BinaryDictionary.java revision f6870cc82ddf394e94155322fcc7e4e2256bea66
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 private static final int MAX_WORD_LENGTH = Constants.Dictionary.MAX_WORD_LENGTH; 45 private static final int MAX_RESULTS = 18; /* Must be identical to MAX_RESULTS in defines.h */ 46 47 private long mNativeDict; 48 private final Locale mLocale; 49 private final int[] mInputCodePoints = new int[MAX_WORD_LENGTH]; 50 private final int[] mOutputCodePoints = new int[MAX_WORD_LENGTH * MAX_RESULTS]; 51 private final int[] mSpaceIndices = new int[MAX_RESULTS]; 52 private final int[] mOutputScores = new int[MAX_RESULTS]; 53 private final int[] mOutputTypes = new int[MAX_RESULTS]; 54 55 private final boolean mUseFullEditDistance; 56 57 private final SparseArray<DicTraverseSession> mDicTraverseSessions = 58 CollectionUtils.newSparseArray(); 59 60 // TODO: There should be a way to remove used DicTraverseSession objects from 61 // {@code mDicTraverseSessions}. 62 private DicTraverseSession getTraverseSession(final int traverseSessionId) { 63 synchronized(mDicTraverseSessions) { 64 DicTraverseSession traverseSession = mDicTraverseSessions.get(traverseSessionId); 65 if (traverseSession == null) { 66 traverseSession = mDicTraverseSessions.get(traverseSessionId); 67 if (traverseSession == null) { 68 traverseSession = new DicTraverseSession(mLocale, mNativeDict); 69 mDicTraverseSessions.put(traverseSessionId, traverseSession); 70 } 71 } 72 return traverseSession; 73 } 74 } 75 76 /** 77 * Constructor for the binary dictionary. This is supposed to be called from the 78 * dictionary factory. 79 * @param filename the name of the file to read through native code. 80 * @param offset the offset of the dictionary data within the file. 81 * @param length the length of the binary data. 82 * @param useFullEditDistance whether to use the full edit distance in suggestions 83 * @param dictType the dictionary type, as a human-readable string 84 */ 85 public BinaryDictionary(final String filename, final long offset, final long length, 86 final boolean useFullEditDistance, final Locale locale, final String dictType) { 87 super(dictType); 88 mLocale = locale; 89 mUseFullEditDistance = useFullEditDistance; 90 loadDictionary(filename, offset, length); 91 } 92 93 static { 94 JniUtils.loadNativeLibrary(); 95 } 96 97 private native long openNative(String sourceDir, long dictOffset, long dictSize, 98 int maxWordLength); 99 private native void closeNative(long dict); 100 private native int getFrequencyNative(long dict, int[] word); 101 private native boolean isValidBigramNative(long dict, int[] word1, int[] word2); 102 private native int getSuggestionsNative(long dict, long proximityInfo, long traverseSession, 103 int[] xCoordinates, int[] yCoordinates, int[] times, int[] pointerIds, 104 int[] inputCodePoints, int codesSize, int commitPoint, boolean isGesture, 105 int[] prevWordCodePointArray, boolean useFullEditDistance, int[] outputCodePoints, 106 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); 109 110 // TODO: Move native dict into session 111 private final void loadDictionary(final String path, final long startOffset, 112 final long length) { 113 mNativeDict = openNative(path, startOffset, length, MAX_WORD_LENGTH); 114 } 115 116 @Override 117 public ArrayList<SuggestedWordInfo> getSuggestions(final WordComposer composer, 118 final String prevWord, final ProximityInfo proximityInfo) { 119 return getSuggestionsWithSessionId(composer, prevWord, proximityInfo, 0); 120 } 121 122 @Override 123 public ArrayList<SuggestedWordInfo> getSuggestionsWithSessionId(final WordComposer composer, 124 final String prevWord, final ProximityInfo proximityInfo, int sessionId) { 125 if (!isValidDictionary()) return null; 126 127 Arrays.fill(mInputCodePoints, Constants.NOT_A_CODE); 128 // TODO: toLowerCase in the native code 129 final int[] prevWordCodePointArray = (null == prevWord) 130 ? null : StringUtils.toCodePointArray(prevWord); 131 final int composerSize = composer.size(); 132 133 final boolean isGesture = composer.isBatchMode(); 134 if (composerSize <= 1 || !isGesture) { 135 if (composerSize > MAX_WORD_LENGTH - 1) return null; 136 for (int i = 0; i < composerSize; i++) { 137 mInputCodePoints[i] = composer.getCodeAt(i); 138 } 139 } 140 141 final InputPointers ips = composer.getInputPointers(); 142 final int codesSize = isGesture ? ips.getPointerSize() : composerSize; 143 // proximityInfo and/or prevWordForBigrams may not be null. 144 final int count = getSuggestionsNative(mNativeDict, proximityInfo.getNativeProximityInfo(), 145 getTraverseSession(sessionId).getSession(), ips.getXCoordinates(), 146 ips.getYCoordinates(), ips.getTimes(), ips.getPointerIds(), mInputCodePoints, 147 codesSize, 0 /* commitPoint */, isGesture, prevWordCodePointArray, 148 mUseFullEditDistance, mOutputCodePoints, mOutputScores, mSpaceIndices, 149 mOutputTypes); 150 final ArrayList<SuggestedWordInfo> suggestions = CollectionUtils.newArrayList(); 151 for (int j = 0; j < count; ++j) { 152 if (composerSize > 0 && mOutputScores[j] < 1) break; 153 final int start = j * MAX_WORD_LENGTH; 154 int len = 0; 155 while (len < MAX_WORD_LENGTH && mOutputCodePoints[start + len] != 0) { 156 ++len; 157 } 158 if (len > 0) { 159 final int score = SuggestedWordInfo.KIND_WHITELIST == mOutputTypes[j] 160 ? SuggestedWordInfo.MAX_SCORE : mOutputScores[j]; 161 suggestions.add(new SuggestedWordInfo(new String(mOutputCodePoints, start, len), 162 score, mOutputTypes[j], mDictType)); 163 } 164 } 165 return suggestions; 166 } 167 168 public boolean isValidDictionary() { 169 return mNativeDict != 0; 170 } 171 172 public static float calcNormalizedScore(final String before, final String after, 173 final int score) { 174 return calcNormalizedScoreNative(StringUtils.toCodePointArray(before), 175 StringUtils.toCodePointArray(after), score); 176 } 177 178 public static int editDistance(final String before, final String after) { 179 if (before == null || after == null) { 180 throw new IllegalArgumentException(); 181 } 182 return editDistanceNative(StringUtils.toCodePointArray(before), 183 StringUtils.toCodePointArray(after)); 184 } 185 186 @Override 187 public boolean isValidWord(final String word) { 188 return getFrequency(word) >= 0; 189 } 190 191 @Override 192 public int getFrequency(final String word) { 193 if (word == null) return -1; 194 int[] codePoints = StringUtils.toCodePointArray(word); 195 return getFrequencyNative(mNativeDict, codePoints); 196 } 197 198 // TODO: Add a batch process version (isValidBigramMultiple?) to avoid excessive numbers of jni 199 // calls when checking for changes in an entire dictionary. 200 public boolean isValidBigram(final String word1, final String word2) { 201 if (TextUtils.isEmpty(word1) || TextUtils.isEmpty(word2)) return false; 202 final int[] codePoints1 = StringUtils.toCodePointArray(word1); 203 final int[] codePoints2 = StringUtils.toCodePointArray(word2); 204 return isValidBigramNative(mNativeDict, codePoints1, codePoints2); 205 } 206 207 @Override 208 public void close() { 209 synchronized (mDicTraverseSessions) { 210 final int sessionsSize = mDicTraverseSessions.size(); 211 for (int index = 0; index < sessionsSize; ++index) { 212 final DicTraverseSession traverseSession = mDicTraverseSessions.valueAt(index); 213 if (traverseSession != null) { 214 traverseSession.close(); 215 } 216 } 217 } 218 closeInternal(); 219 } 220 221 private synchronized void closeInternal() { 222 if (mNativeDict != 0) { 223 closeNative(mNativeDict); 224 mNativeDict = 0; 225 } 226 } 227 228 @Override 229 protected void finalize() throws Throwable { 230 try { 231 closeInternal(); 232 } finally { 233 super.finalize(); 234 } 235 } 236} 237