BinaryDictionary.java revision ecfbf4625c8afd9cde7b79e0c7846b87e20f79e9
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 class BinaryDictionary extends Dictionary { 34 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 = 48; 45 public static final int MAX_WORDS = 18; 46 public static final int MAX_SPACES = 16; 47 48 private static final String TAG = "BinaryDictionary"; 49 private static final int MAX_PREDICTIONS = 60; 50 private static final int MAX_RESULTS = Math.max(MAX_PREDICTIONS, MAX_WORDS); 51 52 private static final int TYPED_LETTER_MULTIPLIER = 2; 53 54 private long mNativeDict; 55 private final Locale mLocale; 56 private final int[] mInputCodePoints = new int[MAX_WORD_LENGTH]; 57 // TODO: The below should be int[] mOutputCodePoints 58 private final char[] mOutputChars = new char[MAX_WORD_LENGTH * MAX_RESULTS]; 59 private final int[] mSpaceIndices = new int[MAX_SPACES]; 60 private final int[] mOutputScores = new int[MAX_RESULTS]; 61 private final int[] mOutputTypes = new int[MAX_RESULTS]; 62 63 private final boolean mUseFullEditDistance; 64 65 private final SparseArray<DicTraverseSession> mDicTraverseSessions = 66 CollectionUtils.newSparseArray(); 67 68 // TODO: There should be a way to remove used DicTraverseSession objects from 69 // {@code mDicTraverseSessions}. 70 private DicTraverseSession getTraverseSession(int traverseSessionId) { 71 synchronized(mDicTraverseSessions) { 72 DicTraverseSession traverseSession = mDicTraverseSessions.get(traverseSessionId); 73 if (traverseSession == null) { 74 traverseSession = mDicTraverseSessions.get(traverseSessionId); 75 if (traverseSession == null) { 76 traverseSession = new DicTraverseSession(mLocale, mNativeDict); 77 mDicTraverseSessions.put(traverseSessionId, traverseSession); 78 } 79 } 80 return traverseSession; 81 } 82 } 83 84 /** 85 * Constructor for the binary dictionary. This is supposed to be called from the 86 * dictionary factory. 87 * All implementations should pass null into flagArray, except for testing purposes. 88 * @param context the context to access the environment from. 89 * @param filename the name of the file to read through native code. 90 * @param offset the offset of the dictionary data within the file. 91 * @param length the length of the binary data. 92 * @param useFullEditDistance whether to use the full edit distance in suggestions 93 * @param dictType the dictionary type, as a human-readable string 94 */ 95 public BinaryDictionary(final Context context, 96 final String filename, final long offset, final long length, 97 final boolean useFullEditDistance, final Locale locale, final String dictType) { 98 super(dictType); 99 mLocale = locale; 100 mUseFullEditDistance = useFullEditDistance; 101 loadDictionary(filename, offset, length); 102 } 103 104 static { 105 JniUtils.loadNativeLibrary(); 106 } 107 108 private native long openNative(String sourceDir, long dictOffset, long dictSize, 109 int typedLetterMultiplier, int fullWordMultiplier, int maxWordLength, int maxWords, 110 int maxPredictions); 111 private native void closeNative(long dict); 112 private native int getFrequencyNative(long dict, int[] word); 113 private native boolean isValidBigramNative(long dict, int[] word1, int[] word2); 114 private native int getSuggestionsNative(long dict, long proximityInfo, long traverseSession, 115 int[] xCoordinates, int[] yCoordinates, int[] times, int[] pointerIds, 116 int[] inputCodePoints, int codesSize, int commitPoint, boolean isGesture, 117 int[] prevWordCodePointArray, boolean useFullEditDistance, char[] outputChars, 118 int[] outputScores, int[] outputIndices, int[] outputTypes); 119 private static native float calcNormalizedScoreNative(char[] before, char[] after, int score); 120 private static native int editDistanceNative(char[] before, char[] after); 121 122 // TODO: Move native dict into session 123 private final void loadDictionary(String path, long startOffset, long length) { 124 mNativeDict = openNative(path, startOffset, length, TYPED_LETTER_MULTIPLIER, 125 FULL_WORD_SCORE_MULTIPLIER, MAX_WORD_LENGTH, MAX_WORDS, MAX_PREDICTIONS); 126 } 127 128 @Override 129 public ArrayList<SuggestedWordInfo> getSuggestions(final WordComposer composer, 130 final CharSequence prevWord, final ProximityInfo proximityInfo) { 131 return getSuggestionsWithSessionId(composer, prevWord, proximityInfo, 0); 132 } 133 134 @Override 135 public ArrayList<SuggestedWordInfo> getSuggestionsWithSessionId(final WordComposer composer, 136 final CharSequence prevWord, final ProximityInfo proximityInfo, int sessionId) { 137 if (!isValidDictionary()) return null; 138 139 Arrays.fill(mInputCodePoints, Constants.NOT_A_CODE); 140 // TODO: toLowerCase in the native code 141 final int[] prevWordCodePointArray = (null == prevWord) 142 ? null : StringUtils.toCodePointArray(prevWord.toString()); 143 final int composerSize = composer.size(); 144 145 final boolean isGesture = composer.isBatchMode(); 146 if (composerSize <= 1 || !isGesture) { 147 if (composerSize > MAX_WORD_LENGTH - 1) return null; 148 for (int i = 0; i < composerSize; i++) { 149 mInputCodePoints[i] = composer.getCodeAt(i); 150 } 151 } 152 153 final InputPointers ips = composer.getInputPointers(); 154 final int codesSize = isGesture ? ips.getPointerSize() : composerSize; 155 // proximityInfo and/or prevWordForBigrams may not be null. 156 final int tmpCount = getSuggestionsNative(mNativeDict, 157 proximityInfo.getNativeProximityInfo(), getTraverseSession(sessionId).getSession(), 158 ips.getXCoordinates(), ips.getYCoordinates(), ips.getTimes(), ips.getPointerIds(), 159 mInputCodePoints, codesSize, 0 /* commitPoint */, isGesture, prevWordCodePointArray, 160 mUseFullEditDistance, mOutputChars, mOutputScores, mSpaceIndices, mOutputTypes); 161 final int count = Math.min(tmpCount, MAX_PREDICTIONS); 162 163 final ArrayList<SuggestedWordInfo> suggestions = CollectionUtils.newArrayList(); 164 for (int j = 0; j < count; ++j) { 165 if (composerSize > 0 && mOutputScores[j] < 1) break; 166 final int start = j * MAX_WORD_LENGTH; 167 int len = 0; 168 while (len < MAX_WORD_LENGTH && mOutputChars[start + len] != 0) { 169 ++len; 170 } 171 if (len > 0) { 172 final int score = SuggestedWordInfo.KIND_WHITELIST == mOutputTypes[j] 173 ? SuggestedWordInfo.MAX_SCORE : mOutputScores[j]; 174 suggestions.add(new SuggestedWordInfo( 175 new String(mOutputChars, start, len), score, mOutputTypes[j], mDictType)); 176 } 177 } 178 return suggestions; 179 } 180 181 /* package for test */ boolean isValidDictionary() { 182 return mNativeDict != 0; 183 } 184 185 public static float calcNormalizedScore(String before, String after, int score) { 186 return calcNormalizedScoreNative(before.toCharArray(), after.toCharArray(), score); 187 } 188 189 public static int editDistance(String before, String after) { 190 return editDistanceNative(before.toCharArray(), after.toCharArray()); 191 } 192 193 @Override 194 public boolean isValidWord(CharSequence word) { 195 return getFrequency(word) >= 0; 196 } 197 198 @Override 199 public int getFrequency(CharSequence word) { 200 if (word == null) return -1; 201 int[] codePoints = StringUtils.toCodePointArray(word.toString()); 202 return getFrequencyNative(mNativeDict, codePoints); 203 } 204 205 // TODO: Add a batch process version (isValidBigramMultiple?) to avoid excessive numbers of jni 206 // calls when checking for changes in an entire dictionary. 207 public boolean isValidBigram(CharSequence word1, CharSequence word2) { 208 if (TextUtils.isEmpty(word1) || TextUtils.isEmpty(word2)) return false; 209 int[] chars1 = StringUtils.toCodePointArray(word1.toString()); 210 int[] chars2 = StringUtils.toCodePointArray(word2.toString()); 211 return isValidBigramNative(mNativeDict, chars1, chars2); 212 } 213 214 @Override 215 public void close() { 216 synchronized (mDicTraverseSessions) { 217 final int sessionsSize = mDicTraverseSessions.size(); 218 for (int index = 0; index < sessionsSize; ++index) { 219 final DicTraverseSession traverseSession = mDicTraverseSessions.valueAt(index); 220 if (traverseSession != null) { 221 traverseSession.close(); 222 } 223 } 224 } 225 closeInternal(); 226 } 227 228 private synchronized void closeInternal() { 229 if (mNativeDict != 0) { 230 closeNative(mNativeDict); 231 mNativeDict = 0; 232 } 233 } 234 235 @Override 236 protected void finalize() throws Throwable { 237 try { 238 closeInternal(); 239 } finally { 240 super.finalize(); 241 } 242 } 243} 244