Code cleanup, release prep

app/src/main/java/com/urik/keyboard/UrikApplication.kt

@@ -1,7 +1,10 @@
 package com.urik.keyboard
 
 import android.app.Application
+import android.database.sqlite.SQLiteDatabaseCorruptException
+import com.urik.keyboard.data.database.KeyboardDatabase
 import com.urik.keyboard.di.ApplicationScope
+import com.urik.keyboard.di.DatabaseModule
 import com.urik.keyboard.service.ClipboardMonitorService
 import com.urik.keyboard.settings.SettingsRepository
 import com.urik.keyboard.utils.ErrorLogger
@@ -11,6 +14,7 @@ import kotlinx.coroutines.CoroutineScope
 import kotlinx.coroutines.flow.distinctUntilChanged
 import kotlinx.coroutines.flow.map
 import kotlinx.coroutines.launch
+import net.zetetic.database.sqlcipher.SQLiteNotADatabaseException
 
 /**
  * SQLCipher is loaded before any database access (must precede Room init).
@@ -43,6 +47,10 @@ class UrikApplication : Application() {
                 context = mapOf("thread" to thread.name)
             )
 
+            if (isDatabaseCorruptionException(throwable)) {
+                recoverFromDatabaseCorruption()
+            }
+
             previousHandler?.uncaughtException(thread, throwable)
         }
 
@@ -71,6 +79,28 @@ class UrikApplication : Application() {
         observeClipboardSettings()
     }
 
+    /**
+     * Room opens its connection lazily on a background coroutine, so SQLCipher corruption
+     * errors can arrive here uncaught instead of via [DatabaseModule.provideKeyboardDatabase]'s
+     * recovery path. Deleting the database files lets the next launch recreate it cleanly.
+     */
+    private fun isDatabaseCorruptionException(throwable: Throwable): Boolean = generateSequence(throwable) { it.cause }
+        .any { it is SQLiteNotADatabaseException || it is SQLiteDatabaseCorruptException }
+
+    private fun recoverFromDatabaseCorruption() {
+        try {
+            KeyboardDatabase.resetInstance()
+            DatabaseModule.deleteDatabaseFiles(applicationContext)
+        } catch (e: Exception) {
+            ErrorLogger.logException(
+                component = "Application",
+                severity = ErrorLogger.Severity.CRITICAL,
+                exception = e,
+                context = mapOf("phase" to "uncaught_corruption_recovery_failed")
+            )
+        }
+    }
+
     private fun observeClipboardSettings() {
         applicationScope.launch {
             settingsRepository.settings

app/src/main/java/com/urik/keyboard/UrikInputMethodService.kt

@@ -75,7 +75,6 @@ import com.urik.keyboard.ui.keyboard.components.SwipeDetector
 import com.urik.keyboard.ui.keyboard.components.SwipeKeyboardView
 import com.urik.keyboard.utils.BackspaceUtils
 import com.urik.keyboard.utils.CacheMemoryManager
-import com.urik.keyboard.utils.CursorEditingUtils
 import com.urik.keyboard.utils.ErrorLogger
 import com.urik.keyboard.utils.KanaTransformUtils
 import com.urik.keyboard.utils.KeyboardModeUtils
@@ -238,8 +237,6 @@ open class UrikInputMethodService :
 
     private fun clearSecureFieldState() = imeStateCoordinator.clearSecureFieldState()
 
-    private fun isValidTextInput(text: String): Boolean = CursorEditingUtils.isValidTextInput(text)
-
     private fun updateScriptContext(locale: ULocale) {
         val currentLayout = viewModel.layout.value
         val isRTL = currentLayout?.isRTL ?: false
@@ -483,7 +480,6 @@ open class UrikInputMethodService :
                 outputBridge = outputBridge,
                 suggestionPipeline = suggestionPipeline,
                 autoCorrectionEngine = autoCorrectionEngine,
-                textInputProcessor = textInputProcessor,
                 swipeSpaceManager = swipeSpaceManager,
                 swipeDetector = swipeDetector,
                 candidateBarController = candidateBarController,
@@ -1425,8 +1421,7 @@ open class UrikInputMethodService :
                 textInputProcessor.removeSuggestion(suggestion)
 
                 withContext(Dispatchers.Main) {
-                    val currentSuggestions = inputState.pendingSuggestions.filter { it != suggestion }
-                    inputState.pendingSuggestions = currentSuggestions
+                    val currentSuggestions = inputState.removeSuggestionFromState(suggestion)
                     if (currentSuggestions.isNotEmpty()) {
                         candidateBarController.updateSuggestions(currentSuggestions)
                     } else {
@@ -1969,10 +1964,7 @@ open class UrikInputMethodService :
     }
 
     private companion object {
-        const val DOUBLE_TAP_SPACE_THRESHOLD_MS = 250L
         const val DOUBLE_SHIFT_THRESHOLD_MS = 400L
-        const val NON_SEQUENTIAL_JUMP_THRESHOLD = 5
-        const val WORD_BOUNDARY_CONTEXT_LENGTH = 64
     }
 }
 

app/src/main/java/com/urik/keyboard/di/DatabaseModule.kt

@@ -68,7 +68,6 @@ object DatabaseModule {
         @ApplicationContext context: Context,
         securityManager: DatabaseSecurityManager
     ): KeyboardDatabase {
-        var passphrase: ByteArray? = null
         var alreadyLogged = false
         try {
             if (securityManager.shouldMigrateToEncrypted(context)) {
@@ -86,7 +85,7 @@ object DatabaseModule {
                 }
             }
 
-            passphrase =
+            val passphrase =
                 try {
                     securityManager.getDatabasePassphrase()
                 } catch (e: Exception) {
@@ -102,7 +101,6 @@ object DatabaseModule {
 
             val passphraseWasAvailable = passphrase != null
             val initialPassphrase = passphrase
-            passphrase = null
             return try {
                 opener.open(context, initialPassphrase)
             } catch (e: Exception) {
@@ -184,8 +182,6 @@ object DatabaseModule {
                 initialPassphrase?.fill(0)
             }
         } catch (e: Exception) {
-            passphrase?.fill(0)
-
             if (!alreadyLogged) {
                 ErrorLogger.logException(
                     component = "DatabaseModule",
@@ -198,7 +194,7 @@ object DatabaseModule {
         }
     }
 
-    private fun deleteDatabaseFiles(context: Context) {
+    internal fun deleteDatabaseFiles(context: Context) {
         val dbPath = context.applicationContext.getDatabasePath(KeyboardDatabase.DATABASE_NAME)
         dbPath.delete()
         if (dbPath.exists()) {

app/src/main/java/com/urik/keyboard/di/KeyboardModule.kt

@@ -7,6 +7,7 @@ import com.urik.keyboard.data.database.KeyboardDatabase
 import com.urik.keyboard.data.database.LearnedWordDao
 import com.urik.keyboard.data.database.UserWordBigramDao
 import com.urik.keyboard.data.database.UserWordFrequencyDao
+import com.urik.keyboard.service.BlacklistRepository
 import com.urik.keyboard.service.CharacterVariationService
 import com.urik.keyboard.service.DictionaryBackupManager
 import com.urik.keyboard.service.EmojiSearchManager
@@ -142,6 +143,7 @@ object KeyboardModule {
         wordLearningEngine: WordLearningEngine,
         wordFrequencyRepository: WordFrequencyRepository,
         cacheMemoryManager: CacheMemoryManager,
+        blacklistRepository: BlacklistRepository,
         wordNormalizer: WordNormalizer,
         fatFingerExpander: FatFingerExpander
     ): SpellCheckManager = SpellCheckManager(
@@ -151,6 +153,7 @@ object KeyboardModule {
         wordFrequencyRepository,
         wordNormalizer,
         cacheMemoryManager,
+        blacklistRepository,
         fatFingerExpander = fatFingerExpander
     )
 

app/src/main/java/com/urik/keyboard/dictionary/BareFormRemovelist.kt

@@ -0,0 +1,58 @@
+package com.urik.keyboard.dictionary
+
+/**
+ * Bare forms present in source frequency corpora only as encoding artifacts of
+ * apostrophe forms (e.g. "im" for "i'm", "jai" for "j'ai") plus OCR garbage
+ * (e.g. "weii" for "well"). Filtered out of dictionary lookups, candidates and
+ * completions at runtime so the apostrophe form wins spell check and autocorrect.
+ *
+ * All entries must be lowercase.
+ */
+object BareFormRemovelist {
+    private val REMOVELIST: Map<String, Set<String>> = mapOf(
+        "en" to setOf(
+            "dont", "wont", "cant", "didnt", "doesnt", "wasnt", "isnt", "arent",
+            "wouldnt", "couldnt", "shouldnt", "hadnt", "hasnt", "havent",
+            "thats", "whats", "whos", "hows", "heres", "theres", "wheres",
+            "hes", "shes", "youre", "theyre", "weve", "theyve",
+            "youll", "theyll", "itll", "youve", "youd", "hed", "shed", "wed",
+            "wouldve", "couldve", "shouldve", "werent", "aint",
+            "howd", "whatre", "lm", "ld", "lll", "lts", "lve", "weii",
+            "im", "ive", "theyd"
+        ),
+        "fr" to setOf(
+            "jai", "cest", "tai", "lai", "nai", "quil", "nest",
+            "aujourdhui", "taime", "cetait"
+        ),
+        "de" to setOf(
+            "gibts",
+            "gehts",
+            "habs",
+            "stimmts",
+            "bins",
+            "sies"
+        ),
+        "el" to setOf(
+            "μένα", "σένα", "κάντο", "σενα", "παρόλα", "βάλτο", "δώστο",
+            "κόφτο", "πάρτο", "γιαυτό", "βούλωστο", "σόλο", "απέξω",
+            "πάρτα", "δώστου", "πάρτον", "δώσμου", "πάρτην", "βάλτα",
+            "φέρτον", "απτο", "γιαυτο", "ρίξτου", "απτην", "ναναι",
+            "σευχαριστώ", "μαρέσει", "σαυτό", "απτη", "θαναι", "σαγαπώ",
+            "απτον", "σουπα", "γιαυτόν", "απαυτά", "απτα", "απότι",
+            "νασαι", "σαρέσει", "απαυτό", "γιαυτήν", "σαυτή", "μαυτό",
+            "σαγαπάω", "ναχει", "θαπρεπε", "σαυτόν", "απόλα", "ναμαι",
+            "γιαυτά", "απτους", "γιαυτή", "θαμαι", "γιαυτούς", "θαθελα",
+            "απτις", "απόσο", "οσι", "σαυτο", "μαρέσουν", "σόλους",
+            "θαρθει", "ναμαστε", "σταλήθεια", "μόλα", "αποτι", "θαταν",
+            "εφόσων", "τοχω", "γιαυτον", "θασαι", "μακούς", "γιαυτα",
+            "μαυτόν", "σαφήσω", "απαυτούς", "σαυτά", "θαχει", "γιαυτην",
+            "τοξερα", "μαυτή", "τόνομα", "θαχεις", "σέχω", "μαυτά",
+            "σέναν", "τοκανες"
+        ),
+        "cs" to setOf("dont", "its"),
+        "nl" to setOf("fotos", "autos"),
+        "it" to setOf("lho", "dacqua")
+    )
+
+    fun forLanguage(languageCode: String): Set<String> = REMOVELIST[languageCode] ?: emptySet()
+}

app/src/main/java/com/urik/keyboard/dictionary/UrikDictionary.kt

@@ -4,7 +4,7 @@ package com.urik.keyboard.dictionary
 
 import java.io.InputStream
 
-class UrikDictionary(inputStream: InputStream) {
+class UrikDictionary(inputStream: InputStream, private val removedWords: Set<String> = emptySet()) {
     private val data: ByteArray = inputStream.readBytes()
 
     val wordCount: Int
@@ -33,7 +33,10 @@ class UrikDictionary(inputStream: InputStream) {
         return UrikFormat.dequantizeFreq(b)
     }
 
+    private fun isRemoved(word: String): Boolean = removedWords.isNotEmpty() && word.lowercase() in removedWords
+
     private fun getFreqByte(word: String): Int? {
+        if (isRemoved(word)) return null
         var stateOffset = stateTableOffset
         for ((idx, ch) in word.withIndex()) {
             val arc = findArc(stateOffset, ch) ?: return null
@@ -67,7 +70,10 @@ class UrikDictionary(inputStream: InputStream) {
         val isFinalDawg = (stateHeader and 0x80) != 0
 
         if (isFinalDawg && auto.isAccepting(autoState)) {
-            results.add(path.toString() to autoState.row.last())
+            val word = path.toString()
+            if (!isRemoved(word)) {
+                results.add(word to autoState.row.last())
+            }
         }
 
         var arcOffset = stateAbsOffset + 1
@@ -113,7 +119,12 @@ class UrikDictionary(inputStream: InputStream) {
         val arcCount = stateHeader and 0x7F
         val isFinal = (stateHeader and 0x80) != 0
 
-        if (isFinal) results.add(path.toString() to UrikFormat.dequantizeFreq(incomingFreqByte))
+        if (isFinal) {
+            val word = path.toString()
+            if (!isRemoved(word)) {
+                results.add(word to UrikFormat.dequantizeFreq(incomingFreqByte))
+            }
+        }
         if (results.size >= maxResults) return
 
         var arcOffset = stateAbsOffset + 1

app/src/main/java/com/urik/keyboard/service/AutoCorrectionEngine.kt

@@ -8,8 +8,8 @@ import javax.inject.Singleton
 sealed class AutocorrectDecision {
     data object None : AutocorrectDecision()
     data class Correct(val suggestion: String) : AutocorrectDecision()
-    data object Pause : AutocorrectDecision()
-    data object ContractionBypass : AutocorrectDecision()
+    data class Pause(val suggestions: List<SpellingSuggestion>) : AutocorrectDecision()
+    data class ContractionBypass(val suggestions: List<SpellingSuggestion>) : AutocorrectDecision()
     data class Suggestions(val list: List<SpellingSuggestion>) : AutocorrectDecision()
 }
 
@@ -45,7 +45,7 @@ constructor(private val textInputProcessor: TextInputProcessor) {
             textInputProcessor.hasDominantContractionForm(buffer)
 
         if (bypassForContraction) {
-            return AutocorrectDecision.ContractionBypass
+            return AutocorrectDecision.ContractionBypass(textInputProcessor.getSuggestions(buffer))
         }
 
         if (isValid) {
@@ -79,7 +79,7 @@ constructor(private val textInputProcessor: TextInputProcessor) {
         val suggestions = textInputProcessor.getSuggestions(buffer)
 
         if (pauseOnMisspelledWord) {
-            return AutocorrectDecision.Pause
+            return AutocorrectDecision.Pause(suggestions)
         }
 
         if (autocorrectionEnabled && suggestions.isNotEmpty() && lastAutocorrection == null) {

app/src/main/java/com/urik/keyboard/service/BlacklistRepository.kt

@@ -0,0 +1,46 @@
+package com.urik.keyboard.service
+
+import android.content.Context
+import androidx.datastore.core.DataStore
+import androidx.datastore.preferences.core.Preferences
+import androidx.datastore.preferences.core.edit
+import androidx.datastore.preferences.core.stringSetPreferencesKey
+import androidx.datastore.preferences.preferencesDataStore
+import dagger.hilt.android.qualifiers.ApplicationContext
+import javax.inject.Inject
+import javax.inject.Singleton
+import kotlinx.coroutines.flow.first
+
+private val Context.blacklistDataStore by preferencesDataStore(name = "blacklist_words")
+
+/**
+ * Persists user-rejected suggestion words so removals survive IME process death.
+ * Global across languages.
+ */
+@Singleton
+class BlacklistRepository internal constructor(private val dataStore: DataStore<Preferences>) {
+    @Inject
+    constructor(
+        @ApplicationContext context: Context
+    ) : this(context.blacklistDataStore)
+
+    private object PreferenceKeys {
+        val BLACKLISTED_WORDS = stringSetPreferencesKey("blacklisted_words")
+    }
+
+    suspend fun getAll(): Set<String> = dataStore.data.first()[PreferenceKeys.BLACKLISTED_WORDS] ?: emptySet()
+
+    suspend fun add(word: String) {
+        dataStore.edit { preferences ->
+            val current = preferences[PreferenceKeys.BLACKLISTED_WORDS] ?: emptySet()
+            preferences[PreferenceKeys.BLACKLISTED_WORDS] = current + word
+        }
+    }
+
+    suspend fun remove(word: String) {
+        dataStore.edit { preferences ->
+            val current = preferences[PreferenceKeys.BLACKLISTED_WORDS] ?: emptySet()
+            preferences[PreferenceKeys.BLACKLISTED_WORDS] = current - word
+        }
+    }
+}

app/src/main/java/com/urik/keyboard/service/CustomKeyMappingService.kt

@@ -74,12 +74,13 @@ constructor(
             if (raw.isBlank()) return emptyList()
 
             return raw
-                .split(LONG_PRESS_DELIMITER)
+                .splitToSequence(LONG_PRESS_DELIMITER)
                 .map { it.trim() }
                 .filter { it.isNotBlank() }
                 .map { Normalizer.normalize(it, Normalizer.Form.NFC) }
                 .distinct()
                 .take(MAX_CUSTOM_SYMBOLS)
+                .toList()
         }
     }
 }

app/src/main/java/com/urik/keyboard/service/FatFingerExpander.kt

@@ -1,6 +1,7 @@
 package com.urik.keyboard.service
 
 import android.graphics.PointF
+import com.urik.keyboard.service.FatFingerExpander.Companion.ADJACENT_KEY_THRESHOLD_MULTIPLIER
 import javax.inject.Inject
 import javax.inject.Singleton
 import kotlin.math.sqrt

app/src/main/java/com/urik/keyboard/service/ImeStateCoordinator.kt

@@ -24,7 +24,6 @@ class ImeStateCoordinator(
         outputBridge.finishComposingText()
 
         spellCheckManager.clearCaches()
-        spellCheckManager.clearBlacklist()
         textInputProcessor.clearCaches()
         wordLearningEngine.clearCurrentLanguageCache()