BinaryDictionary.java revision 4c2767857a02c9cf18a9579aa0391fd09b3fe411
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.NativeSuggestOptions; 25import com.android.inputmethod.latin.utils.CollectionUtils; 26import com.android.inputmethod.latin.utils.JniUtils; 27import com.android.inputmethod.latin.utils.StringUtils; 28 29import java.util.ArrayList; 30import java.util.Arrays; 31import java.util.Locale; 32 33/** 34 * Implements a static, compacted, binary dictionary of standard words. 35 */ 36public final class BinaryDictionary extends Dictionary { 37 private static final String TAG = BinaryDictionary.class.getSimpleName(); 38 39 // Must be equal to MAX_WORD_LENGTH in native/jni/src/defines.h 40 private static final int MAX_WORD_LENGTH = Constants.DICTIONARY_MAX_WORD_LENGTH; 41 // Must be equal to MAX_RESULTS in native/jni/src/defines.h 42 private static final int MAX_RESULTS = 18; 43 44 private long mNativeDict; 45 private final Locale mLocale; 46 private final long mDictSize; 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, mDictSize); 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 mDictSize = length; 90 mNativeSuggestOptions.setUseFullEditDistance(useFullEditDistance); 91 loadDictionary(filename, offset, length, isUpdatable); 92 } 93 94 static { 95 JniUtils.loadNativeLibrary(); 96 } 97 98 private static native long openNative(String sourceDir, long dictOffset, long dictSize, 99 boolean isUpdatable); 100 private static native void closeNative(long dict); 101 private static native int getProbabilityNative(long dict, int[] word); 102 private static native boolean isValidBigramNative(long dict, int[] word0, int[] word1); 103 private static native int getSuggestionsNative(long dict, long proximityInfo, 104 long traverseSession, int[] xCoordinates, int[] yCoordinates, int[] times, 105 int[] pointerIds, int[] inputCodePoints, int inputSize, int commitPoint, 106 int[] suggestOptions, int[] prevWordCodePointArray, 107 int[] outputCodePoints, int[] outputScores, int[] outputIndices, int[] outputTypes); 108 private static native float calcNormalizedScoreNative(int[] before, int[] after, int score); 109 private static native int editDistanceNative(int[] before, int[] after); 110 private static native void addUnigramWordNative(long dict, int[] word, int probability); 111 private static native void addBigramWordsNative(long dict, int[] word0, int[] word1, 112 int probability); 113 private static native void removeBigramWordsNative(long dict, int[] word0, int[] word1); 114 115 // TODO: Move native dict into session 116 private final void loadDictionary(final String path, final long startOffset, 117 final long length, final boolean isUpdatable) { 118 mNativeDict = openNative(path, startOffset, length, isUpdatable); 119 } 120 121 @Override 122 public ArrayList<SuggestedWordInfo> getSuggestions(final WordComposer composer, 123 final String prevWord, final ProximityInfo proximityInfo, 124 final boolean blockOffensiveWords, final int[] additionalFeaturesOptions) { 125 return getSuggestionsWithSessionId(composer, prevWord, proximityInfo, blockOffensiveWords, 126 additionalFeaturesOptions, 0 /* sessionId */); 127 } 128 129 @Override 130 public ArrayList<SuggestedWordInfo> getSuggestionsWithSessionId(final WordComposer composer, 131 final String prevWord, final ProximityInfo proximityInfo, 132 final boolean blockOffensiveWords, final int[] additionalFeaturesOptions, 133 final int sessionId) { 134 if (!isValidDictionary()) return null; 135 136 Arrays.fill(mInputCodePoints, Constants.NOT_A_CODE); 137 // TODO: toLowerCase in the native code 138 final int[] prevWordCodePointArray = (null == prevWord) 139 ? null : StringUtils.toCodePointArray(prevWord); 140 final int composerSize = composer.size(); 141 142 final boolean isGesture = composer.isBatchMode(); 143 if (composerSize <= 1 || !isGesture) { 144 if (composerSize > MAX_WORD_LENGTH - 1) return null; 145 for (int i = 0; i < composerSize; i++) { 146 mInputCodePoints[i] = composer.getCodeAt(i); 147 } 148 } 149 150 final InputPointers ips = composer.getInputPointers(); 151 final int inputSize = isGesture ? ips.getPointerSize() : composerSize; 152 mNativeSuggestOptions.setIsGesture(isGesture); 153 mNativeSuggestOptions.setAdditionalFeaturesOptions(additionalFeaturesOptions); 154 // proximityInfo and/or prevWordForBigrams may not be null. 155 final int count = getSuggestionsNative(mNativeDict, proximityInfo.getNativeProximityInfo(), 156 getTraverseSession(sessionId).getSession(), ips.getXCoordinates(), 157 ips.getYCoordinates(), ips.getTimes(), ips.getPointerIds(), mInputCodePoints, 158 inputSize, 0 /* commitPoint */, mNativeSuggestOptions.getOptions(), 159 prevWordCodePointArray, mOutputCodePoints, mOutputScores, mSpaceIndices, 160 mOutputTypes); 161 final ArrayList<SuggestedWordInfo> suggestions = CollectionUtils.newArrayList(); 162 for (int j = 0; j < count; ++j) { 163 final int start = j * MAX_WORD_LENGTH; 164 int len = 0; 165 while (len < MAX_WORD_LENGTH && mOutputCodePoints[start + len] != 0) { 166 ++len; 167 } 168 if (len > 0) { 169 final int flags = mOutputTypes[j] & SuggestedWordInfo.KIND_MASK_FLAGS; 170 if (blockOffensiveWords 171 && 0 != (flags & SuggestedWordInfo.KIND_FLAG_POSSIBLY_OFFENSIVE) 172 && 0 == (flags & SuggestedWordInfo.KIND_FLAG_EXACT_MATCH)) { 173 // If we block potentially offensive words, and if the word is possibly 174 // offensive, then we don't output it unless it's also an exact match. 175 continue; 176 } 177 final int kind = mOutputTypes[j] & SuggestedWordInfo.KIND_MASK_KIND; 178 final int score = SuggestedWordInfo.KIND_WHITELIST == kind 179 ? SuggestedWordInfo.MAX_SCORE : mOutputScores[j]; 180 // TODO: check that all users of the `kind' parameter are ready to accept 181 // flags too and pass mOutputTypes[j] instead of kind 182 suggestions.add(new SuggestedWordInfo(new String(mOutputCodePoints, start, len), 183 score, kind, this /* sourceDict */, 184 mSpaceIndices[0] /* indexOfTouchPointOfSecondWord */)); 185 } 186 } 187 return suggestions; 188 } 189 190 public boolean isValidDictionary() { 191 return mNativeDict != 0; 192 } 193 194 public static float calcNormalizedScore(final String before, final String after, 195 final int score) { 196 return calcNormalizedScoreNative(StringUtils.toCodePointArray(before), 197 StringUtils.toCodePointArray(after), score); 198 } 199 200 public static int editDistance(final String before, final String after) { 201 if (before == null || after == null) { 202 throw new IllegalArgumentException(); 203 } 204 return editDistanceNative(StringUtils.toCodePointArray(before), 205 StringUtils.toCodePointArray(after)); 206 } 207 208 @Override 209 public boolean isValidWord(final String word) { 210 return getFrequency(word) >= 0; 211 } 212 213 @Override 214 public int getFrequency(final String word) { 215 if (word == null) return -1; 216 int[] codePoints = StringUtils.toCodePointArray(word); 217 return getProbabilityNative(mNativeDict, codePoints); 218 } 219 220 // TODO: Add a batch process version (isValidBigramMultiple?) to avoid excessive numbers of jni 221 // calls when checking for changes in an entire dictionary. 222 public boolean isValidBigram(final String word0, final String word1) { 223 if (TextUtils.isEmpty(word0) || TextUtils.isEmpty(word1)) return false; 224 final int[] codePoints0 = StringUtils.toCodePointArray(word0); 225 final int[] codePoints1 = StringUtils.toCodePointArray(word1); 226 return isValidBigramNative(mNativeDict, codePoints0, codePoints1); 227 } 228 229 // Add a unigram entry to binary dictionary in native code. 230 public void addUnigramWord(final String word, final int probability) { 231 if (TextUtils.isEmpty(word)) { 232 return; 233 } 234 final int[] codePoints = StringUtils.toCodePointArray(word); 235 addUnigramWordNative(mNativeDict, codePoints, probability); 236 } 237 238 // Add a bigram entry to binary dictionary in native code. 239 public void addBigramWords(final String word0, final String word1, final int probability) { 240 if (TextUtils.isEmpty(word0) || TextUtils.isEmpty(word1)) { 241 return; 242 } 243 final int[] codePoints0 = StringUtils.toCodePointArray(word0); 244 final int[] codePoints1 = StringUtils.toCodePointArray(word1); 245 addBigramWordsNative(mNativeDict, codePoints0, codePoints1, probability); 246 } 247 248 // Remove a bigram entry form binary dictionary in native code. 249 public void removeBigramWords(final String word0, final String word1) { 250 if (TextUtils.isEmpty(word0) || TextUtils.isEmpty(word1)) { 251 return; 252 } 253 final int[] codePoints0 = StringUtils.toCodePointArray(word0); 254 final int[] codePoints1 = StringUtils.toCodePointArray(word1); 255 removeBigramWordsNative(mNativeDict, codePoints0, codePoints1); 256 } 257 258 @Override 259 public void close() { 260 synchronized (mDicTraverseSessions) { 261 final int sessionsSize = mDicTraverseSessions.size(); 262 for (int index = 0; index < sessionsSize; ++index) { 263 final DicTraverseSession traverseSession = mDicTraverseSessions.valueAt(index); 264 if (traverseSession != null) { 265 traverseSession.close(); 266 } 267 } 268 } 269 closeInternal(); 270 } 271 272 private synchronized void closeInternal() { 273 if (mNativeDict != 0) { 274 closeNative(mNativeDict); 275 mNativeDict = 0; 276 } 277 } 278 279 @Override 280 protected void finalize() throws Throwable { 281 try { 282 closeInternal(); 283 } finally { 284 super.finalize(); 285 } 286 } 287} 288