UserBinaryDictionary.java revision 61fe28831d93a829f67694a342b6c94bdc7732ea
1/* 2 * Copyright (C) 2012 The Android Open Source Project 3 * 4 * Licensed under the Apache License, Version 2.0 (the "License"); 5 * you may not use this file except in compliance with the License. 6 * You may obtain a copy of 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, 12 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 13 * See the License for the specific language governing permissions and 14 * limitations under the License. 15 */ 16 17package com.android.inputmethod.latin; 18 19import android.content.ContentProviderClient; 20import android.content.ContentResolver; 21import android.content.Context; 22import android.database.ContentObserver; 23import android.database.Cursor; 24import android.database.sqlite.SQLiteException; 25import android.net.Uri; 26import android.os.Build; 27import android.provider.UserDictionary.Words; 28import android.text.TextUtils; 29import android.util.Log; 30 31import com.android.inputmethod.compat.UserDictionaryCompatUtils; 32import com.android.inputmethod.latin.utils.SubtypeLocaleUtils; 33 34import java.io.File; 35import java.util.Arrays; 36import java.util.Locale; 37 38/** 39 * An expandable dictionary that stores the words in the user dictionary provider into a binary 40 * dictionary file to use it from native code. 41 */ 42public class UserBinaryDictionary extends ExpandableBinaryDictionary { 43 private static final String TAG = ExpandableBinaryDictionary.class.getSimpleName(); 44 45 // The user dictionary provider uses an empty string to mean "all languages". 46 private static final String USER_DICTIONARY_ALL_LANGUAGES = ""; 47 private static final int HISTORICAL_DEFAULT_USER_DICTIONARY_FREQUENCY = 250; 48 private static final int LATINIME_DEFAULT_USER_DICTIONARY_FREQUENCY = 160; 49 // Shortcut frequency is 0~15, with 15 = whitelist. We don't want user dictionary entries 50 // to auto-correct, so we set this to the highest frequency that won't, i.e. 14. 51 private static final int USER_DICT_SHORTCUT_FREQUENCY = 14; 52 53 private static final String[] PROJECTION_QUERY_WITH_SHORTCUT = new String[] { 54 Words.WORD, 55 Words.SHORTCUT, 56 Words.FREQUENCY, 57 }; 58 private static final String[] PROJECTION_QUERY_WITHOUT_SHORTCUT = new String[] { 59 Words.WORD, 60 Words.FREQUENCY, 61 }; 62 63 private static final String NAME = "userunigram"; 64 65 private ContentObserver mObserver; 66 final private String mLocale; 67 final private boolean mAlsoUseMoreRestrictiveLocales; 68 69 private UserBinaryDictionary(final Context context, final Locale locale, final File dictFile) { 70 this(context, locale, false /* alsoUseMoreRestrictiveLocales */, dictFile, NAME); 71 } 72 73 protected UserBinaryDictionary(final Context context, final Locale locale, 74 final boolean alsoUseMoreRestrictiveLocales, final File dictFile, final String name) { 75 super(context, getDictName(name, locale, dictFile), locale, Dictionary.TYPE_USER, dictFile); 76 if (null == locale) throw new NullPointerException(); // Catch the error earlier 77 final String localeStr = locale.toString(); 78 if (SubtypeLocaleUtils.NO_LANGUAGE.equals(localeStr)) { 79 // If we don't have a locale, insert into the "all locales" user dictionary. 80 mLocale = USER_DICTIONARY_ALL_LANGUAGES; 81 } else { 82 mLocale = localeStr; 83 } 84 mAlsoUseMoreRestrictiveLocales = alsoUseMoreRestrictiveLocales; 85 ContentResolver cres = context.getContentResolver(); 86 87 mObserver = new ContentObserver(null) { 88 @Override 89 public void onChange(final boolean self) { 90 // This hook is deprecated as of API level 16 (Build.VERSION_CODES.JELLY_BEAN), 91 // but should still be supported for cases where the IME is running on an older 92 // version of the platform. 93 onChange(self, null); 94 } 95 // The following hook is only available as of API level 16 96 // (Build.VERSION_CODES.JELLY_BEAN), and as such it will only work on JellyBean+ 97 // devices. On older versions of the platform, the hook above will be called instead. 98 @Override 99 public void onChange(final boolean self, final Uri uri) { 100 setNeedsToReload(); 101 } 102 }; 103 cres.registerContentObserver(Words.CONTENT_URI, true, mObserver); 104 reloadDictionaryIfRequired(); 105 } 106 107 public static UserBinaryDictionary getDictionary(final Context context, final Locale locale, 108 final File dictFile) { 109 return new UserBinaryDictionary(context, locale, dictFile); 110 } 111 112 @Override 113 public synchronized void close() { 114 if (mObserver != null) { 115 mContext.getContentResolver().unregisterContentObserver(mObserver); 116 mObserver = null; 117 } 118 super.close(); 119 } 120 121 @Override 122 public void loadInitialContentsLocked() { 123 // Split the locale. For example "en" => ["en"], "de_DE" => ["de", "DE"], 124 // "en_US_foo_bar_qux" => ["en", "US", "foo_bar_qux"] because of the limit of 3. 125 // This is correct for locale processing. 126 // For this example, we'll look at the "en_US_POSIX" case. 127 final String[] localeElements = 128 TextUtils.isEmpty(mLocale) ? new String[] {} : mLocale.split("_", 3); 129 final int length = localeElements.length; 130 131 final StringBuilder request = new StringBuilder("(locale is NULL)"); 132 String localeSoFar = ""; 133 // At start, localeElements = ["en", "US", "POSIX"] ; localeSoFar = "" ; 134 // and request = "(locale is NULL)" 135 for (int i = 0; i < length; ++i) { 136 // i | localeSoFar | localeElements 137 // 0 | "" | ["en", "US", "POSIX"] 138 // 1 | "en_" | ["en", "US", "POSIX"] 139 // 2 | "en_US_" | ["en", "en_US", "POSIX"] 140 localeElements[i] = localeSoFar + localeElements[i]; 141 localeSoFar = localeElements[i] + "_"; 142 // i | request 143 // 0 | "(locale is NULL)" 144 // 1 | "(locale is NULL) or (locale=?)" 145 // 2 | "(locale is NULL) or (locale=?) or (locale=?)" 146 request.append(" or (locale=?)"); 147 } 148 // At the end, localeElements = ["en", "en_US", "en_US_POSIX"]; localeSoFar = en_US_POSIX_" 149 // and request = "(locale is NULL) or (locale=?) or (locale=?) or (locale=?)" 150 151 final String[] requestArguments; 152 // If length == 3, we already have all the arguments we need (common prefix is meaningless 153 // inside variants 154 if (mAlsoUseMoreRestrictiveLocales && length < 3) { 155 request.append(" or (locale like ?)"); 156 // The following creates an array with one more (null) position 157 final String[] localeElementsWithMoreRestrictiveLocalesIncluded = 158 Arrays.copyOf(localeElements, length + 1); 159 localeElementsWithMoreRestrictiveLocalesIncluded[length] = 160 localeElements[length - 1] + "_%"; 161 requestArguments = localeElementsWithMoreRestrictiveLocalesIncluded; 162 // If for example localeElements = ["en"] 163 // then requestArguments = ["en", "en_%"] 164 // and request = (locale is NULL) or (locale=?) or (locale like ?) 165 // If localeElements = ["en", "en_US"] 166 // then requestArguments = ["en", "en_US", "en_US_%"] 167 } else { 168 requestArguments = localeElements; 169 } 170 final String requestString = request.toString(); 171 if (Build.VERSION.SDK_INT >= Build.VERSION_CODES.JELLY_BEAN) { 172 try { 173 addWordsFromProjectionLocked(PROJECTION_QUERY_WITH_SHORTCUT, requestString, 174 requestArguments); 175 } catch (IllegalArgumentException e) { 176 // This may happen on some non-compliant devices where the declared API is JB+ but 177 // the SHORTCUT column is not present for some reason. 178 addWordsFromProjectionLocked(PROJECTION_QUERY_WITHOUT_SHORTCUT, requestString, 179 requestArguments); 180 } 181 } else { 182 addWordsFromProjectionLocked(PROJECTION_QUERY_WITHOUT_SHORTCUT, requestString, 183 requestArguments); 184 } 185 } 186 187 private void addWordsFromProjectionLocked(final String[] query, String request, 188 final String[] requestArguments) throws IllegalArgumentException { 189 Cursor cursor = null; 190 try { 191 cursor = mContext.getContentResolver().query( 192 Words.CONTENT_URI, query, request, requestArguments, null); 193 addWordsLocked(cursor); 194 } catch (final SQLiteException e) { 195 Log.e(TAG, "SQLiteException in the remote User dictionary process.", e); 196 } finally { 197 try { 198 if (null != cursor) cursor.close(); 199 } catch (final SQLiteException e) { 200 Log.e(TAG, "SQLiteException in the remote User dictionary process.", e); 201 } 202 } 203 } 204 205 public static boolean isEnabled(final Context context) { 206 final ContentResolver cr = context.getContentResolver(); 207 final ContentProviderClient client = cr.acquireContentProviderClient(Words.CONTENT_URI); 208 if (client != null) { 209 client.release(); 210 return true; 211 } else { 212 return false; 213 } 214 } 215 216 /** 217 * Adds a word to the user dictionary and makes it persistent. 218 * 219 * @param context the context 220 * @param locale the locale 221 * @param word the word to add. If the word is capitalized, then the dictionary will 222 * recognize it as a capitalized word when searched. 223 */ 224 public static void addWordToUserDictionary(final Context context, final Locale locale, 225 final String word) { 226 // Update the user dictionary provider 227 UserDictionaryCompatUtils.addWord(context, word, 228 HISTORICAL_DEFAULT_USER_DICTIONARY_FREQUENCY, null, locale); 229 } 230 231 private int scaleFrequencyFromDefaultToLatinIme(final int defaultFrequency) { 232 // The default frequency for the user dictionary is 250 for historical reasons. 233 // Latin IME considers a good value for the default user dictionary frequency 234 // is about 160 considering the scale we use. So we are scaling down the values. 235 if (defaultFrequency > Integer.MAX_VALUE / LATINIME_DEFAULT_USER_DICTIONARY_FREQUENCY) { 236 return (defaultFrequency / HISTORICAL_DEFAULT_USER_DICTIONARY_FREQUENCY) 237 * LATINIME_DEFAULT_USER_DICTIONARY_FREQUENCY; 238 } else { 239 return (defaultFrequency * LATINIME_DEFAULT_USER_DICTIONARY_FREQUENCY) 240 / HISTORICAL_DEFAULT_USER_DICTIONARY_FREQUENCY; 241 } 242 } 243 244 private void addWordsLocked(final Cursor cursor) { 245 final boolean hasShortcutColumn = Build.VERSION.SDK_INT >= Build.VERSION_CODES.JELLY_BEAN; 246 if (cursor == null) return; 247 if (cursor.moveToFirst()) { 248 final int indexWord = cursor.getColumnIndex(Words.WORD); 249 final int indexShortcut = hasShortcutColumn ? cursor.getColumnIndex(Words.SHORTCUT) : 0; 250 final int indexFrequency = cursor.getColumnIndex(Words.FREQUENCY); 251 while (!cursor.isAfterLast()) { 252 final String word = cursor.getString(indexWord); 253 final String shortcut = hasShortcutColumn ? cursor.getString(indexShortcut) : null; 254 final int frequency = cursor.getInt(indexFrequency); 255 final int adjustedFrequency = scaleFrequencyFromDefaultToLatinIme(frequency); 256 // Safeguard against adding really long words. 257 if (word.length() < MAX_WORD_LENGTH) { 258 runGCIfRequiredLocked(true /* mindsBlockByGC */); 259 addWordDynamicallyLocked(word, adjustedFrequency, null /* shortcutTarget */, 260 0 /* shortcutFreq */, false /* isNotAWord */, 261 false /* isBlacklisted */, BinaryDictionary.NOT_A_VALID_TIMESTAMP); 262 if (null != shortcut && shortcut.length() < MAX_WORD_LENGTH) { 263 runGCIfRequiredLocked(true /* mindsBlockByGC */); 264 addWordDynamicallyLocked(shortcut, adjustedFrequency, word, 265 USER_DICT_SHORTCUT_FREQUENCY, true /* isNotAWord */, 266 false /* isBlacklisted */, BinaryDictionary.NOT_A_VALID_TIMESTAMP); 267 } 268 } 269 cursor.moveToNext(); 270 } 271 } 272 } 273 274 @Override 275 protected boolean haveContentsChanged() { 276 return true; 277 } 278} 279