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.cpp164
1 files changed, 102 insertions, 62 deletions
diff --git a/native/jni/com_android_inputmethod_latin_BinaryDictionary.cpp b/native/jni/com_android_inputmethod_latin_BinaryDictionary.cpp
index 22ad2d0ab..365217a60 100644
--- a/native/jni/com_android_inputmethod_latin_BinaryDictionary.cpp
+++ b/native/jni/com_android_inputmethod_latin_BinaryDictionary.cpp
@@ -32,6 +32,7 @@
#include "suggest/core/suggest_options.h"
#include "suggest/policyimpl/dictionary/structure/dictionary_structure_with_buffer_policy_factory.h"
#include "utils/char_utils.h"
+#include "utils/int_array_view.h"
#include "utils/jni_data_utils.h"
#include "utils/log_utils.h"
#include "utils/time_keeper.h"
@@ -179,9 +180,10 @@ static void latinime_BinaryDictionary_getSuggestions(JNIEnv *env, jclass clazz,
jintArray yCoordinatesArray, jintArray timesArray, jintArray pointerIdsArray,
jintArray inputCodePointsArray, jint inputSize, jintArray suggestOptions,
jobjectArray prevWordCodePointArrays, jbooleanArray isBeginningOfSentenceArray,
- jintArray outSuggestionCount, jintArray outCodePointsArray, jintArray outScoresArray,
- jintArray outSpaceIndicesArray, jintArray outTypesArray,
- jintArray outAutoCommitFirstWordConfidenceArray, jfloatArray inOutLanguageWeight) {
+ jint prevWordCount, jintArray outSuggestionCount, jintArray outCodePointsArray,
+ jintArray outScoresArray, jintArray outSpaceIndicesArray, jintArray outTypesArray,
+ jintArray outAutoCommitFirstWordConfidenceArray,
+ jfloatArray inOutWeightOfLangModelVsSpatialModel) {
Dictionary *dictionary = reinterpret_cast<Dictionary *>(dict);
// Assign 0 to outSuggestionCount here in case of returning earlier in this method.
JniDataUtils::putIntToArray(env, outSuggestionCount, 0 /* index */, 0);
@@ -236,42 +238,44 @@ static void latinime_BinaryDictionary_getSuggestions(JNIEnv *env, jclass clazz,
ASSERT(false);
return;
}
- float languageWeight;
- env->GetFloatArrayRegion(inOutLanguageWeight, 0, 1 /* len */, &languageWeight);
+ float weightOfLangModelVsSpatialModel;
+ env->GetFloatArrayRegion(inOutWeightOfLangModelVsSpatialModel, 0, 1 /* len */,
+ &weightOfLangModelVsSpatialModel);
SuggestionResults suggestionResults(MAX_RESULTS);
const PrevWordsInfo prevWordsInfo = JniDataUtils::constructPrevWordsInfo(env,
- prevWordCodePointArrays, isBeginningOfSentenceArray);
+ prevWordCodePointArrays, isBeginningOfSentenceArray, prevWordCount);
if (givenSuggestOptions.isGesture() || inputSize > 0) {
// TODO: Use SuggestionResults to return suggestions.
dictionary->getSuggestions(pInfo, traverseSession, xCoordinates, yCoordinates,
times, pointerIds, inputCodePoints, inputSize, &prevWordsInfo,
- &givenSuggestOptions, languageWeight, &suggestionResults);
+ &givenSuggestOptions, weightOfLangModelVsSpatialModel, &suggestionResults);
} else {
dictionary->getPredictions(&prevWordsInfo, &suggestionResults);
}
suggestionResults.outputSuggestions(env, outSuggestionCount, outCodePointsArray,
outScoresArray, outSpaceIndicesArray, outTypesArray,
- outAutoCommitFirstWordConfidenceArray, inOutLanguageWeight);
+ outAutoCommitFirstWordConfidenceArray, inOutWeightOfLangModelVsSpatialModel);
}
static jint latinime_BinaryDictionary_getProbability(JNIEnv *env, jclass clazz, jlong dict,
jintArray word) {
Dictionary *dictionary = reinterpret_cast<Dictionary *>(dict);
if (!dictionary) return NOT_A_PROBABILITY;
- const jsize wordLength = env->GetArrayLength(word);
- int codePoints[wordLength];
- env->GetIntArrayRegion(word, 0, wordLength, codePoints);
- return dictionary->getProbability(codePoints, wordLength);
+ const jsize codePointCount = env->GetArrayLength(word);
+ int codePoints[codePointCount];
+ env->GetIntArrayRegion(word, 0, codePointCount, codePoints);
+ return dictionary->getProbability(CodePointArrayView(codePoints, codePointCount));
}
static jint latinime_BinaryDictionary_getMaxProbabilityOfExactMatches(
JNIEnv *env, jclass clazz, jlong dict, jintArray word) {
Dictionary *dictionary = reinterpret_cast<Dictionary *>(dict);
if (!dictionary) return NOT_A_PROBABILITY;
- const jsize wordLength = env->GetArrayLength(word);
- int codePoints[wordLength];
- env->GetIntArrayRegion(word, 0, wordLength, codePoints);
- return dictionary->getMaxProbabilityOfExactMatches(codePoints, wordLength);
+ const jsize codePointCount = env->GetArrayLength(word);
+ int codePoints[codePointCount];
+ env->GetIntArrayRegion(word, 0, codePointCount, codePoints);
+ return dictionary->getMaxProbabilityOfExactMatches(
+ CodePointArrayView(codePoints, codePointCount));
}
static jint latinime_BinaryDictionary_getNgramProbability(JNIEnv *env, jclass clazz,
@@ -283,8 +287,10 @@ static jint latinime_BinaryDictionary_getNgramProbability(JNIEnv *env, jclass cl
int wordCodePoints[wordLength];
env->GetIntArrayRegion(word, 0, wordLength, wordCodePoints);
const PrevWordsInfo prevWordsInfo = JniDataUtils::constructPrevWordsInfo(env,
- prevWordCodePointArrays, isBeginningOfSentenceArray);
- return dictionary->getNgramProbability(&prevWordsInfo, wordCodePoints, wordLength);
+ prevWordCodePointArrays, isBeginningOfSentenceArray,
+ env->GetArrayLength(prevWordCodePointArrays));
+ return dictionary->getNgramProbability(&prevWordsInfo,
+ CodePointArrayView(wordCodePoints, wordLength));
}
// Method to iterate all words in the dictionary for makedict.
@@ -339,7 +345,8 @@ static void latinime_BinaryDictionary_getWordProperty(JNIEnv *env, jclass clazz,
return;
}
}
- const WordProperty wordProperty = dictionary->getWordProperty(wordCodePoints, codePointCount);
+ const WordProperty wordProperty = dictionary->getWordProperty(
+ CodePointArrayView(wordCodePoints, codePointCount));
wordProperty.outputProperties(env, outCodePoints, outFlags, outProbabilityInfo,
outBigramTargets, outBigramProbabilityInfo, outShortcutTargets,
outShortcutProbabilities);
@@ -357,15 +364,19 @@ static bool latinime_BinaryDictionary_addUnigramEntry(JNIEnv *env, jclass clazz,
int codePoints[codePointCount];
env->GetIntArrayRegion(word, 0, codePointCount, codePoints);
std::vector<UnigramProperty::ShortcutProperty> shortcuts;
- std::vector<int> shortcutTargetCodePoints;
- JniDataUtils::jintarrayToVector(env, shortcutTarget, &shortcutTargetCodePoints);
- if (!shortcutTargetCodePoints.empty()) {
- shortcuts.emplace_back(&shortcutTargetCodePoints, shortcutProbability);
+ {
+ std::vector<int> shortcutTargetCodePoints;
+ JniDataUtils::jintarrayToVector(env, shortcutTarget, &shortcutTargetCodePoints);
+ if (!shortcutTargetCodePoints.empty()) {
+ shortcuts.emplace_back(std::move(shortcutTargetCodePoints), shortcutProbability);
+ }
}
// Use 1 for count to indicate the word has inputted.
const UnigramProperty unigramProperty(isBeginningOfSentence, isNotAWord,
- isBlacklisted, probability, timestamp, 0 /* level */, 1 /* count */, &shortcuts);
- return dictionary->addUnigramEntry(codePoints, codePointCount, &unigramProperty);
+ isBlacklisted, probability, HistoricalInfo(timestamp, 0 /* level */, 1 /* count */),
+ std::move(shortcuts));
+ return dictionary->addUnigramEntry(CodePointArrayView(codePoints, codePointCount),
+ &unigramProperty);
}
static bool latinime_BinaryDictionary_removeUnigramEntry(JNIEnv *env, jclass clazz, jlong dict,
@@ -377,7 +388,7 @@ static bool latinime_BinaryDictionary_removeUnigramEntry(JNIEnv *env, jclass cla
jsize codePointCount = env->GetArrayLength(word);
int codePoints[codePointCount];
env->GetIntArrayRegion(word, 0, codePointCount, codePoints);
- return dictionary->removeUnigramEntry(codePoints, codePointCount);
+ return dictionary->removeUnigramEntry(CodePointArrayView(codePoints, codePointCount));
}
static bool latinime_BinaryDictionary_addNgramEntry(JNIEnv *env, jclass clazz, jlong dict,
@@ -388,16 +399,15 @@ static bool latinime_BinaryDictionary_addNgramEntry(JNIEnv *env, jclass clazz, j
return false;
}
const PrevWordsInfo prevWordsInfo = JniDataUtils::constructPrevWordsInfo(env,
- prevWordCodePointArrays, isBeginningOfSentenceArray);
+ prevWordCodePointArrays, isBeginningOfSentenceArray,
+ env->GetArrayLength(prevWordCodePointArrays));
jsize wordLength = env->GetArrayLength(word);
int wordCodePoints[wordLength];
env->GetIntArrayRegion(word, 0, wordLength, wordCodePoints);
- const std::vector<int> bigramTargetCodePoints(
- wordCodePoints, wordCodePoints + wordLength);
- // Use 1 for count to indicate the bigram has inputted.
- const BigramProperty bigramProperty(&bigramTargetCodePoints, probability,
- timestamp, 0 /* level */, 1 /* count */);
- return dictionary->addNgramEntry(&prevWordsInfo, &bigramProperty);
+ // Use 1 for count to indicate the ngram has inputted.
+ const NgramProperty ngramProperty(CodePointArrayView(wordCodePoints, wordLength).toVector(),
+ probability, HistoricalInfo(timestamp, 0 /* level */, 1 /* count */));
+ return dictionary->addNgramEntry(&prevWordsInfo, &ngramProperty);
}
static bool latinime_BinaryDictionary_removeNgramEntry(JNIEnv *env, jclass clazz, jlong dict,
@@ -408,11 +418,32 @@ static bool latinime_BinaryDictionary_removeNgramEntry(JNIEnv *env, jclass clazz
return false;
}
const PrevWordsInfo prevWordsInfo = JniDataUtils::constructPrevWordsInfo(env,
- prevWordCodePointArrays, isBeginningOfSentenceArray);
- jsize wordLength = env->GetArrayLength(word);
- int wordCodePoints[wordLength];
- env->GetIntArrayRegion(word, 0, wordLength, wordCodePoints);
- return dictionary->removeNgramEntry(&prevWordsInfo, wordCodePoints, wordLength);
+ prevWordCodePointArrays, isBeginningOfSentenceArray,
+ env->GetArrayLength(prevWordCodePointArrays));
+ jsize codePointCount = env->GetArrayLength(word);
+ int wordCodePoints[codePointCount];
+ env->GetIntArrayRegion(word, 0, codePointCount, wordCodePoints);
+ return dictionary->removeNgramEntry(&prevWordsInfo,
+ CodePointArrayView(wordCodePoints, codePointCount));
+}
+
+static bool latinime_BinaryDictionary_updateCounter(JNIEnv *env, jclass clazz, jlong dict,
+ jobjectArray prevWordCodePointArrays, jbooleanArray isBeginningOfSentenceArray,
+ jintArray word, jboolean isValidWord, jint count, jint timestamp) {
+ Dictionary *dictionary = reinterpret_cast<Dictionary *>(dict);
+ if (!dictionary) {
+ return false;
+ }
+ const PrevWordsInfo prevWordsInfo = JniDataUtils::constructPrevWordsInfo(env,
+ prevWordCodePointArrays, isBeginningOfSentenceArray,
+ env->GetArrayLength(prevWordCodePointArrays));
+ jsize codePointCount = env->GetArrayLength(word);
+ int wordCodePoints[codePointCount];
+ env->GetIntArrayRegion(word, 0, codePointCount, wordCodePoints);
+ const HistoricalInfo historicalInfo(timestamp, 0 /* level */, count);
+ return dictionary->updateCounter(&prevWordsInfo,
+ CodePointArrayView(wordCodePoints, codePointCount), isValidWord == JNI_TRUE,
+ historicalInfo);
}
// Returns how many language model params are processed.
@@ -472,28 +503,30 @@ static int latinime_BinaryDictionary_addMultipleDictionaryEntries(JNIEnv *env, j
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(&shortcutTargetCodePoints, shortcutProbability);
+ {
+ 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);
+ }
}
// Use 1 for count to indicate the word has inputted.
const UnigramProperty unigramProperty(false /* isBeginningOfSentence */, isNotAWord,
- isBlacklisted, unigramProbability, timestamp, 0 /* level */, 1 /* count */,
- &shortcuts);
- dictionary->addUnigramEntry(word1CodePoints, word1Length, &unigramProperty);
+ 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);
- const std::vector<int> bigramTargetCodePoints(
- word1CodePoints, word1CodePoints + word1Length);
// Use 1 for count to indicate the bigram has inputted.
- const BigramProperty bigramProperty(&bigramTargetCodePoints, bigramProbability,
- timestamp, 0 /* level */, 1 /* count */);
+ const NgramProperty ngramProperty(
+ CodePointArrayView(word1CodePoints, word1Length).toVector(),
+ bigramProbability, HistoricalInfo(timestamp, 0 /* level */, 1 /* count */));
const PrevWordsInfo prevWordsInfo(word0CodePoints, word0Length,
false /* isBeginningOfSentence */);
- dictionary->addNgramEntry(&prevWordsInfo, &bigramProperty);
+ dictionary->addNgramEntry(&prevWordsInfo, &ngramProperty);
}
if (dictionary->needsToRunGC(true /* mindsBlockByGC */)) {
return i + 1;
@@ -567,8 +600,8 @@ static bool latinime_BinaryDictionary_migrateNative(JNIEnv *env, jclass clazz, j
// Add unigrams.
do {
token = dictionary->getNextWordAndNextToken(token, wordCodePoints, &wordCodePointCount);
- const WordProperty wordProperty = dictionary->getWordProperty(wordCodePoints,
- wordCodePointCount);
+ const WordProperty wordProperty = dictionary->getWordProperty(
+ CodePointArrayView(wordCodePoints, wordCodePointCount));
if (wordCodePoints[0] == CODE_POINT_BEGINNING_OF_SENTENCE) {
// Skip beginning-of-sentence unigram.
continue;
@@ -581,18 +614,20 @@ static bool latinime_BinaryDictionary_migrateNative(JNIEnv *env, jclass clazz, j
return false;
}
}
- if (!dictionaryStructureWithBufferPolicy->addUnigramEntry(wordCodePoints,
- wordCodePointCount, wordProperty.getUnigramProperty())) {
+ if (!dictionaryStructureWithBufferPolicy->addUnigramEntry(
+ CodePointArrayView(wordCodePoints, wordCodePointCount),
+ wordProperty.getUnigramProperty())) {
LogUtils::logToJava(env, "Cannot add unigram to the new dict.");
return false;
}
} while (token != 0);
// Add bigrams.
+ // TODO: Support ngrams.
do {
token = dictionary->getNextWordAndNextToken(token, wordCodePoints, &wordCodePointCount);
- const WordProperty wordProperty = dictionary->getWordProperty(wordCodePoints,
- wordCodePointCount);
+ const WordProperty wordProperty = dictionary->getWordProperty(
+ CodePointArrayView(wordCodePoints, wordCodePointCount));
if (dictionaryStructureWithBufferPolicy->needsToRunGC(true /* mindsBlockByGC */)) {
dictionaryStructureWithBufferPolicy = runGCAndGetNewStructurePolicy(
std::move(dictionaryStructureWithBufferPolicy), dictFilePathChars);
@@ -603,10 +638,10 @@ static bool latinime_BinaryDictionary_migrateNative(JNIEnv *env, jclass clazz, j
}
const PrevWordsInfo prevWordsInfo(wordCodePoints, wordCodePointCount,
wordProperty.getUnigramProperty()->representsBeginningOfSentence());
- for (const BigramProperty &bigramProperty : *wordProperty.getBigramProperties()) {
+ for (const NgramProperty &ngramProperty : *wordProperty.getNgramProperties()) {
if (!dictionaryStructureWithBufferPolicy->addNgramEntry(&prevWordsInfo,
- &bigramProperty)) {
- LogUtils::logToJava(env, "Cannot add bigram to the new dict.");
+ &ngramProperty)) {
+ LogUtils::logToJava(env, "Cannot add ngram to the new dict.");
return false;
}
}
@@ -659,7 +694,7 @@ static const JNINativeMethod sMethods[] = {
},
{
const_cast<char *>("getSuggestionsNative"),
- const_cast<char *>("(JJJ[I[I[I[I[II[I[[I[Z[I[I[I[I[I[I[F)V"),
+ const_cast<char *>("(JJJ[I[I[I[I[II[I[[I[ZI[I[I[I[I[I[I[F)V"),
reinterpret_cast<void *>(latinime_BinaryDictionary_getSuggestions)
},
{
@@ -709,6 +744,11 @@ static const JNINativeMethod sMethods[] = {
reinterpret_cast<void *>(latinime_BinaryDictionary_removeNgramEntry)
},
{
+ const_cast<char *>("updateCounterNative"),
+ const_cast<char *>("(J[[I[Z[IZII)Z"),
+ reinterpret_cast<void *>(latinime_BinaryDictionary_updateCounter)
+ },
+ {
const_cast<char *>("addMultipleDictionaryEntriesNative"),
const_cast<char *>(
"(J[Lcom/android/inputmethod/latin/utils/LanguageModelParam;I)I"),
generated by cgit v1.2.3-83-g751a (git 2.47.1) at 2025-05-29 15:13:30 +0000