()
       Self.current = self
-      self.inputHandler = InputHandler(
+      self.inputHandler = InputHandler(lm: self.inputMode.langModel, pref: PrefMgr.shared)
-        lm: self.inputMode.langModel, uom: self.inputMode.uom, pref: PrefMgr.shared
-      )
       self.inputHandler?.delegate = self
       self.syncBaseLMPrefs()
       // 下述兩行很有必要，否則輸入法會在手動重啟之後無法立刻生效。
@@ -313,9 +310,7 @@ public extension SessionCtl {
       if self.isActivated { return }
 
       // 這裡不需要 setValue()，因為 IMK 會在自動呼叫 activateServer() 之後自動執行 setValue()。
-      self.inputHandler = InputHandler(
+      self.inputHandler = InputHandler(lm: self.inputMode.langModel, pref: PrefMgr.shared)
-        lm: self.inputMode.langModel, uom: self.inputMode.uom, pref: PrefMgr.shared
-      )
       self.inputHandler?.delegate = self
       self.syncBaseLMPrefs()
 

Packages/vChewing_MainAssembly/Sources/MainAssembly/Settings/SettingsCocoa/VwrSettingsPaneCocoaPhrases.swift

@@ -140,7 +140,7 @@ extension SettingsPanesCocoa.Phrases: NSTextViewDelegate, NSTextFieldDelegate {
     }
   }
 
-  var selUserDataType: vChewingLM.ReplacableUserDataType {
+  var selUserDataType: LMAssembly.ReplacableUserDataType {
     switch cmbPEDataTypeMenu.selectedTag() {
     case 0: return .thePhrases
     case 1: return .theFilter
@@ -238,7 +238,7 @@ extension SettingsPanesCocoa.Phrases: NSTextViewDelegate, NSTextFieldDelegate {
     // 嚴重警告：NSMenu.items 在 macOS 10.13 為止的系統下是唯讀的！！
     // 往這個 property 裡面直接寫東西會導致整個視窗叫不出來！！！
     cmbPEDataTypeMenu.menu?.appendItems {
-      for (tag, neta) in vChewingLM.ReplacableUserDataType.allCases.enumerated() {
+      for (tag, neta) in LMAssembly.ReplacableUserDataType.allCases.enumerated() {
         NSMenu.Item(verbatim: neta.localizedDescription)?.tag(tag)
       }
     }
@@ -332,7 +332,7 @@ extension SettingsPanesCocoa.Phrases: NSTextViewDelegate, NSTextFieldDelegate {
     DispatchQueue.main.async { [weak self] in
       guard let self = self else { return }
       self.isLoading = true
-      vChewingLM.LMConsolidator.consolidate(text: &self.tfdPETextEditor.string, pragma: false)
+      LMAssembly.LMConsolidator.consolidate(text: &self.tfdPETextEditor.string, pragma: false)
       if self.selUserDataType == .thePhrases {
         LMMgr.shared.tagOverrides(in: &self.tfdPETextEditor.string, mode: self.selInputMode)
       }
@@ -416,7 +416,7 @@ private enum PETerminology {
     case weightInputBox =
       "If not filling the weight, it will be 0.0, the maximum one. An ideal weight situates in [-9.5, 0], making itself can be captured by the walking algorithm. The exception is -114.514, the disciplinary weight. The walking algorithm will ignore it unless it is the unique result."
 
-    public static func sampleDictionaryContent(for type: vChewingLM.ReplacableUserDataType) -> String {
+    public static func sampleDictionaryContent(for type: LMAssembly.ReplacableUserDataType) -> String {
       var result = ""
       switch type {
       case .thePhrases:

Packages/vChewing_MainAssembly/Tests/MainAssemblyTests/MainAssemblyTests_Core.swift

@@ -34,16 +34,15 @@ func vCTestLog(_ str: String) {
 /// 該單元測試使用獨立的語彙資料，因此會在選字時的候選字
 /// 順序等方面與威注音輸入法實際使用時的體驗有差異。
 class MainAssemblyTests: XCTestCase {
-  let testUOM = LangModelAssembly.vChewingLM.LMUserOverride(dataURL: .init(fileURLWithPath: "/dev/null"))
+  var testLM = LMAssembly.LMInstantiator.construct { _ in
-  var testLM = LangModelAssembly.vChewingLM.LMInstantiator.construct { _ in
+    LMAssembly.LMInstantiator.connectToTestSQLDB()
-    vChewingLM.LMInstantiator.connectToTestSQLDB()
   }
 
   static let testServer = IMKServer(name: "org.atelierInmu.vChewing.MainAssembly.UnitTests_Connection", bundleIdentifier: "org.atelierInmu.vChewing.MainAssembly.UnitTests")
 
   static var _testHandler: InputHandler?
   var testHandler: InputHandler {
-    let result = Self._testHandler ?? InputHandler(lm: testLM, uom: testUOM, pref: PrefMgr.shared)
+    let result = Self._testHandler ?? InputHandler(lm: testLM, pref: PrefMgr.shared)
     if Self._testHandler == nil { Self._testHandler = result }
     return result
   }
@@ -65,7 +64,7 @@ class MainAssemblyTests: XCTestCase {
   let dataTab = NSEvent.KeyEventData(chars: NSEvent.SpecialKey.tab.unicodeScalar.description, keyCode: KeyCode.kTab.rawValue)
 
   func clearTestUOM() {
-    testUOM.clearData(withURL: URL(fileURLWithPath: "/dev/null"))
+    testLM.clearUOMData()
   }
 
   func typeSentenceOrCandidates(_ sequence: String) {
@@ -106,11 +105,11 @@ class MainAssemblyTests: XCTestCase {
   }
 }
 
-extension vChewingLM.LMInstantiator {
+extension LMAssembly.LMInstantiator {
   static func construct(
-    isCHS: Bool = false, completionHandler: @escaping (_ this: vChewingLM.LMInstantiator) -> Void
+    isCHS: Bool = false, completionHandler: @escaping (_ this: LMAssembly.LMInstantiator) -> Void
-  ) -> vChewingLM.LMInstantiator {
+  ) -> LMAssembly.LMInstantiator {
-    let this = vChewingLM.LMInstantiator(isCHS: isCHS)
+    let this = LMAssembly.LMInstantiator(isCHS: isCHS)
     completionHandler(this)
     return this
   }

Packages/vChewing_PhraseEditorUI/Sources/PhraseEditorUI/PhraseEditorDelegate.swift

@@ -12,11 +12,11 @@ import Shared
 
 public protocol PhraseEditorDelegate {
   var currentInputMode: Shared.InputMode { get }
-  func retrieveData(mode: Shared.InputMode, type: vChewingLM.ReplacableUserDataType) -> String
+  func retrieveData(mode: Shared.InputMode, type: LMAssembly.ReplacableUserDataType) -> String
-  @discardableResult func saveData(mode: Shared.InputMode, type: vChewingLM.ReplacableUserDataType, data: String)
+  @discardableResult func saveData(mode: Shared.InputMode, type: LMAssembly.ReplacableUserDataType, data: String)
     -> String
   func checkIfPhrasePairExists(userPhrase: String, mode: Shared.InputMode, key unigramKey: String) -> Bool
   func consolidate(text strProcessed: inout String, pragma shouldCheckPragma: Bool)
-  func openPhraseFile(mode: Shared.InputMode, type: vChewingLM.ReplacableUserDataType, using: FileOpenMethod)
+  func openPhraseFile(mode: Shared.InputMode, type: LMAssembly.ReplacableUserDataType, using: FileOpenMethod)
   func tagOverrides(in strProcessed: inout String, mode: Shared.InputMode)
 }

Packages/vChewing_PhraseEditorUI/Sources/PhraseEditorUI/PhraseEditorUI.swift

@@ -37,7 +37,7 @@ public struct VwrPhraseEditorUI: View {
   @State var txtAddPhraseField3 = ""
   @State var txtAddPhraseField4 = ""
   @State public var selInputMode: Shared.InputMode = .imeModeNULL
-  @State public var selUserDataType: vChewingLM.ReplacableUserDataType = .thePhrases
+  @State public var selUserDataType: LMAssembly.ReplacableUserDataType = .thePhrases
   @State private var isLoading = false
   @State private var textEditorTooltip = PETerms.TooltipTexts.sampleDictionaryContent(for: .thePhrases)
   public weak var window: NSWindow?
@@ -212,16 +212,16 @@ public struct VwrPhraseEditorUI: View {
         }
         .labelsHidden()
         Picker("", selection: $selUserDataType.didChange { dropDownMenuDidChange() }) {
-          Text(vChewingLM.ReplacableUserDataType.thePhrases.localizedDescription).tag(
+          Text(LMAssembly.ReplacableUserDataType.thePhrases.localizedDescription).tag(
-            vChewingLM.ReplacableUserDataType.thePhrases)
+            LMAssembly.ReplacableUserDataType.thePhrases)
-          Text(vChewingLM.ReplacableUserDataType.theFilter.localizedDescription).tag(
+          Text(LMAssembly.ReplacableUserDataType.theFilter.localizedDescription).tag(
-            vChewingLM.ReplacableUserDataType.theFilter)
+            LMAssembly.ReplacableUserDataType.theFilter)
-          Text(vChewingLM.ReplacableUserDataType.theReplacements.localizedDescription).tag(
+          Text(LMAssembly.ReplacableUserDataType.theReplacements.localizedDescription).tag(
-            vChewingLM.ReplacableUserDataType.theReplacements)
+            LMAssembly.ReplacableUserDataType.theReplacements)
-          Text(vChewingLM.ReplacableUserDataType.theAssociates.localizedDescription).tag(
+          Text(LMAssembly.ReplacableUserDataType.theAssociates.localizedDescription).tag(
-            vChewingLM.ReplacableUserDataType.theAssociates)
+            LMAssembly.ReplacableUserDataType.theAssociates)
-          Text(vChewingLM.ReplacableUserDataType.theSymbols.localizedDescription).tag(
+          Text(LMAssembly.ReplacableUserDataType.theSymbols.localizedDescription).tag(
-            vChewingLM.ReplacableUserDataType.theSymbols)
+            LMAssembly.ReplacableUserDataType.theSymbols)
         }
         .labelsHidden()
         Button("Reload") {
@@ -340,7 +340,7 @@ public enum PETerms {
     case weightInputBox =
       "If not filling the weight, it will be 0.0, the maximum one. An ideal weight situates in [-9.5, 0], making itself can be captured by the walking algorithm. The exception is -114.514, the disciplinary weight. The walking algorithm will ignore it unless it is the unique result."
 
-    public static func sampleDictionaryContent(for type: vChewingLM.ReplacableUserDataType) -> String {
+    public static func sampleDictionaryContent(for type: LMAssembly.ReplacableUserDataType) -> String {
       var result = ""
       switch type {
       case .thePhrases: