BinaryDictionary.java revision 338d3ec725a952cbe603ac8b2d49c337463f4093
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.content.res.AssetFileDescriptor; 21import android.content.res.Resources; 22 23import com.android.inputmethod.keyboard.ProximityInfo; 24import com.android.inputmethod.latin.LocaleUtils.RunInLocale; 25 26import java.util.Arrays; 27import java.util.Locale; 28 29/** 30 * Implements a static, compacted, binary dictionary of standard words. 31 */ 32public class BinaryDictionary extends Dictionary { 33 34 public static final String DICTIONARY_PACK_AUTHORITY = 35 "com.android.inputmethod.latin.dictionarypack"; 36 37 /** 38 * There is a difference between what java and native code can handle. 39 * This value should only be used in BinaryDictionary.java 40 * It is necessary to keep it at this value because some languages e.g. German have 41 * really long words. 42 */ 43 public static final int MAX_WORD_LENGTH = 48; 44 public static final int MAX_WORDS = 18; 45 46 private static final String TAG = "BinaryDictionary"; 47 private static final int MAX_BIGRAMS = 60; 48 49 private static final int TYPED_LETTER_MULTIPLIER = 2; 50 51 private int mDicTypeId; 52 private long mNativeDict; 53 private final int[] mInputCodes = new int[MAX_WORD_LENGTH]; 54 private final char[] mOutputChars = new char[MAX_WORD_LENGTH * MAX_WORDS]; 55 private final char[] mOutputChars_bigrams = new char[MAX_WORD_LENGTH * MAX_BIGRAMS]; 56 private final int[] mScores = new int[MAX_WORDS]; 57 private final int[] mBigramScores = new int[MAX_BIGRAMS]; 58 59 public static final Flag FLAG_REQUIRES_GERMAN_UMLAUT_PROCESSING = 60 new Flag(R.bool.config_require_umlaut_processing, 0x1); 61 public static final Flag FLAG_REQUIRES_FRENCH_LIGATURES_PROCESSING = 62 new Flag(R.bool.config_require_ligatures_processing, 0x4); 63 64 // Can create a new flag from extravalue : 65 // public static final Flag FLAG_MYFLAG = 66 // new Flag("my_flag", 0x02); 67 68 // ALL_CONFIG_FLAGS is a collection of flags that enable reading all flags from configuration. 69 // This is but a mask - it does not mean the flags will be on, only that the configuration 70 // will be read for this particular flag. 71 public static final Flag[] ALL_CONFIG_FLAGS = { 72 // Here should reside all flags that trigger some special processing 73 // These *must* match the definition in UnigramDictionary enum in 74 // unigram_dictionary.h so please update both at the same time. 75 // Please note that flags created with a resource are of type CONFIG while flags 76 // created with a string are of type EXTRAVALUE. These behave like masks, and the 77 // actual value will be read from the configuration/extra value at run time for 78 // the configuration at dictionary creation time. 79 FLAG_REQUIRES_GERMAN_UMLAUT_PROCESSING, 80 FLAG_REQUIRES_FRENCH_LIGATURES_PROCESSING, 81 }; 82 83 private final boolean mUseFullEditDistance; 84 85 /** 86 * Constructor for the binary dictionary. This is supposed to be called from the 87 * dictionary factory. 88 * All implementations should pass null into flagArray, except for testing purposes. 89 * @param context the context to access the environment from. 90 * @param filename the name of the file to read through native code. 91 * @param offset the offset of the dictionary data within the file. 92 * @param length the length of the binary data. 93 * @param useFullEditDistance whether to use the full edit distance in suggestions 94 */ 95 public BinaryDictionary(final Context context, 96 final String filename, final long offset, final long length, 97 final boolean useFullEditDistance, final Locale locale) { 98 // Note: at the moment a binary dictionary is always of the "main" type. 99 // Initializing this here will help transitioning out of the scheme where 100 // the Suggest class knows everything about every single dictionary. 101 mDicTypeId = Suggest.DIC_MAIN; 102 mUseFullEditDistance = useFullEditDistance; 103 loadDictionary(filename, offset, length); 104 } 105 106 static { 107 JniUtils.loadNativeLibrary(); 108 } 109 110 private native long openNative(String sourceDir, long dictOffset, long dictSize, 111 int typedLetterMultiplier, int fullWordMultiplier, int maxWordLength, int maxWords); 112 private native void closeNative(long dict); 113 private native boolean isValidWordNative(long dict, char[] word, int wordLength); 114 private native int getSuggestionsNative(long dict, long proximityInfo, int[] xCoordinates, 115 int[] yCoordinates, int[] inputCodes, int codesSize, boolean useFullEditDistance, 116 char[] outputChars, int[] scores); 117 private native int getBigramsNative(long dict, char[] prevWord, int prevWordLength, 118 int[] inputCodes, int inputCodesLength, char[] outputChars, int[] scores, 119 int maxWordLength, int maxBigrams); 120 private static native double calcNormalizedScoreNative( 121 char[] before, int beforeLength, char[] after, int afterLength, int score); 122 private static native int editDistanceNative( 123 char[] before, int beforeLength, char[] after, int afterLength); 124 125 private final void loadDictionary(String path, long startOffset, long length) { 126 mNativeDict = openNative(path, startOffset, length, 127 TYPED_LETTER_MULTIPLIER, FULL_WORD_SCORE_MULTIPLIER, MAX_WORD_LENGTH, MAX_WORDS); 128 } 129 130 @Override 131 public void getBigrams(final WordComposer codes, final CharSequence previousWord, 132 final WordCallback callback) { 133 if (mNativeDict == 0) return; 134 135 char[] chars = previousWord.toString().toCharArray(); 136 Arrays.fill(mOutputChars_bigrams, (char) 0); 137 Arrays.fill(mBigramScores, 0); 138 139 int codesSize = codes.size(); 140 Arrays.fill(mInputCodes, -1); 141 if (codesSize > 0) { 142 mInputCodes[0] = codes.getCodeAt(0); 143 } 144 145 int count = getBigramsNative(mNativeDict, chars, chars.length, mInputCodes, codesSize, 146 mOutputChars_bigrams, mBigramScores, MAX_WORD_LENGTH, MAX_BIGRAMS); 147 if (count > MAX_BIGRAMS) { 148 count = MAX_BIGRAMS; 149 } 150 151 for (int j = 0; j < count; ++j) { 152 if (codesSize > 0 && mBigramScores[j] < 1) break; 153 final int start = j * MAX_WORD_LENGTH; 154 int len = 0; 155 while (len < MAX_WORD_LENGTH && mOutputChars_bigrams[start + len] != 0) { 156 ++len; 157 } 158 if (len > 0) { 159 callback.addWord(mOutputChars_bigrams, start, len, mBigramScores[j], 160 mDicTypeId, Dictionary.BIGRAM); 161 } 162 } 163 } 164 165 // proximityInfo may not be null. 166 @Override 167 public void getWords(final WordComposer codes, final WordCallback callback, 168 final ProximityInfo proximityInfo) { 169 final int count = getSuggestions(codes, proximityInfo, mOutputChars, mScores); 170 171 for (int j = 0; j < count; ++j) { 172 if (mScores[j] < 1) break; 173 final int start = j * MAX_WORD_LENGTH; 174 int len = 0; 175 while (len < MAX_WORD_LENGTH && mOutputChars[start + len] != 0) { 176 ++len; 177 } 178 if (len > 0) { 179 callback.addWord(mOutputChars, start, len, mScores[j], mDicTypeId, 180 Dictionary.UNIGRAM); 181 } 182 } 183 } 184 185 /* package for test */ boolean isValidDictionary() { 186 return mNativeDict != 0; 187 } 188 189 // proximityInfo may not be null. 190 /* package for test */ int getSuggestions(final WordComposer codes, 191 final ProximityInfo proximityInfo, char[] outputChars, int[] scores) { 192 if (!isValidDictionary()) return -1; 193 194 final int codesSize = codes.size(); 195 // Won't deal with really long words. 196 if (codesSize > MAX_WORD_LENGTH - 1) return -1; 197 198 Arrays.fill(mInputCodes, WordComposer.NOT_A_CODE); 199 for (int i = 0; i < codesSize; i++) { 200 mInputCodes[i] = codes.getCodeAt(i); 201 } 202 Arrays.fill(outputChars, (char) 0); 203 Arrays.fill(scores, 0); 204 205 return getSuggestionsNative( 206 mNativeDict, proximityInfo.getNativeProximityInfo(), 207 codes.getXCoordinates(), codes.getYCoordinates(), mInputCodes, codesSize, 208 mUseFullEditDistance, outputChars, scores); 209 } 210 211 public static double calcNormalizedScore(String before, String after, int score) { 212 return calcNormalizedScoreNative(before.toCharArray(), before.length(), 213 after.toCharArray(), after.length(), score); 214 } 215 216 public static int editDistance(String before, String after) { 217 return editDistanceNative( 218 before.toCharArray(), before.length(), after.toCharArray(), after.length()); 219 } 220 221 @Override 222 public boolean isValidWord(CharSequence word) { 223 if (word == null) return false; 224 char[] chars = word.toString().toCharArray(); 225 return isValidWordNative(mNativeDict, chars, chars.length); 226 } 227 228 @Override 229 public synchronized void close() { 230 closeInternal(); 231 } 232 233 private 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