BinaryDictionary.java revision e8ef09567077211da034a77b457fd5f87e70f6f0
1/* 2 * Copyright (C) 2008 The Android Open Source Project 3 * 4 * Licensed under the Apache License, Version 2.0 (the "License"); 5 * you may not use this file except in compliance with the License. 6 * You may obtain a copy of the License at 7 * 8 * http://www.apache.org/licenses/LICENSE-2.0 9 * 10 * Unless required by applicable law or agreed to in writing, software 11 * distributed under the License is distributed on an "AS IS" BASIS, 12 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 13 * See the License for the specific language governing permissions and 14 * limitations under the License. 15 */ 16 17package com.android.inputmethod.latin; 18 19import android.text.TextUtils; 20import android.util.SparseArray; 21 22import com.android.inputmethod.keyboard.ProximityInfo; 23import com.android.inputmethod.latin.SuggestedWords.SuggestedWordInfo; 24import com.android.inputmethod.latin.settings.AdditionalFeaturesSettingUtils; 25import com.android.inputmethod.latin.settings.NativeSuggestOptions; 26import com.android.inputmethod.latin.utils.CollectionUtils; 27import com.android.inputmethod.latin.utils.JniUtils; 28import com.android.inputmethod.latin.utils.StringUtils; 29 30import java.util.ArrayList; 31import java.util.Arrays; 32import java.util.Locale; 33 34/** 35 * Implements a static, compacted, binary dictionary of standard words. 36 */ 37public final class BinaryDictionary extends Dictionary { 38 private static final String TAG = BinaryDictionary.class.getSimpleName(); 39 40 // Must be equal to MAX_WORD_LENGTH in native/jni/src/defines.h 41 private static final int MAX_WORD_LENGTH = Constants.DICTIONARY_MAX_WORD_LENGTH; 42 // Must be equal to MAX_RESULTS in native/jni/src/defines.h 43 private static final int MAX_RESULTS = 18; 44 45 private long mNativeDict; 46 private final Locale mLocale; 47 private final int[] mInputCodePoints = new int[MAX_WORD_LENGTH]; 48 private final int[] mOutputCodePoints = new int[MAX_WORD_LENGTH * MAX_RESULTS]; 49 private final int[] mSpaceIndices = new int[MAX_RESULTS]; 50 private final int[] mOutputScores = new int[MAX_RESULTS]; 51 private final int[] mOutputTypes = new int[MAX_RESULTS]; 52 53 private final NativeSuggestOptions mNativeSuggestOptions = new NativeSuggestOptions(); 54 55 private final SparseArray<DicTraverseSession> mDicTraverseSessions = 56 CollectionUtils.newSparseArray(); 57 58 // TODO: There should be a way to remove used DicTraverseSession objects from 59 // {@code mDicTraverseSessions}. 60 private DicTraverseSession getTraverseSession(final int traverseSessionId) { 61 synchronized(mDicTraverseSessions) { 62 DicTraverseSession traverseSession = mDicTraverseSessions.get(traverseSessionId); 63 if (traverseSession == null) { 64 traverseSession = mDicTraverseSessions.get(traverseSessionId); 65 if (traverseSession == null) { 66 traverseSession = new DicTraverseSession(mLocale, mNativeDict); 67 mDicTraverseSessions.put(traverseSessionId, traverseSession); 68 } 69 } 70 return traverseSession; 71 } 72 } 73 74 /** 75 * Constructor for the binary dictionary. This is supposed to be called from the 76 * dictionary factory. 77 * @param filename the name of the file to read through native code. 78 * @param offset the offset of the dictionary data within the file. 79 * @param length the length of the binary data. 80 * @param useFullEditDistance whether to use the full edit distance in suggestions 81 * @param dictType the dictionary type, as a human-readable string 82 * @param isUpdatable whether to open the dictionary file in writable mode. 83 */ 84 public BinaryDictionary(final String filename, final long offset, final long length, 85 final boolean useFullEditDistance, final Locale locale, final String dictType, 86 final boolean isUpdatable) { 87 super(dictType); 88 mLocale = locale; 89 mNativeSuggestOptions.setUseFullEditDistance(useFullEditDistance); 90 loadDictionary(filename, offset, length, isUpdatable); 91 } 92 93 static { 94 JniUtils.loadNativeLibrary(); 95 } 96 97 private static native long openNative(String sourceDir, long dictOffset, long dictSize, 98 boolean isUpdatable); 99 private static native void closeNative(long dict); 100 private static native int getProbabilityNative(long dict, int[] word); 101 private static native boolean isValidBigramNative(long dict, int[] word0, int[] word1); 102 private static native int getSuggestionsNative(long dict, long proximityInfo, 103 long traverseSession, int[] xCoordinates, int[] yCoordinates, int[] times, 104 int[] pointerIds, int[] inputCodePoints, int inputSize, int commitPoint, 105 int[] suggestOptions, int[] prevWordCodePointArray, 106 int[] outputCodePoints, int[] outputScores, int[] outputIndices, int[] outputTypes); 107 private static native float calcNormalizedScoreNative(int[] before, int[] after, int score); 108 private static native int editDistanceNative(int[] before, int[] after); 109 private static native void addUnigramWordNative(long dict, int[] word, int probability); 110 private static native void addBigramWordsNative(long dict, int[] word0, int[] word1, 111 int probability); 112 private static native void removeBigramWordsNative(long dict, int[] word0, int[] word1); 113 114 // TODO: Move native dict into session 115 private final void loadDictionary(final String path, final long startOffset, 116 final long length, final boolean isUpdatable) { 117 mNativeDict = openNative(path, startOffset, length, isUpdatable); 118 } 119 120 @Override 121 public ArrayList<SuggestedWordInfo> getSuggestions(final WordComposer composer, 122 final String prevWord, final ProximityInfo proximityInfo, 123 final boolean blockOffensiveWords) { 124 return getSuggestionsWithSessionId(composer, prevWord, proximityInfo, blockOffensiveWords, 125 0 /* sessionId */); 126 } 127 128 @Override 129 public ArrayList<SuggestedWordInfo> getSuggestionsWithSessionId(final WordComposer composer, 130 final String prevWord, final ProximityInfo proximityInfo, 131 final boolean blockOffensiveWords, final int sessionId) { 132 if (!isValidDictionary()) return null; 133 134 Arrays.fill(mInputCodePoints, Constants.NOT_A_CODE); 135 // TODO: toLowerCase in the native code 136 final int[] prevWordCodePointArray = (null == prevWord) 137 ? null : StringUtils.toCodePointArray(prevWord); 138 final int composerSize = composer.size(); 139 140 final boolean isGesture = composer.isBatchMode(); 141 if (composerSize <= 1 || !isGesture) { 142 if (composerSize > MAX_WORD_LENGTH - 1) return null; 143 for (int i = 0; i < composerSize; i++) { 144 mInputCodePoints[i] = composer.getCodeAt(i); 145 } 146 } 147 148 final InputPointers ips = composer.getInputPointers(); 149 final int inputSize = isGesture ? ips.getPointerSize() : composerSize; 150 mNativeSuggestOptions.setIsGesture(isGesture); 151 mNativeSuggestOptions.setAdditionalFeaturesOptions( 152 AdditionalFeaturesSettingUtils.getAdditionalNativeSuggestOptions()); 153 // proximityInfo and/or prevWordForBigrams may not be null. 154 final int count = getSuggestionsNative(mNativeDict, proximityInfo.getNativeProximityInfo(), 155 getTraverseSession(sessionId).getSession(), ips.getXCoordinates(), 156 ips.getYCoordinates(), ips.getTimes(), ips.getPointerIds(), mInputCodePoints, 157 inputSize, 0 /* commitPoint */, mNativeSuggestOptions.getOptions(), 158 prevWordCodePointArray, mOutputCodePoints, mOutputScores, mSpaceIndices, 159 mOutputTypes); 160 final ArrayList<SuggestedWordInfo> suggestions = CollectionUtils.newArrayList(); 161 for (int j = 0; j < count; ++j) { 162 final int start = j * MAX_WORD_LENGTH; 163 int len = 0; 164 while (len < MAX_WORD_LENGTH && mOutputCodePoints[start + len] != 0) { 165 ++len; 166 } 167 if (len > 0) { 168 final int flags = mOutputTypes[j] & SuggestedWordInfo.KIND_MASK_FLAGS; 169 if (blockOffensiveWords 170 && 0 != (flags & SuggestedWordInfo.KIND_FLAG_POSSIBLY_OFFENSIVE) 171 && 0 == (flags & SuggestedWordInfo.KIND_FLAG_EXACT_MATCH)) { 172 // If we block potentially offensive words, and if the word is possibly 173 // offensive, then we don't output it unless it's also an exact match. 174 continue; 175 } 176 final int kind = mOutputTypes[j] & SuggestedWordInfo.KIND_MASK_KIND; 177 final int score = SuggestedWordInfo.KIND_WHITELIST == kind 178 ? SuggestedWordInfo.MAX_SCORE : mOutputScores[j]; 179 // TODO: check that all users of the `kind' parameter are ready to accept 180 // flags too and pass mOutputTypes[j] instead of kind 181 suggestions.add(new SuggestedWordInfo(new String(mOutputCodePoints, start, len), 182 score, kind, this /* sourceDict */, 183 mSpaceIndices[0] /* indexOfTouchPointOfSecondWord */)); 184 } 185 } 186 return suggestions; 187 } 188 189 public boolean isValidDictionary() { 190 return mNativeDict != 0; 191 } 192 193 public static float calcNormalizedScore(final String before, final String after, 194 final int score) { 195 return calcNormalizedScoreNative(StringUtils.toCodePointArray(before), 196 StringUtils.toCodePointArray(after), score); 197 } 198 199 public static int editDistance(final String before, final String after) { 200 if (before == null || after == null) { 201 throw new IllegalArgumentException(); 202 } 203 return editDistanceNative(StringUtils.toCodePointArray(before), 204 StringUtils.toCodePointArray(after)); 205 } 206 207 @Override 208 public boolean isValidWord(final String word) { 209 return getFrequency(word) >= 0; 210 } 211 212 @Override 213 public int getFrequency(final String word) { 214 if (word == null) return -1; 215 int[] codePoints = StringUtils.toCodePointArray(word); 216 return getProbabilityNative(mNativeDict, codePoints); 217 } 218 219 // TODO: Add a batch process version (isValidBigramMultiple?) to avoid excessive numbers of jni 220 // calls when checking for changes in an entire dictionary. 221 public boolean isValidBigram(final String word0, final String word1) { 222 if (TextUtils.isEmpty(word0) || TextUtils.isEmpty(word1)) return false; 223 final int[] codePoints0 = StringUtils.toCodePointArray(word0); 224 final int[] codePoints1 = StringUtils.toCodePointArray(word1); 225 return isValidBigramNative(mNativeDict, codePoints0, codePoints1); 226 } 227 228 // Add a unigram entry to binary dictionary in native code. 229 public void addUnigramWord(final String word, final int probability) { 230 if (TextUtils.isEmpty(word)) { 231 return; 232 } 233 final int[] codePoints = StringUtils.toCodePointArray(word); 234 addUnigramWordNative(mNativeDict, codePoints, probability); 235 } 236 237 // Add a bigram entry to binary dictionary in native code. 238 public void addBigramWords(final String word0, final String word1, final int probability) { 239 if (TextUtils.isEmpty(word0) || TextUtils.isEmpty(word1)) { 240 return; 241 } 242 final int[] codePoints0 = StringUtils.toCodePointArray(word0); 243 final int[] codePoints1 = StringUtils.toCodePointArray(word1); 244 addBigramWordsNative(mNativeDict, codePoints0, codePoints1, probability); 245 } 246 247 // Remove a bigram entry form binary dictionary in native code. 248 public void removeBigramWords(final String word0, final String word1) { 249 if (TextUtils.isEmpty(word0) || TextUtils.isEmpty(word1)) { 250 return; 251 } 252 final int[] codePoints0 = StringUtils.toCodePointArray(word0); 253 final int[] codePoints1 = StringUtils.toCodePointArray(word1); 254 removeBigramWordsNative(mNativeDict, codePoints0, codePoints1); 255 } 256 257 @Override 258 public void close() { 259 synchronized (mDicTraverseSessions) { 260 final int sessionsSize = mDicTraverseSessions.size(); 261 for (int index = 0; index < sessionsSize; ++index) { 262 final DicTraverseSession traverseSession = mDicTraverseSessions.valueAt(index); 263 if (traverseSession != null) { 264 traverseSession.close(); 265 } 266 } 267 } 268 closeInternal(); 269 } 270 271 private synchronized void closeInternal() { 272 if (mNativeDict != 0) { 273 closeNative(mNativeDict); 274 mNativeDict = 0; 275 } 276 } 277 278 @Override 279 protected void finalize() throws Throwable { 280 try { 281 closeInternal(); 282 } finally { 283 super.finalize(); 284 } 285 } 286} 287