BinaryDictionary.java revision fe9ec6bc7d58464f3117c3fe9372c2994861998b
1/* 2 * Copyright (C) 2008 The Android Open Source Project 3 * 4 * Licensed under the Apache License, Version 2.0 (the "License"); you may not 5 * use this file except in compliance with the License. You may obtain a copy of 6 * 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, WITHOUT 12 * WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the 13 * License for the specific language governing permissions and limitations under 14 * the License. 15 */ 16 17package com.android.inputmethod.latin; 18 19import android.content.Context; 20import android.text.TextUtils; 21import android.util.SparseArray; 22 23import com.android.inputmethod.keyboard.ProximityInfo; 24import com.android.inputmethod.latin.SuggestedWords.SuggestedWordInfo; 25 26import java.util.ArrayList; 27import java.util.Arrays; 28import java.util.Locale; 29 30/** 31 * Implements a static, compacted, binary dictionary of standard words. 32 */ 33public final class BinaryDictionary extends Dictionary { 34 private static final String TAG = BinaryDictionary.class.getSimpleName(); 35 public static final String DICTIONARY_PACK_AUTHORITY = 36 "com.android.inputmethod.latin.dictionarypack"; 37 38 /** 39 * There is a difference between what java and native code can handle. 40 * This value should only be used in BinaryDictionary.java 41 * It is necessary to keep it at this value because some languages e.g. German have 42 * really long words. 43 */ 44 public static final int MAX_WORD_LENGTH = Constants.Dictionary.MAX_WORD_LENGTH; 45 public static final int MAX_WORDS = 18; 46 public static final int MAX_SPACES = 16; 47 48 private static final int MAX_PREDICTIONS = 60; 49 private static final int MAX_RESULTS = Math.max(MAX_PREDICTIONS, MAX_WORDS); 50 51 private long mNativeDict; 52 private final Locale mLocale; 53 private final int[] mInputCodePoints = new int[MAX_WORD_LENGTH]; 54 private final int[] mOutputCodePoints = new int[MAX_WORD_LENGTH * MAX_RESULTS]; 55 private final int[] mSpaceIndices = new int[MAX_SPACES]; 56 private final int[] mOutputScores = new int[MAX_RESULTS]; 57 private final int[] mOutputTypes = new int[MAX_RESULTS]; 58 59 private final boolean mUseFullEditDistance; 60 61 private final SparseArray<DicTraverseSession> mDicTraverseSessions = 62 CollectionUtils.newSparseArray(); 63 64 // TODO: There should be a way to remove used DicTraverseSession objects from 65 // {@code mDicTraverseSessions}. 66 private DicTraverseSession getTraverseSession(final int traverseSessionId) { 67 synchronized(mDicTraverseSessions) { 68 DicTraverseSession traverseSession = mDicTraverseSessions.get(traverseSessionId); 69 if (traverseSession == null) { 70 traverseSession = mDicTraverseSessions.get(traverseSessionId); 71 if (traverseSession == null) { 72 traverseSession = new DicTraverseSession(mLocale, mNativeDict); 73 mDicTraverseSessions.put(traverseSessionId, traverseSession); 74 } 75 } 76 return traverseSession; 77 } 78 } 79 80 /** 81 * Constructor for the binary dictionary. This is supposed to be called from the 82 * dictionary factory. 83 * @param context the context to access the environment from. 84 * @param filename the name of the file to read through native code. 85 * @param offset the offset of the dictionary data within the file. 86 * @param length the length of the binary data. 87 * @param useFullEditDistance whether to use the full edit distance in suggestions 88 * @param dictType the dictionary type, as a human-readable string 89 */ 90 public BinaryDictionary(final Context context, final String filename, final long offset, 91 final long length, final boolean useFullEditDistance, final Locale locale, 92 final String dictType) { 93 super(dictType); 94 mLocale = locale; 95 mUseFullEditDistance = useFullEditDistance; 96 loadDictionary(filename, offset, length); 97 } 98 99 static { 100 JniUtils.loadNativeLibrary(); 101 } 102 103 private native long openNative(String sourceDir, long dictOffset, long dictSize, 104 int maxWordLength, int maxWords, int maxPredictions); 105 private native void closeNative(long dict); 106 private native int getFrequencyNative(long dict, int[] word); 107 private native boolean isValidBigramNative(long dict, int[] word1, int[] word2); 108 private native int getSuggestionsNative(long dict, long proximityInfo, long traverseSession, 109 int[] xCoordinates, int[] yCoordinates, int[] times, int[] pointerIds, 110 int[] inputCodePoints, int codesSize, int commitPoint, boolean isGesture, 111 int[] prevWordCodePointArray, boolean useFullEditDistance, int[] outputCodePoints, 112 int[] outputScores, int[] outputIndices, int[] outputTypes); 113 private static native float calcNormalizedScoreNative(int[] before, int[] after, int score); 114 private static native int editDistanceNative(int[] before, int[] after); 115 116 // TODO: Move native dict into session 117 private final void loadDictionary(final String path, final long startOffset, 118 final long length) { 119 mNativeDict = openNative(path, startOffset, length, MAX_WORD_LENGTH, MAX_WORDS, 120 MAX_PREDICTIONS); 121 } 122 123 @Override 124 public ArrayList<SuggestedWordInfo> getSuggestions(final WordComposer composer, 125 final String prevWord, final ProximityInfo proximityInfo) { 126 return getSuggestionsWithSessionId(composer, prevWord, proximityInfo, 0); 127 } 128 129 @Override 130 public ArrayList<SuggestedWordInfo> getSuggestionsWithSessionId(final WordComposer composer, 131 final String prevWord, final ProximityInfo proximityInfo, 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 codesSize = isGesture ? ips.getPointerSize() : composerSize; 150 // proximityInfo and/or prevWordForBigrams may not be null. 151 final int tmpCount = getSuggestionsNative(mNativeDict, 152 proximityInfo.getNativeProximityInfo(), getTraverseSession(sessionId).getSession(), 153 ips.getXCoordinates(), ips.getYCoordinates(), ips.getTimes(), ips.getPointerIds(), 154 mInputCodePoints, codesSize, 0 /* commitPoint */, isGesture, prevWordCodePointArray, 155 mUseFullEditDistance, mOutputCodePoints, mOutputScores, mSpaceIndices, 156 mOutputTypes); 157 final int count = Math.min(tmpCount, MAX_PREDICTIONS); 158 159 final ArrayList<SuggestedWordInfo> suggestions = CollectionUtils.newArrayList(); 160 for (int j = 0; j < count; ++j) { 161 if (composerSize > 0 && mOutputScores[j] < 1) break; 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 score = SuggestedWordInfo.KIND_WHITELIST == mOutputTypes[j] 169 ? SuggestedWordInfo.MAX_SCORE : mOutputScores[j]; 170 suggestions.add(new SuggestedWordInfo(new String(mOutputCodePoints, start, len), 171 score, mOutputTypes[j], mDictType)); 172 } 173 } 174 return suggestions; 175 } 176 177 public boolean isValidDictionary() { 178 return mNativeDict != 0; 179 } 180 181 public static float calcNormalizedScore(final String before, final String after, 182 final int score) { 183 return calcNormalizedScoreNative(StringUtils.toCodePointArray(before), 184 StringUtils.toCodePointArray(after), score); 185 } 186 187 public static int editDistance(final String before, final String after) { 188 if (before == null || after == null) { 189 throw new IllegalArgumentException(); 190 } 191 return editDistanceNative(StringUtils.toCodePointArray(before), 192 StringUtils.toCodePointArray(after)); 193 } 194 195 @Override 196 public boolean isValidWord(final String word) { 197 return getFrequency(word) >= 0; 198 } 199 200 @Override 201 public int getFrequency(final String word) { 202 if (word == null) return -1; 203 int[] codePoints = StringUtils.toCodePointArray(word); 204 return getFrequencyNative(mNativeDict, codePoints); 205 } 206 207 // TODO: Add a batch process version (isValidBigramMultiple?) to avoid excessive numbers of jni 208 // calls when checking for changes in an entire dictionary. 209 public boolean isValidBigram(final String word1, final String word2) { 210 if (TextUtils.isEmpty(word1) || TextUtils.isEmpty(word2)) return false; 211 final int[] codePoints1 = StringUtils.toCodePointArray(word1); 212 final int[] codePoints2 = StringUtils.toCodePointArray(word2); 213 return isValidBigramNative(mNativeDict, codePoints1, codePoints2); 214 } 215 216 @Override 217 public void close() { 218 synchronized (mDicTraverseSessions) { 219 final int sessionsSize = mDicTraverseSessions.size(); 220 for (int index = 0; index < sessionsSize; ++index) { 221 final DicTraverseSession traverseSession = mDicTraverseSessions.valueAt(index); 222 if (traverseSession != null) { 223 traverseSession.close(); 224 } 225 } 226 } 227 closeInternal(); 228 } 229 230 private synchronized void closeInternal() { 231 if (mNativeDict != 0) { 232 closeNative(mNativeDict); 233 mNativeDict = 0; 234 } 235 } 236 237 @Override 238 protected void finalize() throws Throwable { 239 try { 240 closeInternal(); 241 } finally { 242 super.finalize(); 243 } 244 } 245} 246