aboutsummaryrefslogtreecommitdiffstats
path: root/native/jni/com_android_inputmethod_latin_BinaryDictionary.cpp
diff options
context:
space:
mode:
Diffstat (limited to 'native/jni/com_android_inputmethod_latin_BinaryDictionary.cpp')
-rw-r--r--native/jni/com_android_inputmethod_latin_BinaryDictionary.cpp153
1 files changed, 57 insertions, 96 deletions
diff --git a/native/jni/com_android_inputmethod_latin_BinaryDictionary.cpp b/native/jni/com_android_inputmethod_latin_BinaryDictionary.cpp
index f8dadb488..118f600bb 100644
--- a/native/jni/com_android_inputmethod_latin_BinaryDictionary.cpp
+++ b/native/jni/com_android_inputmethod_latin_BinaryDictionary.cpp
@@ -327,8 +327,8 @@ static jint latinime_BinaryDictionary_getNextWord(JNIEnv *env, jclass clazz,
static void latinime_BinaryDictionary_getWordProperty(JNIEnv *env, jclass clazz,
jlong dict, jintArray word, jboolean isBeginningOfSentence, jintArray outCodePoints,
- jbooleanArray outFlags, jintArray outProbabilityInfo, jobject /* outNgramPrevWordsArray */,
- jobject /* outNgramPrevWordIsBeginningOfSentenceArray */, jobject outNgramTargets,
+ jbooleanArray outFlags, jintArray outProbabilityInfo, jobject outNgramPrevWordsArray,
+ jobject outNgramPrevWordIsBeginningOfSentenceArray, jobject outNgramTargets,
jobject outNgramProbabilityInfo, jobject outShortcutTargets,
jobject outShortcutProbabilities) {
Dictionary *dictionary = reinterpret_cast<Dictionary *>(dict);
@@ -352,13 +352,14 @@ static void latinime_BinaryDictionary_getWordProperty(JNIEnv *env, jclass clazz,
const WordProperty wordProperty = dictionary->getWordProperty(
CodePointArrayView(wordCodePoints, codePointCount));
wordProperty.outputProperties(env, outCodePoints, outFlags, outProbabilityInfo,
+ outNgramPrevWordsArray, outNgramPrevWordIsBeginningOfSentenceArray,
outNgramTargets, outNgramProbabilityInfo, outShortcutTargets,
outShortcutProbabilities);
}
static bool latinime_BinaryDictionary_addUnigramEntry(JNIEnv *env, jclass clazz, jlong dict,
jintArray word, jint probability, jintArray shortcutTarget, jint shortcutProbability,
- jboolean isBeginningOfSentence, jboolean isNotAWord, jboolean isBlacklisted,
+ jboolean isBeginningOfSentence, jboolean isNotAWord, jboolean isPossiblyOffensive,
jint timestamp) {
Dictionary *dictionary = reinterpret_cast<Dictionary *>(dict);
if (!dictionary) {
@@ -377,8 +378,8 @@ static bool latinime_BinaryDictionary_addUnigramEntry(JNIEnv *env, jclass clazz,
}
// Use 1 for count to indicate the word has inputted.
const UnigramProperty unigramProperty(isBeginningOfSentence, isNotAWord,
- isBlacklisted, probability, HistoricalInfo(timestamp, 0 /* level */, 1 /* count */),
- std::move(shortcuts));
+ isPossiblyOffensive, probability, HistoricalInfo(timestamp, 0 /* level */,
+ 1 /* count */), std::move(shortcuts));
return dictionary->addUnigramEntry(CodePointArrayView(codePoints, codePointCount),
&unigramProperty);
}
@@ -409,9 +410,10 @@ static bool latinime_BinaryDictionary_addNgramEntry(JNIEnv *env, jclass clazz, j
int wordCodePoints[wordLength];
env->GetIntArrayRegion(word, 0, wordLength, wordCodePoints);
// Use 1 for count to indicate the ngram has inputted.
- const NgramProperty ngramProperty(CodePointArrayView(wordCodePoints, wordLength).toVector(),
+ const NgramProperty ngramProperty(ngramContext,
+ CodePointArrayView(wordCodePoints, wordLength).toVector(),
probability, HistoricalInfo(timestamp, 0 /* level */, 1 /* count */));
- return dictionary->addNgramEntry(&ngramContext, &ngramProperty);
+ return dictionary->addNgramEntry(&ngramProperty);
}
static bool latinime_BinaryDictionary_removeNgramEntry(JNIEnv *env, jclass clazz, jlong dict,
@@ -451,97 +453,60 @@ static bool latinime_BinaryDictionary_updateEntriesForWordWithNgramContext(JNIEn
historicalInfo);
}
-// Returns how many language model params are processed.
-static int latinime_BinaryDictionary_addMultipleDictionaryEntries(JNIEnv *env, jclass clazz,
- jlong dict, jobjectArray languageModelParams, jint startIndex) {
+// Returns how many input events are processed.
+static int latinime_BinaryDictionary_updateEntriesForInputEvents(JNIEnv *env, jclass clazz,
+ jlong dict, jobjectArray inputEvents, jint startIndex) {
Dictionary *dictionary = reinterpret_cast<Dictionary *>(dict);
if (!dictionary) {
return 0;
}
- jsize languageModelParamCount = env->GetArrayLength(languageModelParams);
- if (languageModelParamCount == 0 || startIndex >= languageModelParamCount) {
+ jsize inputEventCount = env->GetArrayLength(inputEvents);
+ if (inputEventCount == 0 || startIndex >= inputEventCount) {
return 0;
}
- jobject languageModelParam = env->GetObjectArrayElement(languageModelParams, 0);
- jclass languageModelParamClass = env->GetObjectClass(languageModelParam);
- env->DeleteLocalRef(languageModelParam);
-
- jfieldID word0FieldId = env->GetFieldID(languageModelParamClass, "mWord0", "[I");
- jfieldID word1FieldId = env->GetFieldID(languageModelParamClass, "mWord1", "[I");
- jfieldID unigramProbabilityFieldId =
- env->GetFieldID(languageModelParamClass, "mUnigramProbability", "I");
- jfieldID bigramProbabilityFieldId =
- env->GetFieldID(languageModelParamClass, "mBigramProbability", "I");
- jfieldID timestampFieldId =
- env->GetFieldID(languageModelParamClass, "mTimestamp", "I");
- jfieldID shortcutTargetFieldId =
- env->GetFieldID(languageModelParamClass, "mShortcutTarget", "[I");
- jfieldID shortcutProbabilityFieldId =
- env->GetFieldID(languageModelParamClass, "mShortcutProbability", "I");
- jfieldID isNotAWordFieldId =
- env->GetFieldID(languageModelParamClass, "mIsNotAWord", "Z");
- jfieldID isBlacklistedFieldId =
- env->GetFieldID(languageModelParamClass, "mIsBlacklisted", "Z");
- env->DeleteLocalRef(languageModelParamClass);
-
- for (int i = startIndex; i < languageModelParamCount; ++i) {
- jobject languageModelParam = env->GetObjectArrayElement(languageModelParams, i);
- // languageModelParam is a set of params for word1; thus, word1 cannot be null. On the
- // other hand, word0 can be null and then it means the set of params doesn't contain bigram
- // information.
- jintArray word0 = static_cast<jintArray>(
- env->GetObjectField(languageModelParam, word0FieldId));
- jsize word0Length = word0 ? env->GetArrayLength(word0) : 0;
- int word0CodePoints[word0Length];
- if (word0) {
- env->GetIntArrayRegion(word0, 0, word0Length, word0CodePoints);
- }
- jintArray word1 = static_cast<jintArray>(
- env->GetObjectField(languageModelParam, word1FieldId));
- jsize word1Length = env->GetArrayLength(word1);
- int word1CodePoints[word1Length];
- env->GetIntArrayRegion(word1, 0, word1Length, word1CodePoints);
- jint unigramProbability = env->GetIntField(languageModelParam, unigramProbabilityFieldId);
- jint timestamp = env->GetIntField(languageModelParam, timestampFieldId);
- jboolean isNotAWord = env->GetBooleanField(languageModelParam, isNotAWordFieldId);
- jboolean isBlacklisted = env->GetBooleanField(languageModelParam, isBlacklistedFieldId);
- jintArray shortcutTarget = static_cast<jintArray>(
- env->GetObjectField(languageModelParam, shortcutTargetFieldId));
- std::vector<UnigramProperty::ShortcutProperty> shortcuts;
- {
- std::vector<int> shortcutTargetCodePoints;
- JniDataUtils::jintarrayToVector(env, shortcutTarget, &shortcutTargetCodePoints);
- if (!shortcutTargetCodePoints.empty()) {
- jint shortcutProbability =
- env->GetIntField(languageModelParam, shortcutProbabilityFieldId);
- shortcuts.emplace_back(std::move(shortcutTargetCodePoints), shortcutProbability);
- }
- }
+ jobject inputEvent = env->GetObjectArrayElement(inputEvents, 0);
+ jclass wordInputEventClass = env->GetObjectClass(inputEvent);
+ env->DeleteLocalRef(inputEvent);
+
+ jfieldID targetWordFieldId = env->GetFieldID(wordInputEventClass, "mTargetWord", "[I");
+ jfieldID prevWordCountFieldId = env->GetFieldID(wordInputEventClass, "mPrevWordsCount", "I");
+ jfieldID prevWordArrayFieldId = env->GetFieldID(wordInputEventClass, "mPrevWordArray", "[[I");
+ jfieldID isPrevWordBoSArrayFieldId =
+ env->GetFieldID(wordInputEventClass, "mIsPrevWordBeginningOfSentenceArray", "[Z");
+ jfieldID isValidFieldId = env->GetFieldID(wordInputEventClass, "mIsValid", "Z");
+ jfieldID timestampFieldId = env->GetFieldID(wordInputEventClass, "mTimestamp", "I");
+ env->DeleteLocalRef(wordInputEventClass);
+
+ for (int i = startIndex; i < inputEventCount; ++i) {
+ jobject inputEvent = env->GetObjectArrayElement(inputEvents, i);
+ jintArray targetWord = static_cast<jintArray>(
+ env->GetObjectField(inputEvent, targetWordFieldId));
+ jsize wordLength = env->GetArrayLength(targetWord);
+ int wordCodePoints[wordLength];
+ env->GetIntArrayRegion(targetWord, 0, wordLength, wordCodePoints);
+ env->DeleteLocalRef(targetWord);
+
+ jint prevWordCount = env->GetIntField(inputEvent, prevWordCountFieldId);
+ jobjectArray prevWordArray =
+ static_cast<jobjectArray>(env->GetObjectField(inputEvent, prevWordArrayFieldId));
+ jbooleanArray isPrevWordBeginningOfSentenceArray = static_cast<jbooleanArray>(
+ env->GetObjectField(inputEvent, isPrevWordBoSArrayFieldId));
+ jboolean isValid = env->GetBooleanField(inputEvent, isValidFieldId);
+ jint timestamp = env->GetIntField(inputEvent, timestampFieldId);
+ const NgramContext ngramContext = JniDataUtils::constructNgramContext(env,
+ prevWordArray, isPrevWordBeginningOfSentenceArray, prevWordCount);
// Use 1 for count to indicate the word has inputted.
- const UnigramProperty unigramProperty(false /* isBeginningOfSentence */, isNotAWord,
- isBlacklisted, unigramProbability,
- HistoricalInfo(timestamp, 0 /* level */, 1 /* count */), std::move(shortcuts));
- dictionary->addUnigramEntry(CodePointArrayView(word1CodePoints, word1Length),
- &unigramProperty);
- if (word0) {
- jint bigramProbability = env->GetIntField(languageModelParam, bigramProbabilityFieldId);
- // Use 1 for count to indicate the bigram has inputted.
- const NgramProperty ngramProperty(
- CodePointArrayView(word1CodePoints, word1Length).toVector(),
- bigramProbability, HistoricalInfo(timestamp, 0 /* level */, 1 /* count */));
- const NgramContext ngramContext(word0CodePoints, word0Length,
- false /* isBeginningOfSentence */);
- dictionary->addNgramEntry(&ngramContext, &ngramProperty);
- }
+ dictionary->updateEntriesForWordWithNgramContext(&ngramContext,
+ CodePointArrayView(wordCodePoints, wordLength), isValid,
+ HistoricalInfo(timestamp, 0 /* level */, 1 /* count */));
if (dictionary->needsToRunGC(true /* mindsBlockByGC */)) {
return i + 1;
}
- env->DeleteLocalRef(word0);
- env->DeleteLocalRef(word1);
- env->DeleteLocalRef(shortcutTarget);
- env->DeleteLocalRef(languageModelParam);
+ env->DeleteLocalRef(prevWordArray);
+ env->DeleteLocalRef(isPrevWordBeginningOfSentenceArray);
+ env->DeleteLocalRef(inputEvent);
}
- return languageModelParamCount;
+ return inputEventCount;
}
static jstring latinime_BinaryDictionary_getProperty(JNIEnv *env, jclass clazz, jlong dict,
@@ -627,8 +592,7 @@ static bool latinime_BinaryDictionary_migrateNative(JNIEnv *env, jclass clazz, j
}
} while (token != 0);
- // Add bigrams.
- // TODO: Support ngrams.
+ // Add ngrams.
do {
token = dictionary->getNextWordAndNextToken(token, wordCodePoints, &wordCodePointCount);
const WordProperty wordProperty = dictionary->getWordProperty(
@@ -641,11 +605,8 @@ static bool latinime_BinaryDictionary_migrateNative(JNIEnv *env, jclass clazz, j
return false;
}
}
- const NgramContext ngramContext(wordCodePoints, wordCodePointCount,
- wordProperty.getUnigramProperty()->representsBeginningOfSentence());
for (const NgramProperty &ngramProperty : *wordProperty.getNgramProperties()) {
- if (!dictionaryStructureWithBufferPolicy->addNgramEntry(&ngramContext,
- &ngramProperty)) {
+ if (!dictionaryStructureWithBufferPolicy->addNgramEntry(&ngramProperty)) {
LogUtils::logToJava(env, "Cannot add ngram to the new dict.");
return false;
}
@@ -755,10 +716,10 @@ static const JNINativeMethod sMethods[] = {
reinterpret_cast<void *>(latinime_BinaryDictionary_updateEntriesForWordWithNgramContext)
},
{
- const_cast<char *>("addMultipleDictionaryEntriesNative"),
+ const_cast<char *>("updateEntriesForInputEventsNative"),
const_cast<char *>(
- "(J[Lcom/android/inputmethod/latin/utils/LanguageModelParam;I)I"),
- reinterpret_cast<void *>(latinime_BinaryDictionary_addMultipleDictionaryEntries)
+ "(J[Lcom/android/inputmethod/latin/utils/WordInputEventForPersonalization;I)I"),
+ reinterpret_cast<void *>(latinime_BinaryDictionary_updateEntriesForInputEvents)
},
{
const_cast<char *>("getPropertyNative"),
generated by cgit v1.2.3-83-g751a (git 2.47.1) at 2025-06-05 05:35:59 +0000