BinaryDictionary.java revision 73680097996ea2ddbca3f84144a00ce3ba66b763
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; 21 22import com.android.inputmethod.keyboard.ProximityInfo; 23 24import java.util.Arrays; 25import java.util.Locale; 26 27/** 28 * Implements a static, compacted, binary dictionary of standard words. 29 */ 30public class BinaryDictionary extends Dictionary { 31 32 public static final String DICTIONARY_PACK_AUTHORITY = 33 "com.android.inputmethod.latin.dictionarypack"; 34 35 /** 36 * There is a difference between what java and native code can handle. 37 * This value should only be used in BinaryDictionary.java 38 * It is necessary to keep it at this value because some languages e.g. German have 39 * really long words. 40 */ 41 public static final int MAX_WORD_LENGTH = 48; 42 public static final int MAX_WORDS = 18; 43 public static final int MAX_SPACES = 16; 44 45 private static final String TAG = "BinaryDictionary"; 46 private static final int MAX_BIGRAMS = 60; 47 48 private static final int TYPED_LETTER_MULTIPLIER = 2; 49 50 private int mDicTypeId; 51 private long mNativeDict; 52 private final int[] mInputCodes = new int[MAX_WORD_LENGTH]; 53 private final char[] mOutputChars = new char[MAX_WORD_LENGTH * MAX_WORDS]; 54 private final char[] mOutputChars_bigrams = new char[MAX_WORD_LENGTH * MAX_BIGRAMS]; 55 private final int[] mSpaceIndices = new int[MAX_SPACES]; 56 private final int[] mScores = new int[MAX_WORDS]; 57 private final int[] mBigramScores = new int[MAX_BIGRAMS]; 58 59 private final boolean mUseFullEditDistance; 60 61 /** 62 * Constructor for the binary dictionary. This is supposed to be called from the 63 * dictionary factory. 64 * All implementations should pass null into flagArray, except for testing purposes. 65 * @param context the context to access the environment from. 66 * @param filename the name of the file to read through native code. 67 * @param offset the offset of the dictionary data within the file. 68 * @param length the length of the binary data. 69 * @param useFullEditDistance whether to use the full edit distance in suggestions 70 * @param dicTypeId the dictionary type id of the dictionary 71 */ 72 public BinaryDictionary(final Context context, 73 final String filename, final long offset, final long length, 74 final boolean useFullEditDistance, final Locale locale, final int dicTypeId) { 75 mDicTypeId = dicTypeId; 76 mUseFullEditDistance = useFullEditDistance; 77 loadDictionary(filename, offset, length); 78 } 79 80 static { 81 JniUtils.loadNativeLibrary(); 82 } 83 84 private native long openNative(String sourceDir, long dictOffset, long dictSize, 85 int typedLetterMultiplier, int fullWordMultiplier, int maxWordLength, int maxWords); 86 private native void closeNative(long dict); 87 private native int getFrequencyNative(long dict, int[] word, int wordLength); 88 private native boolean isValidBigramNative(long dict, int[] word1, int[] word2); 89 private native int getSuggestionsNative(long dict, long proximityInfo, int[] xCoordinates, 90 int[] yCoordinates, int[] times, int[] pointerIds, int[] inputCodes, int codesSize, 91 int commitPoint, boolean isGesture, int dicTypeId, 92 int[] prevWordCodePointArray, boolean useFullEditDistance, char[] outputChars, 93 int[] scores, int[] outputIndices); 94 private native int getBigramsNative(long dict, int[] prevWord, int prevWordLength, 95 int[] inputCodes, int inputCodesLength, char[] outputChars, int[] scores, 96 int maxWordLength, int maxBigrams); 97 private static native float calcNormalizedScoreNative( 98 char[] before, int beforeLength, char[] after, int afterLength, int score); 99 private static native int editDistanceNative( 100 char[] before, int beforeLength, char[] after, int afterLength); 101 102 private final void loadDictionary(String path, long startOffset, long length) { 103 mNativeDict = openNative(path, startOffset, length, 104 TYPED_LETTER_MULTIPLIER, FULL_WORD_SCORE_MULTIPLIER, MAX_WORD_LENGTH, MAX_WORDS); 105 } 106 107 @Override 108 public void getBigrams(final WordComposer codes, final CharSequence previousWord, 109 final WordCallback callback) { 110 if (mNativeDict == 0) return; 111 112 int[] codePoints = StringUtils.toCodePointArray(previousWord.toString()); 113 Arrays.fill(mOutputChars_bigrams, (char) 0); 114 Arrays.fill(mBigramScores, 0); 115 116 int codesSize = codes.size(); 117 Arrays.fill(mInputCodes, -1); 118 if (codesSize > 0) { 119 mInputCodes[0] = codes.getCodeAt(0); 120 } 121 122 int count = getBigramsNative(mNativeDict, codePoints, codePoints.length, mInputCodes, 123 codesSize, mOutputChars_bigrams, mBigramScores, MAX_WORD_LENGTH, MAX_BIGRAMS); 124 if (count > MAX_BIGRAMS) { 125 count = MAX_BIGRAMS; 126 } 127 128 for (int j = 0; j < count; ++j) { 129 if (codesSize > 0 && mBigramScores[j] < 1) break; 130 final int start = j * MAX_WORD_LENGTH; 131 int len = 0; 132 while (len < MAX_WORD_LENGTH && mOutputChars_bigrams[start + len] != 0) { 133 ++len; 134 } 135 if (len > 0) { 136 callback.addWord(mOutputChars_bigrams, null, start, len, mBigramScores[j], 137 mDicTypeId, Dictionary.BIGRAM); 138 } 139 } 140 } 141 142 // proximityInfo and/or prevWordForBigrams may not be null. 143 @Override 144 public void getWords(final WordComposer codes, final CharSequence prevWordForBigrams, 145 final WordCallback callback, final ProximityInfo proximityInfo) { 146 147 final int count = getSuggestions(codes, prevWordForBigrams, proximityInfo, mOutputChars, 148 mScores, mSpaceIndices); 149 for (int j = 0; j < count; ++j) { 150 if (mScores[j] < 1) break; 151 final int start = j * MAX_WORD_LENGTH; 152 int len = 0; 153 while (len < MAX_WORD_LENGTH && mOutputChars[start + len] != 0) { 154 ++len; 155 } 156 if (len > 0) { 157 callback.addWord(mOutputChars, null, start, len, mScores[j], mDicTypeId, 158 Dictionary.UNIGRAM); 159 } 160 } 161 } 162 163 /* package for test */ boolean isValidDictionary() { 164 return mNativeDict != 0; 165 } 166 167 // proximityInfo may not be null. 168 /* package for test */ int getSuggestions(final WordComposer codes, 169 final CharSequence prevWordForBigrams, final ProximityInfo proximityInfo, 170 char[] outputChars, int[] scores, int[] spaceIndices) { 171 if (!isValidDictionary()) return -1; 172 173 final int codesSize = codes.size(); 174 // Won't deal with really long words. 175 if (codesSize > MAX_WORD_LENGTH - 1) return -1; 176 177 Arrays.fill(mInputCodes, WordComposer.NOT_A_CODE); 178 for (int i = 0; i < codesSize; i++) { 179 mInputCodes[i] = codes.getCodeAt(i); 180 } 181 Arrays.fill(outputChars, (char) 0); 182 Arrays.fill(scores, 0); 183 184 // TODO: toLowerCase in the native code 185 final int[] prevWordCodePointArray = (null == prevWordForBigrams) 186 ? null : StringUtils.toCodePointArray(prevWordForBigrams.toString()); 187 188 int[] emptyArray = new int[codesSize]; 189 Arrays.fill(emptyArray, 0); 190 191 //final int commitPoint = codes.getCommitPoint(); 192 //codes.clearCommitPoint(); 193 194 return getSuggestionsNative(mNativeDict, proximityInfo.getNativeProximityInfo(), 195 codes.getXCoordinates(), codes.getYCoordinates(), emptyArray, emptyArray, mInputCodes, 196 codesSize, 0 /* unused */, false, mDicTypeId, 197 prevWordCodePointArray, mUseFullEditDistance, 198 outputChars, scores, spaceIndices); 199 } 200 201 public static float calcNormalizedScore(String before, String after, int score) { 202 return calcNormalizedScoreNative(before.toCharArray(), before.length(), 203 after.toCharArray(), after.length(), score); 204 } 205 206 public static int editDistance(String before, String after) { 207 return editDistanceNative( 208 before.toCharArray(), before.length(), after.toCharArray(), after.length()); 209 } 210 211 @Override 212 public boolean isValidWord(CharSequence word) { 213 return getFrequency(word) >= 0; 214 } 215 216 @Override 217 public int getFrequency(CharSequence word) { 218 if (word == null) return -1; 219 int[] chars = StringUtils.toCodePointArray(word.toString()); 220 return getFrequencyNative(mNativeDict, chars, chars.length); 221 } 222 223 // TODO: Add a batch process version (isValidBigramMultiple?) to avoid excessive numbers of jni 224 // calls when checking for changes in an entire dictionary. 225 public boolean isValidBigram(CharSequence word1, CharSequence word2) { 226 if (TextUtils.isEmpty(word1) || TextUtils.isEmpty(word2)) return false; 227 int[] chars1 = StringUtils.toCodePointArray(word1.toString()); 228 int[] chars2 = StringUtils.toCodePointArray(word2.toString()); 229 return isValidBigramNative(mNativeDict, chars1, chars2); 230 } 231 232 @Override 233 public synchronized void close() { 234 closeInternal(); 235 } 236 237 private void closeInternal() { 238 if (mNativeDict != 0) { 239 closeNative(mNativeDict); 240 mNativeDict = 0; 241 } 242 } 243 244 @Override 245 protected void finalize() throws Throwable { 246 try { 247 closeInternal(); 248 } finally { 249 super.finalize(); 250 } 251 } 252} 253