BinaryDictionary.java revision 5bf1be71629607e7206e6203489cf742d2f8ed79
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.utils.AdditionalFeaturesSettingUtils; 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 int[] mInputCodePoints = new int[MAX_WORD_LENGTH]; 47 private final int[] mOutputCodePoints = new int[MAX_WORD_LENGTH * MAX_RESULTS]; 48 private final int[] mSpaceIndices = new int[MAX_RESULTS]; 49 private final int[] mOutputScores = new int[MAX_RESULTS]; 50 private final int[] mOutputTypes = new int[MAX_RESULTS]; 51 52 private final NativeSuggestOptions mNativeSuggestOptions = new NativeSuggestOptions(); 53 54 private final SparseArray<DicTraverseSession> mDicTraverseSessions = 55 CollectionUtils.newSparseArray(); 56 57 // TODO: There should be a way to remove used DicTraverseSession objects from 58 // {@code mDicTraverseSessions}. 59 private DicTraverseSession getTraverseSession(final int traverseSessionId) { 60 synchronized(mDicTraverseSessions) { 61 DicTraverseSession traverseSession = mDicTraverseSessions.get(traverseSessionId); 62 if (traverseSession == null) { 63 traverseSession = mDicTraverseSessions.get(traverseSessionId); 64 if (traverseSession == null) { 65 traverseSession = new DicTraverseSession(mLocale, mNativeDict); 66 mDicTraverseSessions.put(traverseSessionId, traverseSession); 67 } 68 } 69 return traverseSession; 70 } 71 } 72 73 /** 74 * Constructor for the binary dictionary. This is supposed to be called from the 75 * dictionary factory. 76 * @param filename the name of the file to read through native code. 77 * @param offset the offset of the dictionary data within the file. 78 * @param length the length of the binary data. 79 * @param useFullEditDistance whether to use the full edit distance in suggestions 80 * @param dictType the dictionary type, as a human-readable string 81 * @param isUpdatable whether to open the dictionary file in writable mode. 82 */ 83 public BinaryDictionary(final String filename, final long offset, final long length, 84 final boolean useFullEditDistance, final Locale locale, final String dictType, 85 final boolean isUpdatable) { 86 super(dictType); 87 mLocale = locale; 88 mNativeSuggestOptions.setUseFullEditDistance(useFullEditDistance); 89 loadDictionary(filename, offset, length, isUpdatable); 90 } 91 92 static { 93 JniUtils.loadNativeLibrary(); 94 } 95 96 private static native long openNative(String sourceDir, long dictOffset, long dictSize, 97 boolean isUpdatable); 98 private static native void closeNative(long dict); 99 private static native int getProbabilityNative(long dict, int[] word); 100 private static native boolean isValidBigramNative(long dict, int[] word0, int[] word1); 101 private static native int getSuggestionsNative(long dict, long proximityInfo, 102 long traverseSession, int[] xCoordinates, int[] yCoordinates, int[] times, 103 int[] pointerIds, int[] inputCodePoints, int inputSize, int commitPoint, 104 int[] suggestOptions, int[] prevWordCodePointArray, 105 int[] outputCodePoints, int[] outputScores, int[] outputIndices, int[] outputTypes); 106 private static native float calcNormalizedScoreNative(int[] before, int[] after, int score); 107 private static native int editDistanceNative(int[] before, int[] after); 108 private static native void addUnigramWordNative(long dict, int[] word, int probability); 109 private static native void addBigramWordsNative(long dict, int[] word0, int[] word1, 110 int probability); 111 private static native void removeBigramWordsNative(long dict, int[] word0, int[] word1); 112 113 // TODO: Move native dict into session 114 private final void loadDictionary(final String path, final long startOffset, 115 final long length, final boolean isUpdatable) { 116 mNativeDict = openNative(path, startOffset, length, isUpdatable); 117 } 118 119 @Override 120 public ArrayList<SuggestedWordInfo> getSuggestions(final WordComposer composer, 121 final String prevWord, final ProximityInfo proximityInfo, 122 final boolean blockOffensiveWords) { 123 return getSuggestionsWithSessionId(composer, prevWord, proximityInfo, blockOffensiveWords, 124 0 /* sessionId */); 125 } 126 127 @Override 128 public ArrayList<SuggestedWordInfo> getSuggestionsWithSessionId(final WordComposer composer, 129 final String prevWord, final ProximityInfo proximityInfo, 130 final boolean blockOffensiveWords, final int sessionId) { 131 if (!isValidDictionary()) return null; 132 133 Arrays.fill(mInputCodePoints, Constants.NOT_A_CODE); 134 // TODO: toLowerCase in the native code 135 final int[] prevWordCodePointArray = (null == prevWord) 136 ? null : StringUtils.toCodePointArray(prevWord); 137 final int composerSize = composer.size(); 138 139 final boolean isGesture = composer.isBatchMode(); 140 if (composerSize <= 1 || !isGesture) { 141 if (composerSize > MAX_WORD_LENGTH - 1) return null; 142 for (int i = 0; i < composerSize; i++) { 143 mInputCodePoints[i] = composer.getCodeAt(i); 144 } 145 } 146 147 final InputPointers ips = composer.getInputPointers(); 148 final int inputSize = isGesture ? ips.getPointerSize() : composerSize; 149 mNativeSuggestOptions.setIsGesture(isGesture); 150 mNativeSuggestOptions.setAdditionalFeaturesOptions( 151 AdditionalFeaturesSettingUtils.getAdditionalNativeSuggestOptions()); 152 // proximityInfo and/or prevWordForBigrams may not be null. 153 final int count = getSuggestionsNative(mNativeDict, proximityInfo.getNativeProximityInfo(), 154 getTraverseSession(sessionId).getSession(), ips.getXCoordinates(), 155 ips.getYCoordinates(), ips.getTimes(), ips.getPointerIds(), mInputCodePoints, 156 inputSize, 0 /* commitPoint */, mNativeSuggestOptions.getOptions(), 157 prevWordCodePointArray, mOutputCodePoints, mOutputScores, mSpaceIndices, 158 mOutputTypes); 159 final ArrayList<SuggestedWordInfo> suggestions = CollectionUtils.newArrayList(); 160 for (int j = 0; j < count; ++j) { 161 final int start = j * MAX_WORD_LENGTH; 162 int len = 0; 163 while (len < MAX_WORD_LENGTH && mOutputCodePoints[start + len] != 0) { 164 ++len; 165 } 166 if (len > 0) { 167 final int flags = mOutputTypes[j] & SuggestedWordInfo.KIND_MASK_FLAGS; 168 if (blockOffensiveWords 169 && 0 != (flags & SuggestedWordInfo.KIND_FLAG_POSSIBLY_OFFENSIVE) 170 && 0 == (flags & SuggestedWordInfo.KIND_FLAG_EXACT_MATCH)) { 171 // If we block potentially offensive words, and if the word is possibly 172 // offensive, then we don't output it unless it's also an exact match. 173 continue; 174 } 175 final int kind = mOutputTypes[j] & SuggestedWordInfo.KIND_MASK_KIND; 176 final int score = SuggestedWordInfo.KIND_WHITELIST == kind 177 ? SuggestedWordInfo.MAX_SCORE : mOutputScores[j]; 178 // TODO: check that all users of the `kind' parameter are ready to accept 179 // flags too and pass mOutputTypes[j] instead of kind 180 suggestions.add(new SuggestedWordInfo(new String(mOutputCodePoints, start, len), 181 score, kind, mDictType)); 182 } 183 } 184 return suggestions; 185 } 186 187 public boolean isValidDictionary() { 188 return mNativeDict != 0; 189 } 190 191 public static float calcNormalizedScore(final String before, final String after, 192 final int score) { 193 return calcNormalizedScoreNative(StringUtils.toCodePointArray(before), 194 StringUtils.toCodePointArray(after), score); 195 } 196 197 public static int editDistance(final String before, final String after) { 198 if (before == null || after == null) { 199 throw new IllegalArgumentException(); 200 } 201 return editDistanceNative(StringUtils.toCodePointArray(before), 202 StringUtils.toCodePointArray(after)); 203 } 204 205 @Override 206 public boolean isValidWord(final String word) { 207 return getFrequency(word) >= 0; 208 } 209 210 @Override 211 public int getFrequency(final String word) { 212 if (word == null) return -1; 213 int[] codePoints = StringUtils.toCodePointArray(word); 214 return getProbabilityNative(mNativeDict, codePoints); 215 } 216 217 // TODO: Add a batch process version (isValidBigramMultiple?) to avoid excessive numbers of jni 218 // calls when checking for changes in an entire dictionary. 219 public boolean isValidBigram(final String word1, final String word2) { 220 if (TextUtils.isEmpty(word1) || TextUtils.isEmpty(word2)) return false; 221 final int[] codePoints1 = StringUtils.toCodePointArray(word1); 222 final int[] codePoints2 = StringUtils.toCodePointArray(word2); 223 return isValidBigramNative(mNativeDict, codePoints1, codePoints2); 224 } 225 226 @Override 227 public void close() { 228 synchronized (mDicTraverseSessions) { 229 final int sessionsSize = mDicTraverseSessions.size(); 230 for (int index = 0; index < sessionsSize; ++index) { 231 final DicTraverseSession traverseSession = mDicTraverseSessions.valueAt(index); 232 if (traverseSession != null) { 233 traverseSession.close(); 234 } 235 } 236 } 237 closeInternal(); 238 } 239 240 private synchronized void closeInternal() { 241 if (mNativeDict != 0) { 242 closeNative(mNativeDict); 243 mNativeDict = 0; 244 } 245 } 246 247 @Override 248 protected void finalize() throws Throwable { 249 try { 250 closeInternal(); 251 } finally { 252 super.finalize(); 253 } 254 } 255} 256