BinaryDictionary.java revision e28eba5074664d5716b8e58b8d0a235746b261eb
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 */ 82 public BinaryDictionary(final String filename, final long offset, final long length, 83 final boolean useFullEditDistance, final Locale locale, final String dictType) { 84 super(dictType); 85 mLocale = locale; 86 mNativeSuggestOptions.setUseFullEditDistance(useFullEditDistance); 87 loadDictionary(filename, offset, length); 88 } 89 90 static { 91 JniUtils.loadNativeLibrary(); 92 } 93 94 private static native long openNative(String sourceDir, long dictOffset, long dictSize); 95 private static native void closeNative(long dict); 96 private static native int getProbabilityNative(long dict, int[] word); 97 private static native boolean isValidBigramNative(long dict, int[] word1, int[] word2); 98 private static native int getSuggestionsNative(long dict, long proximityInfo, 99 long traverseSession, int[] xCoordinates, int[] yCoordinates, int[] times, 100 int[] pointerIds, int[] inputCodePoints, int inputSize, int commitPoint, 101 int[] suggestOptions, int[] prevWordCodePointArray, 102 int[] outputCodePoints, int[] outputScores, int[] outputIndices, int[] outputTypes); 103 private static native float calcNormalizedScoreNative(int[] before, int[] after, int score); 104 private static native int editDistanceNative(int[] before, int[] after); 105 106 // TODO: Move native dict into session 107 private final void loadDictionary(final String path, final long startOffset, 108 final long length) { 109 mNativeDict = openNative(path, startOffset, length); 110 } 111 112 @Override 113 public ArrayList<SuggestedWordInfo> getSuggestions(final WordComposer composer, 114 final String prevWord, final ProximityInfo proximityInfo, 115 final boolean blockOffensiveWords) { 116 return getSuggestionsWithSessionId(composer, prevWord, proximityInfo, blockOffensiveWords, 117 0 /* sessionId */); 118 } 119 120 @Override 121 public ArrayList<SuggestedWordInfo> getSuggestionsWithSessionId(final WordComposer composer, 122 final String prevWord, final ProximityInfo proximityInfo, 123 final boolean blockOffensiveWords, final int sessionId) { 124 if (!isValidDictionary()) return null; 125 126 Arrays.fill(mInputCodePoints, Constants.NOT_A_CODE); 127 // TODO: toLowerCase in the native code 128 final int[] prevWordCodePointArray = (null == prevWord) 129 ? null : StringUtils.toCodePointArray(prevWord); 130 final int composerSize = composer.size(); 131 132 final boolean isGesture = composer.isBatchMode(); 133 if (composerSize <= 1 || !isGesture) { 134 if (composerSize > MAX_WORD_LENGTH - 1) return null; 135 for (int i = 0; i < composerSize; i++) { 136 mInputCodePoints[i] = composer.getCodeAt(i); 137 } 138 } 139 140 final InputPointers ips = composer.getInputPointers(); 141 final int inputSize = isGesture ? ips.getPointerSize() : composerSize; 142 mNativeSuggestOptions.setIsGesture(isGesture); 143 mNativeSuggestOptions.setAdditionalFeaturesOptions( 144 AdditionalFeaturesSettingUtils.getAdditionalNativeSuggestOptions()); 145 // proximityInfo and/or prevWordForBigrams may not be null. 146 final int count = getSuggestionsNative(mNativeDict, proximityInfo.getNativeProximityInfo(), 147 getTraverseSession(sessionId).getSession(), ips.getXCoordinates(), 148 ips.getYCoordinates(), ips.getTimes(), ips.getPointerIds(), mInputCodePoints, 149 inputSize, 0 /* commitPoint */, mNativeSuggestOptions.getOptions(), 150 prevWordCodePointArray, mOutputCodePoints, mOutputScores, mSpaceIndices, 151 mOutputTypes); 152 final ArrayList<SuggestedWordInfo> suggestions = CollectionUtils.newArrayList(); 153 for (int j = 0; j < count; ++j) { 154 final int start = j * MAX_WORD_LENGTH; 155 int len = 0; 156 while (len < MAX_WORD_LENGTH && mOutputCodePoints[start + len] != 0) { 157 ++len; 158 } 159 if (len > 0) { 160 final int flags = mOutputTypes[j] & SuggestedWordInfo.KIND_MASK_FLAGS; 161 if (blockOffensiveWords 162 && 0 != (flags & SuggestedWordInfo.KIND_FLAG_POSSIBLY_OFFENSIVE) 163 && 0 == (flags & SuggestedWordInfo.KIND_FLAG_EXACT_MATCH)) { 164 // If we block potentially offensive words, and if the word is possibly 165 // offensive, then we don't output it unless it's also an exact match. 166 continue; 167 } 168 final int kind = mOutputTypes[j] & SuggestedWordInfo.KIND_MASK_KIND; 169 final int score = SuggestedWordInfo.KIND_WHITELIST == kind 170 ? SuggestedWordInfo.MAX_SCORE : mOutputScores[j]; 171 // TODO: check that all users of the `kind' parameter are ready to accept 172 // flags too and pass mOutputTypes[j] instead of kind 173 suggestions.add(new SuggestedWordInfo(new String(mOutputCodePoints, start, len), 174 score, kind, mDictType)); 175 } 176 } 177 return suggestions; 178 } 179 180 public boolean isValidDictionary() { 181 return mNativeDict != 0; 182 } 183 184 public static float calcNormalizedScore(final String before, final String after, 185 final int score) { 186 return calcNormalizedScoreNative(StringUtils.toCodePointArray(before), 187 StringUtils.toCodePointArray(after), score); 188 } 189 190 public static int editDistance(final String before, final String after) { 191 if (before == null || after == null) { 192 throw new IllegalArgumentException(); 193 } 194 return editDistanceNative(StringUtils.toCodePointArray(before), 195 StringUtils.toCodePointArray(after)); 196 } 197 198 @Override 199 public boolean isValidWord(final String word) { 200 return getFrequency(word) >= 0; 201 } 202 203 @Override 204 public int getFrequency(final String word) { 205 if (word == null) return -1; 206 int[] codePoints = StringUtils.toCodePointArray(word); 207 return getProbabilityNative(mNativeDict, codePoints); 208 } 209 210 // TODO: Add a batch process version (isValidBigramMultiple?) to avoid excessive numbers of jni 211 // calls when checking for changes in an entire dictionary. 212 public boolean isValidBigram(final String word1, final String word2) { 213 if (TextUtils.isEmpty(word1) || TextUtils.isEmpty(word2)) return false; 214 final int[] codePoints1 = StringUtils.toCodePointArray(word1); 215 final int[] codePoints2 = StringUtils.toCodePointArray(word2); 216 return isValidBigramNative(mNativeDict, codePoints1, codePoints2); 217 } 218 219 @Override 220 public void close() { 221 synchronized (mDicTraverseSessions) { 222 final int sessionsSize = mDicTraverseSessions.size(); 223 for (int index = 0; index < sessionsSize; ++index) { 224 final DicTraverseSession traverseSession = mDicTraverseSessions.valueAt(index); 225 if (traverseSession != null) { 226 traverseSession.close(); 227 } 228 } 229 } 230 closeInternal(); 231 } 232 233 private synchronized void closeInternal() { 234 if (mNativeDict != 0) { 235 closeNative(mNativeDict); 236 mNativeDict = 0; 237 } 238 } 239 240 @Override 241 protected void finalize() throws Throwable { 242 try { 243 closeInternal(); 244 } finally { 245 super.finalize(); 246 } 247 } 248} 249