/* * Copyright (C) 2011 The Android Open Source Project * * Licensed under the Apache License, Version 2.0 (the "License"); you may not * use this file except in compliance with the License. You may obtain a copy of * the License at * * http://www.apache.org/licenses/LICENSE-2.0 * * Unless required by applicable law or agreed to in writing, software * distributed under the License is distributed on an "AS IS" BASIS, WITHOUT * WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the * License for the specific language governing permissions and limitations under * the License. */ package com.android.inputmethod.latin; import com.android.inputmethod.latin.define.ProductionFlag; import com.android.inputmethod.latin.makedict.BinaryDictInputOutput; import com.android.inputmethod.latin.makedict.FormatSpec; import android.content.Context; import android.content.SharedPreferences; import android.content.pm.PackageManager.NameNotFoundException; import android.content.res.AssetFileDescriptor; import android.util.Log; import java.io.File; import java.io.FileInputStream; import java.io.IOException; import java.nio.BufferUnderflowException; import java.nio.channels.FileChannel; import java.util.ArrayList; import java.util.HashMap; import java.util.Locale; /** * Helper class to get the address of a mmap'able dictionary file. */ final class BinaryDictionaryGetter { /** * Used for Log actions from this class */ private static final String TAG = BinaryDictionaryGetter.class.getSimpleName(); /** * Used to return empty lists */ private static final File[] EMPTY_FILE_ARRAY = new File[0]; /** * Name of the common preferences name to know which word list are on and which are off. */ private static final String COMMON_PREFERENCES_NAME = "LatinImeDictPrefs"; // Name of the category for the main dictionary private static final String MAIN_DICTIONARY_CATEGORY = "main"; public static final String ID_CATEGORY_SEPARATOR = ":"; // The key considered to read the version attribute in a dictionary file. private static String VERSION_KEY = "version"; // Prevents this from being instantiated private BinaryDictionaryGetter() {} /** * Returns whether we may want to use this character as part of a file name. * * This basically only accepts ascii letters and numbers, and rejects everything else. */ private static boolean isFileNameCharacter(int codePoint) { if (codePoint >= 0x30 && codePoint <= 0x39) return true; // Digit if (codePoint >= 0x41 && codePoint <= 0x5A) return true; // Uppercase if (codePoint >= 0x61 && codePoint <= 0x7A) return true; // Lowercase return codePoint == '_'; // Underscore } /** * Escapes a string for any characters that may be suspicious for a file or directory name. * * Concretely this does a sort of URL-encoding except it will encode everything that's not * alphanumeric or underscore. (true URL-encoding leaves alone characters like '*', which * we cannot allow here) */ // TODO: create a unit test for this method private static String replaceFileNameDangerousCharacters(final String name) { // This assumes '%' is fully available as a non-separator, normal // character in a file name. This is probably true for all file systems. final StringBuilder sb = new StringBuilder(); final int nameLength = name.length(); for (int i = 0; i < nameLength; i = name.offsetByCodePoints(i, 1)) { final int codePoint = name.codePointAt(i); if (isFileNameCharacter(codePoint)) { sb.appendCodePoint(codePoint); } else { // 6 digits - unicode is limited to 21 bits sb.append(String.format((Locale)null, "%%%1$06x", codePoint)); } } return sb.toString(); } /** * Reverse escaping done by replaceFileNameDangerousCharacters. */ private static String getWordListIdFromFileName(final String fname) { final StringBuilder sb = new StringBuilder(); final int fnameLength = fname.length(); for (int i = 0; i < fnameLength; i = fname.offsetByCodePoints(i, 1)) { final int codePoint = fname.codePointAt(i); if ('%' != codePoint) { sb.appendCodePoint(codePoint); } else { final int encodedCodePoint = Integer.parseInt(fname.substring(i + 1, i + 7), 16); i += 6; sb.appendCodePoint(encodedCodePoint); } } return sb.toString(); } /** * Helper method to get the top level cache directory. */ private static String getWordListCacheDirectory(final Context context) { return context.getFilesDir() + File.separator + "dicts"; } /** * Find out the cache directory associated with a specific locale. */ private static String getCacheDirectoryForLocale(final String locale, final Context context) { final String relativeDirectoryName = replaceFileNameDangerousCharacters(locale); final String absoluteDirectoryName = getWordListCacheDirectory(context) + File.separator + relativeDirectoryName; final File directory = new File(absoluteDirectoryName); if (!directory.exists()) { if (!directory.mkdirs()) { Log.e(TAG, "Could not create the directory for locale" + locale); } } return absoluteDirectoryName; } /** * Generates a file name for the id and locale passed as an argument. * * In the current implementation the file name returned will always be unique for * any id/locale pair, but please do not expect that the id can be the same for * different dictionaries with different locales. An id should be unique for any * dictionary. * The file name is pretty much an URL-encoded version of the id inside a directory * named like the locale, except it will also escape characters that look dangerous * to some file systems. * @param id the id of the dictionary for which to get a file name * @param locale the locale for which to get the file name as a string * @param context the context to use for getting the directory * @return the name of the file to be created */ public static String getCacheFileName(String id, String locale, Context context) { final String fileName = replaceFileNameDangerousCharacters(id); return getCacheDirectoryForLocale(locale, context) + File.separator + fileName; } /** * Generates a unique temporary file name in the app cache directory. */ public static String getTempFileName(String id, Context context) throws IOException { return File.createTempFile(replaceFileNameDangerousCharacters(id), null).getAbsolutePath(); } /** * Returns a file address from a resource, or null if it cannot be opened. */ private static AssetFileAddress loadFallbackResource(final Context context, final int fallbackResId) { final AssetFileDescriptor afd = context.getResources().openRawResourceFd(fallbackResId); if (afd == null) { Log.e(TAG, "Found the resource but cannot read it. Is it compressed? resId=" + fallbackResId); return null; } return AssetFileAddress.makeFromFileNameAndOffset( context.getApplicationInfo().sourceDir, afd.getStartOffset(), afd.getLength()); } private static final class DictPackSettings { final SharedPreferences mDictPreferences; public DictPackSettings(final Context context) { Context dictPackContext = null; try { final String dictPackName = context.getString(R.string.dictionary_pack_package_name); dictPackContext = context.createPackageContext(dictPackName, 0); } catch (NameNotFoundException e) { // The dictionary pack is not installed... // TODO: fallback on the built-in dict, see the TODO above Log.e(TAG, "Could not find a dictionary pack"); } mDictPreferences = null == dictPackContext ? null : dictPackContext.getSharedPreferences(COMMON_PREFERENCES_NAME, Context.MODE_WORLD_READABLE | Context.MODE_MULTI_PROCESS); } public boolean isWordListActive(final String dictId) { if (null == mDictPreferences) { // If we don't have preferences it basically means we can't find the dictionary // pack - either it's not installed, or it's disabled, or there is some strange // bug. Either way, a word list with no settings should be on by default: default // dictionaries in LatinIME are on if there is no settings at all, and if for some // reason some dictionaries have been installed BUT the dictionary pack can't be // found anymore it's safer to actually supply installed dictionaries. return true; } else { // The default is true here for the same reasons as above. We got the dictionary // pack but if we don't have any settings for it it means the user has never been // to the settings yet. So by default, the main dictionaries should be on. return mDictPreferences.getBoolean(dictId, true); } } } /** * Helper method to the list of cache directories, one for each distinct locale. */ private static File[] getCachedDirectoryList(final Context context) { return new File(getWordListCacheDirectory(context)).listFiles(); } /** * Returns the category for a given file name. * * This parses the file name, extracts the category, and returns it. See * {@link #getMainDictId(Locale)} and {@link #isMainWordListId(String)}. * @return The category as a string or null if it can't be found in the file name. */ private static String getCategoryFromFileName(final String fileName) { final String id = getWordListIdFromFileName(fileName); final String[] idArray = id.split(ID_CATEGORY_SEPARATOR); if (2 != idArray.length) return null; return idArray[0]; } /** * Utility class for the {@link #getCachedWordLists} method */ private static final class FileAndMatchLevel { final File mFile; final int mMatchLevel; public FileAndMatchLevel(final File file, final int matchLevel) { mFile = file; mMatchLevel = matchLevel; } } /** * Returns the list of cached files for a specific locale, one for each category. * * This will return exactly one file for each word list category that matches * the passed locale. If several files match the locale for any given category, * this returns the file with the closest match to the locale. For example, if * the passed word list is en_US, and for a category we have an en and an en_US * word list available, we'll return only the en_US one. * Thus, the list will contain as many files as there are categories. * * @param locale the locale to find the dictionary files for, as a string. * @param context the context on which to open the files upon. * @return an array of binary dictionary files, which may be empty but may not be null. */ private static File[] getCachedWordLists(final String locale, final Context context) { final File[] directoryList = getCachedDirectoryList(context); if (null == directoryList) return EMPTY_FILE_ARRAY; final HashMap cacheFiles = CollectionUtils.newHashMap(); for (File directory : directoryList) { if (!directory.isDirectory()) continue; final String dirLocale = getWordListIdFromFileName(directory.getName()); final int matchLevel = LocaleUtils.getMatchLevel(dirLocale, locale); if (LocaleUtils.isMatch(matchLevel)) { final File[] wordLists = directory.listFiles(); if (null != wordLists) { for (File wordList : wordLists) { final String category = getCategoryFromFileName(wordList.getName()); final FileAndMatchLevel currentBestMatch = cacheFiles.get(category); if (null == currentBestMatch || currentBestMatch.mMatchLevel < matchLevel) { cacheFiles.put(category, new FileAndMatchLevel(wordList, matchLevel)); } } } } } if (cacheFiles.isEmpty()) return EMPTY_FILE_ARRAY; final File[] result = new File[cacheFiles.size()]; int index = 0; for (final FileAndMatchLevel entry : cacheFiles.values()) { result[index++] = entry.mFile; } return result; } /** * Remove all files with the passed id, except the passed file. * * If a dictionary with a given ID has a metadata change that causes it to change * path, we need to remove the old version. The only way to do this is to check all * installed files for a matching ID in a different directory. */ public static void removeFilesWithIdExcept(final Context context, final String id, final File fileToKeep) { try { final File canonicalFileToKeep = fileToKeep.getCanonicalFile(); final File[] directoryList = getCachedDirectoryList(context); if (null == directoryList) return; for (File directory : directoryList) { // There is one directory per locale. See #getCachedDirectoryList if (!directory.isDirectory()) continue; final File[] wordLists = directory.listFiles(); if (null == wordLists) continue; for (File wordList : wordLists) { final String fileId = getWordListIdFromFileName(wordList.getName()); if (fileId.equals(id)) { if (!canonicalFileToKeep.equals(wordList.getCanonicalFile())) { wordList.delete(); } } } } } catch (java.io.IOException e) { Log.e(TAG, "IOException trying to cleanup files : " + e); } } /** * Returns the id associated with the main word list for a specified locale. * * Word lists stored in Android Keyboard's resources are referred to as the "main" * word lists. Since they can be updated like any other list, we need to assign a * unique ID to them. This ID is just the name of the language (locale-wise) they * are for, and this method returns this ID. */ private static String getMainDictId(final Locale locale) { // This works because we don't include by default different dictionaries for // different countries. This actually needs to return the id that we would // like to use for word lists included in resources, and the following is okay. return MAIN_DICTIONARY_CATEGORY + ID_CATEGORY_SEPARATOR + locale.getLanguage().toString(); } private static boolean isMainWordListId(final String id) { final String[] idArray = id.split(ID_CATEGORY_SEPARATOR); if (2 != idArray.length) return false; return MAIN_DICTIONARY_CATEGORY.equals(idArray[0]); } // ## HACK ## we prevent usage of a dictionary before version 18 for English only. The reason // for this is, since those do not include whitelist entries, the new code with an old version // of the dictionary would lose whitelist functionality. private static boolean hackCanUseDictionaryFile(final Locale locale, final File f) { // Only for English - other languages didn't have a whitelist, hence this // ad-hoc ## HACK ## if (!Locale.ENGLISH.getLanguage().equals(locale.getLanguage())) return true; FileInputStream inStream = null; try { // Read the version of the file inStream = new FileInputStream(f); final BinaryDictInputOutput.ByteBufferWrapper buffer = new BinaryDictInputOutput.ByteBufferWrapper(inStream.getChannel().map( FileChannel.MapMode.READ_ONLY, 0, f.length())); final int magic = buffer.readInt(); if (magic != FormatSpec.VERSION_2_MAGIC_NUMBER) { return false; } final int formatVersion = buffer.readInt(); final int headerSize = buffer.readInt(); final HashMap options = CollectionUtils.newHashMap(); BinaryDictInputOutput.populateOptions(buffer, headerSize, options); final String version = options.get(VERSION_KEY); if (null == version) { // No version in the options : the format is unexpected return false; } // Version 18 is the first one to include the whitelist // Obviously this is a big ## HACK ## return Integer.parseInt(version) >= 18; } catch (java.io.FileNotFoundException e) { return false; } catch (java.io.IOException e) { return false; } catch (NumberFormatException e) { return false; } catch (BufferUnderflowException e) { return false; } finally { if (inStream != null) { try { inStream.close(); } catch (IOException e) { // do nothing } } } } /** * Returns a list of file addresses for a given locale, trying relevant methods in order. * * Tries to get binary dictionaries from various sources, in order: * - Uses a content provider to get a public dictionary set, as per the protocol described * in BinaryDictionaryFileDumper. * If that fails: * - Gets a file name from the built-in dictionary for this locale, if any. * If that fails: * - Returns null. * @return The list of addresses of valid dictionary files, or null. */ public static ArrayList getDictionaryFiles(final Locale locale, final Context context) { final boolean hasDefaultWordList = DictionaryFactory.isDictionaryAvailable(context, locale); // cacheWordListsFromContentProvider returns the list of files it copied to local // storage, but we don't really care about what was copied NOW: what we want is the // list of everything we ever cached, so we ignore the return value. // TODO: The experimental version is not supported by the Dictionary Pack Service yet if (!ProductionFlag.IS_EXPERIMENTAL) { BinaryDictionaryFileDumper.cacheWordListsFromContentProvider(locale, context, hasDefaultWordList); } final File[] cachedWordLists = getCachedWordLists(locale.toString(), context); final String mainDictId = getMainDictId(locale); final DictPackSettings dictPackSettings = new DictPackSettings(context); boolean foundMainDict = false; final ArrayList fileList = CollectionUtils.newArrayList(); // cachedWordLists may not be null, see doc for getCachedDictionaryList for (final File f : cachedWordLists) { final String wordListId = getWordListIdFromFileName(f.getName()); final boolean canUse = f.canRead() && hackCanUseDictionaryFile(locale, f); if (canUse && isMainWordListId(wordListId)) { foundMainDict = true; } if (!dictPackSettings.isWordListActive(wordListId)) continue; if (canUse) { fileList.add(AssetFileAddress.makeFromFileName(f.getPath())); } else { Log.e(TAG, "Found a cached dictionary file but cannot read or use it"); } } if (!foundMainDict && dictPackSettings.isWordListActive(mainDictId)) { final int fallbackResId = DictionaryFactory.getMainDictionaryResourceId(context.getResources(), locale); final AssetFileAddress fallbackAsset = loadFallbackResource(context, fallbackResId); if (null != fallbackAsset) { fileList.add(fallbackAsset); } } return fileList; } }