diff options
Diffstat (limited to 'java/src/com/android/inputmethod/latin')
38 files changed, 3469 insertions, 2754 deletions
diff --git a/java/src/com/android/inputmethod/latin/AbstractDictionaryWriter.java b/java/src/com/android/inputmethod/latin/AbstractDictionaryWriter.java index ebbcedc96..269b3a299 100644 --- a/java/src/com/android/inputmethod/latin/AbstractDictionaryWriter.java +++ b/java/src/com/android/inputmethod/latin/AbstractDictionaryWriter.java @@ -42,8 +42,10 @@ abstract public class AbstractDictionaryWriter extends Dictionary { abstract public void addUnigramWord(final String word, final String shortcutTarget, final int frequency, final boolean isNotAWord); + // TODO: Remove lastModifiedTime after making binary dictionary support forgetting curve. abstract public void addBigramWords(final String word0, final String word1, - final int frequency, final boolean isValid); + final int frequency, final boolean isValid, + final long lastModifiedTime); abstract public void removeBigramWords(final String word0, final String word1); diff --git a/java/src/com/android/inputmethod/latin/AudioAndHapticFeedbackManager.java b/java/src/com/android/inputmethod/latin/AudioAndHapticFeedbackManager.java index 42c57946d..54bc29559 100644 --- a/java/src/com/android/inputmethod/latin/AudioAndHapticFeedbackManager.java +++ b/java/src/com/android/inputmethod/latin/AudioAndHapticFeedbackManager.java @@ -57,10 +57,10 @@ public final class AudioAndHapticFeedbackManager { mVibrator = (Vibrator) context.getSystemService(Context.VIBRATOR_SERVICE); } - public void hapticAndAudioFeedback(final int primaryCode, + public void performHapticAndAudioFeedback(final int code, final View viewToPerformHapticFeedbackOn) { - vibrateInternal(viewToPerformHapticFeedbackOn); - playKeyClick(primaryCode); + performHapticFeedback(viewToPerformHapticFeedbackOn); + performAudioFeedback(code); } public boolean hasVibrator() { @@ -81,14 +81,14 @@ public final class AudioAndHapticFeedbackManager { return mAudioManager.getRingerMode() == AudioManager.RINGER_MODE_NORMAL; } - private void playKeyClick(final int primaryCode) { + public void performAudioFeedback(final int code) { // if mAudioManager is null, we can't play a sound anyway, so return if (mAudioManager == null) { return; } if (mSoundOn) { final int sound; - switch (primaryCode) { + switch (code) { case Constants.CODE_DELETE: sound = AudioManager.FX_KEYPRESS_DELETE; break; @@ -106,7 +106,7 @@ public final class AudioAndHapticFeedbackManager { } } - private void vibrateInternal(final View viewToPerformHapticFeedbackOn) { + public void performHapticFeedback(final View viewToPerformHapticFeedbackOn) { if (!mSettingsValues.mVibrateOn) { return; } diff --git a/java/src/com/android/inputmethod/latin/BinaryDictionaryGetter.java b/java/src/com/android/inputmethod/latin/BinaryDictionaryGetter.java index fa301b5a6..f9f22ecb4 100644 --- a/java/src/com/android/inputmethod/latin/BinaryDictionaryGetter.java +++ b/java/src/com/android/inputmethod/latin/BinaryDictionaryGetter.java @@ -21,7 +21,7 @@ import android.content.SharedPreferences; import android.content.res.AssetFileDescriptor; import android.util.Log; -import com.android.inputmethod.latin.makedict.BinaryDictInputOutput; +import com.android.inputmethod.latin.makedict.BinaryDictDecoderUtils; import com.android.inputmethod.latin.makedict.FormatSpec; import com.android.inputmethod.latin.utils.CollectionUtils; import com.android.inputmethod.latin.utils.DictionaryInfoUtils; @@ -231,17 +231,17 @@ final public class BinaryDictionaryGetter { try { // Read the version of the file inStream = new FileInputStream(f); - final BinaryDictInputOutput.ByteBufferWrapper buffer = - new BinaryDictInputOutput.ByteBufferWrapper(inStream.getChannel().map( + final BinaryDictDecoderUtils.ByteBufferDictBuffer dictBuffer = + new BinaryDictDecoderUtils.ByteBufferDictBuffer(inStream.getChannel().map( FileChannel.MapMode.READ_ONLY, 0, f.length())); - final int magic = buffer.readInt(); - if (magic != FormatSpec.VERSION_2_MAGIC_NUMBER) { + final int magic = dictBuffer.readInt(); + if (magic != FormatSpec.MAGIC_NUMBER) { return false; } - final int formatVersion = buffer.readInt(); - final int headerSize = buffer.readInt(); + final int formatVersion = dictBuffer.readInt(); + final int headerSize = dictBuffer.readInt(); final HashMap<String, String> options = CollectionUtils.newHashMap(); - BinaryDictInputOutput.populateOptions(buffer, headerSize, options); + BinaryDictDecoderUtils.populateOptions(dictBuffer, headerSize, options); final String version = options.get(VERSION_KEY); if (null == version) { diff --git a/java/src/com/android/inputmethod/latin/ContactsBinaryDictionary.java b/java/src/com/android/inputmethod/latin/ContactsBinaryDictionary.java index c99d0e2ea..67eb7f3dd 100644 --- a/java/src/com/android/inputmethod/latin/ContactsBinaryDictionary.java +++ b/java/src/com/android/inputmethod/latin/ContactsBinaryDictionary.java @@ -70,7 +70,8 @@ public class ContactsBinaryDictionary extends ExpandableBinaryDictionary { private final boolean mUseFirstLastBigrams; public ContactsBinaryDictionary(final Context context, final Locale locale) { - super(context, getFilenameWithLocale(NAME, locale.toString()), Dictionary.TYPE_CONTACTS); + super(context, getFilenameWithLocale(NAME, locale.toString()), Dictionary.TYPE_CONTACTS, + false /* isUpdatable */); mLocale = locale; mUseFirstLastBigrams = useFirstLastBigramsForLocale(locale); registerObserver(context); @@ -208,7 +209,8 @@ public class ContactsBinaryDictionary extends ExpandableBinaryDictionary { false /* isNotAWord */); if (!TextUtils.isEmpty(prevWord)) { if (mUseFirstLastBigrams) { - super.setBigram(prevWord, word, FREQUENCY_FOR_CONTACTS_BIGRAM); + super.addBigram(prevWord, word, FREQUENCY_FOR_CONTACTS_BIGRAM, + 0 /* lastModifiedTime */); } } prevWord = word; diff --git a/java/src/com/android/inputmethod/latin/DictionaryWriter.java b/java/src/com/android/inputmethod/latin/DictionaryWriter.java index 47151bf61..1ececd5c1 100644 --- a/java/src/com/android/inputmethod/latin/DictionaryWriter.java +++ b/java/src/com/android/inputmethod/latin/DictionaryWriter.java @@ -20,10 +20,10 @@ import android.content.Context; import com.android.inputmethod.keyboard.ProximityInfo; import com.android.inputmethod.latin.SuggestedWords.SuggestedWordInfo; -import com.android.inputmethod.latin.makedict.BinaryDictInputOutput; +import com.android.inputmethod.latin.makedict.BinaryDictEncoder; import com.android.inputmethod.latin.makedict.FormatSpec; import com.android.inputmethod.latin.makedict.FusionDictionary; -import com.android.inputmethod.latin.makedict.FusionDictionary.Node; +import com.android.inputmethod.latin.makedict.FusionDictionary.PtNodeArray; import com.android.inputmethod.latin.makedict.FusionDictionary.WeightedString; import com.android.inputmethod.latin.makedict.UnsupportedFormatException; import com.android.inputmethod.latin.utils.CollectionUtils; @@ -51,7 +51,7 @@ public class DictionaryWriter extends AbstractDictionaryWriter { @Override public void clear() { final HashMap<String, String> attributes = CollectionUtils.newHashMap(); - mFusionDictionary = new FusionDictionary(new Node(), + mFusionDictionary = new FusionDictionary(new PtNodeArray(), new FusionDictionary.DictionaryOptions(attributes, false, false)); } @@ -75,7 +75,7 @@ public class DictionaryWriter extends AbstractDictionaryWriter { @Override public void addBigramWords(final String word0, final String word1, final int frequency, - final boolean isValid) { + final boolean isValid, final long lastModifiedTime) { mFusionDictionary.setBigram(word0, word1, frequency); } @@ -87,7 +87,7 @@ public class DictionaryWriter extends AbstractDictionaryWriter { @Override protected void writeBinaryDictionary(final FileOutputStream out) throws IOException, UnsupportedFormatException { - BinaryDictInputOutput.writeDictionaryBinary(out, mFusionDictionary, FORMAT_OPTIONS); + BinaryDictEncoder.writeDictionaryBinary(out, mFusionDictionary, FORMAT_OPTIONS); } @Override diff --git a/java/src/com/android/inputmethod/latin/ExpandableBinaryDictionary.java b/java/src/com/android/inputmethod/latin/ExpandableBinaryDictionary.java index 3f11391ba..37256770a 100644 --- a/java/src/com/android/inputmethod/latin/ExpandableBinaryDictionary.java +++ b/java/src/com/android/inputmethod/latin/ExpandableBinaryDictionary.java @@ -20,6 +20,7 @@ import android.content.Context; import android.os.SystemClock; import android.util.Log; +import com.android.inputmethod.annotations.UsedForTesting; import com.android.inputmethod.keyboard.ProximityInfo; import com.android.inputmethod.latin.SuggestedWords.SuggestedWordInfo; import com.android.inputmethod.latin.utils.CollectionUtils; @@ -78,12 +79,18 @@ abstract public class ExpandableBinaryDictionary extends Dictionary { */ private final String mFilename; + /** Whether to support dynamically updating the dictionary */ + private final boolean mIsUpdatable; + /** Controls access to the shared binary dictionary file across multiple instances. */ private final DictionaryController mSharedDictionaryController; /** Controls access to the local binary dictionary for this instance. */ private final DictionaryController mLocalDictionaryController = new DictionaryController(); + /* A extension for a binary dictionary file. */ + public static final String DICT_FILE_EXTENSION = ".dict"; + /** * Abstract method for loading the unigrams and bigrams of a given dictionary in a background * thread. @@ -110,6 +117,16 @@ abstract public class ExpandableBinaryDictionary extends Dictionary { return controller; } + private static AbstractDictionaryWriter getDictionaryWriter(final Context context, + final String dictType, final boolean isUpdatable) { + if (isUpdatable) { + // TODO: Employ dynamically updatable DictionaryWriter. + return new DictionaryWriter(context, dictType); + } else { + return new DictionaryWriter(context, dictType); + } + } + /** * Creates a new expandable binary dictionary. * @@ -117,19 +134,22 @@ abstract public class ExpandableBinaryDictionary extends Dictionary { * @param filename The filename for this binary dictionary. Multiple dictionaries with the same * filename is supported. * @param dictType the dictionary type, as a human-readable string + * @param isUpdatable whether to support dynamically updating the dictionary. Please note that + * dynamic dictionary has negative effects on memory space and computation time. */ - public ExpandableBinaryDictionary( - final Context context, final String filename, final String dictType) { + public ExpandableBinaryDictionary(final Context context, final String filename, + final String dictType, final boolean isUpdatable) { super(dictType); mFilename = filename; mContext = context; + mIsUpdatable = isUpdatable; mBinaryDictionary = null; mSharedDictionaryController = getSharedDictionaryController(filename); - mDictionaryWriter = new DictionaryWriter(context, dictType); + mDictionaryWriter = getDictionaryWriter(context, dictType, isUpdatable); } protected static String getFilenameWithLocale(final String name, final String localeStr) { - return name + "." + localeStr + ".dict"; + return name + "." + localeStr + DICT_FILE_EXTENSION; } /** @@ -137,6 +157,16 @@ abstract public class ExpandableBinaryDictionary extends Dictionary { */ @Override public void close() { + closeBinaryDictionary(); + mLocalDictionaryController.writeLock().lock(); + try { + mDictionaryWriter.close(); + } finally { + mLocalDictionaryController.writeLock().unlock(); + } + } + + protected void closeBinaryDictionary() { // Ensure that no other threads are accessing the local binary dictionary. mLocalDictionaryController.writeLock().lock(); try { @@ -144,7 +174,6 @@ abstract public class ExpandableBinaryDictionary extends Dictionary { mBinaryDictionary.close(); mBinaryDictionary = null; } - mDictionaryWriter.close(); } finally { mLocalDictionaryController.writeLock().unlock(); } @@ -159,35 +188,70 @@ abstract public class ExpandableBinaryDictionary extends Dictionary { } /** - * Sets a word bigram in the dictionary. Used for loading a dictionary. + * Adds a word bigram in the dictionary. Used for loading a dictionary. */ - protected void setBigram(final String prevWord, final String word, final int frequency) { - mDictionaryWriter.addBigramWords(prevWord, word, frequency, true /* isValid */); + protected void addBigram(final String prevWord, final String word, final int frequency, + final long lastModifiedTime) { + mDictionaryWriter.addBigramWords(prevWord, word, frequency, true /* isValid */, + lastModifiedTime); } /** - * Dynamically adds a word unigram to the dictionary. + * Dynamically adds a word unigram to the dictionary. May overwrite an existing entry. */ protected void addWordDynamically(final String word, final String shortcutTarget, final int frequency, final boolean isNotAWord) { - mLocalDictionaryController.writeLock().lock(); - try { - mDictionaryWriter.addUnigramWord(word, shortcutTarget, frequency, isNotAWord); - } finally { - mLocalDictionaryController.writeLock().unlock(); + if (!mIsUpdatable) { + Log.w(TAG, "addWordDynamically is called for non-updatable dictionary: " + mFilename); + return; + } + // TODO: Use a queue to reflect what needs to be reflected. + if (mLocalDictionaryController.writeLock().tryLock()) { + try { + mDictionaryWriter.addUnigramWord(word, shortcutTarget, frequency, isNotAWord); + } finally { + mLocalDictionaryController.writeLock().unlock(); + } } } /** - * Dynamically sets a word bigram in the dictionary. + * Dynamically adds a word bigram in the dictionary. May overwrite an existing entry. */ - protected void setBigramDynamically(final String prevWord, final String word, - final int frequency) { - mLocalDictionaryController.writeLock().lock(); - try { - mDictionaryWriter.addBigramWords(prevWord, word, frequency, true /* isValid */); - } finally { - mLocalDictionaryController.writeLock().unlock(); + protected void addBigramDynamically(final String word0, final String word1, + final int frequency, final boolean isValid) { + if (!mIsUpdatable) { + Log.w(TAG, "addBigramDynamically is called for non-updatable dictionary: " + + mFilename); + return; + } + // TODO: Use a queue to reflect what needs to be reflected. + if (mLocalDictionaryController.writeLock().tryLock()) { + try { + mDictionaryWriter.addBigramWords(word0, word1, frequency, isValid, + 0 /* lastTouchedTime */); + } finally { + mLocalDictionaryController.writeLock().unlock(); + } + } + } + + /** + * Dynamically remove a word bigram in the dictionary. + */ + protected void removeBigramDynamically(final String word0, final String word1) { + if (!mIsUpdatable) { + Log.w(TAG, "removeBigramDynamically is called for non-updatable dictionary: " + + mFilename); + return; + } + // TODO: Use a queue to reflect what needs to be reflected. + if (mLocalDictionaryController.writeLock().tryLock()) { + try { + mDictionaryWriter.removeBigramWords(word0, word1); + } finally { + mLocalDictionaryController.writeLock().unlock(); + } } } @@ -277,7 +341,7 @@ abstract public class ExpandableBinaryDictionary extends Dictionary { // Build the new binary dictionary final BinaryDictionary newBinaryDictionary = new BinaryDictionary(filename, 0, length, - true /* useFullEditDistance */, null, mDictType, false /* isUpdatable */); + true /* useFullEditDistance */, null, mDictType, mIsUpdatable); if (mBinaryDictionary != null) { // Ensure all threads accessing the current dictionary have finished before swapping in @@ -302,9 +366,9 @@ abstract public class ExpandableBinaryDictionary extends Dictionary { abstract protected boolean needsToReloadBeforeWriting(); /** - * Generates and writes a new binary dictionary based on the contents of the fusion dictionary. + * Writes a new binary dictionary based on the contents of the fusion dictionary. */ - private void generateBinaryDictionary() { + private void writeBinaryDictionary() { if (DEBUG) { Log.d(TAG, "Generating binary dictionary: " + mFilename + " request=" + mSharedDictionaryController.mLastUpdateRequestTime + " update=" @@ -367,41 +431,47 @@ abstract public class ExpandableBinaryDictionary extends Dictionary { private final void syncReloadDictionaryInternal() { // Ensure that only one thread attempts to read or write to the shared binary dictionary // file at the same time. - mSharedDictionaryController.writeLock().lock(); + mLocalDictionaryController.writeLock().lock(); try { - final long time = SystemClock.uptimeMillis(); - final boolean dictionaryFileExists = dictionaryFileExists(); - if (mSharedDictionaryController.isOutOfDate() || !dictionaryFileExists) { - // If the shared dictionary file does not exist or is out of date, the first - // instance that acquires the lock will generate a new one. - if (hasContentChanged() || !dictionaryFileExists) { - // If the source content has changed or the dictionary does not exist, rebuild - // the binary dictionary. Empty dictionaries are supported (in the case where - // loadDictionaryAsync() adds nothing) in order to provide a uniform framework. + mSharedDictionaryController.writeLock().lock(); + try { + final long time = SystemClock.uptimeMillis(); + final boolean dictionaryFileExists = dictionaryFileExists(); + if (mSharedDictionaryController.isOutOfDate() || !dictionaryFileExists) { + // If the shared dictionary file does not exist or is out of date, the first + // instance that acquires the lock will generate a new one. + if (hasContentChanged() || !dictionaryFileExists) { + // If the source content has changed or the dictionary does not exist, + // rebuild the binary dictionary. Empty dictionaries are supported (in the + // case where loadDictionaryAsync() adds nothing) in order to provide a + // uniform framework. + mSharedDictionaryController.mLastUpdateTime = time; + writeBinaryDictionary(); + loadBinaryDictionary(); + } else { + // If not, the reload request was unnecessary so revert + // LastUpdateRequestTime to LastUpdateTime. + mSharedDictionaryController.mLastUpdateRequestTime = + mSharedDictionaryController.mLastUpdateTime; + } + } else if (mBinaryDictionary == null || mLocalDictionaryController.mLastUpdateTime + < mSharedDictionaryController.mLastUpdateTime) { + // Otherwise, if the local dictionary is older than the shared dictionary, load + // the shared dictionary. + loadBinaryDictionary(); + } + if (mBinaryDictionary != null && !mBinaryDictionary.isValidDictionary()) { + // Binary dictionary is not valid. Regenerate the dictionary file. mSharedDictionaryController.mLastUpdateTime = time; - generateBinaryDictionary(); + writeBinaryDictionary(); loadBinaryDictionary(); - } else { - // If not, the reload request was unnecessary so revert LastUpdateRequestTime - // to LastUpdateTime. - mSharedDictionaryController.mLastUpdateRequestTime = - mSharedDictionaryController.mLastUpdateTime; } - } else if (mBinaryDictionary == null || mLocalDictionaryController.mLastUpdateTime - < mSharedDictionaryController.mLastUpdateTime) { - // Otherwise, if the local dictionary is older than the shared dictionary, load the - // shared dictionary. - loadBinaryDictionary(); - } - if (mBinaryDictionary != null && !mBinaryDictionary.isValidDictionary()) { - // Binary dictionary is not valid. Regenerate the dictionary file. - mSharedDictionaryController.mLastUpdateTime = time; - generateBinaryDictionary(); - loadBinaryDictionary(); + mLocalDictionaryController.mLastUpdateTime = time; + } finally { + mSharedDictionaryController.writeLock().unlock(); } - mLocalDictionaryController.mLastUpdateTime = time; } finally { - mSharedDictionaryController.writeLock().unlock(); + mLocalDictionaryController.writeLock().unlock(); } } @@ -434,4 +504,45 @@ abstract public class ExpandableBinaryDictionary extends Dictionary { return (mLastUpdateRequestTime > mLastUpdateTime); } } + + /** + * Dynamically adds a word unigram to the dictionary for testing with blocking-lock. + */ + @UsedForTesting + protected void addWordDynamicallyForTests(final String word, final String shortcutTarget, + final int frequency, final boolean isNotAWord) { + mLocalDictionaryController.writeLock().lock(); + try { + addWordDynamically(word, shortcutTarget, frequency, isNotAWord); + } finally { + mLocalDictionaryController.writeLock().unlock(); + } + } + + /** + * Dynamically adds a word bigram in the dictionary for testing with blocking-lock. + */ + @UsedForTesting + protected void addBigramDynamicallyForTests(final String word0, final String word1, + final int frequency, final boolean isValid) { + mLocalDictionaryController.writeLock().lock(); + try { + addBigramDynamically(word0, word1, frequency, isValid); + } finally { + mLocalDictionaryController.writeLock().unlock(); + } + } + + /** + * Dynamically remove a word bigram in the dictionary for testing with blocking-lock. + */ + @UsedForTesting + protected void removeBigramDynamicallyForTests(final String word0, final String word1) { + mLocalDictionaryController.writeLock().lock(); + try { + removeBigramDynamically(word0, word1); + } finally { + mLocalDictionaryController.writeLock().unlock(); + } + } } diff --git a/java/src/com/android/inputmethod/latin/LatinIME.java b/java/src/com/android/inputmethod/latin/LatinIME.java index 5c5b7b7c0..ee7478ca2 100644 --- a/java/src/com/android/inputmethod/latin/LatinIME.java +++ b/java/src/com/android/inputmethod/latin/LatinIME.java @@ -31,7 +31,6 @@ import android.content.SharedPreferences; import android.content.pm.PackageInfo; import android.content.res.Configuration; import android.content.res.Resources; -import android.graphics.Rect; import android.inputmethodservice.InputMethodService; import android.media.AudioManager; import android.net.ConnectivityManager; @@ -51,7 +50,6 @@ import android.util.Printer; import android.view.KeyCharacterMap; import android.view.KeyEvent; import android.view.View; -import android.view.ViewGroup.LayoutParams; import android.view.Window; import android.view.WindowManager; import android.view.inputmethod.CompletionInfo; @@ -76,6 +74,7 @@ import com.android.inputmethod.keyboard.MainKeyboardView; import com.android.inputmethod.latin.SuggestedWords.SuggestedWordInfo; import com.android.inputmethod.latin.define.ProductionFlag; import com.android.inputmethod.latin.personalization.PersonalizationDictionaryHelper; +import com.android.inputmethod.latin.personalization.PersonalizationDictionarySessionRegister; import com.android.inputmethod.latin.personalization.PersonalizationPredictionDictionary; import com.android.inputmethod.latin.personalization.UserHistoryPredictionDictionary; import com.android.inputmethod.latin.settings.Settings; @@ -123,6 +122,8 @@ public class LatinIME extends InputMethodService implements KeyboardActionListen private static final int PENDING_IMS_CALLBACK_DURATION = 800; + private static final int PERIOD_FOR_AUDIO_AND_HAPTIC_FEEDBACK_IN_KEY_REPEAT = 2; + /** * The name of the scheme used by the Package Manager to warn of a new package installation, * replacement or removal. @@ -150,8 +151,8 @@ public class LatinIME extends InputMethodService implements KeyboardActionListen private final Settings mSettings; - private View mExtractArea; - private View mKeyPreviewBackingView; + private View mInputView; + private int mInputViewMinHeight; private SuggestionStripView mSuggestionStripView; // Never null private SuggestedWords mSuggestedWords = SuggestedWords.EMPTY; @@ -472,6 +473,7 @@ public class LatinIME extends InputMethodService implements KeyboardActionListen KeyboardSwitcher.init(this); AudioAndHapticFeedbackManager.init(this); AccessibilityUtils.init(this); + PersonalizationDictionarySessionRegister.init(this); super.onCreate(); @@ -652,6 +654,7 @@ public class LatinIME extends InputMethodService implements KeyboardActionListen mOptionsDialog.dismiss(); } } + PersonalizationDictionarySessionRegister.onConfigurationChanged(this, conf); super.onConfigurationChanged(conf); } @@ -660,17 +663,25 @@ public class LatinIME extends InputMethodService implements KeyboardActionListen return mKeyboardSwitcher.onCreateInputView(mIsHardwareAcceleratedDrawingEnabled); } + private void setInputViewMinHeight(final int minHeight) { + if (mInputView != null && mInputViewMinHeight != minHeight) { + mInputView.setMinimumHeight(minHeight); + mInputViewMinHeight = minHeight; + } + } + @Override - public void setInputView(final View view) { - super.setInputView(view); - mExtractArea = getWindow().getWindow().getDecorView() - .findViewById(android.R.id.extractArea); - mKeyPreviewBackingView = view.findViewById(R.id.key_preview_backing); - mSuggestionStripView = (SuggestionStripView)view.findViewById(R.id.suggestion_strip_view); - if (mSuggestionStripView != null) - mSuggestionStripView.setListener(this, view); + public void setInputView(final View inputView) { + super.setInputView(inputView); + mInputView = inputView; + setInputViewMinHeight(0); + mSuggestionStripView = (SuggestionStripView)inputView.findViewById( + R.id.suggestion_strip_view); + if (mSuggestionStripView != null) { + mSuggestionStripView.setListener(this, inputView); + } if (LatinImeLogger.sVISUALDEBUG) { - mKeyPreviewBackingView.setBackgroundColor(0x10FF0000); + inputView.setBackgroundColor(0x10FF0000); } } @@ -1122,6 +1133,11 @@ public class LatinIME extends InputMethodService implements KeyboardActionListen mSuggestionStripView.setVisibility( shouldShowSuggestions ? View.VISIBLE : View.INVISIBLE); } + if (shouldShowSuggestions && mainKeyboardView != null) { + final int remainingHeight = getWindow().getWindow().getDecorView().getHeight() + - mainKeyboardView.getHeight() - mSuggestionStripView.getHeight(); + mSuggestionStripView.setMoreSuggestionsHeight(remainingHeight); + } } } @@ -1129,31 +1145,6 @@ public class LatinIME extends InputMethodService implements KeyboardActionListen setSuggestionStripShownInternal(shown, /* needsInputViewShown */true); } - private int getAdjustedBackingViewHeight() { - final int currentHeight = mKeyPreviewBackingView.getHeight(); - if (currentHeight > 0) { - return currentHeight; - } - - final MainKeyboardView mainKeyboardView = mKeyboardSwitcher.getMainKeyboardView(); - if (mainKeyboardView == null) { - return 0; - } - final int keyboardHeight = mainKeyboardView.getHeight(); - final int suggestionsHeight = mSuggestionStripView.getHeight(); - final int displayHeight = getResources().getDisplayMetrics().heightPixels; - final Rect rect = new Rect(); - mKeyPreviewBackingView.getWindowVisibleDisplayFrame(rect); - final int notificationBarHeight = rect.top; - final int remainingHeight = displayHeight - notificationBarHeight - suggestionsHeight - - keyboardHeight; - - final LayoutParams params = mKeyPreviewBackingView.getLayoutParams(); - params.height = mSuggestionStripView.setMoreSuggestionsHeight(remainingHeight); - mKeyPreviewBackingView.setLayoutParams(params); - return params.height; - } - @Override public void onComputeInsets(final InputMethodService.Insets outInsets) { super.onComputeInsets(outInsets); @@ -1161,32 +1152,30 @@ public class LatinIME extends InputMethodService implements KeyboardActionListen if (mainKeyboardView == null || mSuggestionStripView == null) { return; } - final int adjustedBackingHeight = getAdjustedBackingViewHeight(); - final boolean backingGone = (mKeyPreviewBackingView.getVisibility() == View.GONE); - final int backingHeight = backingGone ? 0 : adjustedBackingHeight; - // In fullscreen mode, the height of the extract area managed by InputMethodService should - // be considered. - // See {@link android.inputmethodservice.InputMethodService#onComputeInsets}. - final int extractHeight = isFullscreenMode() ? mExtractArea.getHeight() : 0; - final int suggestionsHeight = (mSuggestionStripView.getVisibility() == View.GONE) ? 0 - : mSuggestionStripView.getHeight(); - final int extraHeight = extractHeight + backingHeight + suggestionsHeight; - int visibleTopY = extraHeight; + // This method is never called when in fullscreen mode. + // The contentTop is the top coordinate of the keyboard. The application behind will be + // resized/panned above this coordibnate to be able to show an input field. + final int contentTop = mInputView.getHeight() - mainKeyboardView.getHeight(); + final int suggestionsHeight = (mSuggestionStripView.getVisibility() == View.VISIBLE) + ? mSuggestionStripView.getHeight() : 0; + // The visibleTop is the top coordinates of the visible part of this IME. The application + // behind will never be resized, but may be panned or scrolled. + final int visibleTop = mainKeyboardView.isShowingMoreKeysPanel() ? 0 + : contentTop - suggestionsHeight; + outInsets.contentTopInsets = contentTop; + outInsets.visibleTopInsets = visibleTop; // Need to set touchable region only if input view is being shown if (mainKeyboardView.isShown()) { - if (mSuggestionStripView.getVisibility() == View.VISIBLE) { - visibleTopY -= suggestionsHeight; - } - final int touchY = mainKeyboardView.isShowingMoreKeysPanel() ? 0 : visibleTopY; - final int touchWidth = mainKeyboardView.getWidth(); - final int touchHeight = mainKeyboardView.getHeight() + extraHeight + final int touchLeft = 0; + final int touchTop = visibleTop; + final int touchRight = touchLeft + mainKeyboardView.getWidth(); + final int touchBottom = contentTop + mainKeyboardView.getHeight() // Extend touchable region below the keyboard. + EXTENDED_TOUCHABLE_REGION_HEIGHT; + // The touch event on touchableRegion will be delivered to this IME. + outInsets.touchableRegion.set(touchLeft, touchTop, touchRight, touchBottom); outInsets.touchableInsets = InputMethodService.Insets.TOUCHABLE_INSETS_REGION; - outInsets.touchableRegion.set(0, touchY, touchWidth, touchHeight); } - outInsets.contentTopInsets = visibleTopY; - outInsets.visibleTopInsets = visibleTopY; } @Override @@ -1209,11 +1198,11 @@ public class LatinIME extends InputMethodService implements KeyboardActionListen @Override public void updateFullscreenMode() { super.updateFullscreenMode(); - - if (mKeyPreviewBackingView == null) return; - // In fullscreen mode, no need to have extra space to show the key preview. - // If not, we should have extra space above the keyboard to show the key preview. - mKeyPreviewBackingView.setVisibility(isFullscreenMode() ? View.GONE : View.VISIBLE); + if (!isFullscreenMode()) { + // Expand the input view to cover entire display to be able to show key previews and + // more suggestions view that may be displayed above the keyboard. + setInputViewMinHeight(getResources().getDisplayMetrics().heightPixels); + } } // This will reset the whole input state to the starting state. It will clear @@ -1368,10 +1357,6 @@ public class LatinIME extends InputMethodService implements KeyboardActionListen } } - private static boolean isAlphabet(final int code) { - return Character.isLetter(code); - } - private void onSettingsKeyPressed() { if (isShowingOptionDialog()) return; showSubtypeSelectorAndSettings(); @@ -1479,7 +1464,7 @@ public class LatinIME extends InputMethodService implements KeyboardActionListen break; case Constants.CODE_SHIFT: // Note: Calling back to the keyboard on Shift key is handled in - // {@link #onPressKey(int,boolean)} and {@link #onReleaseKey(int,boolean)}. + // {@link #onPressKey(int,int,boolean)} and {@link #onReleaseKey(int,boolean)}. final Keyboard currentKeyboard = switcher.getKeyboard(); if (null != currentKeyboard && currentKeyboard.mId.isAlphabetKeyboard()) { // TODO: Instead of checking for alphabetic keyboard here, separate keycodes for @@ -1493,7 +1478,7 @@ public class LatinIME extends InputMethodService implements KeyboardActionListen break; case Constants.CODE_SWITCH_ALPHA_SYMBOL: // Note: Calling back to the keyboard on symbol key is handled in - // {@link #onPressKey(int,boolean)} and {@link #onReleaseKey(int,boolean)}. + // {@link #onPressKey(int,int,boolean)} and {@link #onReleaseKey(int,boolean)}. break; case Constants.CODE_SETTINGS: onSettingsKeyPressed(); @@ -1861,23 +1846,23 @@ public class LatinIME extends InputMethodService implements KeyboardActionListen // When we exit this if-clause, mWordComposer.isComposingWord() will return false. } if (mWordComposer.isComposingWord()) { - final int length = mWordComposer.size(); - if (length > 0) { - if (mWordComposer.isBatchMode()) { - if (ProductionFlag.USES_DEVELOPMENT_ONLY_DIAGNOSTICS) { - final String word = mWordComposer.getTypedWord(); - ResearchLogger.latinIME_handleBackspace_batch(word, 1); - } - final String rejectedSuggestion = mWordComposer.getTypedWord(); - mWordComposer.reset(); - mWordComposer.setRejectedBatchModeSuggestion(rejectedSuggestion); - } else { - mWordComposer.deleteLast(); + if (mWordComposer.isBatchMode()) { + if (ProductionFlag.USES_DEVELOPMENT_ONLY_DIAGNOSTICS) { + final String word = mWordComposer.getTypedWord(); + ResearchLogger.latinIME_handleBackspace_batch(word, 1); } - mConnection.setComposingText(getTextWithUnderline(mWordComposer.getTypedWord()), 1); - mHandler.postUpdateSuggestionStrip(); + final String rejectedSuggestion = mWordComposer.getTypedWord(); + mWordComposer.reset(); + mWordComposer.setRejectedBatchModeSuggestion(rejectedSuggestion); } else { - mConnection.deleteSurroundingText(1, 0); + mWordComposer.deleteLast(); + } + mConnection.setComposingText(getTextWithUnderline(mWordComposer.getTypedWord()), 1); + mHandler.postUpdateSuggestionStrip(); + if (!mWordComposer.isComposingWord()) { + // If we just removed the last character, auto-caps mode may have changed so we + // need to re-evaluate. + mKeyboardSwitcher.updateShiftState(); } } else { final SettingsValues currentSettings = mSettings.getCurrent(); @@ -1892,8 +1877,7 @@ public class LatinIME extends InputMethodService implements KeyboardActionListen // Cancel multi-character input: remove the text we just entered. // This is triggered on backspace after a key that inputs multiple characters, // like the smiley key or the .com key. - final int length = mEnteredText.length(); - mConnection.deleteSurroundingText(length, 0); + mConnection.deleteSurroundingText(mEnteredText.length(), 0); if (ProductionFlag.USES_DEVELOPMENT_ONLY_DIAGNOSTICS) { ResearchLogger.latinIME_handleBackspace_cancelTextInput(mEnteredText); } @@ -1939,6 +1923,8 @@ public class LatinIME extends InputMethodService implements KeyboardActionListen // This should never happen. Log.e(TAG, "Backspace when we don't know the selection position"); } + final int lengthToDelete = Character.isSupplementaryCodePoint( + mConnection.getCodePointBeforeCursor()) ? 2 : 1; if (mAppWorkAroundsUtils.isBeforeJellyBean()) { // Backward compatibility mode. Before Jelly bean, the keyboard would simulate // a hardware keyboard event on pressing enter or delete. This is bad for many @@ -1946,22 +1932,28 @@ public class LatinIME extends InputMethodService implements KeyboardActionListen // relying on this behavior so we continue to support it for older apps. sendDownUpKeyEventForBackwardCompatibility(KeyEvent.KEYCODE_DEL); } else { - mConnection.deleteSurroundingText(1, 0); + mConnection.deleteSurroundingText(lengthToDelete, 0); } if (ProductionFlag.USES_DEVELOPMENT_ONLY_DIAGNOSTICS) { - ResearchLogger.latinIME_handleBackspace(1, true /* shouldUncommitLogUnit */); + ResearchLogger.latinIME_handleBackspace(lengthToDelete, + true /* shouldUncommitLogUnit */); } if (mDeleteCount > DELETE_ACCELERATE_AT) { - mConnection.deleteSurroundingText(1, 0); + final int lengthToDeleteAgain = Character.isSupplementaryCodePoint( + mConnection.getCodePointBeforeCursor()) ? 2 : 1; + mConnection.deleteSurroundingText(lengthToDeleteAgain, 0); if (ProductionFlag.USES_DEVELOPMENT_ONLY_DIAGNOSTICS) { - ResearchLogger.latinIME_handleBackspace(1, + ResearchLogger.latinIME_handleBackspace(lengthToDeleteAgain, true /* shouldUncommitLogUnit */); } } } - if (currentSettings.isSuggestionsRequested(mDisplayOrientation)) { + if (currentSettings.isSuggestionsRequested(mDisplayOrientation) + && currentSettings.mCurrentLanguageHasSpaces) { restartSuggestionsOnWordBeforeCursorIfAtEndOfWord(); } + // We just removed a character. We need to update the auto-caps state. + mKeyboardSwitcher.updateShiftState(); } } @@ -1986,6 +1978,9 @@ public class LatinIME extends InputMethodService implements KeyboardActionListen private void handleCharacter(final int primaryCode, final int x, final int y, final int spaceState) { + // TODO: refactor this method to stop flipping isComposingWord around all the time, and + // make it shorter (possibly cut into several pieces). Also factor handleNonSpecialCharacter + // which has the same name as other handle* methods but is not the same. boolean isComposingWord = mWordComposer.isComposingWord(); // TODO: remove isWordConnector() and use isUsuallyFollowedBySpace() instead. @@ -2005,13 +2000,20 @@ public class LatinIME extends InputMethodService implements KeyboardActionListen resetEntireInputState(mLastSelectionStart); isComposingWord = false; } - // NOTE: isCursorTouchingWord() is a blocking IPC call, so it often takes several - // dozen milliseconds. Avoid calling it as much as possible, since we are on the UI - // thread here. - if (!isComposingWord && (isAlphabet(primaryCode) - || currentSettings.isWordConnector(primaryCode)) + // We want to find out whether to start composing a new word with this character. If so, + // we need to reset the composing state and switch isComposingWord. The order of the + // tests is important for good performance. + // We only start composing if we're not already composing. + if (!isComposingWord + // We only start composing if this is a word code point. Essentially that means it's a + // a letter or a word connector. + && currentSettings.isWordCodePoint(primaryCode) + // We never go into composing state if suggestions are not requested. && currentSettings.isSuggestionsRequested(mDisplayOrientation) && - !mConnection.isCursorTouchingWord(currentSettings)) { + // In languages with spaces, we only start composing a word when we are not already + // touching a word. In languages without spaces, the above conditions are sufficient. + (!mConnection.isCursorTouchingWord(currentSettings) + || !currentSettings.mCurrentLanguageHasSpaces)) { // Reset entirely the composing state anyway, then start composing a new word unless // the character is a single quote. The idea here is, single quote is not a // separator and it should be treated as a normal character, except in the first @@ -2099,16 +2101,20 @@ public class LatinIME extends InputMethodService implements KeyboardActionListen private boolean handleSeparator(final int primaryCode, final int x, final int y, final int spaceState) { boolean didAutoCorrect = false; + final SettingsValues currentSettings = mSettings.getCurrent(); + // We avoid sending spaces in languages without spaces if we were composing. + final boolean shouldAvoidSendingCode = Constants.CODE_SPACE == primaryCode + && !currentSettings.mCurrentLanguageHasSpaces && mWordComposer.isComposingWord(); if (mWordComposer.isCursorFrontOrMiddleOfComposingWord()) { // If we are in the middle of a recorrection, we need to commit the recorrection // first so that we can insert the separator at the current cursor position. resetEntireInputState(mLastSelectionStart); } - final SettingsValues currentSettings = mSettings.getCurrent(); - if (mWordComposer.isComposingWord()) { + if (mWordComposer.isComposingWord()) { // May have changed since we stored wasComposing if (currentSettings.mCorrectionEnabled) { - // TODO: maybe cache Strings in an <String> sparse array or something - commitCurrentAutoCorrection(new String(new int[]{primaryCode}, 0, 1)); + final String separator = shouldAvoidSendingCode ? LastComposedWord.NOT_A_SEPARATOR + : new String(new int[] { primaryCode }, 0, 1); + commitCurrentAutoCorrection(separator); didAutoCorrect = true; } else { commitTyped(new String(new int[]{primaryCode}, 0, 1)); @@ -2125,7 +2131,10 @@ public class LatinIME extends InputMethodService implements KeyboardActionListen if (ProductionFlag.USES_DEVELOPMENT_ONLY_DIAGNOSTICS) { ResearchLogger.latinIME_handleSeparator(primaryCode, mWordComposer.isComposingWord()); } - sendKeyCodePoint(primaryCode); + + if (!shouldAvoidSendingCode) { + sendKeyCodePoint(primaryCode); + } if (Constants.CODE_SPACE == primaryCode) { if (currentSettings.isSuggestionsRequested(mDisplayOrientation)) { @@ -2270,11 +2279,17 @@ public class LatinIME extends InputMethodService implements KeyboardActionListen // Get the word on which we should search the bigrams. If we are composing a word, it's // whatever is *before* the half-committed word in the buffer, hence 2; if we aren't, we // should just skip whitespace if any, so 1. - // TODO: this is slow (2-way IPC) - we should probably cache this instead. final SettingsValues currentSettings = mSettings.getCurrent(); - final String prevWord = - mConnection.getNthPreviousWord(currentSettings.mWordSeparators, - mWordComposer.isComposingWord() ? 2 : 1); + final String prevWord; + if (currentSettings.mCurrentLanguageHasSpaces) { + // If we are typing in a language with spaces we can just look up the previous + // word from textview. + prevWord = mConnection.getNthPreviousWord(currentSettings.mWordSeparators, + mWordComposer.isComposingWord() ? 2 : 1); + } else { + prevWord = LastComposedWord.NOT_A_COMPOSED_WORD == mLastComposedWord ? null + : mLastComposedWord.mCommittedWord; + } return suggest.getSuggestedWords(mWordComposer, prevWord, keyboard.getProximityInfo(), currentSettings.mBlockPotentiallyOffensive, currentSettings.mCorrectionEnabled, sessionId); @@ -2544,6 +2559,9 @@ public class LatinIME extends InputMethodService implements KeyboardActionListen // recorrection. This is a temporary, stopgap measure that will be removed later. // TODO: remove this. if (mAppWorkAroundsUtils.isBrokenByRecorrection()) return; + // Recorrection is not supported in languages without spaces because we don't know + // how to segment them yet. + if (!mSettings.getCurrent().mCurrentLanguageHasSpaces) return; // If the cursor is not touching a word, or if there is a selection, return right away. if (mLastSelectionStart != mLastSelectionEnd) return; // If we don't know the cursor location, return. @@ -2571,8 +2589,8 @@ public class LatinIME extends InputMethodService implements KeyboardActionListen } } mWordComposer.setComposingWord(typedWord, mKeyboardSwitcher.getKeyboard()); - // TODO: this is in chars but the callee expects code points! - mWordComposer.setCursorPositionWithinWord(numberOfCharsInWordBeforeCursor); + mWordComposer.setCursorPositionWithinWord( + typedWord.codePointCount(0, numberOfCharsInWordBeforeCursor)); mConnection.setComposingRegion( mLastSelectionStart - numberOfCharsInWordBeforeCursor, mLastSelectionEnd + range.getNumberOfCharsInWordAfterCursor()); @@ -2666,7 +2684,18 @@ public class LatinIME extends InputMethodService implements KeyboardActionListen if (!TextUtils.isEmpty(previousWord) && !TextUtils.isEmpty(committedWord)) { mUserHistoryPredictionDictionary.cancelAddingUserHistory(previousWord, committedWord); } - mConnection.commitText(originallyTypedWord + mLastComposedWord.mSeparatorString, 1); + final String stringToCommit = originallyTypedWord + mLastComposedWord.mSeparatorString; + if (mSettings.getCurrent().mCurrentLanguageHasSpaces) { + // For languages with spaces, we revert to the typed string, but the cursor is still + // after the separator so we don't resume suggestions. If the user wants to correct + // the word, they have to press backspace again. + mConnection.commitText(stringToCommit, 1); + } else { + // For languages without spaces, we revert the typed string but the cursor is flush + // with the typed word, so we need to resume suggestions right away. + mWordComposer.setComposingWord(stringToCommit, mKeyboardSwitcher.getKeyboard()); + mConnection.setComposingText(stringToCommit, 1); + } if (mSettings.isInternal()) { LatinImeLoggerUtils.onSeparator(mLastComposedWord.mSeparatorString, Constants.NOT_A_COORDINATE, Constants.NOT_A_COORDINATE); @@ -2684,7 +2713,9 @@ public class LatinIME extends InputMethodService implements KeyboardActionListen // This essentially inserts a space, and that's it. public void promotePhantomSpace() { - if (mSettings.getCurrent().shouldInsertSpacesAutomatically() + final SettingsValues currentSettings = mSettings.getCurrent(); + if (currentSettings.shouldInsertSpacesAutomatically() + && currentSettings.mCurrentLanguageHasSpaces && !mConnection.textBeforeCursorLooksLikeURL()) { if (ProductionFlag.USES_DEVELOPMENT_ONLY_DIAGNOSTICS) { ResearchLogger.latinIME_promotePhantomSpace(); @@ -2710,30 +2741,43 @@ public class LatinIME extends InputMethodService implements KeyboardActionListen } } - private void hapticAndAudioFeedback(final int code, final boolean isRepeatKey) { + private void hapticAndAudioFeedback(final int code, final int repeatCount) { final MainKeyboardView keyboardView = mKeyboardSwitcher.getMainKeyboardView(); if (keyboardView != null && keyboardView.isInSlidingKeyInput()) { // No need to feedback while sliding input. return; } - if (isRepeatKey && code == Constants.CODE_DELETE && !mConnection.canDeleteCharacters()) { - // No need to feedback when repeating delete key will have no effect. - return; + if (repeatCount > 0) { + if (code == Constants.CODE_DELETE && !mConnection.canDeleteCharacters()) { + // No need to feedback when repeat delete key will have no effect. + return; + } + // TODO: Use event time that the last feedback has been generated instead of relying on + // a repeat count to thin out feedback. + if (repeatCount % PERIOD_FOR_AUDIO_AND_HAPTIC_FEEDBACK_IN_KEY_REPEAT == 0) { + return; + } + } + final AudioAndHapticFeedbackManager feedbackManager = + AudioAndHapticFeedbackManager.getInstance(); + if (repeatCount == 0) { + // TODO: Reconsider how to perform haptic feedback when repeating key. + feedbackManager.performHapticFeedback(keyboardView); } - AudioAndHapticFeedbackManager.getInstance().hapticAndAudioFeedback(code, keyboardView); + feedbackManager.performAudioFeedback(code); } // Callback of the {@link KeyboardActionListener}. This is called when a key is depressed; // release matching call is {@link #onReleaseKey(int,boolean)} below. @Override - public void onPressKey(final int primaryCode, final boolean isRepeatKey, + public void onPressKey(final int primaryCode, final int repeatCount, final boolean isSinglePointer) { mKeyboardSwitcher.onPressKey(primaryCode, isSinglePointer); - hapticAndAudioFeedback(primaryCode, isRepeatKey); + hapticAndAudioFeedback(primaryCode, repeatCount); } // Callback of the {@link KeyboardActionListener}. This is called when a key is released; - // press matching call is {@link #onPressKey(int,boolean,boolean)} above. + // press matching call is {@link #onPressKey(int,int,boolean)} above. @Override public void onReleaseKey(final int primaryCode, final boolean withSliding) { mKeyboardSwitcher.onReleaseKey(primaryCode, withSliding); @@ -2749,17 +2793,6 @@ public class LatinIME extends InputMethodService implements KeyboardActionListen break; } } - - if (Constants.CODE_DELETE == primaryCode) { - // This is a stopgap solution to avoid leaving a high surrogate alone in a text view. - // In the future, we need to deprecate deteleSurroundingText() and have a surrogate - // pair-friendly way of deleting characters in InputConnection. - // TODO: use getCodePointBeforeCursor instead to improve performance - final CharSequence lastChar = mConnection.getTextBeforeCursor(1, 0); - if (!TextUtils.isEmpty(lastChar) && Character.isHighSurrogate(lastChar.charAt(0))) { - mConnection.deleteSurroundingText(1, 0); - } - } } // Hooks for hardware keyboard @@ -2895,6 +2928,12 @@ public class LatinIME extends InputMethodService implements KeyboardActionListen return mSuggest.hasMainDictionary(); } + // DO NOT USE THIS for any other purpose than testing. This can break the keyboard badly. + @UsedForTesting + /* package for test */ void replaceMainDictionaryForTest(final Locale locale) { + mSuggest.resetMainDict(this, locale, null); + } + public void debugDumpStateAndCrashWithException(final String context) { final StringBuilder s = new StringBuilder(mAppWorkAroundsUtils.toString()); s.append("\nAttributes : ").append(mSettings.getCurrent().mInputAttributes) diff --git a/java/src/com/android/inputmethod/latin/Suggest.java b/java/src/com/android/inputmethod/latin/Suggest.java index c2fdcb552..55b70c6f8 100644 --- a/java/src/com/android/inputmethod/latin/Suggest.java +++ b/java/src/com/android/inputmethod/latin/Suggest.java @@ -107,7 +107,7 @@ public final class Suggest { } private void addOrReplaceDictionaryInternal(final String key, final Dictionary dict) { - if (mOnlyDictionarySetForDebug != null && mOnlyDictionarySetForDebug.contains(key)) { + if (mOnlyDictionarySetForDebug != null && !mOnlyDictionarySetForDebug.contains(key)) { Log.w(TAG, "Ignore add " + key + " dictionary for debug."); return; } diff --git a/java/src/com/android/inputmethod/latin/UserBinaryDictionary.java b/java/src/com/android/inputmethod/latin/UserBinaryDictionary.java index ed6fefae4..b2bb61596 100644 --- a/java/src/com/android/inputmethod/latin/UserBinaryDictionary.java +++ b/java/src/com/android/inputmethod/latin/UserBinaryDictionary.java @@ -75,7 +75,8 @@ public class UserBinaryDictionary extends ExpandableBinaryDictionary { public UserBinaryDictionary(final Context context, final String locale, final boolean alsoUseMoreRestrictiveLocales) { - super(context, getFilenameWithLocale(NAME, locale), Dictionary.TYPE_USER); + super(context, getFilenameWithLocale(NAME, locale), Dictionary.TYPE_USER, + false /* isUpdatable */); if (null == locale) throw new NullPointerException(); // Catch the error earlier if (SubtypeLocaleUtils.NO_LANGUAGE.equals(locale)) { // If we don't have a locale, insert into the "all locales" user dictionary. diff --git a/java/src/com/android/inputmethod/latin/WordComposer.java b/java/src/com/android/inputmethod/latin/WordComposer.java index a09ca605c..039dadc66 100644 --- a/java/src/com/android/inputmethod/latin/WordComposer.java +++ b/java/src/com/android/inputmethod/latin/WordComposer.java @@ -272,8 +272,8 @@ public final class WordComposer { final int x, y; final Key key; if (keyboard != null && (key = keyboard.getKey(codePoint)) != null) { - x = key.mX + key.mWidth / 2; - y = key.mY + key.mHeight / 2; + x = key.getX() + key.getWidth() / 2; + y = key.getY() + key.getHeight() / 2; } else { x = Constants.NOT_A_COORDINATE; y = Constants.NOT_A_COORDINATE; diff --git a/java/src/com/android/inputmethod/latin/makedict/BinaryDictDecoder.java b/java/src/com/android/inputmethod/latin/makedict/BinaryDictDecoder.java new file mode 100644 index 000000000..b86dfe552 --- /dev/null +++ b/java/src/com/android/inputmethod/latin/makedict/BinaryDictDecoder.java @@ -0,0 +1,189 @@ +/* + * Copyright (C) 2013 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.makedict; + +import com.android.inputmethod.annotations.UsedForTesting; +import com.android.inputmethod.latin.makedict.BinaryDictDecoderUtils.CharEncoding; +import com.android.inputmethod.latin.makedict.BinaryDictDecoderUtils.DictBuffer; +import com.android.inputmethod.latin.makedict.decoder.HeaderReader; +import com.android.inputmethod.latin.utils.ByteArrayDictBuffer; +import com.android.inputmethod.latin.utils.JniUtils; + +import java.io.File; +import java.io.FileInputStream; +import java.io.FileNotFoundException; +import java.io.IOException; +import java.io.RandomAccessFile; +import java.nio.ByteBuffer; +import java.nio.channels.FileChannel; +import java.util.HashMap; + +@UsedForTesting +public class BinaryDictDecoder implements HeaderReader { + + static { + JniUtils.loadNativeLibrary(); + } + + // TODO: implement something sensical instead of just a phony method + private static native int doNothing(); + + public interface DictionaryBufferFactory { + public DictBuffer getDictionaryBuffer(final File file) + throws FileNotFoundException, IOException; + } + + /** + * Creates DictionaryBuffer using a ByteBuffer + * + * This class uses less memory than DictionaryBufferFromByteArrayFactory, + * but doesn't perform as fast. + * When operating on a big dictionary, this class is preferred. + */ + public static final class DictionaryBufferFromReadOnlyByteBufferFactory + implements DictionaryBufferFactory { + @Override + public DictBuffer getDictionaryBuffer(final File file) + throws FileNotFoundException, IOException { + FileInputStream inStream = null; + ByteBuffer buffer = null; + try { + inStream = new FileInputStream(file); + buffer = inStream.getChannel().map(FileChannel.MapMode.READ_ONLY, + 0, file.length()); + } finally { + if (inStream != null) { + inStream.close(); + } + } + if (buffer != null) { + return new BinaryDictDecoderUtils.ByteBufferDictBuffer(buffer); + } + return null; + } + } + + /** + * Creates DictionaryBuffer using a byte array + * + * This class performs faster than other classes, but consumes more memory. + * When operating on a small dictionary, this class is preferred. + */ + public static final class DictionaryBufferFromByteArrayFactory + implements DictionaryBufferFactory { + @Override + public DictBuffer getDictionaryBuffer(final File file) + throws FileNotFoundException, IOException { + FileInputStream inStream = null; + try { + inStream = new FileInputStream(file); + final byte[] array = new byte[(int) file.length()]; + inStream.read(array); + return new ByteArrayDictBuffer(array); + } finally { + if (inStream != null) { + inStream.close(); + } + } + } + } + + /** + * Creates DictionaryBuffer using a writable ByteBuffer and a RandomAccessFile. + * + * This class doesn't perform as fast as other classes, + * but this class is the only option available for destructive operations (insert or delete) + * on a dictionary. + */ + @UsedForTesting + public static final class DictionaryBufferFromWritableByteBufferFactory + implements DictionaryBufferFactory { + @Override + public DictBuffer getDictionaryBuffer(final File file) + throws FileNotFoundException, IOException { + RandomAccessFile raFile = null; + ByteBuffer buffer = null; + try { + raFile = new RandomAccessFile(file, "rw"); + buffer = raFile.getChannel().map(FileChannel.MapMode.READ_WRITE, 0, file.length()); + } finally { + if (raFile != null) { + raFile.close(); + } + } + if (buffer != null) { + return new BinaryDictDecoderUtils.ByteBufferDictBuffer(buffer); + } + return null; + } + } + + private final File mDictionaryBinaryFile; + private DictBuffer mDictBuffer; + + public BinaryDictDecoder(final File file) { + mDictionaryBinaryFile = file; + mDictBuffer = null; + } + + public void openDictBuffer(final DictionaryBufferFactory factory) + throws FileNotFoundException, IOException { + mDictBuffer = factory.getDictionaryBuffer(mDictionaryBinaryFile); + } + + public DictBuffer getDictBuffer() { + return mDictBuffer; + } + + @UsedForTesting + public DictBuffer openAndGetDictBuffer( + final DictionaryBufferFactory factory) + throws FileNotFoundException, IOException { + openDictBuffer(factory); + return getDictBuffer(); + } + + // The implementation of HeaderReader + @Override + public int readVersion() throws IOException, UnsupportedFormatException { + return BinaryDictDecoderUtils.checkFormatVersion(mDictBuffer); + } + + @Override + public int readOptionFlags() { + return mDictBuffer.readUnsignedShort(); + } + + @Override + public int readHeaderSize() { + return mDictBuffer.readInt(); + } + + @Override + public HashMap<String, String> readAttributes(final int headerSize) { + final HashMap<String, String> attributes = new HashMap<String, String>(); + while (mDictBuffer.position() < headerSize) { + // We can avoid infinite loop here since mFusionDictonary.position() is always increased + // by calling CharEncoding.readString. + final String key = CharEncoding.readString(mDictBuffer); + final String value = CharEncoding.readString(mDictBuffer); + attributes.put(key, value); + } + mDictBuffer.position(headerSize); + return attributes; + } +} diff --git a/java/src/com/android/inputmethod/latin/makedict/BinaryDictDecoderUtils.java b/java/src/com/android/inputmethod/latin/makedict/BinaryDictDecoderUtils.java new file mode 100644 index 000000000..efa491099 --- /dev/null +++ b/java/src/com/android/inputmethod/latin/makedict/BinaryDictDecoderUtils.java @@ -0,0 +1,777 @@ +/* + * Copyright (C) 2013 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.makedict; + +import com.android.inputmethod.annotations.UsedForTesting; +import com.android.inputmethod.latin.makedict.FormatSpec.FileHeader; +import com.android.inputmethod.latin.makedict.FormatSpec.FormatOptions; +import com.android.inputmethod.latin.makedict.FusionDictionary.CharGroup; +import com.android.inputmethod.latin.makedict.FusionDictionary.PtNodeArray; +import com.android.inputmethod.latin.makedict.FusionDictionary.WeightedString; +import com.android.inputmethod.latin.makedict.decoder.HeaderReader; + +import java.io.ByteArrayOutputStream; +import java.io.File; +import java.io.FileInputStream; +import java.io.FileNotFoundException; +import java.io.IOException; +import java.nio.ByteBuffer; +import java.nio.channels.FileChannel; +import java.util.ArrayList; +import java.util.Arrays; +import java.util.HashMap; +import java.util.Map; +import java.util.TreeMap; + +/** + * Decodes binary files for a FusionDictionary. + * + * All the methods in this class are static. + * + * TODO: Remove calls from classes except BinaryDictDecoder + * TODO: Move this file to makedict/internal. + */ +public final class BinaryDictDecoderUtils { + + private static final boolean DBG = MakedictLog.DBG; + + private BinaryDictDecoderUtils() { + // This utility class is not publicly instantiable. + } + + private static final int MAX_JUMPS = 12; + + @UsedForTesting + public interface DictBuffer { + public int readUnsignedByte(); + public int readUnsignedShort(); + public int readUnsignedInt24(); + public int readInt(); + public int position(); + public void position(int newPosition); + public void put(final byte b); + public int limit(); + @UsedForTesting + public int capacity(); + } + + public static final class ByteBufferDictBuffer implements DictBuffer { + private ByteBuffer mBuffer; + + public ByteBufferDictBuffer(final ByteBuffer buffer) { + mBuffer = buffer; + } + + @Override + public int readUnsignedByte() { + return mBuffer.get() & 0xFF; + } + + @Override + public int readUnsignedShort() { + return mBuffer.getShort() & 0xFFFF; + } + + @Override + public int readUnsignedInt24() { + final int retval = readUnsignedByte(); + return (retval << 16) + readUnsignedShort(); + } + + @Override + public int readInt() { + return mBuffer.getInt(); + } + + @Override + public int position() { + return mBuffer.position(); + } + + @Override + public void position(int newPos) { + mBuffer.position(newPos); + } + + @Override + public void put(final byte b) { + mBuffer.put(b); + } + + @Override + public int limit() { + return mBuffer.limit(); + } + + @Override + public int capacity() { + return mBuffer.capacity(); + } + } + + /** + * A class grouping utility function for our specific character encoding. + */ + static final class CharEncoding { + private static final int MINIMAL_ONE_BYTE_CHARACTER_VALUE = 0x20; + private static final int MAXIMAL_ONE_BYTE_CHARACTER_VALUE = 0xFF; + + /** + * Helper method to find out whether this code fits on one byte + */ + private static boolean fitsOnOneByte(final int character) { + return character >= MINIMAL_ONE_BYTE_CHARACTER_VALUE + && character <= MAXIMAL_ONE_BYTE_CHARACTER_VALUE; + } + + /** + * Compute the size of a character given its character code. + * + * Char format is: + * 1 byte = bbbbbbbb match + * case 000xxxxx: xxxxx << 16 + next byte << 8 + next byte + * else: if 00011111 (= 0x1F) : this is the terminator. This is a relevant choice because + * unicode code points range from 0 to 0x10FFFF, so any 3-byte value starting with + * 00011111 would be outside unicode. + * else: iso-latin-1 code + * This allows for the whole unicode range to be encoded, including chars outside of + * the BMP. Also everything in the iso-latin-1 charset is only 1 byte, except control + * characters which should never happen anyway (and still work, but take 3 bytes). + * + * @param character the character code. + * @return the size in binary encoded-form, either 1 or 3 bytes. + */ + static int getCharSize(final int character) { + // See char encoding in FusionDictionary.java + if (fitsOnOneByte(character)) return 1; + if (FormatSpec.INVALID_CHARACTER == character) return 1; + return 3; + } + + /** + * Compute the byte size of a character array. + */ + static int getCharArraySize(final int[] chars) { + int size = 0; + for (int character : chars) size += getCharSize(character); + return size; + } + + /** + * Writes a char array to a byte buffer. + * + * @param codePoints the code point array to write. + * @param buffer the byte buffer to write to. + * @param index the index in buffer to write the character array to. + * @return the index after the last character. + */ + static int writeCharArray(final int[] codePoints, final byte[] buffer, int index) { + for (int codePoint : codePoints) { + if (1 == getCharSize(codePoint)) { + buffer[index++] = (byte)codePoint; + } else { + buffer[index++] = (byte)(0xFF & (codePoint >> 16)); + buffer[index++] = (byte)(0xFF & (codePoint >> 8)); + buffer[index++] = (byte)(0xFF & codePoint); + } + } + return index; + } + + /** + * Writes a string with our character format to a byte buffer. + * + * This will also write the terminator byte. + * + * @param buffer the byte buffer to write to. + * @param origin the offset to write from. + * @param word the string to write. + * @return the size written, in bytes. + */ + static int writeString(final byte[] buffer, final int origin, + final String word) { + final int length = word.length(); + int index = origin; + for (int i = 0; i < length; i = word.offsetByCodePoints(i, 1)) { + final int codePoint = word.codePointAt(i); + if (1 == getCharSize(codePoint)) { + buffer[index++] = (byte)codePoint; + } else { + buffer[index++] = (byte)(0xFF & (codePoint >> 16)); + buffer[index++] = (byte)(0xFF & (codePoint >> 8)); + buffer[index++] = (byte)(0xFF & codePoint); + } + } + buffer[index++] = FormatSpec.GROUP_CHARACTERS_TERMINATOR; + return index - origin; + } + + /** + * Writes a string with our character format to a ByteArrayOutputStream. + * + * This will also write the terminator byte. + * + * @param buffer the ByteArrayOutputStream to write to. + * @param word the string to write. + */ + static void writeString(final ByteArrayOutputStream buffer, final String word) { + final int length = word.length(); + for (int i = 0; i < length; i = word.offsetByCodePoints(i, 1)) { + final int codePoint = word.codePointAt(i); + if (1 == getCharSize(codePoint)) { + buffer.write((byte) codePoint); + } else { + buffer.write((byte) (0xFF & (codePoint >> 16))); + buffer.write((byte) (0xFF & (codePoint >> 8))); + buffer.write((byte) (0xFF & codePoint)); + } + } + buffer.write(FormatSpec.GROUP_CHARACTERS_TERMINATOR); + } + + /** + * Reads a string from a DictBuffer. This is the converse of the above method. + */ + static String readString(final DictBuffer dictBuffer) { + final StringBuilder s = new StringBuilder(); + int character = readChar(dictBuffer); + while (character != FormatSpec.INVALID_CHARACTER) { + s.appendCodePoint(character); + character = readChar(dictBuffer); + } + return s.toString(); + } + + /** + * Reads a character from the buffer. + * + * This follows the character format documented earlier in this source file. + * + * @param dictBuffer the buffer, positioned over an encoded character. + * @return the character code. + */ + static int readChar(final DictBuffer dictBuffer) { + int character = dictBuffer.readUnsignedByte(); + if (!fitsOnOneByte(character)) { + if (FormatSpec.GROUP_CHARACTERS_TERMINATOR == character) { + return FormatSpec.INVALID_CHARACTER; + } + character <<= 16; + character += dictBuffer.readUnsignedShort(); + } + return character; + } + } + + // Input methods: Read a binary dictionary to memory. + // readDictionaryBinary is the public entry point for them. + + static int readChildrenAddress(final DictBuffer dictBuffer, + final int optionFlags, final FormatOptions options) { + if (options.mSupportsDynamicUpdate) { + final int address = dictBuffer.readUnsignedInt24(); + if (address == 0) return FormatSpec.NO_CHILDREN_ADDRESS; + if ((address & FormatSpec.MSB24) != 0) { + return -(address & FormatSpec.SINT24_MAX); + } else { + return address; + } + } + int address; + switch (optionFlags & FormatSpec.MASK_GROUP_ADDRESS_TYPE) { + case FormatSpec.FLAG_GROUP_ADDRESS_TYPE_ONEBYTE: + return dictBuffer.readUnsignedByte(); + case FormatSpec.FLAG_GROUP_ADDRESS_TYPE_TWOBYTES: + return dictBuffer.readUnsignedShort(); + case FormatSpec.FLAG_GROUP_ADDRESS_TYPE_THREEBYTES: + return dictBuffer.readUnsignedInt24(); + case FormatSpec.FLAG_GROUP_ADDRESS_TYPE_NOADDRESS: + default: + return FormatSpec.NO_CHILDREN_ADDRESS; + } + } + + static int readParentAddress(final DictBuffer dictBuffer, + final FormatOptions formatOptions) { + if (BinaryDictIOUtils.supportsDynamicUpdate(formatOptions)) { + final int parentAddress = dictBuffer.readUnsignedInt24(); + final int sign = ((parentAddress & FormatSpec.MSB24) != 0) ? -1 : 1; + return sign * (parentAddress & FormatSpec.SINT24_MAX); + } else { + return FormatSpec.NO_PARENT_ADDRESS; + } + } + + private static final int[] CHARACTER_BUFFER = new int[FormatSpec.MAX_WORD_LENGTH]; + public static CharGroupInfo readCharGroup(final DictBuffer dictBuffer, + final int originalGroupAddress, final FormatOptions options) { + int addressPointer = originalGroupAddress; + final int flags = dictBuffer.readUnsignedByte(); + ++addressPointer; + + final int parentAddress = readParentAddress(dictBuffer, options); + if (BinaryDictIOUtils.supportsDynamicUpdate(options)) { + addressPointer += 3; + } + + final int characters[]; + if (0 != (flags & FormatSpec.FLAG_HAS_MULTIPLE_CHARS)) { + int index = 0; + int character = CharEncoding.readChar(dictBuffer); + addressPointer += CharEncoding.getCharSize(character); + while (-1 != character) { + // FusionDictionary is making sure that the length of the word is smaller than + // MAX_WORD_LENGTH. + // So we'll never write past the end of CHARACTER_BUFFER. + CHARACTER_BUFFER[index++] = character; + character = CharEncoding.readChar(dictBuffer); + addressPointer += CharEncoding.getCharSize(character); + } + characters = Arrays.copyOfRange(CHARACTER_BUFFER, 0, index); + } else { + final int character = CharEncoding.readChar(dictBuffer); + addressPointer += CharEncoding.getCharSize(character); + characters = new int[] { character }; + } + final int frequency; + if (0 != (FormatSpec.FLAG_IS_TERMINAL & flags)) { + ++addressPointer; + frequency = dictBuffer.readUnsignedByte(); + } else { + frequency = CharGroup.NOT_A_TERMINAL; + } + int childrenAddress = readChildrenAddress(dictBuffer, flags, options); + if (childrenAddress != FormatSpec.NO_CHILDREN_ADDRESS) { + childrenAddress += addressPointer; + } + addressPointer += BinaryDictIOUtils.getChildrenAddressSize(flags, options); + ArrayList<WeightedString> shortcutTargets = null; + if (0 != (flags & FormatSpec.FLAG_HAS_SHORTCUT_TARGETS)) { + final int pointerBefore = dictBuffer.position(); + shortcutTargets = new ArrayList<WeightedString>(); + dictBuffer.readUnsignedShort(); // Skip the size + while (true) { + final int targetFlags = dictBuffer.readUnsignedByte(); + final String word = CharEncoding.readString(dictBuffer); + shortcutTargets.add(new WeightedString(word, + targetFlags & FormatSpec.FLAG_ATTRIBUTE_FREQUENCY)); + if (0 == (targetFlags & FormatSpec.FLAG_ATTRIBUTE_HAS_NEXT)) break; + } + addressPointer += dictBuffer.position() - pointerBefore; + } + ArrayList<PendingAttribute> bigrams = null; + if (0 != (flags & FormatSpec.FLAG_HAS_BIGRAMS)) { + bigrams = new ArrayList<PendingAttribute>(); + int bigramCount = 0; + while (bigramCount++ < FormatSpec.MAX_BIGRAMS_IN_A_GROUP) { + final int bigramFlags = dictBuffer.readUnsignedByte(); + ++addressPointer; + final int sign = 0 == (bigramFlags & FormatSpec.FLAG_ATTRIBUTE_OFFSET_NEGATIVE) + ? 1 : -1; + int bigramAddress = addressPointer; + switch (bigramFlags & FormatSpec.MASK_ATTRIBUTE_ADDRESS_TYPE) { + case FormatSpec.FLAG_ATTRIBUTE_ADDRESS_TYPE_ONEBYTE: + bigramAddress += sign * dictBuffer.readUnsignedByte(); + addressPointer += 1; + break; + case FormatSpec.FLAG_ATTRIBUTE_ADDRESS_TYPE_TWOBYTES: + bigramAddress += sign * dictBuffer.readUnsignedShort(); + addressPointer += 2; + break; + case FormatSpec.FLAG_ATTRIBUTE_ADDRESS_TYPE_THREEBYTES: + final int offset = (dictBuffer.readUnsignedByte() << 16) + + dictBuffer.readUnsignedShort(); + bigramAddress += sign * offset; + addressPointer += 3; + break; + default: + throw new RuntimeException("Has bigrams with no address"); + } + bigrams.add(new PendingAttribute(bigramFlags & FormatSpec.FLAG_ATTRIBUTE_FREQUENCY, + bigramAddress)); + if (0 == (bigramFlags & FormatSpec.FLAG_ATTRIBUTE_HAS_NEXT)) break; + } + if (bigramCount >= FormatSpec.MAX_BIGRAMS_IN_A_GROUP) { + MakedictLog.d("too many bigrams in a group."); + } + } + return new CharGroupInfo(originalGroupAddress, addressPointer, flags, characters, frequency, + parentAddress, childrenAddress, shortcutTargets, bigrams); + } + + /** + * Reads and returns the char group count out of a buffer and forwards the pointer. + */ + public static int readCharGroupCount(final DictBuffer dictBuffer) { + final int msb = dictBuffer.readUnsignedByte(); + if (FormatSpec.MAX_CHARGROUPS_FOR_ONE_BYTE_CHARGROUP_COUNT >= msb) { + return msb; + } else { + return ((FormatSpec.MAX_CHARGROUPS_FOR_ONE_BYTE_CHARGROUP_COUNT & msb) << 8) + + dictBuffer.readUnsignedByte(); + } + } + + /** + * Finds, as a string, the word at the address passed as an argument. + * + * @param dictBuffer the buffer to read from. + * @param headerSize the size of the header. + * @param address the address to seek. + * @param formatOptions file format options. + * @return the word with its frequency, as a weighted string. + */ + /* package for tests */ static WeightedString getWordAtAddress( + final DictBuffer dictBuffer, final int headerSize, final int address, + final FormatOptions formatOptions) { + final WeightedString result; + final int originalPointer = dictBuffer.position(); + dictBuffer.position(address); + + if (BinaryDictIOUtils.supportsDynamicUpdate(formatOptions)) { + result = getWordAtAddressWithParentAddress(dictBuffer, headerSize, address, + formatOptions); + } else { + result = getWordAtAddressWithoutParentAddress(dictBuffer, headerSize, address, + formatOptions); + } + + dictBuffer.position(originalPointer); + return result; + } + + @SuppressWarnings("unused") + private static WeightedString getWordAtAddressWithParentAddress( + final DictBuffer dictBuffer, final int headerSize, final int address, + final FormatOptions options) { + int currentAddress = address; + int frequency = Integer.MIN_VALUE; + final StringBuilder builder = new StringBuilder(); + // the length of the path from the root to the leaf is limited by MAX_WORD_LENGTH + for (int count = 0; count < FormatSpec.MAX_WORD_LENGTH; ++count) { + CharGroupInfo currentInfo; + int loopCounter = 0; + do { + dictBuffer.position(currentAddress + headerSize); + currentInfo = readCharGroup(dictBuffer, currentAddress, options); + if (BinaryDictIOUtils.isMovedGroup(currentInfo.mFlags, options)) { + currentAddress = currentInfo.mParentAddress + currentInfo.mOriginalAddress; + } + if (DBG && loopCounter++ > MAX_JUMPS) { + MakedictLog.d("Too many jumps - probably a bug"); + } + } while (BinaryDictIOUtils.isMovedGroup(currentInfo.mFlags, options)); + if (Integer.MIN_VALUE == frequency) frequency = currentInfo.mFrequency; + builder.insert(0, + new String(currentInfo.mCharacters, 0, currentInfo.mCharacters.length)); + if (currentInfo.mParentAddress == FormatSpec.NO_PARENT_ADDRESS) break; + currentAddress = currentInfo.mParentAddress + currentInfo.mOriginalAddress; + } + return new WeightedString(builder.toString(), frequency); + } + + private static WeightedString getWordAtAddressWithoutParentAddress( + final DictBuffer dictBuffer, final int headerSize, final int address, + final FormatOptions options) { + dictBuffer.position(headerSize); + final int count = readCharGroupCount(dictBuffer); + int groupOffset = BinaryDictIOUtils.getGroupCountSize(count); + final StringBuilder builder = new StringBuilder(); + WeightedString result = null; + + CharGroupInfo last = null; + for (int i = count - 1; i >= 0; --i) { + CharGroupInfo info = readCharGroup(dictBuffer, groupOffset, options); + groupOffset = info.mEndAddress; + if (info.mOriginalAddress == address) { + builder.append(new String(info.mCharacters, 0, info.mCharacters.length)); + result = new WeightedString(builder.toString(), info.mFrequency); + break; // and return + } + if (BinaryDictIOUtils.hasChildrenAddress(info.mChildrenAddress)) { + if (info.mChildrenAddress > address) { + if (null == last) continue; + builder.append(new String(last.mCharacters, 0, last.mCharacters.length)); + dictBuffer.position(last.mChildrenAddress + headerSize); + i = readCharGroupCount(dictBuffer); + groupOffset = last.mChildrenAddress + BinaryDictIOUtils.getGroupCountSize(i); + last = null; + continue; + } + last = info; + } + if (0 == i && BinaryDictIOUtils.hasChildrenAddress(last.mChildrenAddress)) { + builder.append(new String(last.mCharacters, 0, last.mCharacters.length)); + dictBuffer.position(last.mChildrenAddress + headerSize); + i = readCharGroupCount(dictBuffer); + groupOffset = last.mChildrenAddress + BinaryDictIOUtils.getGroupCountSize(i); + last = null; + continue; + } + } + return result; + } + + /** + * Reads a single node array from a buffer. + * + * This methods reads the file at the current position. A node array is fully expected to start + * at the current position. + * This will recursively read other node arrays into the structure, populating the reverse + * maps on the fly and using them to keep track of already read nodes. + * + * @param dictBuffer the buffer, correctly positioned at the start of a node array. + * @param headerSize the size, in bytes, of the file header. + * @param reverseNodeArrayMap a mapping from addresses to already read node arrays. + * @param reverseGroupMap a mapping from addresses to already read character groups. + * @param options file format options. + * @return the read node array with all his children already read. + */ + private static PtNodeArray readNodeArray(final DictBuffer dictBuffer, + final int headerSize, final Map<Integer, PtNodeArray> reverseNodeArrayMap, + final Map<Integer, CharGroup> reverseGroupMap, final FormatOptions options) + throws IOException { + final ArrayList<CharGroup> nodeArrayContents = new ArrayList<CharGroup>(); + final int nodeArrayOrigin = dictBuffer.position() - headerSize; + + do { // Scan the linked-list node. + final int nodeArrayHeadPosition = dictBuffer.position() - headerSize; + final int count = readCharGroupCount(dictBuffer); + int groupOffset = nodeArrayHeadPosition + BinaryDictIOUtils.getGroupCountSize(count); + for (int i = count; i > 0; --i) { // Scan the array of CharGroup. + CharGroupInfo info = readCharGroup(dictBuffer, groupOffset, options); + if (BinaryDictIOUtils.isMovedGroup(info.mFlags, options)) continue; + ArrayList<WeightedString> shortcutTargets = info.mShortcutTargets; + ArrayList<WeightedString> bigrams = null; + if (null != info.mBigrams) { + bigrams = new ArrayList<WeightedString>(); + for (PendingAttribute bigram : info.mBigrams) { + final WeightedString word = getWordAtAddress( + dictBuffer, headerSize, bigram.mAddress, options); + final int reconstructedFrequency = + BinaryDictIOUtils.reconstructBigramFrequency(word.mFrequency, + bigram.mFrequency); + bigrams.add(new WeightedString(word.mWord, reconstructedFrequency)); + } + } + if (BinaryDictIOUtils.hasChildrenAddress(info.mChildrenAddress)) { + PtNodeArray children = reverseNodeArrayMap.get(info.mChildrenAddress); + if (null == children) { + final int currentPosition = dictBuffer.position(); + dictBuffer.position(info.mChildrenAddress + headerSize); + children = readNodeArray(dictBuffer, headerSize, reverseNodeArrayMap, + reverseGroupMap, options); + dictBuffer.position(currentPosition); + } + nodeArrayContents.add( + new CharGroup(info.mCharacters, shortcutTargets, bigrams, + info.mFrequency, + 0 != (info.mFlags & FormatSpec.FLAG_IS_NOT_A_WORD), + 0 != (info.mFlags & FormatSpec.FLAG_IS_BLACKLISTED), children)); + } else { + nodeArrayContents.add( + new CharGroup(info.mCharacters, shortcutTargets, bigrams, + info.mFrequency, + 0 != (info.mFlags & FormatSpec.FLAG_IS_NOT_A_WORD), + 0 != (info.mFlags & FormatSpec.FLAG_IS_BLACKLISTED))); + } + groupOffset = info.mEndAddress; + } + + // reach the end of the array. + if (options.mSupportsDynamicUpdate) { + final int nextAddress = dictBuffer.readUnsignedInt24(); + if (nextAddress >= 0 && nextAddress < dictBuffer.limit()) { + dictBuffer.position(nextAddress); + } else { + break; + } + } + } while (options.mSupportsDynamicUpdate && + dictBuffer.position() != FormatSpec.NO_FORWARD_LINK_ADDRESS); + + final PtNodeArray nodeArray = new PtNodeArray(nodeArrayContents); + nodeArray.mCachedAddressBeforeUpdate = nodeArrayOrigin; + nodeArray.mCachedAddressAfterUpdate = nodeArrayOrigin; + reverseNodeArrayMap.put(nodeArray.mCachedAddressAfterUpdate, nodeArray); + return nodeArray; + } + + /** + * Helper function to get the binary format version from the header. + * @throws IOException + */ + private static int getFormatVersion(final DictBuffer dictBuffer) + throws IOException { + final int magic = dictBuffer.readInt(); + if (FormatSpec.MAGIC_NUMBER == magic) return dictBuffer.readUnsignedShort(); + return FormatSpec.NOT_A_VERSION_NUMBER; + } + + /** + * Helper function to get and validate the binary format version. + * @throws UnsupportedFormatException + * @throws IOException + */ + static int checkFormatVersion(final DictBuffer dictBuffer) + throws IOException, UnsupportedFormatException { + final int version = getFormatVersion(dictBuffer); + if (version < FormatSpec.MINIMUM_SUPPORTED_VERSION + || version > FormatSpec.MAXIMUM_SUPPORTED_VERSION) { + throw new UnsupportedFormatException("This file has version " + version + + ", but this implementation does not support versions above " + + FormatSpec.MAXIMUM_SUPPORTED_VERSION); + } + return version; + } + + /** + * Reads a header from a buffer. + * @param headerReader the header reader + * @throws IOException + * @throws UnsupportedFormatException + */ + public static FileHeader readHeader(final HeaderReader headerReader) + throws IOException, UnsupportedFormatException { + final int version = headerReader.readVersion(); + final int optionsFlags = headerReader.readOptionFlags(); + + final int headerSize = headerReader.readHeaderSize(); + + if (headerSize < 0) { + throw new UnsupportedFormatException("header size can't be negative."); + } + + final HashMap<String, String> attributes = headerReader.readAttributes(headerSize); + + final FileHeader header = new FileHeader(headerSize, + new FusionDictionary.DictionaryOptions(attributes, + 0 != (optionsFlags & FormatSpec.GERMAN_UMLAUT_PROCESSING_FLAG), + 0 != (optionsFlags & FormatSpec.FRENCH_LIGATURE_PROCESSING_FLAG)), + new FormatOptions(version, + 0 != (optionsFlags & FormatSpec.SUPPORTS_DYNAMIC_UPDATE))); + return header; + } + + /** + * Reads options from a buffer and populate a map with their contents. + * + * The buffer is read at the current position, so the caller must take care the pointer + * is in the right place before calling this. + */ + public static void populateOptions(final DictBuffer dictBuffer, + final int headerSize, final HashMap<String, String> options) { + while (dictBuffer.position() < headerSize) { + final String key = CharEncoding.readString(dictBuffer); + final String value = CharEncoding.readString(dictBuffer); + options.put(key, value); + } + } + + /** + * Reads a buffer and returns the memory representation of the dictionary. + * + * This high-level method takes a buffer and reads its contents, populating a + * FusionDictionary structure. The optional dict argument is an existing dictionary to + * which words from the buffer should be added. If it is null, a new dictionary is created. + * + * @param dictDecoder the dict decoder. + * @param dict an optional dictionary to add words to, or null. + * @return the created (or merged) dictionary. + */ + @UsedForTesting + public static FusionDictionary readDictionaryBinary(final BinaryDictDecoder dictDecoder, + final FusionDictionary dict) throws FileNotFoundException, IOException, + UnsupportedFormatException { + + // if the buffer has not been opened, open the buffer with bytebuffer. + if (dictDecoder.getDictBuffer() == null) dictDecoder.openDictBuffer( + new BinaryDictDecoder.DictionaryBufferFromReadOnlyByteBufferFactory()); + if (dictDecoder.getDictBuffer() == null) { + MakedictLog.e("Cannot open the buffer"); + } + + // Read header + final FileHeader fileHeader = readHeader(dictDecoder); + + Map<Integer, PtNodeArray> reverseNodeArrayMapping = new TreeMap<Integer, PtNodeArray>(); + Map<Integer, CharGroup> reverseGroupMapping = new TreeMap<Integer, CharGroup>(); + final PtNodeArray root = readNodeArray(dictDecoder.getDictBuffer(), fileHeader.mHeaderSize, + reverseNodeArrayMapping, reverseGroupMapping, fileHeader.mFormatOptions); + + FusionDictionary newDict = new FusionDictionary(root, fileHeader.mDictionaryOptions); + if (null != dict) { + for (final Word w : dict) { + if (w.mIsBlacklistEntry) { + newDict.addBlacklistEntry(w.mWord, w.mShortcutTargets, w.mIsNotAWord); + } else { + newDict.add(w.mWord, w.mFrequency, w.mShortcutTargets, w.mIsNotAWord); + } + } + for (final Word w : dict) { + // By construction a binary dictionary may not have bigrams pointing to + // words that are not also registered as unigrams so we don't have to avoid + // them explicitly here. + for (final WeightedString bigram : w.mBigrams) { + newDict.setBigram(w.mWord, bigram.mWord, bigram.mFrequency); + } + } + } + + return newDict; + } + + /** + * Helper method to pass a file name instead of a File object to isBinaryDictionary. + */ + public static boolean isBinaryDictionary(final String filename) { + final File file = new File(filename); + return isBinaryDictionary(file); + } + + /** + * Basic test to find out whether the file is a binary dictionary or not. + * + * Concretely this only tests the magic number. + * + * @param file The file to test. + * @return true if it's a binary dictionary, false otherwise + */ + public static boolean isBinaryDictionary(final File file) { + FileInputStream inStream = null; + try { + inStream = new FileInputStream(file); + final ByteBuffer buffer = inStream.getChannel().map( + FileChannel.MapMode.READ_ONLY, 0, file.length()); + final int version = getFormatVersion(new ByteBufferDictBuffer(buffer)); + return (version >= FormatSpec.MINIMUM_SUPPORTED_VERSION + && version <= FormatSpec.MAXIMUM_SUPPORTED_VERSION); + } catch (FileNotFoundException e) { + return false; + } catch (IOException e) { + return false; + } finally { + if (inStream != null) { + try { + inStream.close(); + } catch (IOException e) { + // do nothing + } + } + } + } +} diff --git a/java/src/com/android/inputmethod/latin/makedict/BinaryDictEncoder.java b/java/src/com/android/inputmethod/latin/makedict/BinaryDictEncoder.java new file mode 100644 index 000000000..ff11cde39 --- /dev/null +++ b/java/src/com/android/inputmethod/latin/makedict/BinaryDictEncoder.java @@ -0,0 +1,1006 @@ +/* + * Copyright (C) 2013 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.makedict; + +import com.android.inputmethod.latin.makedict.BinaryDictDecoderUtils.CharEncoding; +import com.android.inputmethod.latin.makedict.FormatSpec.FormatOptions; +import com.android.inputmethod.latin.makedict.FusionDictionary.CharGroup; +import com.android.inputmethod.latin.makedict.FusionDictionary.DictionaryOptions; +import com.android.inputmethod.latin.makedict.FusionDictionary.PtNodeArray; +import com.android.inputmethod.latin.makedict.FusionDictionary.WeightedString; + +import java.io.ByteArrayOutputStream; +import java.io.IOException; +import java.io.OutputStream; +import java.util.ArrayList; +import java.util.Iterator; + +/** + * Encodes binary files for a FusionDictionary. + * + * All the methods in this class are static. + */ +public class BinaryDictEncoder { + + private static final boolean DBG = MakedictLog.DBG; + + private BinaryDictEncoder() { + // This utility class is not publicly instantiable. + } + + // Arbitrary limit to how much passes we consider address size compression should + // terminate in. At the time of this writing, our largest dictionary completes + // compression in five passes. + // If the number of passes exceeds this number, makedict bails with an exception on + // suspicion that a bug might be causing an infinite loop. + private static final int MAX_PASSES = 24; + + /** + * Compute the binary size of the character array. + * + * If only one character, this is the size of this character. If many, it's the sum of their + * sizes + 1 byte for the terminator. + * + * @param characters the character array + * @return the size of the char array, including the terminator if any + */ + static int getGroupCharactersSize(final int[] characters) { + int size = CharEncoding.getCharArraySize(characters); + if (characters.length > 1) size += FormatSpec.GROUP_TERMINATOR_SIZE; + return size; + } + + /** + * Compute the binary size of the character array in a group + * + * If only one character, this is the size of this character. If many, it's the sum of their + * sizes + 1 byte for the terminator. + * + * @param group the group + * @return the size of the char array, including the terminator if any + */ + private static int getGroupCharactersSize(final CharGroup group) { + return getGroupCharactersSize(group.mChars); + } + + /** + * Compute the binary size of the group count for a node array. + * @param nodeArray the nodeArray + * @return the size of the group count, either 1 or 2 bytes. + */ + private static int getGroupCountSize(final PtNodeArray nodeArray) { + return BinaryDictIOUtils.getGroupCountSize(nodeArray.mData.size()); + } + + /** + * Compute the size of a shortcut in bytes. + */ + private static int getShortcutSize(final WeightedString shortcut) { + int size = FormatSpec.GROUP_ATTRIBUTE_FLAGS_SIZE; + final String word = shortcut.mWord; + final int length = word.length(); + for (int i = 0; i < length; i = word.offsetByCodePoints(i, 1)) { + final int codePoint = word.codePointAt(i); + size += CharEncoding.getCharSize(codePoint); + } + size += FormatSpec.GROUP_TERMINATOR_SIZE; + return size; + } + + /** + * Compute the size of a shortcut list in bytes. + * + * This is known in advance and does not change according to position in the file + * like address lists do. + */ + static int getShortcutListSize(final ArrayList<WeightedString> shortcutList) { + if (null == shortcutList) return 0; + int size = FormatSpec.GROUP_SHORTCUT_LIST_SIZE_SIZE; + for (final WeightedString shortcut : shortcutList) { + size += getShortcutSize(shortcut); + } + return size; + } + + /** + * Compute the maximum size of a CharGroup, assuming 3-byte addresses for everything. + * + * @param group the CharGroup to compute the size of. + * @param options file format options. + * @return the maximum size of the group. + */ + private static int getCharGroupMaximumSize(final CharGroup group, final FormatOptions options) { + int size = getGroupHeaderSize(group, options); + // If terminal, one byte for the frequency + if (group.isTerminal()) size += FormatSpec.GROUP_FREQUENCY_SIZE; + size += FormatSpec.GROUP_MAX_ADDRESS_SIZE; // For children address + size += getShortcutListSize(group.mShortcutTargets); + if (null != group.mBigrams) { + size += (FormatSpec.GROUP_ATTRIBUTE_FLAGS_SIZE + + FormatSpec.GROUP_ATTRIBUTE_MAX_ADDRESS_SIZE) + * group.mBigrams.size(); + } + return size; + } + + /** + * Compute the maximum size of each node of a node array, assuming 3-byte addresses for + * everything, and caches it in the `mCachedSize' member of the nodes; deduce the size of + * the containing node array, and cache it it its 'mCachedSize' member. + * + * @param nodeArray the node array to compute the maximum size of. + * @param options file format options. + */ + private static void calculateNodeArrayMaximumSize(final PtNodeArray nodeArray, + final FormatOptions options) { + int size = getGroupCountSize(nodeArray); + for (CharGroup g : nodeArray.mData) { + final int groupSize = getCharGroupMaximumSize(g, options); + g.mCachedSize = groupSize; + size += groupSize; + } + if (options.mSupportsDynamicUpdate) { + size += FormatSpec.FORWARD_LINK_ADDRESS_SIZE; + } + nodeArray.mCachedSize = size; + } + + /** + * Compute the size of the header (flag + [parent address] + characters size) of a CharGroup. + * + * @param group the group of which to compute the size of the header + * @param options file format options. + */ + private static int getGroupHeaderSize(final CharGroup group, final FormatOptions options) { + if (BinaryDictIOUtils.supportsDynamicUpdate(options)) { + return FormatSpec.GROUP_FLAGS_SIZE + FormatSpec.PARENT_ADDRESS_SIZE + + getGroupCharactersSize(group); + } else { + return FormatSpec.GROUP_FLAGS_SIZE + getGroupCharactersSize(group); + } + } + + /** + * Compute the size, in bytes, that an address will occupy. + * + * This can be used either for children addresses (which are always positive) or for + * attribute, which may be positive or negative but + * store their sign bit separately. + * + * @param address the address + * @return the byte size. + */ + static int getByteSize(final int address) { + assert(address <= FormatSpec.UINT24_MAX); + if (!BinaryDictIOUtils.hasChildrenAddress(address)) { + return 0; + } else if (Math.abs(address) <= FormatSpec.UINT8_MAX) { + return 1; + } else if (Math.abs(address) <= FormatSpec.UINT16_MAX) { + return 2; + } else { + return 3; + } + } + + // End utility methods + + // This method is responsible for finding a nice ordering of the nodes that favors run-time + // cache performance and dictionary size. + /* package for tests */ static ArrayList<PtNodeArray> flattenTree( + final PtNodeArray rootNodeArray) { + final int treeSize = FusionDictionary.countCharGroups(rootNodeArray); + MakedictLog.i("Counted nodes : " + treeSize); + final ArrayList<PtNodeArray> flatTree = new ArrayList<PtNodeArray>(treeSize); + return flattenTreeInner(flatTree, rootNodeArray); + } + + private static ArrayList<PtNodeArray> flattenTreeInner(final ArrayList<PtNodeArray> list, + final PtNodeArray nodeArray) { + // Removing the node is necessary if the tails are merged, because we would then + // add the same node several times when we only want it once. A number of places in + // the code also depends on any node being only once in the list. + // Merging tails can only be done if there are no attributes. Searching for attributes + // in LatinIME code depends on a total breadth-first ordering, which merging tails + // breaks. If there are no attributes, it should be fine (and reduce the file size) + // to merge tails, and removing the node from the list would be necessary. However, + // we don't merge tails because breaking the breadth-first ordering would result in + // extreme overhead at bigram lookup time (it would make the search function O(n) instead + // of the current O(log(n)), where n=number of nodes in the dictionary which is pretty + // high). + // If no nodes are ever merged, we can't have the same node twice in the list, hence + // searching for duplicates in unnecessary. It is also very performance consuming, + // since `list' is an ArrayList so it's an O(n) operation that runs on all nodes, making + // this simple list.remove operation O(n*n) overall. On Android this overhead is very + // high. + // For future reference, the code to remove duplicate is a simple : list.remove(node); + list.add(nodeArray); + final ArrayList<CharGroup> branches = nodeArray.mData; + final int nodeSize = branches.size(); + for (CharGroup group : branches) { + if (null != group.mChildren) flattenTreeInner(list, group.mChildren); + } + return list; + } + + /** + * Get the offset from a position inside a current node array to a target node array, during + * update. + * + * If the current node array is before the target node array, the target node array has not + * been updated yet, so we should return the offset from the old position of the current node + * array to the old position of the target node array. If on the other hand the target is + * before the current node array, it already has been updated, so we should return the offset + * from the new position in the current node array to the new position in the target node + * array. + * + * @param currentNodeArray node array containing the CharGroup where the offset will be written + * @param offsetFromStartOfCurrentNodeArray offset, in bytes, from the start of currentNodeArray + * @param targetNodeArray the target node array to get the offset to + * @return the offset to the target node array + */ + private static int getOffsetToTargetNodeArrayDuringUpdate(final PtNodeArray currentNodeArray, + final int offsetFromStartOfCurrentNodeArray, final PtNodeArray targetNodeArray) { + final boolean isTargetBeforeCurrent = (targetNodeArray.mCachedAddressBeforeUpdate + < currentNodeArray.mCachedAddressBeforeUpdate); + if (isTargetBeforeCurrent) { + return targetNodeArray.mCachedAddressAfterUpdate + - (currentNodeArray.mCachedAddressAfterUpdate + + offsetFromStartOfCurrentNodeArray); + } else { + return targetNodeArray.mCachedAddressBeforeUpdate + - (currentNodeArray.mCachedAddressBeforeUpdate + + offsetFromStartOfCurrentNodeArray); + } + } + + /** + * Get the offset from a position inside a current node array to a target CharGroup, during + * update. + * + * @param currentNodeArray node array containing the CharGroup where the offset will be written + * @param offsetFromStartOfCurrentNodeArray offset, in bytes, from the start of currentNodeArray + * @param targetCharGroup the target CharGroup to get the offset to + * @return the offset to the target CharGroup + */ + // TODO: is there any way to factorize this method with the one above? + private static int getOffsetToTargetCharGroupDuringUpdate(final PtNodeArray currentNodeArray, + final int offsetFromStartOfCurrentNodeArray, final CharGroup targetCharGroup) { + final int oldOffsetBasePoint = currentNodeArray.mCachedAddressBeforeUpdate + + offsetFromStartOfCurrentNodeArray; + final boolean isTargetBeforeCurrent = (targetCharGroup.mCachedAddressBeforeUpdate + < oldOffsetBasePoint); + // If the target is before the current node array, then its address has already been + // updated. We can use the AfterUpdate member, and compare it to our own member after + // update. Otherwise, the AfterUpdate member is not updated yet, so we need to use the + // BeforeUpdate member, and of course we have to compare this to our own address before + // update. + if (isTargetBeforeCurrent) { + final int newOffsetBasePoint = currentNodeArray.mCachedAddressAfterUpdate + + offsetFromStartOfCurrentNodeArray; + return targetCharGroup.mCachedAddressAfterUpdate - newOffsetBasePoint; + } else { + return targetCharGroup.mCachedAddressBeforeUpdate - oldOffsetBasePoint; + } + } + + /** + * Computes the actual node array size, based on the cached addresses of the children nodes. + * + * Each node array stores its tentative address. During dictionary address computing, these + * are not final, but they can be used to compute the node array size (the node array size + * depends on the address of the children because the number of bytes necessary to store an + * address depends on its numeric value. The return value indicates whether the node array + * contents (as in, any of the addresses stored in the cache fields) have changed with + * respect to their previous value. + * + * @param nodeArray the node array to compute the size of. + * @param dict the dictionary in which the word/attributes are to be found. + * @param formatOptions file format options. + * @return false if none of the cached addresses inside the node array changed, true otherwise. + */ + private static boolean computeActualNodeArraySize(final PtNodeArray nodeArray, + final FusionDictionary dict, final FormatOptions formatOptions) { + boolean changed = false; + int size = getGroupCountSize(nodeArray); + for (CharGroup group : nodeArray.mData) { + group.mCachedAddressAfterUpdate = nodeArray.mCachedAddressAfterUpdate + size; + if (group.mCachedAddressAfterUpdate != group.mCachedAddressBeforeUpdate) { + changed = true; + } + int groupSize = getGroupHeaderSize(group, formatOptions); + if (group.isTerminal()) groupSize += FormatSpec.GROUP_FREQUENCY_SIZE; + if (null == group.mChildren && formatOptions.mSupportsDynamicUpdate) { + groupSize += FormatSpec.SIGNED_CHILDREN_ADDRESS_SIZE; + } else if (null != group.mChildren) { + if (formatOptions.mSupportsDynamicUpdate) { + groupSize += FormatSpec.SIGNED_CHILDREN_ADDRESS_SIZE; + } else { + groupSize += getByteSize(getOffsetToTargetNodeArrayDuringUpdate(nodeArray, + groupSize + size, group.mChildren)); + } + } + groupSize += getShortcutListSize(group.mShortcutTargets); + if (null != group.mBigrams) { + for (WeightedString bigram : group.mBigrams) { + final int offset = getOffsetToTargetCharGroupDuringUpdate(nodeArray, + groupSize + size + FormatSpec.GROUP_FLAGS_SIZE, + FusionDictionary.findWordInTree(dict.mRootNodeArray, bigram.mWord)); + groupSize += getByteSize(offset) + FormatSpec.GROUP_FLAGS_SIZE; + } + } + group.mCachedSize = groupSize; + size += groupSize; + } + if (formatOptions.mSupportsDynamicUpdate) { + size += FormatSpec.FORWARD_LINK_ADDRESS_SIZE; + } + if (nodeArray.mCachedSize != size) { + nodeArray.mCachedSize = size; + changed = true; + } + return changed; + } + + /** + * Initializes the cached addresses of node arrays and their containing nodes from their size. + * + * @param flatNodes the list of node arrays. + * @param formatOptions file format options. + * @return the byte size of the entire stack. + */ + private static int initializeNodeArraysCachedAddresses(final ArrayList<PtNodeArray> flatNodes, + final FormatOptions formatOptions) { + int nodeArrayOffset = 0; + for (final PtNodeArray nodeArray : flatNodes) { + nodeArray.mCachedAddressBeforeUpdate = nodeArrayOffset; + int groupCountSize = getGroupCountSize(nodeArray); + int groupOffset = 0; + for (final CharGroup g : nodeArray.mData) { + g.mCachedAddressBeforeUpdate = g.mCachedAddressAfterUpdate = + groupCountSize + nodeArrayOffset + groupOffset; + groupOffset += g.mCachedSize; + } + final int nodeSize = groupCountSize + groupOffset + + (formatOptions.mSupportsDynamicUpdate + ? FormatSpec.FORWARD_LINK_ADDRESS_SIZE : 0); + nodeArrayOffset += nodeArray.mCachedSize; + } + return nodeArrayOffset; + } + + /** + * Updates the cached addresses of node arrays after recomputing their new positions. + * + * @param flatNodes the list of node arrays. + */ + private static void updateNodeArraysCachedAddresses(final ArrayList<PtNodeArray> flatNodes) { + for (final PtNodeArray nodeArray : flatNodes) { + nodeArray.mCachedAddressBeforeUpdate = nodeArray.mCachedAddressAfterUpdate; + for (final CharGroup g : nodeArray.mData) { + g.mCachedAddressBeforeUpdate = g.mCachedAddressAfterUpdate; + } + } + } + + /** + * Compute the cached parent addresses after all has been updated. + * + * The parent addresses are used by some binary formats at write-to-disk time. Not all formats + * need them. In particular, version 2 does not need them, and version 3 does. + * + * @param flatNodes the flat array of node arrays to fill in + */ + private static void computeParentAddresses(final ArrayList<PtNodeArray> flatNodes) { + for (final PtNodeArray nodeArray : flatNodes) { + for (final CharGroup group : nodeArray.mData) { + if (null != group.mChildren) { + // Assign my address to children's parent address + // Here BeforeUpdate and AfterUpdate addresses have the same value, so it + // does not matter which we use. + group.mChildren.mCachedParentAddress = group.mCachedAddressAfterUpdate + - group.mChildren.mCachedAddressAfterUpdate; + } + } + } + } + + /** + * Compute the addresses and sizes of an ordered list of node arrays. + * + * This method takes a list of node arrays and will update their cached address and size + * values so that they can be written into a file. It determines the smallest size each of the + * nodes arrays can be given the addresses of its children and attributes, and store that into + * each node. + * The order of the node is given by the order of the array. This method makes no effort + * to find a good order; it only mechanically computes the size this order results in. + * + * @param dict the dictionary + * @param flatNodes the ordered list of nodes arrays + * @param formatOptions file format options. + * @return the same array it was passed. The nodes have been updated for address and size. + */ + private static ArrayList<PtNodeArray> computeAddresses(final FusionDictionary dict, + final ArrayList<PtNodeArray> flatNodes, final FormatOptions formatOptions) { + // First get the worst possible sizes and offsets + for (final PtNodeArray n : flatNodes) calculateNodeArrayMaximumSize(n, formatOptions); + final int offset = initializeNodeArraysCachedAddresses(flatNodes, formatOptions); + + MakedictLog.i("Compressing the array addresses. Original size : " + offset); + MakedictLog.i("(Recursively seen size : " + offset + ")"); + + int passes = 0; + boolean changesDone = false; + do { + changesDone = false; + int nodeArrayStartOffset = 0; + for (final PtNodeArray nodeArray : flatNodes) { + nodeArray.mCachedAddressAfterUpdate = nodeArrayStartOffset; + final int oldNodeArraySize = nodeArray.mCachedSize; + final boolean changed = computeActualNodeArraySize(nodeArray, dict, formatOptions); + final int newNodeArraySize = nodeArray.mCachedSize; + if (oldNodeArraySize < newNodeArraySize) { + throw new RuntimeException("Increased size ?!"); + } + nodeArrayStartOffset += newNodeArraySize; + changesDone |= changed; + } + updateNodeArraysCachedAddresses(flatNodes); + ++passes; + if (passes > MAX_PASSES) throw new RuntimeException("Too many passes - probably a bug"); + } while (changesDone); + + if (formatOptions.mSupportsDynamicUpdate) { + computeParentAddresses(flatNodes); + } + final PtNodeArray lastNodeArray = flatNodes.get(flatNodes.size() - 1); + MakedictLog.i("Compression complete in " + passes + " passes."); + MakedictLog.i("After address compression : " + + (lastNodeArray.mCachedAddressAfterUpdate + lastNodeArray.mCachedSize)); + + return flatNodes; + } + + /** + * Sanity-checking method. + * + * This method checks a list of node arrays for juxtaposition, that is, it will do + * nothing if each node array's cached address is actually the previous node array's address + * plus the previous node's size. + * If this is not the case, it will throw an exception. + * + * @param arrays the list of node arrays to check + */ + private static void checkFlatNodeArrayList(final ArrayList<PtNodeArray> arrays) { + int offset = 0; + int index = 0; + for (final PtNodeArray nodeArray : arrays) { + // BeforeUpdate and AfterUpdate addresses are the same here, so it does not matter + // which we use. + if (nodeArray.mCachedAddressAfterUpdate != offset) { + throw new RuntimeException("Wrong address for node " + index + + " : expected " + offset + ", got " + nodeArray.mCachedAddressAfterUpdate); + } + ++index; + offset += nodeArray.mCachedSize; + } + } + + /** + * Helper method to write a variable-size address to a file. + * + * @param buffer the buffer to write to. + * @param index the index in the buffer to write the address to. + * @param address the address to write. + * @return the size in bytes the address actually took. + */ + private static int writeVariableAddress(final byte[] buffer, int index, final int address) { + switch (getByteSize(address)) { + case 1: + buffer[index++] = (byte)address; + return 1; + case 2: + buffer[index++] = (byte)(0xFF & (address >> 8)); + buffer[index++] = (byte)(0xFF & address); + return 2; + case 3: + buffer[index++] = (byte)(0xFF & (address >> 16)); + buffer[index++] = (byte)(0xFF & (address >> 8)); + buffer[index++] = (byte)(0xFF & address); + return 3; + case 0: + return 0; + default: + throw new RuntimeException("Address " + address + " has a strange size"); + } + } + + /** + * Helper method to write a variable-size signed address to a file. + * + * @param buffer the buffer to write to. + * @param index the index in the buffer to write the address to. + * @param address the address to write. + * @return the size in bytes the address actually took. + */ + private static int writeVariableSignedAddress(final byte[] buffer, int index, + final int address) { + if (!BinaryDictIOUtils.hasChildrenAddress(address)) { + buffer[index] = buffer[index + 1] = buffer[index + 2] = 0; + } else { + final int absAddress = Math.abs(address); + buffer[index++] = + (byte)((address < 0 ? FormatSpec.MSB8 : 0) | (0xFF & (absAddress >> 16))); + buffer[index++] = (byte)(0xFF & (absAddress >> 8)); + buffer[index++] = (byte)(0xFF & absAddress); + } + return 3; + } + + /** + * Makes the flag value for a char group. + * + * @param hasMultipleChars whether the group has multiple chars. + * @param isTerminal whether the group is terminal. + * @param childrenAddressSize the size of a children address. + * @param hasShortcuts whether the group has shortcuts. + * @param hasBigrams whether the group has bigrams. + * @param isNotAWord whether the group is not a word. + * @param isBlackListEntry whether the group is a blacklist entry. + * @param formatOptions file format options. + * @return the flags + */ + static int makeCharGroupFlags(final boolean hasMultipleChars, final boolean isTerminal, + final int childrenAddressSize, final boolean hasShortcuts, final boolean hasBigrams, + final boolean isNotAWord, final boolean isBlackListEntry, + final FormatOptions formatOptions) { + byte flags = 0; + if (hasMultipleChars) flags |= FormatSpec.FLAG_HAS_MULTIPLE_CHARS; + if (isTerminal) flags |= FormatSpec.FLAG_IS_TERMINAL; + if (formatOptions.mSupportsDynamicUpdate) { + flags |= FormatSpec.FLAG_IS_NOT_MOVED; + } else if (true) { + switch (childrenAddressSize) { + case 1: + flags |= FormatSpec.FLAG_GROUP_ADDRESS_TYPE_ONEBYTE; + break; + case 2: + flags |= FormatSpec.FLAG_GROUP_ADDRESS_TYPE_TWOBYTES; + break; + case 3: + flags |= FormatSpec.FLAG_GROUP_ADDRESS_TYPE_THREEBYTES; + break; + case 0: + flags |= FormatSpec.FLAG_GROUP_ADDRESS_TYPE_NOADDRESS; + break; + default: + throw new RuntimeException("Node with a strange address"); + } + } + if (hasShortcuts) flags |= FormatSpec.FLAG_HAS_SHORTCUT_TARGETS; + if (hasBigrams) flags |= FormatSpec.FLAG_HAS_BIGRAMS; + if (isNotAWord) flags |= FormatSpec.FLAG_IS_NOT_A_WORD; + if (isBlackListEntry) flags |= FormatSpec.FLAG_IS_BLACKLISTED; + return flags; + } + + private static byte makeCharGroupFlags(final CharGroup group, final int groupAddress, + final int childrenOffset, final FormatOptions formatOptions) { + return (byte) makeCharGroupFlags(group.mChars.length > 1, group.mFrequency >= 0, + getByteSize(childrenOffset), group.mShortcutTargets != null, group.mBigrams != null, + group.mIsNotAWord, group.mIsBlacklistEntry, formatOptions); + } + + /** + * Makes the flag value for a bigram. + * + * @param more whether there are more bigrams after this one. + * @param offset the offset of the bigram. + * @param bigramFrequency the frequency of the bigram, 0..255. + * @param unigramFrequency the unigram frequency of the same word, 0..255. + * @param word the second bigram, for debugging purposes + * @return the flags + */ + private static final int makeBigramFlags(final boolean more, final int offset, + int bigramFrequency, final int unigramFrequency, final String word) { + int bigramFlags = (more ? FormatSpec.FLAG_ATTRIBUTE_HAS_NEXT : 0) + + (offset < 0 ? FormatSpec.FLAG_ATTRIBUTE_OFFSET_NEGATIVE : 0); + switch (getByteSize(offset)) { + case 1: + bigramFlags |= FormatSpec.FLAG_ATTRIBUTE_ADDRESS_TYPE_ONEBYTE; + break; + case 2: + bigramFlags |= FormatSpec.FLAG_ATTRIBUTE_ADDRESS_TYPE_TWOBYTES; + break; + case 3: + bigramFlags |= FormatSpec.FLAG_ATTRIBUTE_ADDRESS_TYPE_THREEBYTES; + break; + default: + throw new RuntimeException("Strange offset size"); + } + if (unigramFrequency > bigramFrequency) { + MakedictLog.e("Unigram freq is superior to bigram freq for \"" + word + + "\". Bigram freq is " + bigramFrequency + ", unigram freq for " + + word + " is " + unigramFrequency); + bigramFrequency = unigramFrequency; + } + // We compute the difference between 255 (which means probability = 1) and the + // unigram score. We split this into a number of discrete steps. + // Now, the steps are numbered 0~15; 0 represents an increase of 1 step while 15 + // represents an increase of 16 steps: a value of 15 will be interpreted as the median + // value of the 16th step. In all justice, if the bigram frequency is low enough to be + // rounded below the first step (which means it is less than half a step higher than the + // unigram frequency) then the unigram frequency itself is the best approximation of the + // bigram freq that we could possibly supply, hence we should *not* include this bigram + // in the file at all. + // until this is done, we'll write 0 and slightly overestimate this case. + // In other words, 0 means "between 0.5 step and 1.5 step", 1 means "between 1.5 step + // and 2.5 steps", and 15 means "between 15.5 steps and 16.5 steps". So we want to + // divide our range [unigramFreq..MAX_TERMINAL_FREQUENCY] in 16.5 steps to get the + // step size. Then we compute the start of the first step (the one where value 0 starts) + // by adding half-a-step to the unigramFrequency. From there, we compute the integer + // number of steps to the bigramFrequency. One last thing: we want our steps to include + // their lower bound and exclude their higher bound so we need to have the first step + // start at exactly 1 unit higher than floor(unigramFreq + half a step). + // Note : to reconstruct the score, the dictionary reader will need to divide + // MAX_TERMINAL_FREQUENCY - unigramFreq by 16.5 likewise to get the value of the step, + // and add (discretizedFrequency + 0.5 + 0.5) times this value to get the best + // approximation. (0.5 to get the first step start, and 0.5 to get the middle of the + // step pointed by the discretized frequency. + final float stepSize = + (FormatSpec.MAX_TERMINAL_FREQUENCY - unigramFrequency) + / (1.5f + FormatSpec.MAX_BIGRAM_FREQUENCY); + final float firstStepStart = 1 + unigramFrequency + (stepSize / 2.0f); + final int discretizedFrequency = (int)((bigramFrequency - firstStepStart) / stepSize); + // If the bigram freq is less than half-a-step higher than the unigram freq, we get -1 + // here. The best approximation would be the unigram freq itself, so we should not + // include this bigram in the dictionary. For now, register as 0, and live with the + // small over-estimation that we get in this case. TODO: actually remove this bigram + // if discretizedFrequency < 0. + final int finalBigramFrequency = discretizedFrequency > 0 ? discretizedFrequency : 0; + bigramFlags += finalBigramFrequency & FormatSpec.FLAG_ATTRIBUTE_FREQUENCY; + return bigramFlags; + } + + /** + * Makes the 2-byte value for options flags. + */ + private static final int makeOptionsValue(final FusionDictionary dictionary, + final FormatOptions formatOptions) { + final DictionaryOptions options = dictionary.mOptions; + final boolean hasBigrams = dictionary.hasBigrams(); + return (options.mFrenchLigatureProcessing ? FormatSpec.FRENCH_LIGATURE_PROCESSING_FLAG : 0) + + (options.mGermanUmlautProcessing ? FormatSpec.GERMAN_UMLAUT_PROCESSING_FLAG : 0) + + (hasBigrams ? FormatSpec.CONTAINS_BIGRAMS_FLAG : 0) + + (formatOptions.mSupportsDynamicUpdate ? FormatSpec.SUPPORTS_DYNAMIC_UPDATE : 0); + } + + /** + * Makes the flag value for a shortcut. + * + * @param more whether there are more attributes after this one. + * @param frequency the frequency of the attribute, 0..15 + * @return the flags + */ + static final int makeShortcutFlags(final boolean more, final int frequency) { + return (more ? FormatSpec.FLAG_ATTRIBUTE_HAS_NEXT : 0) + + (frequency & FormatSpec.FLAG_ATTRIBUTE_FREQUENCY); + } + + private static final int writeParentAddress(final byte[] buffer, final int index, + final int address, final FormatOptions formatOptions) { + if (BinaryDictIOUtils.supportsDynamicUpdate(formatOptions)) { + if (address == FormatSpec.NO_PARENT_ADDRESS) { + buffer[index] = buffer[index + 1] = buffer[index + 2] = 0; + } else { + final int absAddress = Math.abs(address); + assert(absAddress <= FormatSpec.SINT24_MAX); + buffer[index] = (byte)((address < 0 ? FormatSpec.MSB8 : 0) + | ((absAddress >> 16) & 0xFF)); + buffer[index + 1] = (byte)((absAddress >> 8) & 0xFF); + buffer[index + 2] = (byte)(absAddress & 0xFF); + } + return index + 3; + } else { + return index; + } + } + + /** + * Write a node array to memory. The node array is expected to have its final position cached. + * + * @param dict the dictionary the node array is a part of (for relative offsets). + * @param buffer the memory buffer to write to. + * @param nodeArray the node array to write. + * @param formatOptions file format options. + * @return the address of the END of the node. + */ + @SuppressWarnings("unused") + private static int writePlacedNode(final FusionDictionary dict, byte[] buffer, + final PtNodeArray nodeArray, final FormatOptions formatOptions) { + // TODO: Make the code in common with BinaryDictIOUtils#writeCharGroup + int index = nodeArray.mCachedAddressAfterUpdate; + + final int groupCount = nodeArray.mData.size(); + final int countSize = getGroupCountSize(nodeArray); + final int parentAddress = nodeArray.mCachedParentAddress; + if (1 == countSize) { + buffer[index++] = (byte)groupCount; + } else if (2 == countSize) { + // We need to signal 2-byte size by setting the top bit of the MSB to 1, so + // we | 0x80 to do this. + buffer[index++] = (byte)((groupCount >> 8) | 0x80); + buffer[index++] = (byte)(groupCount & 0xFF); + } else { + throw new RuntimeException("Strange size from getGroupCountSize : " + countSize); + } + int groupAddress = index; + for (int i = 0; i < groupCount; ++i) { + final CharGroup group = nodeArray.mData.get(i); + if (index != group.mCachedAddressAfterUpdate) { + throw new RuntimeException("Bug: write index is not the same as the cached address " + + "of the group : " + index + " <> " + group.mCachedAddressAfterUpdate); + } + groupAddress += getGroupHeaderSize(group, formatOptions); + // Sanity checks. + if (DBG && group.mFrequency > FormatSpec.MAX_TERMINAL_FREQUENCY) { + throw new RuntimeException("A node has a frequency > " + + FormatSpec.MAX_TERMINAL_FREQUENCY + + " : " + group.mFrequency); + } + if (group.mFrequency >= 0) groupAddress += FormatSpec.GROUP_FREQUENCY_SIZE; + final int childrenOffset = null == group.mChildren + ? FormatSpec.NO_CHILDREN_ADDRESS + : group.mChildren.mCachedAddressAfterUpdate - groupAddress; + buffer[index++] = + makeCharGroupFlags(group, groupAddress, childrenOffset, formatOptions); + + if (parentAddress == FormatSpec.NO_PARENT_ADDRESS) { + index = writeParentAddress(buffer, index, parentAddress, formatOptions); + } else { + index = writeParentAddress(buffer, index, parentAddress + + (nodeArray.mCachedAddressAfterUpdate - group.mCachedAddressAfterUpdate), + formatOptions); + } + + index = CharEncoding.writeCharArray(group.mChars, buffer, index); + if (group.hasSeveralChars()) { + buffer[index++] = FormatSpec.GROUP_CHARACTERS_TERMINATOR; + } + if (group.mFrequency >= 0) { + buffer[index++] = (byte) group.mFrequency; + } + + final int shift; + if (formatOptions.mSupportsDynamicUpdate) { + shift = writeVariableSignedAddress(buffer, index, childrenOffset); + } else { + shift = writeVariableAddress(buffer, index, childrenOffset); + } + index += shift; + groupAddress += shift; + + // Write shortcuts + if (null != group.mShortcutTargets) { + final int indexOfShortcutByteSize = index; + index += FormatSpec.GROUP_SHORTCUT_LIST_SIZE_SIZE; + groupAddress += FormatSpec.GROUP_SHORTCUT_LIST_SIZE_SIZE; + final Iterator<WeightedString> shortcutIterator = group.mShortcutTargets.iterator(); + while (shortcutIterator.hasNext()) { + final WeightedString target = shortcutIterator.next(); + ++groupAddress; + int shortcutFlags = makeShortcutFlags(shortcutIterator.hasNext(), + target.mFrequency); + buffer[index++] = (byte)shortcutFlags; + final int shortcutShift = CharEncoding.writeString(buffer, index, target.mWord); + index += shortcutShift; + groupAddress += shortcutShift; + } + final int shortcutByteSize = index - indexOfShortcutByteSize; + if (shortcutByteSize > 0xFFFF) { + throw new RuntimeException("Shortcut list too large"); + } + buffer[indexOfShortcutByteSize] = (byte)(shortcutByteSize >> 8); + buffer[indexOfShortcutByteSize + 1] = (byte)(shortcutByteSize & 0xFF); + } + // Write bigrams + if (null != group.mBigrams) { + final Iterator<WeightedString> bigramIterator = group.mBigrams.iterator(); + while (bigramIterator.hasNext()) { + final WeightedString bigram = bigramIterator.next(); + final CharGroup target = + FusionDictionary.findWordInTree(dict.mRootNodeArray, bigram.mWord); + final int addressOfBigram = target.mCachedAddressAfterUpdate; + final int unigramFrequencyForThisWord = target.mFrequency; + ++groupAddress; + final int offset = addressOfBigram - groupAddress; + int bigramFlags = makeBigramFlags(bigramIterator.hasNext(), offset, + bigram.mFrequency, unigramFrequencyForThisWord, bigram.mWord); + buffer[index++] = (byte)bigramFlags; + final int bigramShift = writeVariableAddress(buffer, index, Math.abs(offset)); + index += bigramShift; + groupAddress += bigramShift; + } + } + + } + if (formatOptions.mSupportsDynamicUpdate) { + buffer[index] = buffer[index + 1] = buffer[index + 2] + = FormatSpec.NO_FORWARD_LINK_ADDRESS; + index += FormatSpec.FORWARD_LINK_ADDRESS_SIZE; + } + if (index != nodeArray.mCachedAddressAfterUpdate + nodeArray.mCachedSize) { + throw new RuntimeException( + "Not the same size : written " + (index - nodeArray.mCachedAddressAfterUpdate) + + " bytes from a node that should have " + nodeArray.mCachedSize + " bytes"); + } + return index; + } + + /** + * Dumps a collection of useful statistics about a list of node arrays. + * + * This prints purely informative stuff, like the total estimated file size, the + * number of node arrays, of character groups, the repartition of each address size, etc + * + * @param nodeArrays the list of node arrays. + */ + private static void showStatistics(ArrayList<PtNodeArray> nodeArrays) { + int firstTerminalAddress = Integer.MAX_VALUE; + int lastTerminalAddress = Integer.MIN_VALUE; + int size = 0; + int charGroups = 0; + int maxGroups = 0; + int maxRuns = 0; + for (final PtNodeArray nodeArray : nodeArrays) { + if (maxGroups < nodeArray.mData.size()) maxGroups = nodeArray.mData.size(); + for (final CharGroup cg : nodeArray.mData) { + ++charGroups; + if (cg.mChars.length > maxRuns) maxRuns = cg.mChars.length; + if (cg.mFrequency >= 0) { + if (nodeArray.mCachedAddressAfterUpdate < firstTerminalAddress) + firstTerminalAddress = nodeArray.mCachedAddressAfterUpdate; + if (nodeArray.mCachedAddressAfterUpdate > lastTerminalAddress) + lastTerminalAddress = nodeArray.mCachedAddressAfterUpdate; + } + } + if (nodeArray.mCachedAddressAfterUpdate + nodeArray.mCachedSize > size) { + size = nodeArray.mCachedAddressAfterUpdate + nodeArray.mCachedSize; + } + } + final int[] groupCounts = new int[maxGroups + 1]; + final int[] runCounts = new int[maxRuns + 1]; + for (final PtNodeArray nodeArray : nodeArrays) { + ++groupCounts[nodeArray.mData.size()]; + for (final CharGroup cg : nodeArray.mData) { + ++runCounts[cg.mChars.length]; + } + } + + MakedictLog.i("Statistics:\n" + + " total file size " + size + "\n" + + " " + nodeArrays.size() + " node arrays\n" + + " " + charGroups + " groups (" + ((float)charGroups / nodeArrays.size()) + + " groups per node)\n" + + " first terminal at " + firstTerminalAddress + "\n" + + " last terminal at " + lastTerminalAddress + "\n" + + " Group stats : max = " + maxGroups); + for (int i = 0; i < groupCounts.length; ++i) { + MakedictLog.i(" " + i + " : " + groupCounts[i]); + } + MakedictLog.i(" Character run stats : max = " + maxRuns); + for (int i = 0; i < runCounts.length; ++i) { + MakedictLog.i(" " + i + " : " + runCounts[i]); + } + } + + /** + * Dumps a FusionDictionary to a file. + * + * This is the public entry point to write a dictionary to a file. + * + * @param destination the stream to write the binary data to. + * @param dict the dictionary to write. + * @param formatOptions file format options. + */ + public static void writeDictionaryBinary(final OutputStream destination, + final FusionDictionary dict, final FormatOptions formatOptions) + throws IOException, UnsupportedFormatException { + + // Addresses are limited to 3 bytes, but since addresses can be relative to each node + // array, the structure itself is not limited to 16MB. However, if it is over 16MB deciding + // the order of the node arrays becomes a quite complicated problem, because though the + // dictionary itself does not have a size limit, each node array must still be within 16MB + // of all its children and parents. As long as this is ensured, the dictionary file may + // grow to any size. + + final int version = formatOptions.mVersion; + if (version < FormatSpec.MINIMUM_SUPPORTED_VERSION + || version > FormatSpec.MAXIMUM_SUPPORTED_VERSION) { + throw new UnsupportedFormatException("Requested file format version " + version + + ", but this implementation only supports versions " + + FormatSpec.MINIMUM_SUPPORTED_VERSION + " through " + + FormatSpec.MAXIMUM_SUPPORTED_VERSION); + } + + ByteArrayOutputStream headerBuffer = new ByteArrayOutputStream(256); + + // The magic number in big-endian order. + // Magic number for all versions. + headerBuffer.write((byte) (0xFF & (FormatSpec.MAGIC_NUMBER >> 24))); + headerBuffer.write((byte) (0xFF & (FormatSpec.MAGIC_NUMBER >> 16))); + headerBuffer.write((byte) (0xFF & (FormatSpec.MAGIC_NUMBER >> 8))); + headerBuffer.write((byte) (0xFF & FormatSpec.MAGIC_NUMBER)); + // Dictionary version. + headerBuffer.write((byte) (0xFF & (version >> 8))); + headerBuffer.write((byte) (0xFF & version)); + + // Options flags + final int options = makeOptionsValue(dict, formatOptions); + headerBuffer.write((byte) (0xFF & (options >> 8))); + headerBuffer.write((byte) (0xFF & options)); + final int headerSizeOffset = headerBuffer.size(); + // Placeholder to be written later with header size. + for (int i = 0; i < 4; ++i) { + headerBuffer.write(0); + } + // Write out the options. + for (final String key : dict.mOptions.mAttributes.keySet()) { + final String value = dict.mOptions.mAttributes.get(key); + CharEncoding.writeString(headerBuffer, key); + CharEncoding.writeString(headerBuffer, value); + } + final int size = headerBuffer.size(); + final byte[] bytes = headerBuffer.toByteArray(); + // Write out the header size. + bytes[headerSizeOffset] = (byte) (0xFF & (size >> 24)); + bytes[headerSizeOffset + 1] = (byte) (0xFF & (size >> 16)); + bytes[headerSizeOffset + 2] = (byte) (0xFF & (size >> 8)); + bytes[headerSizeOffset + 3] = (byte) (0xFF & (size >> 0)); + destination.write(bytes); + + headerBuffer.close(); + + // Leave the choice of the optimal node order to the flattenTree function. + MakedictLog.i("Flattening the tree..."); + ArrayList<PtNodeArray> flatNodes = flattenTree(dict.mRootNodeArray); + + MakedictLog.i("Computing addresses..."); + computeAddresses(dict, flatNodes, formatOptions); + MakedictLog.i("Checking array..."); + if (DBG) checkFlatNodeArrayList(flatNodes); + + // Create a buffer that matches the final dictionary size. + final PtNodeArray lastNodeArray = flatNodes.get(flatNodes.size() - 1); + final int bufferSize = lastNodeArray.mCachedAddressAfterUpdate + lastNodeArray.mCachedSize; + final byte[] buffer = new byte[bufferSize]; + int index = 0; + + MakedictLog.i("Writing file..."); + int dataEndOffset = 0; + for (PtNodeArray nodeArray : flatNodes) { + dataEndOffset = writePlacedNode(dict, buffer, nodeArray, formatOptions); + } + + if (DBG) showStatistics(flatNodes); + + destination.write(buffer, 0, dataEndOffset); + + destination.close(); + MakedictLog.i("Done"); + } +} diff --git a/java/src/com/android/inputmethod/latin/makedict/BinaryDictIOUtils.java b/java/src/com/android/inputmethod/latin/makedict/BinaryDictIOUtils.java index 167c6915c..4cf72e915 100644 --- a/java/src/com/android/inputmethod/latin/makedict/BinaryDictIOUtils.java +++ b/java/src/com/android/inputmethod/latin/makedict/BinaryDictIOUtils.java @@ -18,30 +18,26 @@ package com.android.inputmethod.latin.makedict; import com.android.inputmethod.annotations.UsedForTesting; import com.android.inputmethod.latin.Constants; -import com.android.inputmethod.latin.makedict.BinaryDictInputOutput.CharEncoding; -import com.android.inputmethod.latin.makedict.BinaryDictInputOutput.FusionDictionaryBufferInterface; +import com.android.inputmethod.latin.makedict.BinaryDictDecoderUtils.CharEncoding; +import com.android.inputmethod.latin.makedict.BinaryDictDecoderUtils.DictBuffer; import com.android.inputmethod.latin.makedict.FormatSpec.FileHeader; import com.android.inputmethod.latin.makedict.FormatSpec.FormatOptions; import com.android.inputmethod.latin.makedict.FusionDictionary.CharGroup; import com.android.inputmethod.latin.makedict.FusionDictionary.WeightedString; +import com.android.inputmethod.latin.utils.ByteArrayDictBuffer; import java.io.File; import java.io.FileInputStream; import java.io.FileNotFoundException; import java.io.IOException; import java.io.OutputStream; -import java.nio.channels.FileChannel; import java.util.ArrayList; -import java.util.Arrays; import java.util.Iterator; import java.util.Map; import java.util.Stack; public final class BinaryDictIOUtils { private static final boolean DBG = false; - private static final int MSB24 = 0x800000; - private static final int SINT24_MAX = 0x7FFFFF; - private static final int MAX_JUMPS = 10000; private BinaryDictIOUtils() { // This utility class is not publicly instantiable. @@ -63,10 +59,10 @@ public final class BinaryDictIOUtils { } /** - * Tours all node without recursive call. + * Retrieves all node arrays without recursive call. */ private static void readUnigramsAndBigramsBinaryInner( - final FusionDictionaryBufferInterface buffer, final int headerSize, + final DictBuffer dictBuffer, final int headerSize, final Map<Integer, String> words, final Map<Integer, Integer> frequencies, final Map<Integer, ArrayList<PendingAttribute>> bigrams, final FormatOptions formatOptions) { @@ -86,28 +82,28 @@ public final class BinaryDictIOUtils { p.mNumOfCharGroup + ", position=" + p.mPosition + ", length=" + p.mLength); } - if (buffer.position() != p.mAddress) buffer.position(p.mAddress); + if (dictBuffer.position() != p.mAddress) dictBuffer.position(p.mAddress); if (index != p.mLength) index = p.mLength; if (p.mNumOfCharGroup == Position.NOT_READ_GROUPCOUNT) { - p.mNumOfCharGroup = BinaryDictInputOutput.readCharGroupCount(buffer); - p.mAddress += BinaryDictInputOutput.getGroupCountSize(p.mNumOfCharGroup); + p.mNumOfCharGroup = BinaryDictDecoderUtils.readCharGroupCount(dictBuffer); + p.mAddress += getGroupCountSize(p.mNumOfCharGroup); p.mPosition = 0; } if (p.mNumOfCharGroup == 0) { stack.pop(); continue; } - CharGroupInfo info = BinaryDictInputOutput.readCharGroup(buffer, + CharGroupInfo info = BinaryDictDecoderUtils.readCharGroup(dictBuffer, p.mAddress - headerSize, formatOptions); for (int i = 0; i < info.mCharacters.length; ++i) { pushedChars[index++] = info.mCharacters[i]; } p.mPosition++; - final boolean isMovedGroup = BinaryDictInputOutput.isMovedGroup(info.mFlags, + final boolean isMovedGroup = isMovedGroup(info.mFlags, formatOptions); - final boolean isDeletedGroup = BinaryDictInputOutput.isDeletedGroup(info.mFlags, + final boolean isDeletedGroup = isDeletedGroup(info.mFlags, formatOptions); if (!isMovedGroup && !isDeletedGroup && info.mFrequency != FusionDictionary.CharGroup.NOT_A_TERMINAL) {// found word @@ -118,9 +114,9 @@ public final class BinaryDictIOUtils { if (p.mPosition == p.mNumOfCharGroup) { if (formatOptions.mSupportsDynamicUpdate) { - final int forwardLinkAddress = buffer.readUnsignedInt24(); + final int forwardLinkAddress = dictBuffer.readUnsignedInt24(); if (forwardLinkAddress != FormatSpec.NO_FORWARD_LINK_ADDRESS) { - // the node has a forward link. + // The node array has a forward link. p.mNumOfCharGroup = Position.NOT_READ_GROUPCOUNT; p.mAddress = forwardLinkAddress; } else { @@ -130,11 +126,11 @@ public final class BinaryDictIOUtils { stack.pop(); } } else { - // the node has more groups. - p.mAddress = buffer.position(); + // The node array has more groups. + p.mAddress = dictBuffer.position(); } - if (!isMovedGroup && BinaryDictInputOutput.hasChildrenAddress(info.mChildrenAddress)) { + if (!isMovedGroup && hasChildrenAddress(info.mChildrenAddress)) { Position childrenPos = new Position(info.mChildrenAddress + headerSize, index); stack.push(childrenPos); } @@ -143,60 +139,59 @@ public final class BinaryDictIOUtils { /** * Reads unigrams and bigrams from the binary file. - * Doesn't make the memory representation of the dictionary. + * Doesn't store a full memory representation of the dictionary. * - * @param buffer the buffer to read. + * @param dictDecoder the dict decoder. * @param words the map to store the address as a key and the word as a value. * @param frequencies the map to store the address as a key and the frequency as a value. * @param bigrams the map to store the address as a key and the list of address as a value. - * @throws IOException - * @throws UnsupportedFormatException + * @throws IOException if the file can't be read. + * @throws UnsupportedFormatException if the format of the file is not recognized. */ - public static void readUnigramsAndBigramsBinary(final FusionDictionaryBufferInterface buffer, + public static void readUnigramsAndBigramsBinary(final BinaryDictDecoder dictDecoder, final Map<Integer, String> words, final Map<Integer, Integer> frequencies, final Map<Integer, ArrayList<PendingAttribute>> bigrams) throws IOException, UnsupportedFormatException { // Read header - final FileHeader header = BinaryDictInputOutput.readHeader(buffer); - readUnigramsAndBigramsBinaryInner(buffer, header.mHeaderSize, words, frequencies, bigrams, - header.mFormatOptions); + final FileHeader header = BinaryDictDecoderUtils.readHeader(dictDecoder); + readUnigramsAndBigramsBinaryInner(dictDecoder.getDictBuffer(), header.mHeaderSize, words, + frequencies, bigrams, header.mFormatOptions); } /** * Gets the address of the last CharGroup of the exact matching word in the dictionary. * If no match is found, returns NOT_VALID_WORD. * - * @param buffer the buffer to read. + * @param dictDecoder the dict decoder. * @param word the word we search for. * @return the address of the terminal node. - * @throws IOException - * @throws UnsupportedFormatException + * @throws IOException if the file can't be read. + * @throws UnsupportedFormatException if the format of the file is not recognized. */ @UsedForTesting - public static int getTerminalPosition(final FusionDictionaryBufferInterface buffer, + public static int getTerminalPosition(final BinaryDictDecoder dictDecoder, final String word) throws IOException, UnsupportedFormatException { + final DictBuffer dictBuffer = dictDecoder.getDictBuffer(); if (word == null) return FormatSpec.NOT_VALID_WORD; - if (buffer.position() != 0) buffer.position(0); + if (dictBuffer.position() != 0) dictBuffer.position(0); - final FileHeader header = BinaryDictInputOutput.readHeader(buffer); + final FileHeader header = BinaryDictDecoderUtils.readHeader(dictDecoder); int wordPos = 0; final int wordLen = word.codePointCount(0, word.length()); for (int depth = 0; depth < Constants.DICTIONARY_MAX_WORD_LENGTH; ++depth) { if (wordPos >= wordLen) return FormatSpec.NOT_VALID_WORD; do { - final int charGroupCount = BinaryDictInputOutput.readCharGroupCount(buffer); + final int charGroupCount = BinaryDictDecoderUtils.readCharGroupCount(dictBuffer); boolean foundNextCharGroup = false; for (int i = 0; i < charGroupCount; ++i) { - final int charGroupPos = buffer.position(); - final CharGroupInfo currentInfo = BinaryDictInputOutput.readCharGroup(buffer, - buffer.position(), header.mFormatOptions); - final boolean isMovedGroup = - BinaryDictInputOutput.isMovedGroup(currentInfo.mFlags, - header.mFormatOptions); - final boolean isDeletedGroup = - BinaryDictInputOutput.isDeletedGroup(currentInfo.mFlags, - header.mFormatOptions); + final int charGroupPos = dictBuffer.position(); + final CharGroupInfo currentInfo = BinaryDictDecoderUtils.readCharGroup( + dictBuffer, dictBuffer.position(), header.mFormatOptions); + final boolean isMovedGroup = isMovedGroup(currentInfo.mFlags, + header.mFormatOptions); + final boolean isDeletedGroup = isDeletedGroup(currentInfo.mFlags, + header.mFormatOptions); if (isMovedGroup) continue; boolean same = true; for (int p = 0, j = word.offsetByCodePoints(0, wordPos); @@ -224,72 +219,46 @@ public final class BinaryDictIOUtils { return FormatSpec.NOT_VALID_WORD; } foundNextCharGroup = true; - buffer.position(currentInfo.mChildrenAddress); + dictBuffer.position(currentInfo.mChildrenAddress); break; } } // If we found the next char group, it is under the file pointer. - // But if not, we are at the end of this node so we expect to have + // But if not, we are at the end of this node array so we expect to have // a forward link address that we need to consult and possibly resume - // search on the next node in the linked list. + // search on the next node array in the linked list. if (foundNextCharGroup) break; if (!header.mFormatOptions.mSupportsDynamicUpdate) { return FormatSpec.NOT_VALID_WORD; } - final int forwardLinkAddress = buffer.readUnsignedInt24(); + final int forwardLinkAddress = dictBuffer.readUnsignedInt24(); if (forwardLinkAddress == FormatSpec.NO_FORWARD_LINK_ADDRESS) { return FormatSpec.NOT_VALID_WORD; } - buffer.position(forwardLinkAddress); + dictBuffer.position(forwardLinkAddress); } while(true); } return FormatSpec.NOT_VALID_WORD; } - private static int markAsDeleted(final int flags) { - return (flags & (~FormatSpec.MASK_GROUP_ADDRESS_TYPE)) | FormatSpec.FLAG_IS_DELETED; - } - - /** - * Delete the word from the binary file. - * - * @param buffer the buffer to write. - * @param word the word we delete - * @throws IOException - * @throws UnsupportedFormatException - */ - @UsedForTesting - public static void deleteWord(final FusionDictionaryBufferInterface buffer, - final String word) throws IOException, UnsupportedFormatException { - buffer.position(0); - final FileHeader header = BinaryDictInputOutput.readHeader(buffer); - final int wordPosition = getTerminalPosition(buffer, word); - if (wordPosition == FormatSpec.NOT_VALID_WORD) return; - - buffer.position(wordPosition); - final int flags = buffer.readUnsignedByte(); - buffer.position(wordPosition); - buffer.put((byte)markAsDeleted(flags)); - } - /** * @return the size written, in bytes. Always 3 bytes. */ - private static int writeSInt24ToBuffer(final FusionDictionaryBufferInterface buffer, + static int writeSInt24ToBuffer(final DictBuffer dictBuffer, final int value) { final int absValue = Math.abs(value); - buffer.put((byte)(((value < 0 ? 0x80 : 0) | (absValue >> 16)) & 0xFF)); - buffer.put((byte)((absValue >> 8) & 0xFF)); - buffer.put((byte)(absValue & 0xFF)); + dictBuffer.put((byte)(((value < 0 ? 0x80 : 0) | (absValue >> 16)) & 0xFF)); + dictBuffer.put((byte)((absValue >> 8) & 0xFF)); + dictBuffer.put((byte)(absValue & 0xFF)); return 3; } /** * @return the size written, in bytes. Always 3 bytes. */ - private static int writeSInt24ToStream(final OutputStream destination, final int value) + static int writeSInt24ToStream(final OutputStream destination, final int value) throws IOException { final int absValue = Math.abs(value); destination.write((byte)(((value < 0 ? 0x80 : 0) | (absValue >> 16)) & 0xFF)); @@ -303,7 +272,7 @@ public final class BinaryDictIOUtils { */ private static int writeVariableAddress(final OutputStream destination, final int value) throws IOException { - switch (BinaryDictInputOutput.getByteSize(value)) { + switch (BinaryDictEncoder.getByteSize(value)) { case 1: destination.write((byte)value); break; @@ -317,66 +286,34 @@ public final class BinaryDictIOUtils { destination.write((byte)(0xFF & value)); break; } - return BinaryDictInputOutput.getByteSize(value); + return BinaryDictEncoder.getByteSize(value); } - /** - * Update a parent address in a CharGroup that is referred to by groupOriginAddress. - * - * @param buffer the buffer to write. - * @param groupOriginAddress the address of the group. - * @param newParentAddress the absolute address of the parent. - * @param formatOptions file format options. - */ - public static void updateParentAddress(final FusionDictionaryBufferInterface buffer, - final int groupOriginAddress, final int newParentAddress, - final FormatOptions formatOptions) { - final int originalPosition = buffer.position(); - buffer.position(groupOriginAddress); - if (!formatOptions.mSupportsDynamicUpdate) { - throw new RuntimeException("this file format does not support parent addresses"); - } - final int flags = buffer.readUnsignedByte(); - if (BinaryDictInputOutput.isMovedGroup(flags, formatOptions)) { - // if the group is moved, the parent address is stored in the destination group. - // We are guaranteed to process the destination group later, so there is no need to - // update anything here. - buffer.position(originalPosition); - return; - } - if (DBG) { - MakedictLog.d("update parent address flags=" + flags + ", " + groupOriginAddress); - } - final int parentOffset = newParentAddress - groupOriginAddress; - writeSInt24ToBuffer(buffer, parentOffset); - buffer.position(originalPosition); - } - - private static void skipCharGroup(final FusionDictionaryBufferInterface buffer, + static void skipCharGroup(final DictBuffer dictBuffer, final FormatOptions formatOptions) { - final int flags = buffer.readUnsignedByte(); - BinaryDictInputOutput.readParentAddress(buffer, formatOptions); - skipString(buffer, (flags & FormatSpec.FLAG_HAS_MULTIPLE_CHARS) != 0); - BinaryDictInputOutput.readChildrenAddress(buffer, flags, formatOptions); - if ((flags & FormatSpec.FLAG_IS_TERMINAL) != 0) buffer.readUnsignedByte(); + final int flags = dictBuffer.readUnsignedByte(); + BinaryDictDecoderUtils.readParentAddress(dictBuffer, formatOptions); + skipString(dictBuffer, (flags & FormatSpec.FLAG_HAS_MULTIPLE_CHARS) != 0); + BinaryDictDecoderUtils.readChildrenAddress(dictBuffer, flags, formatOptions); + if ((flags & FormatSpec.FLAG_IS_TERMINAL) != 0) dictBuffer.readUnsignedByte(); if ((flags & FormatSpec.FLAG_HAS_SHORTCUT_TARGETS) != 0) { - final int shortcutsSize = buffer.readUnsignedShort(); - buffer.position(buffer.position() + shortcutsSize + final int shortcutsSize = dictBuffer.readUnsignedShort(); + dictBuffer.position(dictBuffer.position() + shortcutsSize - FormatSpec.GROUP_SHORTCUT_LIST_SIZE_SIZE); } if ((flags & FormatSpec.FLAG_HAS_BIGRAMS) != 0) { int bigramCount = 0; while (bigramCount++ < FormatSpec.MAX_BIGRAMS_IN_A_GROUP) { - final int bigramFlags = buffer.readUnsignedByte(); + final int bigramFlags = dictBuffer.readUnsignedByte(); switch (bigramFlags & FormatSpec.MASK_ATTRIBUTE_ADDRESS_TYPE) { case FormatSpec.FLAG_ATTRIBUTE_ADDRESS_TYPE_ONEBYTE: - buffer.readUnsignedByte(); + dictBuffer.readUnsignedByte(); break; case FormatSpec.FLAG_ATTRIBUTE_ADDRESS_TYPE_TWOBYTES: - buffer.readUnsignedShort(); + dictBuffer.readUnsignedShort(); break; case FormatSpec.FLAG_ATTRIBUTE_ADDRESS_TYPE_THREEBYTES: - buffer.readUnsignedInt24(); + dictBuffer.readUnsignedInt24(); break; } if ((bigramFlags & FormatSpec.FLAG_ATTRIBUTE_HAS_NEXT) == 0) break; @@ -387,41 +324,15 @@ public final class BinaryDictIOUtils { } } - /** - * Update parent addresses in a Node that is referred to by nodeOriginAddress. - * - * @param buffer the buffer to be modified. - * @param nodeOriginAddress the address of a modified Node. - * @param newParentAddress the address to be written. - * @param formatOptions file format options. - */ - public static void updateParentAddresses(final FusionDictionaryBufferInterface buffer, - final int nodeOriginAddress, final int newParentAddress, - final FormatOptions formatOptions) { - final int originalPosition = buffer.position(); - buffer.position(nodeOriginAddress); - do { - final int count = BinaryDictInputOutput.readCharGroupCount(buffer); - for (int i = 0; i < count; ++i) { - updateParentAddress(buffer, buffer.position(), newParentAddress, formatOptions); - skipCharGroup(buffer, formatOptions); - } - final int forwardLinkAddress = buffer.readUnsignedInt24(); - buffer.position(forwardLinkAddress); - } while (formatOptions.mSupportsDynamicUpdate - && buffer.position() != FormatSpec.NO_FORWARD_LINK_ADDRESS); - buffer.position(originalPosition); - } - - private static void skipString(final FusionDictionaryBufferInterface buffer, + static void skipString(final DictBuffer dictBuffer, final boolean hasMultipleChars) { if (hasMultipleChars) { - int character = CharEncoding.readChar(buffer); + int character = CharEncoding.readChar(dictBuffer); while (character != FormatSpec.INVALID_CHARACTER) { - character = CharEncoding.readChar(buffer); + character = CharEncoding.readChar(dictBuffer); } } else { - CharEncoding.readChar(buffer); + CharEncoding.readChar(dictBuffer); } } @@ -455,32 +366,10 @@ public final class BinaryDictIOUtils { } /** - * Update a children address in a CharGroup that is addressed by groupOriginAddress. - * - * @param buffer the buffer to write. - * @param groupOriginAddress the address of the group. - * @param newChildrenAddress the absolute address of the child. - * @param formatOptions file format options. - */ - public static void updateChildrenAddress(final FusionDictionaryBufferInterface buffer, - final int groupOriginAddress, final int newChildrenAddress, - final FormatOptions formatOptions) { - final int originalPosition = buffer.position(); - buffer.position(groupOriginAddress); - final int flags = buffer.readUnsignedByte(); - final int parentAddress = BinaryDictInputOutput.readParentAddress(buffer, formatOptions); - skipString(buffer, (flags & FormatSpec.FLAG_HAS_MULTIPLE_CHARS) != 0); - if ((flags & FormatSpec.FLAG_IS_TERMINAL) != 0) buffer.readUnsignedByte(); - final int childrenOffset = newChildrenAddress == FormatSpec.NO_CHILDREN_ADDRESS - ? FormatSpec.NO_CHILDREN_ADDRESS : newChildrenAddress - buffer.position(); - writeSInt24ToBuffer(buffer, childrenOffset); - buffer.position(originalPosition); - } - - /** - * Write a char group to an output stream. - * A char group is an in-memory representation of a node in trie. - * A char group info is an on-disk representation of a node. + * Write a char group to an output stream from a CharGroupInfo. + * A char group is an in-memory representation of a node in the patricia trie. + * A char group info is a container for low-level information about how the + * char group is stored in the binary format. * * @param destination the stream to write. * @param info the char group info to be written. @@ -524,14 +413,14 @@ public final class BinaryDictIOUtils { if (info.mShortcutTargets != null && info.mShortcutTargets.size() > 0) { final int shortcutListSize = - BinaryDictInputOutput.getShortcutListSize(info.mShortcutTargets); + BinaryDictEncoder.getShortcutListSize(info.mShortcutTargets); destination.write((byte)(shortcutListSize >> 8)); destination.write((byte)(shortcutListSize & 0xFF)); size += 2; final Iterator<WeightedString> shortcutIterator = info.mShortcutTargets.iterator(); while (shortcutIterator.hasNext()) { final WeightedString target = shortcutIterator.next(); - destination.write((byte)BinaryDictInputOutput.makeShortcutFlags( + destination.write((byte)BinaryDictEncoder.makeShortcutFlags( shortcutIterator.hasNext(), target.mFrequency)); size++; size += writeString(destination, target.mWord); @@ -540,7 +429,7 @@ public final class BinaryDictIOUtils { if (info.mBigrams != null) { // TODO: Consolidate this code with the code that computes the size of the bigram list - // in BinaryDictionaryInputOutput#computeActualNodeSize + // in BinaryDictEncoder#computeActualNodeArraySize for (int i = 0; i < info.mBigrams.size(); ++i) { final int bigramFrequency = info.mBigrams.get(i).mFrequency; @@ -550,7 +439,7 @@ public final class BinaryDictIOUtils { final int bigramOffset = info.mBigrams.get(i).mAddress - (info.mOriginalAddress + size); bigramFlags |= (bigramOffset < 0) ? FormatSpec.FLAG_ATTRIBUTE_OFFSET_NEGATIVE : 0; - switch (BinaryDictInputOutput.getByteSize(bigramOffset)) { + switch (BinaryDictEncoder.getByteSize(bigramOffset)) { case 1: bigramFlags |= FormatSpec.FLAG_ATTRIBUTE_ADDRESS_TYPE_ONEBYTE; break; @@ -569,82 +458,40 @@ public final class BinaryDictIOUtils { return size; } - @SuppressWarnings("unused") - private static void updateForwardLink(final FusionDictionaryBufferInterface buffer, - final int nodeOriginAddress, final int newNodeAddress, - final FormatOptions formatOptions) { - buffer.position(nodeOriginAddress); - int jumpCount = 0; - while (jumpCount++ < MAX_JUMPS) { - final int count = BinaryDictInputOutput.readCharGroupCount(buffer); - for (int i = 0; i < count; ++i) skipCharGroup(buffer, formatOptions); - final int forwardLinkAddress = buffer.readUnsignedInt24(); - if (forwardLinkAddress == FormatSpec.NO_FORWARD_LINK_ADDRESS) { - buffer.position(buffer.position() - FormatSpec.FORWARD_LINK_ADDRESS_SIZE); - writeSInt24ToBuffer(buffer, newNodeAddress); - return; - } - buffer.position(forwardLinkAddress); - } - if (DBG && jumpCount >= MAX_JUMPS) { - throw new RuntimeException("too many jumps, probably a bug."); - } - } - - /** - * Helper method to move a char group to the tail of the file. - */ - private static int moveCharGroup(final OutputStream destination, - final FusionDictionaryBufferInterface buffer, final CharGroupInfo info, - final int nodeOriginAddress, final int oldGroupAddress, - final FormatOptions formatOptions) throws IOException { - updateParentAddress(buffer, oldGroupAddress, buffer.limit() + 1, formatOptions); - buffer.position(oldGroupAddress); - final int currentFlags = buffer.readUnsignedByte(); - buffer.position(oldGroupAddress); - buffer.put((byte)(FormatSpec.FLAG_IS_MOVED | (currentFlags - & (~FormatSpec.MASK_MOVE_AND_DELETE_FLAG)))); - int size = FormatSpec.GROUP_FLAGS_SIZE; - updateForwardLink(buffer, nodeOriginAddress, buffer.limit(), formatOptions); - size += writeNode(destination, new CharGroupInfo[] { info }); - return size; - } - /** * Compute the size of the char group. */ - private static int computeGroupSize(final CharGroupInfo info, - final FormatOptions formatOptions) { + static int computeGroupSize(final CharGroupInfo info, final FormatOptions formatOptions) { int size = FormatSpec.GROUP_FLAGS_SIZE + FormatSpec.PARENT_ADDRESS_SIZE - + BinaryDictInputOutput.getGroupCharactersSize(info.mCharacters) - + BinaryDictInputOutput.getChildrenAddressSize(info.mFlags, formatOptions); + + BinaryDictEncoder.getGroupCharactersSize(info.mCharacters) + + getChildrenAddressSize(info.mFlags, formatOptions); if ((info.mFlags & FormatSpec.FLAG_IS_TERMINAL) != 0) { size += FormatSpec.GROUP_FREQUENCY_SIZE; } if (info.mShortcutTargets != null && !info.mShortcutTargets.isEmpty()) { - size += BinaryDictInputOutput.getShortcutListSize(info.mShortcutTargets); + size += BinaryDictEncoder.getShortcutListSize(info.mShortcutTargets); } if (info.mBigrams != null) { for (final PendingAttribute attr : info.mBigrams) { size += FormatSpec.GROUP_FLAGS_SIZE; - size += BinaryDictInputOutput.getByteSize(attr.mAddress); + size += BinaryDictEncoder.getByteSize(attr.mAddress); } } return size; } /** - * Write a node to the stream. + * Write a node array to the stream. * * @param destination the stream to write. - * @param infos groups to be written. + * @param infos an array of CharGroupInfo to be written. * @return the size written, in bytes. * @throws IOException */ - private static int writeNode(final OutputStream destination, final CharGroupInfo[] infos) + static int writeNodes(final OutputStream destination, final CharGroupInfo[] infos) throws IOException { - int size = BinaryDictInputOutput.getGroupCountSize(infos.length); - switch (BinaryDictInputOutput.getGroupCountSize(infos.length)) { + int size = getGroupCountSize(infos.length); + switch (getGroupCountSize(infos.length)) { case 1: destination.write((byte)infos.length); break; @@ -661,323 +508,25 @@ public final class BinaryDictIOUtils { } /** - * Move a group that is referred to by oldGroupOrigin to the tail of the file. - * And set the children address to the byte after the group. - * - * @param nodeOrigin the address of the tail of the file. - * @param characters - * @param length - * @param flags - * @param frequency - * @param parentAddress - * @param shortcutTargets - * @param bigrams - * @param destination the stream representing the tail of the file. - * @param buffer the buffer representing the (constant-size) body of the file. - * @param oldNodeOrigin - * @param oldGroupOrigin - * @param formatOptions - * @return the size written, in bytes. - * @throws IOException - */ - private static int moveGroup(final int nodeOrigin, final int[] characters, final int length, - final int flags, final int frequency, final int parentAddress, - final ArrayList<WeightedString> shortcutTargets, - final ArrayList<PendingAttribute> bigrams, final OutputStream destination, - final FusionDictionaryBufferInterface buffer, final int oldNodeOrigin, - final int oldGroupOrigin, final FormatOptions formatOptions) throws IOException { - int size = 0; - final int newGroupOrigin = nodeOrigin + 1; - final int[] writtenCharacters = Arrays.copyOfRange(characters, 0, length); - final CharGroupInfo tmpInfo = new CharGroupInfo(newGroupOrigin, -1 /* endAddress */, - flags, writtenCharacters, frequency, parentAddress, FormatSpec.NO_CHILDREN_ADDRESS, - shortcutTargets, bigrams); - size = computeGroupSize(tmpInfo, formatOptions); - final CharGroupInfo newInfo = new CharGroupInfo(newGroupOrigin, newGroupOrigin + size, - flags, writtenCharacters, frequency, parentAddress, - nodeOrigin + 1 + size + FormatSpec.FORWARD_LINK_ADDRESS_SIZE, shortcutTargets, - bigrams); - moveCharGroup(destination, buffer, newInfo, oldNodeOrigin, oldGroupOrigin, formatOptions); - return 1 + size + FormatSpec.FORWARD_LINK_ADDRESS_SIZE; - } - - /** - * Insert a word into a binary dictionary. - * - * @param buffer - * @param destination - * @param word - * @param frequency - * @param bigramStrings - * @param shortcuts - * @throws IOException - * @throws UnsupportedFormatException - */ - // TODO: Support batch insertion. - // TODO: Remove @UsedForTesting once UserHistoryDictionary is implemented by BinaryDictionary. - @UsedForTesting - public static void insertWord(final FusionDictionaryBufferInterface buffer, - final OutputStream destination, final String word, final int frequency, - final ArrayList<WeightedString> bigramStrings, - final ArrayList<WeightedString> shortcuts, final boolean isNotAWord, - final boolean isBlackListEntry) - throws IOException, UnsupportedFormatException { - final ArrayList<PendingAttribute> bigrams = new ArrayList<PendingAttribute>(); - if (bigramStrings != null) { - for (final WeightedString bigram : bigramStrings) { - int position = getTerminalPosition(buffer, bigram.mWord); - if (position == FormatSpec.NOT_VALID_WORD) { - // TODO: figure out what is the correct thing to do here. - } else { - bigrams.add(new PendingAttribute(bigram.mFrequency, position)); - } - } - } - - final boolean isTerminal = true; - final boolean hasBigrams = !bigrams.isEmpty(); - final boolean hasShortcuts = shortcuts != null && !shortcuts.isEmpty(); - - // find the insert position of the word. - if (buffer.position() != 0) buffer.position(0); - final FileHeader header = BinaryDictInputOutput.readHeader(buffer); - - int wordPos = 0, address = buffer.position(), nodeOriginAddress = buffer.position(); - final int[] codePoints = FusionDictionary.getCodePoints(word); - final int wordLen = codePoints.length; - - for (int depth = 0; depth < Constants.DICTIONARY_MAX_WORD_LENGTH; ++depth) { - if (wordPos >= wordLen) break; - nodeOriginAddress = buffer.position(); - int nodeParentAddress = -1; - final int charGroupCount = BinaryDictInputOutput.readCharGroupCount(buffer); - boolean foundNextGroup = false; - - for (int i = 0; i < charGroupCount; ++i) { - address = buffer.position(); - final CharGroupInfo currentInfo = BinaryDictInputOutput.readCharGroup(buffer, - buffer.position(), header.mFormatOptions); - final boolean isMovedGroup = BinaryDictInputOutput.isMovedGroup(currentInfo.mFlags, - header.mFormatOptions); - if (isMovedGroup) continue; - nodeParentAddress = (currentInfo.mParentAddress == FormatSpec.NO_PARENT_ADDRESS) - ? FormatSpec.NO_PARENT_ADDRESS : currentInfo.mParentAddress + address; - boolean matched = true; - for (int p = 0; p < currentInfo.mCharacters.length; ++p) { - if (wordPos + p >= wordLen) { - /* - * splitting - * before - * abcd - ef - * - * insert "abc" - * - * after - * abc - d - ef - */ - final int newNodeAddress = buffer.limit(); - final int flags = BinaryDictInputOutput.makeCharGroupFlags(p > 1, - isTerminal, 0, hasShortcuts, hasBigrams, false /* isNotAWord */, - false /* isBlackListEntry */, header.mFormatOptions); - int written = moveGroup(newNodeAddress, currentInfo.mCharacters, p, flags, - frequency, nodeParentAddress, shortcuts, bigrams, destination, - buffer, nodeOriginAddress, address, header.mFormatOptions); - - final int[] characters2 = Arrays.copyOfRange(currentInfo.mCharacters, p, - currentInfo.mCharacters.length); - if (currentInfo.mChildrenAddress != FormatSpec.NO_CHILDREN_ADDRESS) { - updateParentAddresses(buffer, currentInfo.mChildrenAddress, - newNodeAddress + written + 1, header.mFormatOptions); - } - final CharGroupInfo newInfo2 = new CharGroupInfo( - newNodeAddress + written + 1, -1 /* endAddress */, - currentInfo.mFlags, characters2, currentInfo.mFrequency, - newNodeAddress + 1, currentInfo.mChildrenAddress, - currentInfo.mShortcutTargets, currentInfo.mBigrams); - writeNode(destination, new CharGroupInfo[] { newInfo2 }); - return; - } else if (codePoints[wordPos + p] != currentInfo.mCharacters[p]) { - if (p > 0) { - /* - * splitting - * before - * ab - cd - * - * insert "ac" - * - * after - * a - b - cd - * | - * - c - */ - - final int newNodeAddress = buffer.limit(); - final int childrenAddress = currentInfo.mChildrenAddress; - - // move prefix - final int prefixFlags = BinaryDictInputOutput.makeCharGroupFlags(p > 1, - false /* isTerminal */, 0 /* childrenAddressSize*/, - false /* hasShortcut */, false /* hasBigrams */, - false /* isNotAWord */, false /* isBlackListEntry */, - header.mFormatOptions); - int written = moveGroup(newNodeAddress, currentInfo.mCharacters, p, - prefixFlags, -1 /* frequency */, nodeParentAddress, null, null, - destination, buffer, nodeOriginAddress, address, - header.mFormatOptions); - - final int[] suffixCharacters = Arrays.copyOfRange( - currentInfo.mCharacters, p, currentInfo.mCharacters.length); - if (currentInfo.mChildrenAddress != FormatSpec.NO_CHILDREN_ADDRESS) { - updateParentAddresses(buffer, currentInfo.mChildrenAddress, - newNodeAddress + written + 1, header.mFormatOptions); - } - final int suffixFlags = BinaryDictInputOutput.makeCharGroupFlags( - suffixCharacters.length > 1, - (currentInfo.mFlags & FormatSpec.FLAG_IS_TERMINAL) != 0, - 0 /* childrenAddressSize */, - (currentInfo.mFlags & FormatSpec.FLAG_HAS_SHORTCUT_TARGETS) - != 0, - (currentInfo.mFlags & FormatSpec.FLAG_HAS_BIGRAMS) != 0, - isNotAWord, isBlackListEntry, header.mFormatOptions); - final CharGroupInfo suffixInfo = new CharGroupInfo( - newNodeAddress + written + 1, -1 /* endAddress */, suffixFlags, - suffixCharacters, currentInfo.mFrequency, newNodeAddress + 1, - currentInfo.mChildrenAddress, currentInfo.mShortcutTargets, - currentInfo.mBigrams); - written += computeGroupSize(suffixInfo, header.mFormatOptions) + 1; - - final int[] newCharacters = Arrays.copyOfRange(codePoints, wordPos + p, - codePoints.length); - final int flags = BinaryDictInputOutput.makeCharGroupFlags( - newCharacters.length > 1, isTerminal, - 0 /* childrenAddressSize */, hasShortcuts, hasBigrams, - isNotAWord, isBlackListEntry, header.mFormatOptions); - final CharGroupInfo newInfo = new CharGroupInfo( - newNodeAddress + written, -1 /* endAddress */, flags, - newCharacters, frequency, newNodeAddress + 1, - FormatSpec.NO_CHILDREN_ADDRESS, shortcuts, bigrams); - writeNode(destination, new CharGroupInfo[] { suffixInfo, newInfo }); - return; - } - matched = false; - break; - } - } - - if (matched) { - if (wordPos + currentInfo.mCharacters.length == wordLen) { - // the word exists in the dictionary. - // only update group. - final int newNodeAddress = buffer.limit(); - final boolean hasMultipleChars = currentInfo.mCharacters.length > 1; - final int flags = BinaryDictInputOutput.makeCharGroupFlags(hasMultipleChars, - isTerminal, 0 /* childrenAddressSize */, hasShortcuts, hasBigrams, - isNotAWord, isBlackListEntry, header.mFormatOptions); - final CharGroupInfo newInfo = new CharGroupInfo(newNodeAddress + 1, - -1 /* endAddress */, flags, currentInfo.mCharacters, frequency, - nodeParentAddress, currentInfo.mChildrenAddress, shortcuts, - bigrams); - moveCharGroup(destination, buffer, newInfo, nodeOriginAddress, address, - header.mFormatOptions); - return; - } - wordPos += currentInfo.mCharacters.length; - if (currentInfo.mChildrenAddress == FormatSpec.NO_CHILDREN_ADDRESS) { - /* - * found the prefix of the word. - * make new node and link to the node from this group. - * - * before - * ab - cd - * - * insert "abcde" - * - * after - * ab - cd - e - */ - final int newNodeAddress = buffer.limit(); - updateChildrenAddress(buffer, address, newNodeAddress, - header.mFormatOptions); - final int newGroupAddress = newNodeAddress + 1; - final boolean hasMultipleChars = (wordLen - wordPos) > 1; - final int flags = BinaryDictInputOutput.makeCharGroupFlags(hasMultipleChars, - isTerminal, 0 /* childrenAddressSize */, hasShortcuts, hasBigrams, - isNotAWord, isBlackListEntry, header.mFormatOptions); - final int[] characters = Arrays.copyOfRange(codePoints, wordPos, wordLen); - final CharGroupInfo newInfo = new CharGroupInfo(newGroupAddress, -1, flags, - characters, frequency, address, FormatSpec.NO_CHILDREN_ADDRESS, - shortcuts, bigrams); - writeNode(destination, new CharGroupInfo[] { newInfo }); - return; - } - buffer.position(currentInfo.mChildrenAddress); - foundNextGroup = true; - break; - } - } - - if (foundNextGroup) continue; - - // reached the end of the array. - final int linkAddressPosition = buffer.position(); - int nextLink = buffer.readUnsignedInt24(); - if ((nextLink & MSB24) != 0) { - nextLink = -(nextLink & SINT24_MAX); - } - if (nextLink == FormatSpec.NO_FORWARD_LINK_ADDRESS) { - /* - * expand this node. - * - * before - * ab - cd - * - * insert "abef" - * - * after - * ab - cd - * | - * - ef - */ - - // change the forward link address. - final int newNodeAddress = buffer.limit(); - buffer.position(linkAddressPosition); - writeSInt24ToBuffer(buffer, newNodeAddress); - - final int[] characters = Arrays.copyOfRange(codePoints, wordPos, wordLen); - final int flags = BinaryDictInputOutput.makeCharGroupFlags(characters.length > 1, - isTerminal, 0 /* childrenAddressSize */, hasShortcuts, hasBigrams, - isNotAWord, isBlackListEntry, header.mFormatOptions); - final CharGroupInfo newInfo = new CharGroupInfo(newNodeAddress + 1, - -1 /* endAddress */, flags, characters, frequency, nodeParentAddress, - FormatSpec.NO_CHILDREN_ADDRESS, shortcuts, bigrams); - writeNode(destination, new CharGroupInfo[]{ newInfo }); - return; - } else { - depth--; - buffer.position(nextLink); - } - } - } - - /** - * Find a word from the buffer. + * Find a word using the BinaryDictDecoder. * - * @param buffer the buffer representing the body of the dictionary file. + * @param dictDecoder the dict reader * @param word the word searched * @return the found group * @throws IOException * @throws UnsupportedFormatException */ @UsedForTesting - public static CharGroupInfo findWordFromBuffer(final FusionDictionaryBufferInterface buffer, + public static CharGroupInfo findWordByBinaryDictReader(final BinaryDictDecoder dictDecoder, final String word) throws IOException, UnsupportedFormatException { - int position = getTerminalPosition(buffer, word); + int position = getTerminalPosition(dictDecoder, word); + final DictBuffer dictBuffer = dictDecoder.getDictBuffer(); if (position != FormatSpec.NOT_VALID_WORD) { - buffer.position(0); - final FileHeader header = BinaryDictInputOutput.readHeader(buffer); - buffer.position(position); - return BinaryDictInputOutput.readCharGroup(buffer, position, header.mFormatOptions); + dictBuffer.position(0); + final FileHeader header = BinaryDictDecoderUtils.readHeader(dictDecoder); + dictBuffer.position(position); + return BinaryDictDecoderUtils.readCharGroup(dictBuffer, position, + header.mFormatOptions); } return null; } @@ -996,16 +545,21 @@ public final class BinaryDictIOUtils { final File file, final long offset, final long length) throws FileNotFoundException, IOException, UnsupportedFormatException { final byte[] buffer = new byte[HEADER_READING_BUFFER_SIZE]; - final FileInputStream inStream = new FileInputStream(file); - try { - inStream.read(buffer); - final BinaryDictInputOutput.ByteBufferWrapper wrapper = - new BinaryDictInputOutput.ByteBufferWrapper(inStream.getChannel().map( - FileChannel.MapMode.READ_ONLY, offset, length)); - return BinaryDictInputOutput.readHeader(wrapper); - } finally { - inStream.close(); - } + final BinaryDictDecoder dictDecoder = new BinaryDictDecoder(file); + dictDecoder.openDictBuffer(new BinaryDictDecoder.DictionaryBufferFactory() { + @Override + public DictBuffer getDictionaryBuffer(File file) + throws FileNotFoundException, IOException { + final FileInputStream inStream = new FileInputStream(file); + try { + inStream.read(buffer); + return new ByteArrayDictBuffer(buffer); + } finally { + inStream.close(); + } + } + }); + return BinaryDictDecoderUtils.readHeader(dictDecoder); } public static FileHeader getDictionaryFileHeaderOrNull(final File file, final long offset, @@ -1019,4 +573,83 @@ public final class BinaryDictIOUtils { return null; } } + + /** + * Helper method to hide the actual value of the no children address. + */ + public static boolean hasChildrenAddress(final int address) { + return FormatSpec.NO_CHILDREN_ADDRESS != address; + } + + /** + * Helper method to check whether the group is moved. + */ + public static boolean isMovedGroup(final int flags, final FormatOptions options) { + return options.mSupportsDynamicUpdate + && ((flags & FormatSpec.MASK_GROUP_ADDRESS_TYPE) == FormatSpec.FLAG_IS_MOVED); + } + + /** + * Helper method to check whether the dictionary can be updated dynamically. + */ + public static boolean supportsDynamicUpdate(final FormatOptions options) { + return options.mVersion >= FormatSpec.FIRST_VERSION_WITH_DYNAMIC_UPDATE + && options.mSupportsDynamicUpdate; + } + + /** + * Helper method to check whether the group is deleted. + */ + public static boolean isDeletedGroup(final int flags, final FormatOptions formatOptions) { + return formatOptions.mSupportsDynamicUpdate + && ((flags & FormatSpec.MASK_GROUP_ADDRESS_TYPE) == FormatSpec.FLAG_IS_DELETED); + } + + /** + * Compute the binary size of the group count + * @param count the group count + * @return the size of the group count, either 1 or 2 bytes. + */ + public static int getGroupCountSize(final int count) { + if (FormatSpec.MAX_CHARGROUPS_FOR_ONE_BYTE_CHARGROUP_COUNT >= count) { + return 1; + } else if (FormatSpec.MAX_CHARGROUPS_IN_A_PT_NODE_ARRAY >= count) { + return 2; + } else { + throw new RuntimeException("Can't have more than " + + FormatSpec.MAX_CHARGROUPS_IN_A_PT_NODE_ARRAY + " groups in a node (found " + + count + ")"); + } + } + + static int getChildrenAddressSize(final int optionFlags, + final FormatOptions formatOptions) { + if (formatOptions.mSupportsDynamicUpdate) return FormatSpec.SIGNED_CHILDREN_ADDRESS_SIZE; + switch (optionFlags & FormatSpec.MASK_GROUP_ADDRESS_TYPE) { + case FormatSpec.FLAG_GROUP_ADDRESS_TYPE_ONEBYTE: + return 1; + case FormatSpec.FLAG_GROUP_ADDRESS_TYPE_TWOBYTES: + return 2; + case FormatSpec.FLAG_GROUP_ADDRESS_TYPE_THREEBYTES: + return 3; + case FormatSpec.FLAG_GROUP_ADDRESS_TYPE_NOADDRESS: + default: + return 0; + } + } + + /** + * Calculate bigram frequency from compressed value + * + * @param unigramFrequency + * @param bigramFrequency compressed frequency + * @return approximate bigram frequency + */ + public static int reconstructBigramFrequency(final int unigramFrequency, + final int bigramFrequency) { + final float stepSize = (FormatSpec.MAX_TERMINAL_FREQUENCY - unigramFrequency) + / (1.5f + FormatSpec.MAX_BIGRAM_FREQUENCY); + final float resultFreqFloat = unigramFrequency + stepSize * (bigramFrequency + 1.0f); + return (int)resultFreqFloat; + } } diff --git a/java/src/com/android/inputmethod/latin/makedict/BinaryDictInputOutput.java b/java/src/com/android/inputmethod/latin/makedict/BinaryDictInputOutput.java deleted file mode 100644 index 1b187d85d..000000000 --- a/java/src/com/android/inputmethod/latin/makedict/BinaryDictInputOutput.java +++ /dev/null @@ -1,1838 +0,0 @@ -/* - * 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.makedict; - -import com.android.inputmethod.annotations.UsedForTesting; -import com.android.inputmethod.latin.makedict.FormatSpec.FileHeader; -import com.android.inputmethod.latin.makedict.FormatSpec.FormatOptions; -import com.android.inputmethod.latin.makedict.FusionDictionary.CharGroup; -import com.android.inputmethod.latin.makedict.FusionDictionary.DictionaryOptions; -import com.android.inputmethod.latin.makedict.FusionDictionary.Node; -import com.android.inputmethod.latin.makedict.FusionDictionary.WeightedString; - -import java.io.ByteArrayOutputStream; -import java.io.File; -import java.io.FileInputStream; -import java.io.FileNotFoundException; -import java.io.IOException; -import java.io.OutputStream; -import java.nio.ByteBuffer; -import java.nio.channels.FileChannel; -import java.util.ArrayList; -import java.util.Arrays; -import java.util.HashMap; -import java.util.Iterator; -import java.util.Map; -import java.util.TreeMap; - -/** - * Reads and writes XML files for a FusionDictionary. - * - * All the methods in this class are static. - */ -public final class BinaryDictInputOutput { - - private static final boolean DBG = MakedictLog.DBG; - - // Arbitrary limit to how much passes we consider address size compression should - // terminate in. At the time of this writing, our largest dictionary completes - // compression in five passes. - // If the number of passes exceeds this number, makedict bails with an exception on - // suspicion that a bug might be causing an infinite loop. - private static final int MAX_PASSES = 24; - private static final int MAX_JUMPS = 12; - - @UsedForTesting - public interface FusionDictionaryBufferInterface { - public int readUnsignedByte(); - public int readUnsignedShort(); - public int readUnsignedInt24(); - public int readInt(); - public int position(); - public void position(int newPosition); - public void put(final byte b); - public int limit(); - public int capacity(); - } - - public static final class ByteBufferWrapper implements FusionDictionaryBufferInterface { - private ByteBuffer mBuffer; - - public ByteBufferWrapper(final ByteBuffer buffer) { - mBuffer = buffer; - } - - @Override - public int readUnsignedByte() { - return mBuffer.get() & 0xFF; - } - - @Override - public int readUnsignedShort() { - return mBuffer.getShort() & 0xFFFF; - } - - @Override - public int readUnsignedInt24() { - final int retval = readUnsignedByte(); - return (retval << 16) + readUnsignedShort(); - } - - @Override - public int readInt() { - return mBuffer.getInt(); - } - - @Override - public int position() { - return mBuffer.position(); - } - - @Override - public void position(int newPos) { - mBuffer.position(newPos); - } - - @Override - public void put(final byte b) { - mBuffer.put(b); - } - - @Override - public int limit() { - return mBuffer.limit(); - } - - @Override - public int capacity() { - return mBuffer.capacity(); - } - } - - /** - * A class grouping utility function for our specific character encoding. - */ - static final class CharEncoding { - private static final int MINIMAL_ONE_BYTE_CHARACTER_VALUE = 0x20; - private static final int MAXIMAL_ONE_BYTE_CHARACTER_VALUE = 0xFF; - - /** - * Helper method to find out whether this code fits on one byte - */ - private static boolean fitsOnOneByte(final int character) { - return character >= MINIMAL_ONE_BYTE_CHARACTER_VALUE - && character <= MAXIMAL_ONE_BYTE_CHARACTER_VALUE; - } - - /** - * Compute the size of a character given its character code. - * - * Char format is: - * 1 byte = bbbbbbbb match - * case 000xxxxx: xxxxx << 16 + next byte << 8 + next byte - * else: if 00011111 (= 0x1F) : this is the terminator. This is a relevant choice because - * unicode code points range from 0 to 0x10FFFF, so any 3-byte value starting with - * 00011111 would be outside unicode. - * else: iso-latin-1 code - * This allows for the whole unicode range to be encoded, including chars outside of - * the BMP. Also everything in the iso-latin-1 charset is only 1 byte, except control - * characters which should never happen anyway (and still work, but take 3 bytes). - * - * @param character the character code. - * @return the size in binary encoded-form, either 1 or 3 bytes. - */ - static int getCharSize(final int character) { - // See char encoding in FusionDictionary.java - if (fitsOnOneByte(character)) return 1; - if (FormatSpec.INVALID_CHARACTER == character) return 1; - return 3; - } - - /** - * Compute the byte size of a character array. - */ - private static int getCharArraySize(final int[] chars) { - int size = 0; - for (int character : chars) size += getCharSize(character); - return size; - } - - /** - * Writes a char array to a byte buffer. - * - * @param codePoints the code point array to write. - * @param buffer the byte buffer to write to. - * @param index the index in buffer to write the character array to. - * @return the index after the last character. - */ - private static int writeCharArray(final int[] codePoints, final byte[] buffer, int index) { - for (int codePoint : codePoints) { - if (1 == getCharSize(codePoint)) { - buffer[index++] = (byte)codePoint; - } else { - buffer[index++] = (byte)(0xFF & (codePoint >> 16)); - buffer[index++] = (byte)(0xFF & (codePoint >> 8)); - buffer[index++] = (byte)(0xFF & codePoint); - } - } - return index; - } - - /** - * Writes a string with our character format to a byte buffer. - * - * This will also write the terminator byte. - * - * @param buffer the byte buffer to write to. - * @param origin the offset to write from. - * @param word the string to write. - * @return the size written, in bytes. - */ - private static int writeString(final byte[] buffer, final int origin, - final String word) { - final int length = word.length(); - int index = origin; - for (int i = 0; i < length; i = word.offsetByCodePoints(i, 1)) { - final int codePoint = word.codePointAt(i); - if (1 == getCharSize(codePoint)) { - buffer[index++] = (byte)codePoint; - } else { - buffer[index++] = (byte)(0xFF & (codePoint >> 16)); - buffer[index++] = (byte)(0xFF & (codePoint >> 8)); - buffer[index++] = (byte)(0xFF & codePoint); - } - } - buffer[index++] = FormatSpec.GROUP_CHARACTERS_TERMINATOR; - return index - origin; - } - - /** - * Writes a string with our character format to a ByteArrayOutputStream. - * - * This will also write the terminator byte. - * - * @param buffer the ByteArrayOutputStream to write to. - * @param word the string to write. - */ - private static void writeString(final ByteArrayOutputStream buffer, final String word) { - final int length = word.length(); - for (int i = 0; i < length; i = word.offsetByCodePoints(i, 1)) { - final int codePoint = word.codePointAt(i); - if (1 == getCharSize(codePoint)) { - buffer.write((byte) codePoint); - } else { - buffer.write((byte) (0xFF & (codePoint >> 16))); - buffer.write((byte) (0xFF & (codePoint >> 8))); - buffer.write((byte) (0xFF & codePoint)); - } - } - buffer.write(FormatSpec.GROUP_CHARACTERS_TERMINATOR); - } - - /** - * Reads a string from a buffer. This is the converse of the above method. - */ - private static String readString(final FusionDictionaryBufferInterface buffer) { - final StringBuilder s = new StringBuilder(); - int character = readChar(buffer); - while (character != FormatSpec.INVALID_CHARACTER) { - s.appendCodePoint(character); - character = readChar(buffer); - } - return s.toString(); - } - - /** - * Reads a character from the buffer. - * - * This follows the character format documented earlier in this source file. - * - * @param buffer the buffer, positioned over an encoded character. - * @return the character code. - */ - static int readChar(final FusionDictionaryBufferInterface buffer) { - int character = buffer.readUnsignedByte(); - if (!fitsOnOneByte(character)) { - if (FormatSpec.GROUP_CHARACTERS_TERMINATOR == character) { - return FormatSpec.INVALID_CHARACTER; - } - character <<= 16; - character += buffer.readUnsignedShort(); - } - return character; - } - } - - /** - * Compute the binary size of the character array. - * - * If only one character, this is the size of this character. If many, it's the sum of their - * sizes + 1 byte for the terminator. - * - * @param characters the character array - * @return the size of the char array, including the terminator if any - */ - static int getGroupCharactersSize(final int[] characters) { - int size = CharEncoding.getCharArraySize(characters); - if (characters.length > 1) size += FormatSpec.GROUP_TERMINATOR_SIZE; - return size; - } - - /** - * Compute the binary size of the character array in a group - * - * If only one character, this is the size of this character. If many, it's the sum of their - * sizes + 1 byte for the terminator. - * - * @param group the group - * @return the size of the char array, including the terminator if any - */ - private static int getGroupCharactersSize(final CharGroup group) { - return getGroupCharactersSize(group.mChars); - } - - /** - * Compute the binary size of the group count - * @param count the group count - * @return the size of the group count, either 1 or 2 bytes. - */ - public static int getGroupCountSize(final int count) { - if (FormatSpec.MAX_CHARGROUPS_FOR_ONE_BYTE_CHARGROUP_COUNT >= count) { - return 1; - } else if (FormatSpec.MAX_CHARGROUPS_IN_A_NODE >= count) { - return 2; - } else { - throw new RuntimeException("Can't have more than " - + FormatSpec.MAX_CHARGROUPS_IN_A_NODE + " groups in a node (found " + count - + ")"); - } - } - - /** - * Compute the binary size of the group count for a node - * @param node the node - * @return the size of the group count, either 1 or 2 bytes. - */ - private static int getGroupCountSize(final Node node) { - return getGroupCountSize(node.mData.size()); - } - - /** - * Compute the size of a shortcut in bytes. - */ - private static int getShortcutSize(final WeightedString shortcut) { - int size = FormatSpec.GROUP_ATTRIBUTE_FLAGS_SIZE; - final String word = shortcut.mWord; - final int length = word.length(); - for (int i = 0; i < length; i = word.offsetByCodePoints(i, 1)) { - final int codePoint = word.codePointAt(i); - size += CharEncoding.getCharSize(codePoint); - } - size += FormatSpec.GROUP_TERMINATOR_SIZE; - return size; - } - - /** - * Compute the size of a shortcut list in bytes. - * - * This is known in advance and does not change according to position in the file - * like address lists do. - */ - static int getShortcutListSize(final ArrayList<WeightedString> shortcutList) { - if (null == shortcutList) return 0; - int size = FormatSpec.GROUP_SHORTCUT_LIST_SIZE_SIZE; - for (final WeightedString shortcut : shortcutList) { - size += getShortcutSize(shortcut); - } - return size; - } - - /** - * Compute the maximum size of a CharGroup, assuming 3-byte addresses for everything. - * - * @param group the CharGroup to compute the size of. - * @param options file format options. - * @return the maximum size of the group. - */ - private static int getCharGroupMaximumSize(final CharGroup group, final FormatOptions options) { - int size = getGroupHeaderSize(group, options); - // If terminal, one byte for the frequency - if (group.isTerminal()) size += FormatSpec.GROUP_FREQUENCY_SIZE; - size += FormatSpec.GROUP_MAX_ADDRESS_SIZE; // For children address - size += getShortcutListSize(group.mShortcutTargets); - if (null != group.mBigrams) { - size += (FormatSpec.GROUP_ATTRIBUTE_FLAGS_SIZE - + FormatSpec.GROUP_ATTRIBUTE_MAX_ADDRESS_SIZE) - * group.mBigrams.size(); - } - return size; - } - - /** - * Compute the maximum size of a node, assuming 3-byte addresses for everything, and caches - * it in the 'actualSize' member of the node. - * - * @param node the node to compute the maximum size of. - * @param options file format options. - */ - private static void calculateNodeMaximumSize(final Node node, final FormatOptions options) { - int size = getGroupCountSize(node); - for (CharGroup g : node.mData) { - final int groupSize = getCharGroupMaximumSize(g, options); - g.mCachedSize = groupSize; - size += groupSize; - } - if (options.mSupportsDynamicUpdate) { - size += FormatSpec.FORWARD_LINK_ADDRESS_SIZE; - } - node.mCachedSize = size; - } - - /** - * Helper method to hide the actual value of the no children address. - */ - public static boolean hasChildrenAddress(final int address) { - return FormatSpec.NO_CHILDREN_ADDRESS != address; - } - - /** - * Helper method to check whether the group is moved. - */ - public static boolean isMovedGroup(final int flags, final FormatOptions options) { - return options.mSupportsDynamicUpdate - && ((flags & FormatSpec.MASK_GROUP_ADDRESS_TYPE) == FormatSpec.FLAG_IS_MOVED); - } - - /** - * Helper method to check whether the group is deleted. - */ - public static boolean isDeletedGroup(final int flags, final FormatOptions formatOptions) { - return formatOptions.mSupportsDynamicUpdate - && ((flags & FormatSpec.MASK_GROUP_ADDRESS_TYPE) == FormatSpec.FLAG_IS_DELETED); - } - - /** - * Helper method to check whether the dictionary can be updated dynamically. - */ - public static boolean supportsDynamicUpdate(final FormatOptions options) { - return options.mVersion >= FormatSpec.FIRST_VERSION_WITH_DYNAMIC_UPDATE - && options.mSupportsDynamicUpdate; - } - - /** - * Compute the size of the header (flag + [parent address] + characters size) of a CharGroup. - * - * @param group the group of which to compute the size of the header - * @param options file format options. - */ - private static int getGroupHeaderSize(final CharGroup group, final FormatOptions options) { - if (supportsDynamicUpdate(options)) { - return FormatSpec.GROUP_FLAGS_SIZE + FormatSpec.PARENT_ADDRESS_SIZE - + getGroupCharactersSize(group); - } else { - return FormatSpec.GROUP_FLAGS_SIZE + getGroupCharactersSize(group); - } - } - - private static final int UINT8_MAX = 0xFF; - private static final int UINT16_MAX = 0xFFFF; - private static final int UINT24_MAX = 0xFFFFFF; - - /** - * Compute the size, in bytes, that an address will occupy. - * - * This can be used either for children addresses (which are always positive) or for - * attribute, which may be positive or negative but - * store their sign bit separately. - * - * @param address the address - * @return the byte size. - */ - static int getByteSize(final int address) { - assert(address <= UINT24_MAX); - if (!hasChildrenAddress(address)) { - return 0; - } else if (Math.abs(address) <= UINT8_MAX) { - return 1; - } else if (Math.abs(address) <= UINT16_MAX) { - return 2; - } else { - return 3; - } - } - - private static final int SINT24_MAX = 0x7FFFFF; - private static final int MSB8 = 0x80; - private static final int MSB24 = 0x800000; - - // End utility methods. - - // This method is responsible for finding a nice ordering of the nodes that favors run-time - // cache performance and dictionary size. - /* package for tests */ static ArrayList<Node> flattenTree(final Node root) { - final int treeSize = FusionDictionary.countCharGroups(root); - MakedictLog.i("Counted nodes : " + treeSize); - final ArrayList<Node> flatTree = new ArrayList<Node>(treeSize); - return flattenTreeInner(flatTree, root); - } - - private static ArrayList<Node> flattenTreeInner(final ArrayList<Node> list, final Node node) { - // Removing the node is necessary if the tails are merged, because we would then - // add the same node several times when we only want it once. A number of places in - // the code also depends on any node being only once in the list. - // Merging tails can only be done if there are no attributes. Searching for attributes - // in LatinIME code depends on a total breadth-first ordering, which merging tails - // breaks. If there are no attributes, it should be fine (and reduce the file size) - // to merge tails, and removing the node from the list would be necessary. However, - // we don't merge tails because breaking the breadth-first ordering would result in - // extreme overhead at bigram lookup time (it would make the search function O(n) instead - // of the current O(log(n)), where n=number of nodes in the dictionary which is pretty - // high). - // If no nodes are ever merged, we can't have the same node twice in the list, hence - // searching for duplicates in unnecessary. It is also very performance consuming, - // since `list' is an ArrayList so it's an O(n) operation that runs on all nodes, making - // this simple list.remove operation O(n*n) overall. On Android this overhead is very - // high. - // For future reference, the code to remove duplicate is a simple : list.remove(node); - list.add(node); - final ArrayList<CharGroup> branches = node.mData; - final int nodeSize = branches.size(); - for (CharGroup group : branches) { - if (null != group.mChildren) flattenTreeInner(list, group.mChildren); - } - return list; - } - - /** - * Get the offset from a position inside a current node to a target node, during update. - * - * If the current node is before the target node, the target node has not been updated yet, - * so we should return the offset from the old position of the current node to the old position - * of the target node. If on the other hand the target is before the current node, it already - * has been updated, so we should return the offset from the new position in the current node - * to the new position in the target node. - * @param currentNode the node containing the CharGroup where the offset will be written - * @param offsetFromStartOfCurrentNode the offset, in bytes, from the start of currentNode - * @param targetNode the target node to get the offset to - * @return the offset to the target node - */ - private static int getOffsetToTargetNodeDuringUpdate(final Node currentNode, - final int offsetFromStartOfCurrentNode, final Node targetNode) { - final boolean isTargetBeforeCurrent = (targetNode.mCachedAddressBeforeUpdate - < currentNode.mCachedAddressBeforeUpdate); - if (isTargetBeforeCurrent) { - return targetNode.mCachedAddressAfterUpdate - - (currentNode.mCachedAddressAfterUpdate + offsetFromStartOfCurrentNode); - } else { - return targetNode.mCachedAddressBeforeUpdate - - (currentNode.mCachedAddressBeforeUpdate + offsetFromStartOfCurrentNode); - } - } - - /** - * Get the offset from a position inside a current node to a target CharGroup, during update. - * @param currentNode the node containing the CharGroup where the offset will be written - * @param offsetFromStartOfCurrentNode the offset, in bytes, from the start of currentNode - * @param targetCharGroup the target CharGroup to get the offset to - * @return the offset to the target CharGroup - */ - // TODO: is there any way to factorize this method with the one above? - private static int getOffsetToTargetCharGroupDuringUpdate(final Node currentNode, - final int offsetFromStartOfCurrentNode, final CharGroup targetCharGroup) { - final int oldOffsetBasePoint = currentNode.mCachedAddressBeforeUpdate - + offsetFromStartOfCurrentNode; - final boolean isTargetBeforeCurrent = (targetCharGroup.mCachedAddressBeforeUpdate - < oldOffsetBasePoint); - // If the target is before the current node, then its address has already been updated. - // We can use the AfterUpdate member, and compare it to our own member after update. - // Otherwise, the AfterUpdate member is not updated yet, so we need to use the BeforeUpdate - // member, and of course we have to compare this to our own address before update. - if (isTargetBeforeCurrent) { - final int newOffsetBasePoint = currentNode.mCachedAddressAfterUpdate - + offsetFromStartOfCurrentNode; - return targetCharGroup.mCachedAddressAfterUpdate - newOffsetBasePoint; - } else { - return targetCharGroup.mCachedAddressBeforeUpdate - oldOffsetBasePoint; - } - } - - /** - * Computes the actual node size, based on the cached addresses of the children nodes. - * - * Each node stores its tentative address. During dictionary address computing, these - * are not final, but they can be used to compute the node size (the node size depends - * on the address of the children because the number of bytes necessary to store an - * address depends on its numeric value. The return value indicates whether the node - * contents (as in, any of the addresses stored in the cache fields) have changed with - * respect to their previous value. - * - * @param node the node to compute the size of. - * @param dict the dictionary in which the word/attributes are to be found. - * @param formatOptions file format options. - * @return false if none of the cached addresses inside the node changed, true otherwise. - */ - private static boolean computeActualNodeSize(final Node node, final FusionDictionary dict, - final FormatOptions formatOptions) { - boolean changed = false; - int size = getGroupCountSize(node); - for (CharGroup group : node.mData) { - group.mCachedAddressAfterUpdate = node.mCachedAddressAfterUpdate + size; - if (group.mCachedAddressAfterUpdate != group.mCachedAddressBeforeUpdate) { - changed = true; - } - int groupSize = getGroupHeaderSize(group, formatOptions); - if (group.isTerminal()) groupSize += FormatSpec.GROUP_FREQUENCY_SIZE; - if (null == group.mChildren && formatOptions.mSupportsDynamicUpdate) { - groupSize += FormatSpec.SIGNED_CHILDREN_ADDRESS_SIZE; - } else if (null != group.mChildren) { - if (formatOptions.mSupportsDynamicUpdate) { - groupSize += FormatSpec.SIGNED_CHILDREN_ADDRESS_SIZE; - } else { - groupSize += getByteSize(getOffsetToTargetNodeDuringUpdate(node, - groupSize + size, group.mChildren)); - } - } - groupSize += getShortcutListSize(group.mShortcutTargets); - if (null != group.mBigrams) { - for (WeightedString bigram : group.mBigrams) { - final int offset = getOffsetToTargetCharGroupDuringUpdate(node, - groupSize + size + FormatSpec.GROUP_FLAGS_SIZE, - FusionDictionary.findWordInTree(dict.mRoot, bigram.mWord)); - groupSize += getByteSize(offset) + FormatSpec.GROUP_FLAGS_SIZE; - } - } - group.mCachedSize = groupSize; - size += groupSize; - } - if (formatOptions.mSupportsDynamicUpdate) { - size += FormatSpec.FORWARD_LINK_ADDRESS_SIZE; - } - if (node.mCachedSize != size) { - node.mCachedSize = size; - changed = true; - } - return changed; - } - - /** - * Initializes the cached addresses of nodes from their size. - * - * @param flatNodes the array of nodes. - * @param formatOptions file format options. - * @return the byte size of the entire stack. - */ - private static int initializeNodesCachedAddresses(final ArrayList<Node> flatNodes, - final FormatOptions formatOptions) { - int nodeOffset = 0; - for (final Node n : flatNodes) { - n.mCachedAddressBeforeUpdate = nodeOffset; - int groupCountSize = getGroupCountSize(n); - int groupOffset = 0; - for (final CharGroup g : n.mData) { - g.mCachedAddressBeforeUpdate = g.mCachedAddressAfterUpdate = - groupCountSize + nodeOffset + groupOffset; - groupOffset += g.mCachedSize; - } - final int nodeSize = groupCountSize + groupOffset - + (formatOptions.mSupportsDynamicUpdate - ? FormatSpec.FORWARD_LINK_ADDRESS_SIZE : 0); - nodeOffset += n.mCachedSize; - } - return nodeOffset; - } - - /** - * Updates the cached addresses of nodes after recomputing their new positions. - * - * @param flatNodes the array of nodes. - */ - private static void updateNodeCachedAddresses(final ArrayList<Node> flatNodes) { - for (final Node n : flatNodes) { - n.mCachedAddressBeforeUpdate = n.mCachedAddressAfterUpdate; - for (final CharGroup g : n.mData) { - g.mCachedAddressBeforeUpdate = g.mCachedAddressAfterUpdate; - } - } - } - - /** - * Compute the cached parent addresses after all has been updated. - * - * The parent addresses are used by some binary formats at write-to-disk time. Not all formats - * need them. In particular, version 2 does not need them, and version 3 does. - * - * @param flatNodes the flat array of nodes to fill in - */ - private static void computeParentAddresses(final ArrayList<Node> flatNodes) { - for (final Node node : flatNodes) { - for (final CharGroup group : node.mData) { - if (null != group.mChildren) { - // Assign my address to children's parent address - // Here BeforeUpdate and AfterUpdate addresses have the same value, so it - // does not matter which we use. - group.mChildren.mCachedParentAddress = group.mCachedAddressAfterUpdate - - group.mChildren.mCachedAddressAfterUpdate; - } - } - } - } - - /** - * Compute the addresses and sizes of an ordered node array. - * - * This method takes a node array and will update its cached address and size values - * so that they can be written into a file. It determines the smallest size each of the - * nodes can be given the addresses of its children and attributes, and store that into - * each node. - * The order of the node is given by the order of the array. This method makes no effort - * to find a good order; it only mechanically computes the size this order results in. - * - * @param dict the dictionary - * @param flatNodes the ordered array of nodes - * @param formatOptions file format options. - * @return the same array it was passed. The nodes have been updated for address and size. - */ - private static ArrayList<Node> computeAddresses(final FusionDictionary dict, - final ArrayList<Node> flatNodes, final FormatOptions formatOptions) { - // First get the worst possible sizes and offsets - for (final Node n : flatNodes) calculateNodeMaximumSize(n, formatOptions); - final int offset = initializeNodesCachedAddresses(flatNodes, formatOptions); - - MakedictLog.i("Compressing the array addresses. Original size : " + offset); - MakedictLog.i("(Recursively seen size : " + offset + ")"); - - int passes = 0; - boolean changesDone = false; - do { - changesDone = false; - int nodeStartOffset = 0; - for (final Node n : flatNodes) { - n.mCachedAddressAfterUpdate = nodeStartOffset; - final int oldNodeSize = n.mCachedSize; - final boolean changed = computeActualNodeSize(n, dict, formatOptions); - final int newNodeSize = n.mCachedSize; - if (oldNodeSize < newNodeSize) throw new RuntimeException("Increased size ?!"); - nodeStartOffset += newNodeSize; - changesDone |= changed; - } - updateNodeCachedAddresses(flatNodes); - ++passes; - if (passes > MAX_PASSES) throw new RuntimeException("Too many passes - probably a bug"); - } while (changesDone); - - if (formatOptions.mSupportsDynamicUpdate) { - computeParentAddresses(flatNodes); - } - final Node lastNode = flatNodes.get(flatNodes.size() - 1); - MakedictLog.i("Compression complete in " + passes + " passes."); - MakedictLog.i("After address compression : " - + (lastNode.mCachedAddressAfterUpdate + lastNode.mCachedSize)); - - return flatNodes; - } - - /** - * Sanity-checking method. - * - * This method checks an array of node for juxtaposition, that is, it will do - * nothing if each node's cached address is actually the previous node's address - * plus the previous node's size. - * If this is not the case, it will throw an exception. - * - * @param array the array node to check - */ - private static void checkFlatNodeArray(final ArrayList<Node> array) { - int offset = 0; - int index = 0; - for (final Node n : array) { - // BeforeUpdate and AfterUpdate addresses are the same here, so it does not matter - // which we use. - if (n.mCachedAddressAfterUpdate != offset) { - throw new RuntimeException("Wrong address for node " + index - + " : expected " + offset + ", got " + n.mCachedAddressAfterUpdate); - } - ++index; - offset += n.mCachedSize; - } - } - - /** - * Helper method to write a variable-size address to a file. - * - * @param buffer the buffer to write to. - * @param index the index in the buffer to write the address to. - * @param address the address to write. - * @return the size in bytes the address actually took. - */ - private static int writeVariableAddress(final byte[] buffer, int index, final int address) { - switch (getByteSize(address)) { - case 1: - buffer[index++] = (byte)address; - return 1; - case 2: - buffer[index++] = (byte)(0xFF & (address >> 8)); - buffer[index++] = (byte)(0xFF & address); - return 2; - case 3: - buffer[index++] = (byte)(0xFF & (address >> 16)); - buffer[index++] = (byte)(0xFF & (address >> 8)); - buffer[index++] = (byte)(0xFF & address); - return 3; - case 0: - return 0; - default: - throw new RuntimeException("Address " + address + " has a strange size"); - } - } - - /** - * Helper method to write a variable-size signed address to a file. - * - * @param buffer the buffer to write to. - * @param index the index in the buffer to write the address to. - * @param address the address to write. - * @return the size in bytes the address actually took. - */ - private static int writeVariableSignedAddress(final byte[] buffer, int index, - final int address) { - if (!hasChildrenAddress(address)) { - buffer[index] = buffer[index + 1] = buffer[index + 2] = 0; - } else { - final int absAddress = Math.abs(address); - buffer[index++] = (byte)((address < 0 ? MSB8 : 0) | (0xFF & (absAddress >> 16))); - buffer[index++] = (byte)(0xFF & (absAddress >> 8)); - buffer[index++] = (byte)(0xFF & absAddress); - } - return 3; - } - - /** - * Makes the flag value for a char group. - * - * @param hasMultipleChars whether the group has multiple chars. - * @param isTerminal whether the group is terminal. - * @param childrenAddressSize the size of a children address. - * @param hasShortcuts whether the group has shortcuts. - * @param hasBigrams whether the group has bigrams. - * @param isNotAWord whether the group is not a word. - * @param isBlackListEntry whether the group is a blacklist entry. - * @param formatOptions file format options. - * @return the flags - */ - static int makeCharGroupFlags(final boolean hasMultipleChars, final boolean isTerminal, - final int childrenAddressSize, final boolean hasShortcuts, final boolean hasBigrams, - final boolean isNotAWord, final boolean isBlackListEntry, - final FormatOptions formatOptions) { - byte flags = 0; - if (hasMultipleChars) flags |= FormatSpec.FLAG_HAS_MULTIPLE_CHARS; - if (isTerminal) flags |= FormatSpec.FLAG_IS_TERMINAL; - if (formatOptions.mSupportsDynamicUpdate) { - flags |= FormatSpec.FLAG_IS_NOT_MOVED; - } else if (true) { - switch (childrenAddressSize) { - case 1: - flags |= FormatSpec.FLAG_GROUP_ADDRESS_TYPE_ONEBYTE; - break; - case 2: - flags |= FormatSpec.FLAG_GROUP_ADDRESS_TYPE_TWOBYTES; - break; - case 3: - flags |= FormatSpec.FLAG_GROUP_ADDRESS_TYPE_THREEBYTES; - break; - case 0: - flags |= FormatSpec.FLAG_GROUP_ADDRESS_TYPE_NOADDRESS; - break; - default: - throw new RuntimeException("Node with a strange address"); - } - } - if (hasShortcuts) flags |= FormatSpec.FLAG_HAS_SHORTCUT_TARGETS; - if (hasBigrams) flags |= FormatSpec.FLAG_HAS_BIGRAMS; - if (isNotAWord) flags |= FormatSpec.FLAG_IS_NOT_A_WORD; - if (isBlackListEntry) flags |= FormatSpec.FLAG_IS_BLACKLISTED; - return flags; - } - - private static byte makeCharGroupFlags(final CharGroup group, final int groupAddress, - final int childrenOffset, final FormatOptions formatOptions) { - return (byte) makeCharGroupFlags(group.mChars.length > 1, group.mFrequency >= 0, - getByteSize(childrenOffset), group.mShortcutTargets != null, group.mBigrams != null, - group.mIsNotAWord, group.mIsBlacklistEntry, formatOptions); - } - - /** - * Makes the flag value for a bigram. - * - * @param more whether there are more bigrams after this one. - * @param offset the offset of the bigram. - * @param bigramFrequency the frequency of the bigram, 0..255. - * @param unigramFrequency the unigram frequency of the same word, 0..255. - * @param word the second bigram, for debugging purposes - * @return the flags - */ - private static final int makeBigramFlags(final boolean more, final int offset, - int bigramFrequency, final int unigramFrequency, final String word) { - int bigramFlags = (more ? FormatSpec.FLAG_ATTRIBUTE_HAS_NEXT : 0) - + (offset < 0 ? FormatSpec.FLAG_ATTRIBUTE_OFFSET_NEGATIVE : 0); - switch (getByteSize(offset)) { - case 1: - bigramFlags |= FormatSpec.FLAG_ATTRIBUTE_ADDRESS_TYPE_ONEBYTE; - break; - case 2: - bigramFlags |= FormatSpec.FLAG_ATTRIBUTE_ADDRESS_TYPE_TWOBYTES; - break; - case 3: - bigramFlags |= FormatSpec.FLAG_ATTRIBUTE_ADDRESS_TYPE_THREEBYTES; - break; - default: - throw new RuntimeException("Strange offset size"); - } - if (unigramFrequency > bigramFrequency) { - MakedictLog.e("Unigram freq is superior to bigram freq for \"" + word - + "\". Bigram freq is " + bigramFrequency + ", unigram freq for " - + word + " is " + unigramFrequency); - bigramFrequency = unigramFrequency; - } - // We compute the difference between 255 (which means probability = 1) and the - // unigram score. We split this into a number of discrete steps. - // Now, the steps are numbered 0~15; 0 represents an increase of 1 step while 15 - // represents an increase of 16 steps: a value of 15 will be interpreted as the median - // value of the 16th step. In all justice, if the bigram frequency is low enough to be - // rounded below the first step (which means it is less than half a step higher than the - // unigram frequency) then the unigram frequency itself is the best approximation of the - // bigram freq that we could possibly supply, hence we should *not* include this bigram - // in the file at all. - // until this is done, we'll write 0 and slightly overestimate this case. - // In other words, 0 means "between 0.5 step and 1.5 step", 1 means "between 1.5 step - // and 2.5 steps", and 15 means "between 15.5 steps and 16.5 steps". So we want to - // divide our range [unigramFreq..MAX_TERMINAL_FREQUENCY] in 16.5 steps to get the - // step size. Then we compute the start of the first step (the one where value 0 starts) - // by adding half-a-step to the unigramFrequency. From there, we compute the integer - // number of steps to the bigramFrequency. One last thing: we want our steps to include - // their lower bound and exclude their higher bound so we need to have the first step - // start at exactly 1 unit higher than floor(unigramFreq + half a step). - // Note : to reconstruct the score, the dictionary reader will need to divide - // MAX_TERMINAL_FREQUENCY - unigramFreq by 16.5 likewise to get the value of the step, - // and add (discretizedFrequency + 0.5 + 0.5) times this value to get the best - // approximation. (0.5 to get the first step start, and 0.5 to get the middle of the - // step pointed by the discretized frequency. - final float stepSize = - (FormatSpec.MAX_TERMINAL_FREQUENCY - unigramFrequency) - / (1.5f + FormatSpec.MAX_BIGRAM_FREQUENCY); - final float firstStepStart = 1 + unigramFrequency + (stepSize / 2.0f); - final int discretizedFrequency = (int)((bigramFrequency - firstStepStart) / stepSize); - // If the bigram freq is less than half-a-step higher than the unigram freq, we get -1 - // here. The best approximation would be the unigram freq itself, so we should not - // include this bigram in the dictionary. For now, register as 0, and live with the - // small over-estimation that we get in this case. TODO: actually remove this bigram - // if discretizedFrequency < 0. - final int finalBigramFrequency = discretizedFrequency > 0 ? discretizedFrequency : 0; - bigramFlags += finalBigramFrequency & FormatSpec.FLAG_ATTRIBUTE_FREQUENCY; - return bigramFlags; - } - - /** - * Makes the 2-byte value for options flags. - */ - private static final int makeOptionsValue(final FusionDictionary dictionary, - final FormatOptions formatOptions) { - final DictionaryOptions options = dictionary.mOptions; - final boolean hasBigrams = dictionary.hasBigrams(); - return (options.mFrenchLigatureProcessing ? FormatSpec.FRENCH_LIGATURE_PROCESSING_FLAG : 0) - + (options.mGermanUmlautProcessing ? FormatSpec.GERMAN_UMLAUT_PROCESSING_FLAG : 0) - + (hasBigrams ? FormatSpec.CONTAINS_BIGRAMS_FLAG : 0) - + (formatOptions.mSupportsDynamicUpdate ? FormatSpec.SUPPORTS_DYNAMIC_UPDATE : 0); - } - - /** - * Makes the flag value for a shortcut. - * - * @param more whether there are more attributes after this one. - * @param frequency the frequency of the attribute, 0..15 - * @return the flags - */ - static final int makeShortcutFlags(final boolean more, final int frequency) { - return (more ? FormatSpec.FLAG_ATTRIBUTE_HAS_NEXT : 0) - + (frequency & FormatSpec.FLAG_ATTRIBUTE_FREQUENCY); - } - - private static final int writeParentAddress(final byte[] buffer, final int index, - final int address, final FormatOptions formatOptions) { - if (supportsDynamicUpdate(formatOptions)) { - if (address == FormatSpec.NO_PARENT_ADDRESS) { - buffer[index] = buffer[index + 1] = buffer[index + 2] = 0; - } else { - final int absAddress = Math.abs(address); - assert(absAddress <= SINT24_MAX); - buffer[index] = (byte)((address < 0 ? MSB8 : 0) - | ((absAddress >> 16) & 0xFF)); - buffer[index + 1] = (byte)((absAddress >> 8) & 0xFF); - buffer[index + 2] = (byte)(absAddress & 0xFF); - } - return index + 3; - } else { - return index; - } - } - - /** - * Write a node to memory. The node is expected to have its final position cached. - * - * This can be an empty map, but the more is inside the faster the lookups will be. It can - * be carried on as long as nodes do not move. - * - * @param dict the dictionary the node is a part of (for relative offsets). - * @param buffer the memory buffer to write to. - * @param node the node to write. - * @param formatOptions file format options. - * @return the address of the END of the node. - */ - @SuppressWarnings("unused") - private static int writePlacedNode(final FusionDictionary dict, byte[] buffer, - final Node node, final FormatOptions formatOptions) { - // TODO: Make the code in common with BinaryDictIOUtils#writeCharGroup - int index = node.mCachedAddressAfterUpdate; - - final int groupCount = node.mData.size(); - final int countSize = getGroupCountSize(node); - final int parentAddress = node.mCachedParentAddress; - if (1 == countSize) { - buffer[index++] = (byte)groupCount; - } else if (2 == countSize) { - // We need to signal 2-byte size by setting the top bit of the MSB to 1, so - // we | 0x80 to do this. - buffer[index++] = (byte)((groupCount >> 8) | 0x80); - buffer[index++] = (byte)(groupCount & 0xFF); - } else { - throw new RuntimeException("Strange size from getGroupCountSize : " + countSize); - } - int groupAddress = index; - for (int i = 0; i < groupCount; ++i) { - final CharGroup group = node.mData.get(i); - if (index != group.mCachedAddressAfterUpdate) { - throw new RuntimeException("Bug: write index is not the same as the cached address " - + "of the group : " + index + " <> " + group.mCachedAddressAfterUpdate); - } - groupAddress += getGroupHeaderSize(group, formatOptions); - // Sanity checks. - if (DBG && group.mFrequency > FormatSpec.MAX_TERMINAL_FREQUENCY) { - throw new RuntimeException("A node has a frequency > " - + FormatSpec.MAX_TERMINAL_FREQUENCY - + " : " + group.mFrequency); - } - if (group.mFrequency >= 0) groupAddress += FormatSpec.GROUP_FREQUENCY_SIZE; - final int childrenOffset = null == group.mChildren - ? FormatSpec.NO_CHILDREN_ADDRESS - : group.mChildren.mCachedAddressAfterUpdate - groupAddress; - buffer[index++] = - makeCharGroupFlags(group, groupAddress, childrenOffset, formatOptions); - - if (parentAddress == FormatSpec.NO_PARENT_ADDRESS) { - index = writeParentAddress(buffer, index, parentAddress, formatOptions); - } else { - index = writeParentAddress(buffer, index, parentAddress - + (node.mCachedAddressAfterUpdate - group.mCachedAddressAfterUpdate), - formatOptions); - } - - index = CharEncoding.writeCharArray(group.mChars, buffer, index); - if (group.hasSeveralChars()) { - buffer[index++] = FormatSpec.GROUP_CHARACTERS_TERMINATOR; - } - if (group.mFrequency >= 0) { - buffer[index++] = (byte) group.mFrequency; - } - - final int shift; - if (formatOptions.mSupportsDynamicUpdate) { - shift = writeVariableSignedAddress(buffer, index, childrenOffset); - } else { - shift = writeVariableAddress(buffer, index, childrenOffset); - } - index += shift; - groupAddress += shift; - - // Write shortcuts - if (null != group.mShortcutTargets) { - final int indexOfShortcutByteSize = index; - index += FormatSpec.GROUP_SHORTCUT_LIST_SIZE_SIZE; - groupAddress += FormatSpec.GROUP_SHORTCUT_LIST_SIZE_SIZE; - final Iterator<WeightedString> shortcutIterator = group.mShortcutTargets.iterator(); - while (shortcutIterator.hasNext()) { - final WeightedString target = shortcutIterator.next(); - ++groupAddress; - int shortcutFlags = makeShortcutFlags(shortcutIterator.hasNext(), - target.mFrequency); - buffer[index++] = (byte)shortcutFlags; - final int shortcutShift = CharEncoding.writeString(buffer, index, target.mWord); - index += shortcutShift; - groupAddress += shortcutShift; - } - final int shortcutByteSize = index - indexOfShortcutByteSize; - if (shortcutByteSize > 0xFFFF) { - throw new RuntimeException("Shortcut list too large"); - } - buffer[indexOfShortcutByteSize] = (byte)(shortcutByteSize >> 8); - buffer[indexOfShortcutByteSize + 1] = (byte)(shortcutByteSize & 0xFF); - } - // Write bigrams - if (null != group.mBigrams) { - final Iterator<WeightedString> bigramIterator = group.mBigrams.iterator(); - while (bigramIterator.hasNext()) { - final WeightedString bigram = bigramIterator.next(); - final CharGroup target = - FusionDictionary.findWordInTree(dict.mRoot, bigram.mWord); - final int addressOfBigram = target.mCachedAddressAfterUpdate; - final int unigramFrequencyForThisWord = target.mFrequency; - ++groupAddress; - final int offset = addressOfBigram - groupAddress; - int bigramFlags = makeBigramFlags(bigramIterator.hasNext(), offset, - bigram.mFrequency, unigramFrequencyForThisWord, bigram.mWord); - buffer[index++] = (byte)bigramFlags; - final int bigramShift = writeVariableAddress(buffer, index, Math.abs(offset)); - index += bigramShift; - groupAddress += bigramShift; - } - } - - } - if (formatOptions.mSupportsDynamicUpdate) { - buffer[index] = buffer[index + 1] = buffer[index + 2] - = FormatSpec.NO_FORWARD_LINK_ADDRESS; - index += FormatSpec.FORWARD_LINK_ADDRESS_SIZE; - } - if (index != node.mCachedAddressAfterUpdate + node.mCachedSize) throw new RuntimeException( - "Not the same size : written " - + (index - node.mCachedAddressAfterUpdate) + " bytes from a node that should have " - + node.mCachedSize + " bytes"); - return index; - } - - /** - * Dumps a collection of useful statistics about a node array. - * - * This prints purely informative stuff, like the total estimated file size, the - * number of nodes, of character groups, the repartition of each address size, etc - * - * @param nodes the node array. - */ - private static void showStatistics(ArrayList<Node> nodes) { - int firstTerminalAddress = Integer.MAX_VALUE; - int lastTerminalAddress = Integer.MIN_VALUE; - int size = 0; - int charGroups = 0; - int maxGroups = 0; - int maxRuns = 0; - for (final Node n : nodes) { - if (maxGroups < n.mData.size()) maxGroups = n.mData.size(); - for (final CharGroup cg : n.mData) { - ++charGroups; - if (cg.mChars.length > maxRuns) maxRuns = cg.mChars.length; - if (cg.mFrequency >= 0) { - if (n.mCachedAddressAfterUpdate < firstTerminalAddress) - firstTerminalAddress = n.mCachedAddressAfterUpdate; - if (n.mCachedAddressAfterUpdate > lastTerminalAddress) - lastTerminalAddress = n.mCachedAddressAfterUpdate; - } - } - if (n.mCachedAddressAfterUpdate + n.mCachedSize > size) { - size = n.mCachedAddressAfterUpdate + n.mCachedSize; - } - } - final int[] groupCounts = new int[maxGroups + 1]; - final int[] runCounts = new int[maxRuns + 1]; - for (final Node n : nodes) { - ++groupCounts[n.mData.size()]; - for (final CharGroup cg : n.mData) { - ++runCounts[cg.mChars.length]; - } - } - - MakedictLog.i("Statistics:\n" - + " total file size " + size + "\n" - + " " + nodes.size() + " nodes\n" - + " " + charGroups + " groups (" + ((float)charGroups / nodes.size()) - + " groups per node)\n" - + " first terminal at " + firstTerminalAddress + "\n" - + " last terminal at " + lastTerminalAddress + "\n" - + " Group stats : max = " + maxGroups); - for (int i = 0; i < groupCounts.length; ++i) { - MakedictLog.i(" " + i + " : " + groupCounts[i]); - } - MakedictLog.i(" Character run stats : max = " + maxRuns); - for (int i = 0; i < runCounts.length; ++i) { - MakedictLog.i(" " + i + " : " + runCounts[i]); - } - } - - /** - * Dumps a FusionDictionary to a file. - * - * This is the public entry point to write a dictionary to a file. - * - * @param destination the stream to write the binary data to. - * @param dict the dictionary to write. - * @param formatOptions file format options. - */ - public static void writeDictionaryBinary(final OutputStream destination, - final FusionDictionary dict, final FormatOptions formatOptions) - throws IOException, UnsupportedFormatException { - - // Addresses are limited to 3 bytes, but since addresses can be relative to each node, the - // structure itself is not limited to 16MB. However, if it is over 16MB deciding the order - // of the nodes becomes a quite complicated problem, because though the dictionary itself - // does not have a size limit, each node must still be within 16MB of all its children and - // parents. As long as this is ensured, the dictionary file may grow to any size. - - final int version = formatOptions.mVersion; - if (version < FormatSpec.MINIMUM_SUPPORTED_VERSION - || version > FormatSpec.MAXIMUM_SUPPORTED_VERSION) { - throw new UnsupportedFormatException("Requested file format version " + version - + ", but this implementation only supports versions " - + FormatSpec.MINIMUM_SUPPORTED_VERSION + " through " - + FormatSpec.MAXIMUM_SUPPORTED_VERSION); - } - - ByteArrayOutputStream headerBuffer = new ByteArrayOutputStream(256); - - // The magic number in big-endian order. - if (version >= FormatSpec.FIRST_VERSION_WITH_HEADER_SIZE) { - // Magic number for version 2+. - headerBuffer.write((byte) (0xFF & (FormatSpec.VERSION_2_MAGIC_NUMBER >> 24))); - headerBuffer.write((byte) (0xFF & (FormatSpec.VERSION_2_MAGIC_NUMBER >> 16))); - headerBuffer.write((byte) (0xFF & (FormatSpec.VERSION_2_MAGIC_NUMBER >> 8))); - headerBuffer.write((byte) (0xFF & FormatSpec.VERSION_2_MAGIC_NUMBER)); - // Dictionary version. - headerBuffer.write((byte) (0xFF & (version >> 8))); - headerBuffer.write((byte) (0xFF & version)); - } else { - // Magic number for version 1. - headerBuffer.write((byte) (0xFF & (FormatSpec.VERSION_1_MAGIC_NUMBER >> 8))); - headerBuffer.write((byte) (0xFF & FormatSpec.VERSION_1_MAGIC_NUMBER)); - // Dictionary version. - headerBuffer.write((byte) (0xFF & version)); - } - // Options flags - final int options = makeOptionsValue(dict, formatOptions); - headerBuffer.write((byte) (0xFF & (options >> 8))); - headerBuffer.write((byte) (0xFF & options)); - if (version >= FormatSpec.FIRST_VERSION_WITH_HEADER_SIZE) { - final int headerSizeOffset = headerBuffer.size(); - // Placeholder to be written later with header size. - for (int i = 0; i < 4; ++i) { - headerBuffer.write(0); - } - // Write out the options. - for (final String key : dict.mOptions.mAttributes.keySet()) { - final String value = dict.mOptions.mAttributes.get(key); - CharEncoding.writeString(headerBuffer, key); - CharEncoding.writeString(headerBuffer, value); - } - final int size = headerBuffer.size(); - final byte[] bytes = headerBuffer.toByteArray(); - // Write out the header size. - bytes[headerSizeOffset] = (byte) (0xFF & (size >> 24)); - bytes[headerSizeOffset + 1] = (byte) (0xFF & (size >> 16)); - bytes[headerSizeOffset + 2] = (byte) (0xFF & (size >> 8)); - bytes[headerSizeOffset + 3] = (byte) (0xFF & (size >> 0)); - destination.write(bytes); - } else { - headerBuffer.writeTo(destination); - } - - headerBuffer.close(); - - // Leave the choice of the optimal node order to the flattenTree function. - MakedictLog.i("Flattening the tree..."); - ArrayList<Node> flatNodes = flattenTree(dict.mRoot); - - MakedictLog.i("Computing addresses..."); - computeAddresses(dict, flatNodes, formatOptions); - MakedictLog.i("Checking array..."); - if (DBG) checkFlatNodeArray(flatNodes); - - // Create a buffer that matches the final dictionary size. - final Node lastNode = flatNodes.get(flatNodes.size() - 1); - final int bufferSize = lastNode.mCachedAddressAfterUpdate + lastNode.mCachedSize; - final byte[] buffer = new byte[bufferSize]; - int index = 0; - - MakedictLog.i("Writing file..."); - int dataEndOffset = 0; - for (Node n : flatNodes) { - dataEndOffset = writePlacedNode(dict, buffer, n, formatOptions); - } - - if (DBG) showStatistics(flatNodes); - - destination.write(buffer, 0, dataEndOffset); - - destination.close(); - MakedictLog.i("Done"); - } - - - // Input methods: Read a binary dictionary to memory. - // readDictionaryBinary is the public entry point for them. - - static int getChildrenAddressSize(final int optionFlags, - final FormatOptions formatOptions) { - if (formatOptions.mSupportsDynamicUpdate) return FormatSpec.SIGNED_CHILDREN_ADDRESS_SIZE; - switch (optionFlags & FormatSpec.MASK_GROUP_ADDRESS_TYPE) { - case FormatSpec.FLAG_GROUP_ADDRESS_TYPE_ONEBYTE: - return 1; - case FormatSpec.FLAG_GROUP_ADDRESS_TYPE_TWOBYTES: - return 2; - case FormatSpec.FLAG_GROUP_ADDRESS_TYPE_THREEBYTES: - return 3; - case FormatSpec.FLAG_GROUP_ADDRESS_TYPE_NOADDRESS: - default: - return 0; - } - } - - static int readChildrenAddress(final FusionDictionaryBufferInterface buffer, - final int optionFlags, final FormatOptions options) { - if (options.mSupportsDynamicUpdate) { - final int address = buffer.readUnsignedInt24(); - if (address == 0) return FormatSpec.NO_CHILDREN_ADDRESS; - if ((address & MSB24) != 0) { - return -(address & SINT24_MAX); - } else { - return address; - } - } - int address; - switch (optionFlags & FormatSpec.MASK_GROUP_ADDRESS_TYPE) { - case FormatSpec.FLAG_GROUP_ADDRESS_TYPE_ONEBYTE: - return buffer.readUnsignedByte(); - case FormatSpec.FLAG_GROUP_ADDRESS_TYPE_TWOBYTES: - return buffer.readUnsignedShort(); - case FormatSpec.FLAG_GROUP_ADDRESS_TYPE_THREEBYTES: - return buffer.readUnsignedInt24(); - case FormatSpec.FLAG_GROUP_ADDRESS_TYPE_NOADDRESS: - default: - return FormatSpec.NO_CHILDREN_ADDRESS; - } - } - - static int readParentAddress(final FusionDictionaryBufferInterface buffer, - final FormatOptions formatOptions) { - if (supportsDynamicUpdate(formatOptions)) { - final int parentAddress = buffer.readUnsignedInt24(); - final int sign = ((parentAddress & MSB24) != 0) ? -1 : 1; - return sign * (parentAddress & SINT24_MAX); - } else { - return FormatSpec.NO_PARENT_ADDRESS; - } - } - - private static final int[] CHARACTER_BUFFER = new int[FormatSpec.MAX_WORD_LENGTH]; - public static CharGroupInfo readCharGroup(final FusionDictionaryBufferInterface buffer, - final int originalGroupAddress, final FormatOptions options) { - int addressPointer = originalGroupAddress; - final int flags = buffer.readUnsignedByte(); - ++addressPointer; - - final int parentAddress = readParentAddress(buffer, options); - if (supportsDynamicUpdate(options)) { - addressPointer += 3; - } - - final int characters[]; - if (0 != (flags & FormatSpec.FLAG_HAS_MULTIPLE_CHARS)) { - int index = 0; - int character = CharEncoding.readChar(buffer); - addressPointer += CharEncoding.getCharSize(character); - while (-1 != character) { - // FusionDictionary is making sure that the length of the word is smaller than - // MAX_WORD_LENGTH. - // So we'll never write past the end of CHARACTER_BUFFER. - CHARACTER_BUFFER[index++] = character; - character = CharEncoding.readChar(buffer); - addressPointer += CharEncoding.getCharSize(character); - } - characters = Arrays.copyOfRange(CHARACTER_BUFFER, 0, index); - } else { - final int character = CharEncoding.readChar(buffer); - addressPointer += CharEncoding.getCharSize(character); - characters = new int[] { character }; - } - final int frequency; - if (0 != (FormatSpec.FLAG_IS_TERMINAL & flags)) { - ++addressPointer; - frequency = buffer.readUnsignedByte(); - } else { - frequency = CharGroup.NOT_A_TERMINAL; - } - int childrenAddress = readChildrenAddress(buffer, flags, options); - if (childrenAddress != FormatSpec.NO_CHILDREN_ADDRESS) { - childrenAddress += addressPointer; - } - addressPointer += getChildrenAddressSize(flags, options); - ArrayList<WeightedString> shortcutTargets = null; - if (0 != (flags & FormatSpec.FLAG_HAS_SHORTCUT_TARGETS)) { - final int pointerBefore = buffer.position(); - shortcutTargets = new ArrayList<WeightedString>(); - buffer.readUnsignedShort(); // Skip the size - while (true) { - final int targetFlags = buffer.readUnsignedByte(); - final String word = CharEncoding.readString(buffer); - shortcutTargets.add(new WeightedString(word, - targetFlags & FormatSpec.FLAG_ATTRIBUTE_FREQUENCY)); - if (0 == (targetFlags & FormatSpec.FLAG_ATTRIBUTE_HAS_NEXT)) break; - } - addressPointer += buffer.position() - pointerBefore; - } - ArrayList<PendingAttribute> bigrams = null; - if (0 != (flags & FormatSpec.FLAG_HAS_BIGRAMS)) { - bigrams = new ArrayList<PendingAttribute>(); - int bigramCount = 0; - while (bigramCount++ < FormatSpec.MAX_BIGRAMS_IN_A_GROUP) { - final int bigramFlags = buffer.readUnsignedByte(); - ++addressPointer; - final int sign = 0 == (bigramFlags & FormatSpec.FLAG_ATTRIBUTE_OFFSET_NEGATIVE) - ? 1 : -1; - int bigramAddress = addressPointer; - switch (bigramFlags & FormatSpec.MASK_ATTRIBUTE_ADDRESS_TYPE) { - case FormatSpec.FLAG_ATTRIBUTE_ADDRESS_TYPE_ONEBYTE: - bigramAddress += sign * buffer.readUnsignedByte(); - addressPointer += 1; - break; - case FormatSpec.FLAG_ATTRIBUTE_ADDRESS_TYPE_TWOBYTES: - bigramAddress += sign * buffer.readUnsignedShort(); - addressPointer += 2; - break; - case FormatSpec.FLAG_ATTRIBUTE_ADDRESS_TYPE_THREEBYTES: - final int offset = (buffer.readUnsignedByte() << 16) - + buffer.readUnsignedShort(); - bigramAddress += sign * offset; - addressPointer += 3; - break; - default: - throw new RuntimeException("Has bigrams with no address"); - } - bigrams.add(new PendingAttribute(bigramFlags & FormatSpec.FLAG_ATTRIBUTE_FREQUENCY, - bigramAddress)); - if (0 == (bigramFlags & FormatSpec.FLAG_ATTRIBUTE_HAS_NEXT)) break; - } - if (bigramCount >= FormatSpec.MAX_BIGRAMS_IN_A_GROUP) { - MakedictLog.d("too many bigrams in a group."); - } - } - return new CharGroupInfo(originalGroupAddress, addressPointer, flags, characters, frequency, - parentAddress, childrenAddress, shortcutTargets, bigrams); - } - - /** - * Reads and returns the char group count out of a buffer and forwards the pointer. - */ - public static int readCharGroupCount(final FusionDictionaryBufferInterface buffer) { - final int msb = buffer.readUnsignedByte(); - if (FormatSpec.MAX_CHARGROUPS_FOR_ONE_BYTE_CHARGROUP_COUNT >= msb) { - return msb; - } else { - return ((FormatSpec.MAX_CHARGROUPS_FOR_ONE_BYTE_CHARGROUP_COUNT & msb) << 8) - + buffer.readUnsignedByte(); - } - } - - // The word cache here is a stopgap bandaid to help the catastrophic performance - // of this method. Since it performs direct, unbuffered random access to the file and - // may be called hundreds of thousands of times, the resulting performance is not - // reasonable without some kind of cache. Thus: - private static TreeMap<Integer, WeightedString> wordCache = - new TreeMap<Integer, WeightedString>(); - /** - * Finds, as a string, the word at the address passed as an argument. - * - * @param buffer the buffer to read from. - * @param headerSize the size of the header. - * @param address the address to seek. - * @param formatOptions file format options. - * @return the word with its frequency, as a weighted string. - */ - /* package for tests */ static WeightedString getWordAtAddress( - final FusionDictionaryBufferInterface buffer, final int headerSize, final int address, - final FormatOptions formatOptions) { - final WeightedString cachedString = wordCache.get(address); - if (null != cachedString) return cachedString; - - final WeightedString result; - final int originalPointer = buffer.position(); - buffer.position(address); - - if (supportsDynamicUpdate(formatOptions)) { - result = getWordAtAddressWithParentAddress(buffer, headerSize, address, formatOptions); - } else { - result = getWordAtAddressWithoutParentAddress(buffer, headerSize, address, - formatOptions); - } - - wordCache.put(address, result); - buffer.position(originalPointer); - return result; - } - - // TODO: static!? This will behave erratically when used in multi-threaded code. - // We need to fix this - private static int[] sGetWordBuffer = new int[FormatSpec.MAX_WORD_LENGTH]; - @SuppressWarnings("unused") - private static WeightedString getWordAtAddressWithParentAddress( - final FusionDictionaryBufferInterface buffer, final int headerSize, final int address, - final FormatOptions options) { - int currentAddress = address; - int index = FormatSpec.MAX_WORD_LENGTH - 1; - int frequency = Integer.MIN_VALUE; - // the length of the path from the root to the leaf is limited by MAX_WORD_LENGTH - for (int count = 0; count < FormatSpec.MAX_WORD_LENGTH; ++count) { - CharGroupInfo currentInfo; - int loopCounter = 0; - do { - buffer.position(currentAddress + headerSize); - currentInfo = readCharGroup(buffer, currentAddress, options); - if (isMovedGroup(currentInfo.mFlags, options)) { - currentAddress = currentInfo.mParentAddress + currentInfo.mOriginalAddress; - } - if (DBG && loopCounter++ > MAX_JUMPS) { - MakedictLog.d("Too many jumps - probably a bug"); - } - } while (isMovedGroup(currentInfo.mFlags, options)); - if (Integer.MIN_VALUE == frequency) frequency = currentInfo.mFrequency; - for (int i = 0; i < currentInfo.mCharacters.length; ++i) { - sGetWordBuffer[index--] = - currentInfo.mCharacters[currentInfo.mCharacters.length - i - 1]; - } - if (currentInfo.mParentAddress == FormatSpec.NO_PARENT_ADDRESS) break; - currentAddress = currentInfo.mParentAddress + currentInfo.mOriginalAddress; - } - - return new WeightedString( - new String(sGetWordBuffer, index + 1, FormatSpec.MAX_WORD_LENGTH - index - 1), - frequency); - } - - private static WeightedString getWordAtAddressWithoutParentAddress( - final FusionDictionaryBufferInterface buffer, final int headerSize, final int address, - final FormatOptions options) { - buffer.position(headerSize); - final int count = readCharGroupCount(buffer); - int groupOffset = getGroupCountSize(count); - final StringBuilder builder = new StringBuilder(); - WeightedString result = null; - - CharGroupInfo last = null; - for (int i = count - 1; i >= 0; --i) { - CharGroupInfo info = readCharGroup(buffer, groupOffset, options); - groupOffset = info.mEndAddress; - if (info.mOriginalAddress == address) { - builder.append(new String(info.mCharacters, 0, info.mCharacters.length)); - result = new WeightedString(builder.toString(), info.mFrequency); - break; // and return - } - if (hasChildrenAddress(info.mChildrenAddress)) { - if (info.mChildrenAddress > address) { - if (null == last) continue; - builder.append(new String(last.mCharacters, 0, last.mCharacters.length)); - buffer.position(last.mChildrenAddress + headerSize); - i = readCharGroupCount(buffer); - groupOffset = last.mChildrenAddress + getGroupCountSize(i); - last = null; - continue; - } - last = info; - } - if (0 == i && hasChildrenAddress(last.mChildrenAddress)) { - builder.append(new String(last.mCharacters, 0, last.mCharacters.length)); - buffer.position(last.mChildrenAddress + headerSize); - i = readCharGroupCount(buffer); - groupOffset = last.mChildrenAddress + getGroupCountSize(i); - last = null; - continue; - } - } - return result; - } - - /** - * Reads a single node from a buffer. - * - * This methods reads the file at the current position. A node is fully expected to start at - * the current position. - * This will recursively read other nodes into the structure, populating the reverse - * maps on the fly and using them to keep track of already read nodes. - * - * @param buffer the buffer, correctly positioned at the start of a node. - * @param headerSize the size, in bytes, of the file header. - * @param reverseNodeMap a mapping from addresses to already read nodes. - * @param reverseGroupMap a mapping from addresses to already read character groups. - * @param options file format options. - * @return the read node with all his children already read. - */ - private static Node readNode(final FusionDictionaryBufferInterface buffer, final int headerSize, - final Map<Integer, Node> reverseNodeMap, final Map<Integer, CharGroup> reverseGroupMap, - final FormatOptions options) - throws IOException { - final ArrayList<CharGroup> nodeContents = new ArrayList<CharGroup>(); - final int nodeOrigin = buffer.position() - headerSize; - - do { // Scan the linked-list node. - final int nodeHeadPosition = buffer.position() - headerSize; - final int count = readCharGroupCount(buffer); - int groupOffset = nodeHeadPosition + getGroupCountSize(count); - for (int i = count; i > 0; --i) { // Scan the array of CharGroup. - CharGroupInfo info = readCharGroup(buffer, groupOffset, options); - if (isMovedGroup(info.mFlags, options)) continue; - ArrayList<WeightedString> shortcutTargets = info.mShortcutTargets; - ArrayList<WeightedString> bigrams = null; - if (null != info.mBigrams) { - bigrams = new ArrayList<WeightedString>(); - for (PendingAttribute bigram : info.mBigrams) { - final WeightedString word = getWordAtAddress( - buffer, headerSize, bigram.mAddress, options); - final int reconstructedFrequency = - reconstructBigramFrequency(word.mFrequency, bigram.mFrequency); - bigrams.add(new WeightedString(word.mWord, reconstructedFrequency)); - } - } - if (hasChildrenAddress(info.mChildrenAddress)) { - Node children = reverseNodeMap.get(info.mChildrenAddress); - if (null == children) { - final int currentPosition = buffer.position(); - buffer.position(info.mChildrenAddress + headerSize); - children = readNode( - buffer, headerSize, reverseNodeMap, reverseGroupMap, options); - buffer.position(currentPosition); - } - nodeContents.add( - new CharGroup(info.mCharacters, shortcutTargets, bigrams, - info.mFrequency, - 0 != (info.mFlags & FormatSpec.FLAG_IS_NOT_A_WORD), - 0 != (info.mFlags & FormatSpec.FLAG_IS_BLACKLISTED), children)); - } else { - nodeContents.add( - new CharGroup(info.mCharacters, shortcutTargets, bigrams, - info.mFrequency, - 0 != (info.mFlags & FormatSpec.FLAG_IS_NOT_A_WORD), - 0 != (info.mFlags & FormatSpec.FLAG_IS_BLACKLISTED))); - } - groupOffset = info.mEndAddress; - } - - // reach the end of the array. - if (options.mSupportsDynamicUpdate) { - final int nextAddress = buffer.readUnsignedInt24(); - if (nextAddress >= 0 && nextAddress < buffer.limit()) { - buffer.position(nextAddress); - } else { - break; - } - } - } while (options.mSupportsDynamicUpdate && - buffer.position() != FormatSpec.NO_FORWARD_LINK_ADDRESS); - - final Node node = new Node(nodeContents); - node.mCachedAddressBeforeUpdate = nodeOrigin; - node.mCachedAddressAfterUpdate = nodeOrigin; - reverseNodeMap.put(node.mCachedAddressAfterUpdate, node); - return node; - } - - /** - * Helper function to get the binary format version from the header. - * @throws IOException - */ - private static int getFormatVersion(final FusionDictionaryBufferInterface buffer) - throws IOException { - final int magic_v1 = buffer.readUnsignedShort(); - if (FormatSpec.VERSION_1_MAGIC_NUMBER == magic_v1) return buffer.readUnsignedByte(); - final int magic_v2 = (magic_v1 << 16) + buffer.readUnsignedShort(); - if (FormatSpec.VERSION_2_MAGIC_NUMBER == magic_v2) return buffer.readUnsignedShort(); - return FormatSpec.NOT_A_VERSION_NUMBER; - } - - /** - * Helper function to get and validate the binary format version. - * @throws UnsupportedFormatException - * @throws IOException - */ - private static int checkFormatVersion(final FusionDictionaryBufferInterface buffer) - throws IOException, UnsupportedFormatException { - final int version = getFormatVersion(buffer); - if (version < FormatSpec.MINIMUM_SUPPORTED_VERSION - || version > FormatSpec.MAXIMUM_SUPPORTED_VERSION) { - throw new UnsupportedFormatException("This file has version " + version - + ", but this implementation does not support versions above " - + FormatSpec.MAXIMUM_SUPPORTED_VERSION); - } - return version; - } - - /** - * Reads a header from a buffer. - * @param buffer the buffer to read. - * @throws IOException - * @throws UnsupportedFormatException - */ - public static FileHeader readHeader(final FusionDictionaryBufferInterface buffer) - throws IOException, UnsupportedFormatException { - final int version = checkFormatVersion(buffer); - final int optionsFlags = buffer.readUnsignedShort(); - - final HashMap<String, String> attributes = new HashMap<String, String>(); - final int headerSize; - if (version < FormatSpec.FIRST_VERSION_WITH_HEADER_SIZE) { - headerSize = buffer.position(); - } else { - headerSize = buffer.readInt(); - populateOptions(buffer, headerSize, attributes); - buffer.position(headerSize); - } - - if (headerSize < 0) { - throw new UnsupportedFormatException("header size can't be negative."); - } - - final FileHeader header = new FileHeader(headerSize, - new FusionDictionary.DictionaryOptions(attributes, - 0 != (optionsFlags & FormatSpec.GERMAN_UMLAUT_PROCESSING_FLAG), - 0 != (optionsFlags & FormatSpec.FRENCH_LIGATURE_PROCESSING_FLAG)), - new FormatOptions(version, - 0 != (optionsFlags & FormatSpec.SUPPORTS_DYNAMIC_UPDATE))); - return header; - } - - /** - * Reads options from a buffer and populate a map with their contents. - * - * The buffer is read at the current position, so the caller must take care the pointer - * is in the right place before calling this. - */ - public static void populateOptions(final FusionDictionaryBufferInterface buffer, - final int headerSize, final HashMap<String, String> options) { - while (buffer.position() < headerSize) { - final String key = CharEncoding.readString(buffer); - final String value = CharEncoding.readString(buffer); - options.put(key, value); - } - } - - /** - * Reads a buffer and returns the memory representation of the dictionary. - * - * This high-level method takes a buffer and reads its contents, populating a - * FusionDictionary structure. The optional dict argument is an existing dictionary to - * which words from the buffer should be added. If it is null, a new dictionary is created. - * - * @param buffer the buffer to read. - * @param dict an optional dictionary to add words to, or null. - * @return the created (or merged) dictionary. - */ - @UsedForTesting - public static FusionDictionary readDictionaryBinary( - final FusionDictionaryBufferInterface buffer, final FusionDictionary dict) - throws IOException, UnsupportedFormatException { - // clear cache - wordCache.clear(); - - // Read header - final FileHeader header = readHeader(buffer); - - Map<Integer, Node> reverseNodeMapping = new TreeMap<Integer, Node>(); - Map<Integer, CharGroup> reverseGroupMapping = new TreeMap<Integer, CharGroup>(); - final Node root = readNode(buffer, header.mHeaderSize, reverseNodeMapping, - reverseGroupMapping, header.mFormatOptions); - - FusionDictionary newDict = new FusionDictionary(root, header.mDictionaryOptions); - if (null != dict) { - for (final Word w : dict) { - if (w.mIsBlacklistEntry) { - newDict.addBlacklistEntry(w.mWord, w.mShortcutTargets, w.mIsNotAWord); - } else { - newDict.add(w.mWord, w.mFrequency, w.mShortcutTargets, w.mIsNotAWord); - } - } - for (final Word w : dict) { - // By construction a binary dictionary may not have bigrams pointing to - // words that are not also registered as unigrams so we don't have to avoid - // them explicitly here. - for (final WeightedString bigram : w.mBigrams) { - newDict.setBigram(w.mWord, bigram.mWord, bigram.mFrequency); - } - } - } - - return newDict; - } - - /** - * Helper method to pass a file name instead of a File object to isBinaryDictionary. - */ - public static boolean isBinaryDictionary(final String filename) { - final File file = new File(filename); - return isBinaryDictionary(file); - } - - /** - * Basic test to find out whether the file is a binary dictionary or not. - * - * Concretely this only tests the magic number. - * - * @param file The file to test. - * @return true if it's a binary dictionary, false otherwise - */ - public static boolean isBinaryDictionary(final File file) { - FileInputStream inStream = null; - try { - inStream = new FileInputStream(file); - final ByteBuffer buffer = inStream.getChannel().map( - FileChannel.MapMode.READ_ONLY, 0, file.length()); - final int version = getFormatVersion(new ByteBufferWrapper(buffer)); - return (version >= FormatSpec.MINIMUM_SUPPORTED_VERSION - && version <= FormatSpec.MAXIMUM_SUPPORTED_VERSION); - } catch (FileNotFoundException e) { - return false; - } catch (IOException e) { - return false; - } finally { - if (inStream != null) { - try { - inStream.close(); - } catch (IOException e) { - // do nothing - } - } - } - } - - /** - * Calculate bigram frequency from compressed value - * - * @see #makeBigramFlags - * - * @param unigramFrequency - * @param bigramFrequency compressed frequency - * @return approximate bigram frequency - */ - public static int reconstructBigramFrequency(final int unigramFrequency, - final int bigramFrequency) { - final float stepSize = (FormatSpec.MAX_TERMINAL_FREQUENCY - unigramFrequency) - / (1.5f + FormatSpec.MAX_BIGRAM_FREQUENCY); - final float resultFreqFloat = unigramFrequency + stepSize * (bigramFrequency + 1.0f); - return (int)resultFreqFloat; - } -} diff --git a/java/src/com/android/inputmethod/latin/makedict/DynamicBinaryDictIOUtils.java b/java/src/com/android/inputmethod/latin/makedict/DynamicBinaryDictIOUtils.java new file mode 100644 index 000000000..5361d2eba --- /dev/null +++ b/java/src/com/android/inputmethod/latin/makedict/DynamicBinaryDictIOUtils.java @@ -0,0 +1,502 @@ +/* + * Copyright (C) 2013 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.makedict; + +import com.android.inputmethod.annotations.UsedForTesting; +import com.android.inputmethod.latin.Constants; +import com.android.inputmethod.latin.makedict.BinaryDictDecoderUtils.DictBuffer; +import com.android.inputmethod.latin.makedict.FormatSpec.FileHeader; +import com.android.inputmethod.latin.makedict.FormatSpec.FormatOptions; +import com.android.inputmethod.latin.makedict.FusionDictionary.WeightedString; + +import java.io.IOException; +import java.io.OutputStream; +import java.util.ArrayList; +import java.util.Arrays; + +/** + * The utility class to help dynamic updates on the binary dictionary. + * + * All the methods in this class are static. + */ +@UsedForTesting +public final class DynamicBinaryDictIOUtils { + private static final boolean DBG = false; + private static final int MAX_JUMPS = 10000; + + private DynamicBinaryDictIOUtils() { + // This utility class is not publicly instantiable. + } + + private static int markAsDeleted(final int flags) { + return (flags & (~FormatSpec.MASK_GROUP_ADDRESS_TYPE)) | FormatSpec.FLAG_IS_DELETED; + } + + /** + * Delete the word from the binary file. + * + * @param dictDecoder the dict decoder. + * @param word the word we delete + * @throws IOException + * @throws UnsupportedFormatException + */ + @UsedForTesting + public static void deleteWord(final BinaryDictDecoder dictDecoder, final String word) + throws IOException, UnsupportedFormatException { + final DictBuffer dictBuffer = dictDecoder.getDictBuffer(); + dictBuffer.position(0); + final FileHeader header = BinaryDictDecoderUtils.readHeader(dictDecoder); + final int wordPosition = BinaryDictIOUtils.getTerminalPosition(dictDecoder, word); + if (wordPosition == FormatSpec.NOT_VALID_WORD) return; + + dictBuffer.position(wordPosition); + final int flags = dictBuffer.readUnsignedByte(); + dictBuffer.position(wordPosition); + dictBuffer.put((byte)markAsDeleted(flags)); + } + + /** + * Update a parent address in a CharGroup that is referred to by groupOriginAddress. + * + * @param dictBuffer the DictBuffer to write. + * @param groupOriginAddress the address of the group. + * @param newParentAddress the absolute address of the parent. + * @param formatOptions file format options. + */ + public static void updateParentAddress(final DictBuffer dictBuffer, + final int groupOriginAddress, final int newParentAddress, + final FormatOptions formatOptions) { + final int originalPosition = dictBuffer.position(); + dictBuffer.position(groupOriginAddress); + if (!formatOptions.mSupportsDynamicUpdate) { + throw new RuntimeException("this file format does not support parent addresses"); + } + final int flags = dictBuffer.readUnsignedByte(); + if (BinaryDictIOUtils.isMovedGroup(flags, formatOptions)) { + // If the group is moved, the parent address is stored in the destination group. + // We are guaranteed to process the destination group later, so there is no need to + // update anything here. + dictBuffer.position(originalPosition); + return; + } + if (DBG) { + MakedictLog.d("update parent address flags=" + flags + ", " + groupOriginAddress); + } + final int parentOffset = newParentAddress - groupOriginAddress; + BinaryDictIOUtils.writeSInt24ToBuffer(dictBuffer, parentOffset); + dictBuffer.position(originalPosition); + } + + /** + * Update parent addresses in a node array stored at nodeOriginAddress. + * + * @param dictBuffer the DictBuffer to be modified. + * @param nodeOriginAddress the address of the node array to update. + * @param newParentAddress the address to be written. + * @param formatOptions file format options. + */ + public static void updateParentAddresses(final DictBuffer dictBuffer, + final int nodeOriginAddress, final int newParentAddress, + final FormatOptions formatOptions) { + final int originalPosition = dictBuffer.position(); + dictBuffer.position(nodeOriginAddress); + do { + final int count = BinaryDictDecoderUtils.readCharGroupCount(dictBuffer); + for (int i = 0; i < count; ++i) { + updateParentAddress(dictBuffer, dictBuffer.position(), newParentAddress, + formatOptions); + BinaryDictIOUtils.skipCharGroup(dictBuffer, formatOptions); + } + final int forwardLinkAddress = dictBuffer.readUnsignedInt24(); + dictBuffer.position(forwardLinkAddress); + } while (formatOptions.mSupportsDynamicUpdate + && dictBuffer.position() != FormatSpec.NO_FORWARD_LINK_ADDRESS); + dictBuffer.position(originalPosition); + } + + /** + * Update a children address in a CharGroup that is addressed by groupOriginAddress. + * + * @param dictBuffer the DictBuffer to write. + * @param groupOriginAddress the address of the group. + * @param newChildrenAddress the absolute address of the child. + * @param formatOptions file format options. + */ + public static void updateChildrenAddress(final DictBuffer dictBuffer, + final int groupOriginAddress, final int newChildrenAddress, + final FormatOptions formatOptions) { + final int originalPosition = dictBuffer.position(); + dictBuffer.position(groupOriginAddress); + final int flags = dictBuffer.readUnsignedByte(); + final int parentAddress = BinaryDictDecoderUtils.readParentAddress(dictBuffer, + formatOptions); + BinaryDictIOUtils.skipString(dictBuffer, (flags & FormatSpec.FLAG_HAS_MULTIPLE_CHARS) != 0); + if ((flags & FormatSpec.FLAG_IS_TERMINAL) != 0) dictBuffer.readUnsignedByte(); + final int childrenOffset = newChildrenAddress == FormatSpec.NO_CHILDREN_ADDRESS + ? FormatSpec.NO_CHILDREN_ADDRESS : newChildrenAddress - dictBuffer.position(); + BinaryDictIOUtils.writeSInt24ToBuffer(dictBuffer, childrenOffset); + dictBuffer.position(originalPosition); + } + + /** + * Helper method to move a char group to the tail of the file. + */ + private static int moveCharGroup(final OutputStream destination, + final DictBuffer dictBuffer, final CharGroupInfo info, + final int nodeArrayOriginAddress, final int oldGroupAddress, + final FormatOptions formatOptions) throws IOException { + updateParentAddress(dictBuffer, oldGroupAddress, dictBuffer.limit() + 1, formatOptions); + dictBuffer.position(oldGroupAddress); + final int currentFlags = dictBuffer.readUnsignedByte(); + dictBuffer.position(oldGroupAddress); + dictBuffer.put((byte)(FormatSpec.FLAG_IS_MOVED | (currentFlags + & (~FormatSpec.MASK_MOVE_AND_DELETE_FLAG)))); + int size = FormatSpec.GROUP_FLAGS_SIZE; + updateForwardLink(dictBuffer, nodeArrayOriginAddress, dictBuffer.limit(), formatOptions); + size += BinaryDictIOUtils.writeNodes(destination, new CharGroupInfo[] { info }); + return size; + } + + @SuppressWarnings("unused") + private static void updateForwardLink(final DictBuffer dictBuffer, + final int nodeArrayOriginAddress, final int newNodeArrayAddress, + final FormatOptions formatOptions) { + dictBuffer.position(nodeArrayOriginAddress); + int jumpCount = 0; + while (jumpCount++ < MAX_JUMPS) { + final int count = BinaryDictDecoderUtils.readCharGroupCount(dictBuffer); + for (int i = 0; i < count; ++i) { + BinaryDictIOUtils.skipCharGroup(dictBuffer, formatOptions); + } + final int forwardLinkAddress = dictBuffer.readUnsignedInt24(); + if (forwardLinkAddress == FormatSpec.NO_FORWARD_LINK_ADDRESS) { + dictBuffer.position(dictBuffer.position() - FormatSpec.FORWARD_LINK_ADDRESS_SIZE); + BinaryDictIOUtils.writeSInt24ToBuffer(dictBuffer, newNodeArrayAddress); + return; + } + dictBuffer.position(forwardLinkAddress); + } + if (DBG && jumpCount >= MAX_JUMPS) { + throw new RuntimeException("too many jumps, probably a bug."); + } + } + + /** + * Move a group that is referred to by oldGroupOrigin to the tail of the file, and set the + * children address to the byte after the group + * + * @param fileEndAddress the address of the tail of the file. + * @param codePoints the characters to put inside the group. + * @param length how many code points to read from codePoints. + * @param flags the flags for this group. + * @param frequency the frequency of this terminal. + * @param parentAddress the address of the parent group of this group. + * @param shortcutTargets the shortcut targets for this group. + * @param bigrams the bigrams for this group. + * @param destination the stream representing the tail of the file. + * @param dictBuffer the DictBuffer representing the (constant-size) body of the file. + * @param oldNodeArrayOrigin the origin of the old node array this group was a part of. + * @param oldGroupOrigin the old origin where this group used to be stored. + * @param formatOptions format options for this dictionary. + * @return the size written, in bytes. + * @throws IOException if the file can't be accessed + */ + private static int moveGroup(final int fileEndAddress, final int[] codePoints, + final int length, final int flags, final int frequency, final int parentAddress, + final ArrayList<WeightedString> shortcutTargets, + final ArrayList<PendingAttribute> bigrams, final OutputStream destination, + final DictBuffer dictBuffer, final int oldNodeArrayOrigin, + final int oldGroupOrigin, final FormatOptions formatOptions) throws IOException { + int size = 0; + final int newGroupOrigin = fileEndAddress + 1; + final int[] writtenCharacters = Arrays.copyOfRange(codePoints, 0, length); + final CharGroupInfo tmpInfo = new CharGroupInfo(newGroupOrigin, -1 /* endAddress */, + flags, writtenCharacters, frequency, parentAddress, FormatSpec.NO_CHILDREN_ADDRESS, + shortcutTargets, bigrams); + size = BinaryDictIOUtils.computeGroupSize(tmpInfo, formatOptions); + final CharGroupInfo newInfo = new CharGroupInfo(newGroupOrigin, newGroupOrigin + size, + flags, writtenCharacters, frequency, parentAddress, + fileEndAddress + 1 + size + FormatSpec.FORWARD_LINK_ADDRESS_SIZE, shortcutTargets, + bigrams); + moveCharGroup(destination, dictBuffer, newInfo, oldNodeArrayOrigin, oldGroupOrigin, + formatOptions); + return 1 + size + FormatSpec.FORWARD_LINK_ADDRESS_SIZE; + } + + /** + * Insert a word into a binary dictionary. + * + * @param dictDecoder the dict decoder. + * @param destination a stream to the underlying file, with the pointer at the end of the file. + * @param word the word to insert. + * @param frequency the frequency of the new word. + * @param bigramStrings bigram list, or null if none. + * @param shortcuts shortcut list, or null if none. + * @param isBlackListEntry whether this should be a blacklist entry. + * @throws IOException if the file can't be accessed. + * @throws UnsupportedFormatException if the existing dictionary is in an unexpected format. + */ + // TODO: Support batch insertion. + // TODO: Remove @UsedForTesting once UserHistoryDictionary is implemented by BinaryDictionary. + @UsedForTesting + public static void insertWord(final BinaryDictDecoder dictDecoder, + final OutputStream destination, final String word, final int frequency, + final ArrayList<WeightedString> bigramStrings, + final ArrayList<WeightedString> shortcuts, final boolean isNotAWord, + final boolean isBlackListEntry) + throws IOException, UnsupportedFormatException { + final ArrayList<PendingAttribute> bigrams = new ArrayList<PendingAttribute>(); + final DictBuffer dictBuffer = dictDecoder.getDictBuffer(); + if (bigramStrings != null) { + for (final WeightedString bigram : bigramStrings) { + int position = BinaryDictIOUtils.getTerminalPosition(dictDecoder, bigram.mWord); + if (position == FormatSpec.NOT_VALID_WORD) { + // TODO: figure out what is the correct thing to do here. + } else { + bigrams.add(new PendingAttribute(bigram.mFrequency, position)); + } + } + } + + final boolean isTerminal = true; + final boolean hasBigrams = !bigrams.isEmpty(); + final boolean hasShortcuts = shortcuts != null && !shortcuts.isEmpty(); + + // find the insert position of the word. + if (dictBuffer.position() != 0) dictBuffer.position(0); + final FileHeader fileHeader = BinaryDictDecoderUtils.readHeader(dictDecoder); + + int wordPos = 0, address = dictBuffer.position(), nodeOriginAddress = dictBuffer.position(); + final int[] codePoints = FusionDictionary.getCodePoints(word); + final int wordLen = codePoints.length; + + for (int depth = 0; depth < Constants.DICTIONARY_MAX_WORD_LENGTH; ++depth) { + if (wordPos >= wordLen) break; + nodeOriginAddress = dictBuffer.position(); + int nodeParentAddress = -1; + final int charGroupCount = BinaryDictDecoderUtils.readCharGroupCount(dictBuffer); + boolean foundNextGroup = false; + + for (int i = 0; i < charGroupCount; ++i) { + address = dictBuffer.position(); + final CharGroupInfo currentInfo = BinaryDictDecoderUtils.readCharGroup(dictBuffer, + dictBuffer.position(), fileHeader.mFormatOptions); + final boolean isMovedGroup = BinaryDictIOUtils.isMovedGroup(currentInfo.mFlags, + fileHeader.mFormatOptions); + if (isMovedGroup) continue; + nodeParentAddress = (currentInfo.mParentAddress == FormatSpec.NO_PARENT_ADDRESS) + ? FormatSpec.NO_PARENT_ADDRESS : currentInfo.mParentAddress + address; + boolean matched = true; + for (int p = 0; p < currentInfo.mCharacters.length; ++p) { + if (wordPos + p >= wordLen) { + /* + * splitting + * before + * abcd - ef + * + * insert "abc" + * + * after + * abc - d - ef + */ + final int newNodeAddress = dictBuffer.limit(); + final int flags = BinaryDictEncoder.makeCharGroupFlags(p > 1, + isTerminal, 0, hasShortcuts, hasBigrams, false /* isNotAWord */, + false /* isBlackListEntry */, fileHeader.mFormatOptions); + int written = moveGroup(newNodeAddress, currentInfo.mCharacters, p, flags, + frequency, nodeParentAddress, shortcuts, bigrams, destination, + dictBuffer, nodeOriginAddress, address, fileHeader.mFormatOptions); + + final int[] characters2 = Arrays.copyOfRange(currentInfo.mCharacters, p, + currentInfo.mCharacters.length); + if (currentInfo.mChildrenAddress != FormatSpec.NO_CHILDREN_ADDRESS) { + updateParentAddresses(dictBuffer, currentInfo.mChildrenAddress, + newNodeAddress + written + 1, fileHeader.mFormatOptions); + } + final CharGroupInfo newInfo2 = new CharGroupInfo( + newNodeAddress + written + 1, -1 /* endAddress */, + currentInfo.mFlags, characters2, currentInfo.mFrequency, + newNodeAddress + 1, currentInfo.mChildrenAddress, + currentInfo.mShortcutTargets, currentInfo.mBigrams); + BinaryDictIOUtils.writeNodes(destination, new CharGroupInfo[] { newInfo2 }); + return; + } else if (codePoints[wordPos + p] != currentInfo.mCharacters[p]) { + if (p > 0) { + /* + * splitting + * before + * ab - cd + * + * insert "ac" + * + * after + * a - b - cd + * | + * - c + */ + + final int newNodeAddress = dictBuffer.limit(); + final int childrenAddress = currentInfo.mChildrenAddress; + + // move prefix + final int prefixFlags = BinaryDictEncoder.makeCharGroupFlags(p > 1, + false /* isTerminal */, 0 /* childrenAddressSize*/, + false /* hasShortcut */, false /* hasBigrams */, + false /* isNotAWord */, false /* isBlackListEntry */, + fileHeader.mFormatOptions); + int written = moveGroup(newNodeAddress, currentInfo.mCharacters, p, + prefixFlags, -1 /* frequency */, nodeParentAddress, null, null, + destination, dictBuffer, nodeOriginAddress, address, + fileHeader.mFormatOptions); + + final int[] suffixCharacters = Arrays.copyOfRange( + currentInfo.mCharacters, p, currentInfo.mCharacters.length); + if (currentInfo.mChildrenAddress != FormatSpec.NO_CHILDREN_ADDRESS) { + updateParentAddresses(dictBuffer, currentInfo.mChildrenAddress, + newNodeAddress + written + 1, fileHeader.mFormatOptions); + } + final int suffixFlags = BinaryDictEncoder.makeCharGroupFlags( + suffixCharacters.length > 1, + (currentInfo.mFlags & FormatSpec.FLAG_IS_TERMINAL) != 0, + 0 /* childrenAddressSize */, + (currentInfo.mFlags & FormatSpec.FLAG_HAS_SHORTCUT_TARGETS) + != 0, + (currentInfo.mFlags & FormatSpec.FLAG_HAS_BIGRAMS) != 0, + isNotAWord, isBlackListEntry, fileHeader.mFormatOptions); + final CharGroupInfo suffixInfo = new CharGroupInfo( + newNodeAddress + written + 1, -1 /* endAddress */, suffixFlags, + suffixCharacters, currentInfo.mFrequency, newNodeAddress + 1, + currentInfo.mChildrenAddress, currentInfo.mShortcutTargets, + currentInfo.mBigrams); + written += BinaryDictIOUtils.computeGroupSize(suffixInfo, + fileHeader.mFormatOptions) + 1; + + final int[] newCharacters = Arrays.copyOfRange(codePoints, wordPos + p, + codePoints.length); + final int flags = BinaryDictEncoder.makeCharGroupFlags( + newCharacters.length > 1, isTerminal, + 0 /* childrenAddressSize */, hasShortcuts, hasBigrams, + isNotAWord, isBlackListEntry, fileHeader.mFormatOptions); + final CharGroupInfo newInfo = new CharGroupInfo( + newNodeAddress + written, -1 /* endAddress */, flags, + newCharacters, frequency, newNodeAddress + 1, + FormatSpec.NO_CHILDREN_ADDRESS, shortcuts, bigrams); + BinaryDictIOUtils.writeNodes(destination, + new CharGroupInfo[] { suffixInfo, newInfo }); + return; + } + matched = false; + break; + } + } + + if (matched) { + if (wordPos + currentInfo.mCharacters.length == wordLen) { + // the word exists in the dictionary. + // only update group. + final int newNodeAddress = dictBuffer.limit(); + final boolean hasMultipleChars = currentInfo.mCharacters.length > 1; + final int flags = BinaryDictEncoder.makeCharGroupFlags(hasMultipleChars, + isTerminal, 0 /* childrenAddressSize */, hasShortcuts, hasBigrams, + isNotAWord, isBlackListEntry, fileHeader.mFormatOptions); + final CharGroupInfo newInfo = new CharGroupInfo(newNodeAddress + 1, + -1 /* endAddress */, flags, currentInfo.mCharacters, frequency, + nodeParentAddress, currentInfo.mChildrenAddress, shortcuts, + bigrams); + moveCharGroup(destination, dictBuffer, newInfo, nodeOriginAddress, address, + fileHeader.mFormatOptions); + return; + } + wordPos += currentInfo.mCharacters.length; + if (currentInfo.mChildrenAddress == FormatSpec.NO_CHILDREN_ADDRESS) { + /* + * found the prefix of the word. + * make new node and link to the node from this group. + * + * before + * ab - cd + * + * insert "abcde" + * + * after + * ab - cd - e + */ + final int newNodeAddress = dictBuffer.limit(); + updateChildrenAddress(dictBuffer, address, newNodeAddress, + fileHeader.mFormatOptions); + final int newGroupAddress = newNodeAddress + 1; + final boolean hasMultipleChars = (wordLen - wordPos) > 1; + final int flags = BinaryDictEncoder.makeCharGroupFlags(hasMultipleChars, + isTerminal, 0 /* childrenAddressSize */, hasShortcuts, hasBigrams, + isNotAWord, isBlackListEntry, fileHeader.mFormatOptions); + final int[] characters = Arrays.copyOfRange(codePoints, wordPos, wordLen); + final CharGroupInfo newInfo = new CharGroupInfo(newGroupAddress, -1, flags, + characters, frequency, address, FormatSpec.NO_CHILDREN_ADDRESS, + shortcuts, bigrams); + BinaryDictIOUtils.writeNodes(destination, new CharGroupInfo[] { newInfo }); + return; + } + dictBuffer.position(currentInfo.mChildrenAddress); + foundNextGroup = true; + break; + } + } + + if (foundNextGroup) continue; + + // reached the end of the array. + final int linkAddressPosition = dictBuffer.position(); + int nextLink = dictBuffer.readUnsignedInt24(); + if ((nextLink & FormatSpec.MSB24) != 0) { + nextLink = -(nextLink & FormatSpec.SINT24_MAX); + } + if (nextLink == FormatSpec.NO_FORWARD_LINK_ADDRESS) { + /* + * expand this node. + * + * before + * ab - cd + * + * insert "abef" + * + * after + * ab - cd + * | + * - ef + */ + + // change the forward link address. + final int newNodeAddress = dictBuffer.limit(); + dictBuffer.position(linkAddressPosition); + BinaryDictIOUtils.writeSInt24ToBuffer(dictBuffer, newNodeAddress); + + final int[] characters = Arrays.copyOfRange(codePoints, wordPos, wordLen); + final int flags = BinaryDictEncoder.makeCharGroupFlags(characters.length > 1, + isTerminal, 0 /* childrenAddressSize */, hasShortcuts, hasBigrams, + isNotAWord, isBlackListEntry, fileHeader.mFormatOptions); + final CharGroupInfo newInfo = new CharGroupInfo(newNodeAddress + 1, + -1 /* endAddress */, flags, characters, frequency, nodeParentAddress, + FormatSpec.NO_CHILDREN_ADDRESS, shortcuts, bigrams); + BinaryDictIOUtils.writeNodes(destination, new CharGroupInfo[]{ newInfo }); + return; + } else { + depth--; + dictBuffer.position(nextLink); + } + } + } +} diff --git a/java/src/com/android/inputmethod/latin/makedict/FormatSpec.java b/java/src/com/android/inputmethod/latin/makedict/FormatSpec.java index 46266aa50..5e331219c 100644 --- a/java/src/com/android/inputmethod/latin/makedict/FormatSpec.java +++ b/java/src/com/android/inputmethod/latin/makedict/FormatSpec.java @@ -26,7 +26,41 @@ import com.android.inputmethod.latin.makedict.FusionDictionary.DictionaryOptions public final class FormatSpec { /* - * Array of Node(FusionDictionary.Node) layout is as follows: + * File header layout is as follows: + * + * v | + * e | MAGIC_NUMBER + version of the file format, 2 bytes. + * r | + * sion + * + * o | + * p | not used 4 bits + * t | has bigrams ? 1 bit, 1 = yes, 0 = no : CONTAINS_BIGRAMS_FLAG + * i | FRENCH_LIGATURE_PROCESSING_FLAG + * o | supports dynamic updates ? 1 bit, 1 = yes, 0 = no : SUPPORTS_DYNAMIC_UPDATE + * n | GERMAN_UMLAUT_PROCESSING_FLAG + * f | + * lags + * + * h | + * e | size of the file header, 4bytes + * a | including the size of the magic number, the option flags and the header size + * d | + * ersize + * + * | attributes list + * + * attributes list is: + * <key> = | string of characters at the char format described below, with the terminator used + * | to signal the end of the string. + * <value> = | string of characters at the char format described below, with the terminator used + * | to signal the end of the string. + * if the size of already read < headersize, goto key. + * + */ + + /* + * Node array (FusionDictionary.PtNodeArray) layout is as follows: * * g | * r | the number of groups, 1 or 2 bytes. @@ -52,7 +86,7 @@ public final class FormatSpec { * linkaddress */ - /* Node(CharGroup) layout is as follows: + /* Node (FusionDictionary.CharGroup) layout is as follows: * | IF !SUPPORTS_DYNAMIC_UPDATE * | addressType xx : mask with MASK_GROUP_ADDRESS_TYPE * | 2 bits, 00 = no children : FLAG_GROUP_ADDRESS_TYPE_NOADDRESS @@ -151,12 +185,10 @@ public final class FormatSpec { * if (FLAG_ATTRIBUTE_HAS_NEXT goto flags */ - static final int VERSION_1_MAGIC_NUMBER = 0x78B1; - public static final int VERSION_2_MAGIC_NUMBER = 0x9BC13AFE; - static final int MINIMUM_SUPPORTED_VERSION = 1; + public static final int MAGIC_NUMBER = 0x9BC13AFE; + static final int MINIMUM_SUPPORTED_VERSION = 2; static final int MAXIMUM_SUPPORTED_VERSION = 3; static final int NOT_A_VERSION_NUMBER = -1; - static final int FIRST_VERSION_WITH_HEADER_SIZE = 2; static final int FIRST_VERSION_WITH_DYNAMIC_UPDATE = 3; // These options need to be the same numeric values as the one in the native reading code. @@ -219,7 +251,7 @@ public final class FormatSpec { static final int INVALID_CHARACTER = -1; static final int MAX_CHARGROUPS_FOR_ONE_BYTE_CHARGROUP_COUNT = 0x7F; // 127 - static final int MAX_CHARGROUPS_IN_A_NODE = 0x7FFF; // 32767 + static final int MAX_CHARGROUPS_IN_A_PT_NODE_ARRAY = 0x7FFF; // 32767 static final int MAX_BIGRAMS_IN_A_GROUP = 10000; static final int MAX_TERMINAL_FREQUENCY = 255; @@ -231,6 +263,13 @@ public final class FormatSpec { static final int NOT_VALID_WORD = -99; static final int SIGNED_CHILDREN_ADDRESS_SIZE = 3; + static final int UINT8_MAX = 0xFF; + static final int UINT16_MAX = 0xFFFF; + static final int UINT24_MAX = 0xFFFFFF; + static final int SINT24_MAX = 0x7FFFFF; + static final int MSB8 = 0x80; + static final int MSB24 = 0x800000; + /** * Options about file format. */ diff --git a/java/src/com/android/inputmethod/latin/makedict/FusionDictionary.java b/java/src/com/android/inputmethod/latin/makedict/FusionDictionary.java index 118dc22b8..fce1c5cdd 100644 --- a/java/src/com/android/inputmethod/latin/makedict/FusionDictionary.java +++ b/java/src/com/android/inputmethod/latin/makedict/FusionDictionary.java @@ -37,14 +37,14 @@ public final class FusionDictionary implements Iterable<Word> { private static int CHARACTER_NOT_FOUND_INDEX = -1; /** - * A node of the dictionary, containing several CharGroups. + * A node array of the dictionary, containing several CharGroups. * - * A node is but an ordered array of CharGroups, which essentially contain all the + * A PtNodeArray is but an ordered array of CharGroups, which essentially contain all the * real information. * This class also contains fields to cache size and address, to help with binary * generation. */ - public static final class Node { + public static final class PtNodeArray { ArrayList<CharGroup> mData; // To help with binary generation int mCachedSize = Integer.MIN_VALUE; @@ -57,10 +57,10 @@ public final class FusionDictionary implements Iterable<Word> { int mCachedAddressAfterUpdate = Integer.MIN_VALUE; int mCachedParentAddress = 0; - public Node() { + public PtNodeArray() { mData = new ArrayList<CharGroup>(); } - public Node(ArrayList<CharGroup> data) { + public PtNodeArray(ArrayList<CharGroup> data) { mData = data; } } @@ -98,7 +98,7 @@ public final class FusionDictionary implements Iterable<Word> { * This is the central class of the in-memory representation. A CharGroup is what can * be seen as a traditional "trie node", except it can hold several characters at the * same time. A CharGroup essentially represents one or several characters in the middle - * of the trie trie; as such, it can be a terminal, and it can have children. + * of the trie tree; as such, it can be a terminal, and it can have children. * In this in-memory representation, whether the CharGroup is a terminal or not is represented * in the frequency, where NOT_A_TERMINAL (= -1) means this is not a terminal and any other * value is the frequency of this terminal. A terminal may have non-null shortcuts and/or @@ -110,7 +110,7 @@ public final class FusionDictionary implements Iterable<Word> { ArrayList<WeightedString> mShortcutTargets; ArrayList<WeightedString> mBigrams; int mFrequency; // NOT_A_TERMINAL == mFrequency indicates this is not a terminal. - Node mChildren; + PtNodeArray mChildren; boolean mIsNotAWord; // Only a shortcut boolean mIsBlacklistEntry; // mCachedSize and mCachedAddressBefore/AfterUpdate are helpers for binary dictionary @@ -137,7 +137,8 @@ public final class FusionDictionary implements Iterable<Word> { public CharGroup(final int[] chars, final ArrayList<WeightedString> shortcutTargets, final ArrayList<WeightedString> bigrams, final int frequency, - final boolean isNotAWord, final boolean isBlacklistEntry, final Node children) { + final boolean isNotAWord, final boolean isBlacklistEntry, + final PtNodeArray children) { mChars = chars; mFrequency = frequency; mShortcutTargets = shortcutTargets; @@ -149,7 +150,7 @@ public final class FusionDictionary implements Iterable<Word> { public void addChild(CharGroup n) { if (null == mChildren) { - mChildren = new Node(); + mChildren = new PtNodeArray(); } mChildren.mData.add(n); } @@ -344,10 +345,10 @@ public final class FusionDictionary implements Iterable<Word> { } public final DictionaryOptions mOptions; - public final Node mRoot; + public final PtNodeArray mRootNodeArray; - public FusionDictionary(final Node root, final DictionaryOptions options) { - mRoot = root; + public FusionDictionary(final PtNodeArray rootNodeArray, final DictionaryOptions options) { + mRootNodeArray = rootNodeArray; mOptions = options; } @@ -406,13 +407,13 @@ public final class FusionDictionary implements Iterable<Word> { } /** - * Sanity check for a node. + * Sanity check for a node array. * - * This method checks that all CharGroups in a node are ordered as expected. + * This method checks that all CharGroups in a node array are ordered as expected. * If they are, nothing happens. If they aren't, an exception is thrown. */ - private void checkStack(Node node) { - ArrayList<CharGroup> stack = node.mData; + private void checkStack(PtNodeArray nodeArray) { + ArrayList<CharGroup> stack = nodeArray.mData; int lastValue = -1; for (int i = 0; i < stack.size(); ++i) { int currentValue = stack.get(i).mChars[0]; @@ -431,16 +432,16 @@ public final class FusionDictionary implements Iterable<Word> { * @param frequency the bigram frequency */ public void setBigram(final String word1, final String word2, final int frequency) { - CharGroup charGroup = findWordInTree(mRoot, word1); + CharGroup charGroup = findWordInTree(mRootNodeArray, word1); if (charGroup != null) { - final CharGroup charGroup2 = findWordInTree(mRoot, word2); + final CharGroup charGroup2 = findWordInTree(mRootNodeArray, word2); if (charGroup2 == null) { add(getCodePoints(word2), 0, null, false /* isNotAWord */, false /* isBlacklistEntry */); // The chargroup for the first word may have moved by the above insertion, // if word1 and word2 share a common stem that happens not to have been // a cutting point until now. In this case, we need to refresh charGroup. - charGroup = findWordInTree(mRoot, word1); + charGroup = findWordInTree(mRootNodeArray, word1); } charGroup.addBigram(word2, frequency); } else { @@ -469,38 +470,38 @@ public final class FusionDictionary implements Iterable<Word> { return; } - Node currentNode = mRoot; + PtNodeArray currentNodeArray = mRootNodeArray; int charIndex = 0; CharGroup currentGroup = null; int differentCharIndex = 0; // Set by the loop to the index of the char that differs - int nodeIndex = findIndexOfChar(mRoot, word[charIndex]); + int nodeIndex = findIndexOfChar(mRootNodeArray, word[charIndex]); while (CHARACTER_NOT_FOUND_INDEX != nodeIndex) { - currentGroup = currentNode.mData.get(nodeIndex); - differentCharIndex = compareArrays(currentGroup.mChars, word, charIndex); + currentGroup = currentNodeArray.mData.get(nodeIndex); + differentCharIndex = compareCharArrays(currentGroup.mChars, word, charIndex); if (ARRAYS_ARE_EQUAL != differentCharIndex && differentCharIndex < currentGroup.mChars.length) break; if (null == currentGroup.mChildren) break; charIndex += currentGroup.mChars.length; if (charIndex >= word.length) break; - currentNode = currentGroup.mChildren; - nodeIndex = findIndexOfChar(currentNode, word[charIndex]); + currentNodeArray = currentGroup.mChildren; + nodeIndex = findIndexOfChar(currentNodeArray, word[charIndex]); } if (CHARACTER_NOT_FOUND_INDEX == nodeIndex) { // No node at this point to accept the word. Create one. - final int insertionIndex = findInsertionIndex(currentNode, word[charIndex]); + final int insertionIndex = findInsertionIndex(currentNodeArray, word[charIndex]); final CharGroup newGroup = new CharGroup( Arrays.copyOfRange(word, charIndex, word.length), shortcutTargets, null /* bigrams */, frequency, isNotAWord, isBlacklistEntry); - currentNode.mData.add(insertionIndex, newGroup); - if (DBG) checkStack(currentNode); + currentNodeArray.mData.add(insertionIndex, newGroup); + if (DBG) checkStack(currentNodeArray); } else { // There is a word with a common prefix. if (differentCharIndex == currentGroup.mChars.length) { if (charIndex + differentCharIndex >= word.length) { // The new word is a prefix of an existing word, but the node on which it - // should end already exists as is. Since the old CharNode was not a terminal, + // should end already exists as is. Since the old CharGroup was not a terminal, // make it one by filling in its frequency and other attributes currentGroup.update(frequency, shortcutTargets, null, isNotAWord, isBlacklistEntry); @@ -511,7 +512,7 @@ public final class FusionDictionary implements Iterable<Word> { Arrays.copyOfRange(word, charIndex + differentCharIndex, word.length), shortcutTargets, null /* bigrams */, frequency, isNotAWord, isBlacklistEntry); - currentGroup.mChildren = new Node(); + currentGroup.mChildren = new PtNodeArray(); currentGroup.mChildren.mData.add(newNode); } } else { @@ -524,7 +525,7 @@ public final class FusionDictionary implements Iterable<Word> { } else { // Partial prefix match only. We have to replace the current node with a node // containing the current prefix and create two new ones for the tails. - Node newChildren = new Node(); + PtNodeArray newChildren = new PtNodeArray(); final CharGroup newOldWord = new CharGroup( Arrays.copyOfRange(currentGroup.mChars, differentCharIndex, currentGroup.mChars.length), currentGroup.mShortcutTargets, @@ -552,9 +553,9 @@ public final class FusionDictionary implements Iterable<Word> { > currentGroup.mChars[differentCharIndex] ? 1 : 0; newChildren.mData.add(addIndex, newWord); } - currentNode.mData.set(nodeIndex, newParent); + currentNodeArray.mData.set(nodeIndex, newParent); } - if (DBG) checkStack(currentNode); + if (DBG) checkStack(currentNodeArray); } } } @@ -576,7 +577,7 @@ public final class FusionDictionary implements Iterable<Word> { * @param dstOffset the offset in the right-hand side string. * @return the index at which the strings differ, or ARRAYS_ARE_EQUAL = 0 if they don't. */ - private static int compareArrays(final int[] src, final int[] dst, int dstOffset) { + private static int compareCharArrays(final int[] src, final int[] dst, int dstOffset) { // We do NOT test the first char, because we come from a method that already // tested it. for (int i = 1; i < src.length; ++i) { @@ -603,10 +604,10 @@ public final class FusionDictionary implements Iterable<Word> { final static private CharGroupComparator CHARGROUP_COMPARATOR = new CharGroupComparator(); /** - * Finds the insertion index of a character within a node. + * Finds the insertion index of a character within a node array. */ - private static int findInsertionIndex(final Node node, int character) { - final ArrayList<CharGroup> data = node.mData; + private static int findInsertionIndex(final PtNodeArray nodeArray, int character) { + final ArrayList<CharGroup> data = nodeArray.mData; final CharGroup reference = new CharGroup(new int[] { character }, null /* shortcutTargets */, null /* bigrams */, 0, false /* isNotAWord */, false /* isBlacklistEntry */); @@ -615,16 +616,16 @@ public final class FusionDictionary implements Iterable<Word> { } /** - * Find the index of a char in a node, if it exists. + * Find the index of a char in a node array, if it exists. * - * @param node the node to search in. + * @param nodeArray the node array to search in. * @param character the character to search for. * @return the position of the character if it's there, or CHARACTER_NOT_FOUND_INDEX = -1 else. */ - private static int findIndexOfChar(final Node node, int character) { - final int insertionIndex = findInsertionIndex(node, character); - if (node.mData.size() <= insertionIndex) return CHARACTER_NOT_FOUND_INDEX; - return character == node.mData.get(insertionIndex).mChars[0] ? insertionIndex + private static int findIndexOfChar(final PtNodeArray nodeArray, int character) { + final int insertionIndex = findInsertionIndex(nodeArray, character); + if (nodeArray.mData.size() <= insertionIndex) return CHARACTER_NOT_FOUND_INDEX; + return character == nodeArray.mData.get(insertionIndex).mChars[0] ? insertionIndex : CHARACTER_NOT_FOUND_INDEX; } @@ -632,16 +633,16 @@ public final class FusionDictionary implements Iterable<Word> { * Helper method to find a word in a given branch. */ @SuppressWarnings("unused") - public static CharGroup findWordInTree(Node node, final String string) { + public static CharGroup findWordInTree(PtNodeArray nodeArray, final String string) { int index = 0; final StringBuilder checker = DBG ? new StringBuilder() : null; final int[] codePoints = getCodePoints(string); CharGroup currentGroup; do { - int indexOfGroup = findIndexOfChar(node, codePoints[index]); + int indexOfGroup = findIndexOfChar(nodeArray, codePoints[index]); if (CHARACTER_NOT_FOUND_INDEX == indexOfGroup) return null; - currentGroup = node.mData.get(indexOfGroup); + currentGroup = nodeArray.mData.get(indexOfGroup); if (codePoints.length - index < currentGroup.mChars.length) return null; int newIndex = index; @@ -653,9 +654,9 @@ public final class FusionDictionary implements Iterable<Word> { if (DBG) checker.append(new String(currentGroup.mChars, 0, currentGroup.mChars.length)); if (index < codePoints.length) { - node = currentGroup.mChildren; + nodeArray = currentGroup.mChildren; } - } while (null != node && index < codePoints.length); + } while (null != nodeArray && index < codePoints.length); if (index < codePoints.length) return null; if (!currentGroup.isTerminal()) return null; @@ -670,20 +671,20 @@ public final class FusionDictionary implements Iterable<Word> { if (null == s || "".equals(s)) { throw new RuntimeException("Can't search for a null or empty string"); } - return null != findWordInTree(mRoot, s); + return null != findWordInTree(mRootNodeArray, s); } /** * Recursively count the number of character groups in a given branch of the trie. * - * @param node the parent node. + * @param nodeArray the parent node. * @return the number of char groups in all the branch under this node. */ - public static int countCharGroups(final Node node) { - final int nodeSize = node.mData.size(); + public static int countCharGroups(final PtNodeArray nodeArray) { + final int nodeSize = nodeArray.mData.size(); int size = nodeSize; for (int i = nodeSize - 1; i >= 0; --i) { - CharGroup group = node.mData.get(i); + CharGroup group = nodeArray.mData.get(i); if (null != group.mChildren) size += countCharGroups(group.mChildren); } @@ -693,15 +694,15 @@ public final class FusionDictionary implements Iterable<Word> { /** * Recursively count the number of nodes in a given branch of the trie. * - * @param node the node to count. + * @param nodeArray the node array to count. * @return the number of nodes in this branch. */ - public static int countNodes(final Node node) { + public static int countNodeArrays(final PtNodeArray nodeArray) { int size = 1; - for (int i = node.mData.size() - 1; i >= 0; --i) { - CharGroup group = node.mData.get(i); + for (int i = nodeArray.mData.size() - 1; i >= 0; --i) { + CharGroup group = nodeArray.mData.get(i); if (null != group.mChildren) - size += countNodes(group.mChildren); + size += countNodeArrays(group.mChildren); } return size; } @@ -709,10 +710,10 @@ public final class FusionDictionary implements Iterable<Word> { // Recursively find out whether there are any bigrams. // This can be pretty expensive especially if there aren't any (we return as soon // as we find one, so it's much cheaper if there are bigrams) - private static boolean hasBigramsInternal(final Node node) { - if (null == node) return false; - for (int i = node.mData.size() - 1; i >= 0; --i) { - CharGroup group = node.mData.get(i); + private static boolean hasBigramsInternal(final PtNodeArray nodeArray) { + if (null == nodeArray) return false; + for (int i = nodeArray.mData.size() - 1; i >= 0; --i) { + CharGroup group = nodeArray.mData.get(i); if (null != group.mBigrams) return true; if (hasBigramsInternal(group.mChildren)) return true; } @@ -729,7 +730,7 @@ public final class FusionDictionary implements Iterable<Word> { // find a more efficient way of doing this, without compromising too much on memory // and ease of use. public boolean hasBigrams() { - return hasBigramsInternal(mRoot); + return hasBigramsInternal(mRootNodeArray); } // Historically, the tails of the words were going to be merged to save space. @@ -750,13 +751,13 @@ public final class FusionDictionary implements Iterable<Word> { // MakedictLog.i("Merging nodes. Number of nodes : " + countNodes(root)); // MakedictLog.i("Number of groups : " + countCharGroups(root)); // -// final HashMap<String, ArrayList<Node>> repository = -// new HashMap<String, ArrayList<Node>>(); +// final HashMap<String, ArrayList<PtNodeArray>> repository = +// new HashMap<String, ArrayList<PtNodeArray>>(); // mergeTailsInner(repository, root); // // MakedictLog.i("Number of different pseudohashes : " + repository.size()); // int size = 0; -// for (ArrayList<Node> a : repository.values()) { +// for (ArrayList<PtNodeArray> a : repository.values()) { // size += a.size(); // } // MakedictLog.i("Number of nodes after merge : " + (1 + size)); @@ -764,7 +765,7 @@ public final class FusionDictionary implements Iterable<Word> { } // The following methods are used by the deactivated mergeTails() -// private static boolean isEqual(Node a, Node b) { +// private static boolean isEqual(PtNodeArray a, PtNodeArray b) { // if (null == a && null == b) return true; // if (null == a || null == b) return false; // if (a.data.size() != b.data.size()) return false; @@ -781,21 +782,21 @@ public final class FusionDictionary implements Iterable<Word> { // return true; // } -// static private HashMap<String, ArrayList<Node>> mergeTailsInner( -// final HashMap<String, ArrayList<Node>> map, final Node node) { -// final ArrayList<CharGroup> branches = node.data; +// static private HashMap<String, ArrayList<PtNodeArray>> mergeTailsInner( +// final HashMap<String, ArrayList<PtNodeArray>> map, final PtNodeArray nodeArray) { +// final ArrayList<CharGroup> branches = nodeArray.data; // final int nodeSize = branches.size(); // for (int i = 0; i < nodeSize; ++i) { // CharGroup group = branches.get(i); // if (null != group.children) { // String pseudoHash = getPseudoHash(group.children); -// ArrayList<Node> similarList = map.get(pseudoHash); +// ArrayList<PtNodeArray> similarList = map.get(pseudoHash); // if (null == similarList) { -// similarList = new ArrayList<Node>(); +// similarList = new ArrayList<PtNodeArray>(); // map.put(pseudoHash, similarList); // } // boolean merged = false; -// for (Node similar : similarList) { +// for (PtNodeArray similar : similarList) { // if (isEqual(group.children, similar)) { // group.children = similar; // merged = true; @@ -811,9 +812,9 @@ public final class FusionDictionary implements Iterable<Word> { // return map; // } -// private static String getPseudoHash(final Node node) { +// private static String getPseudoHash(final PtNodeArray nodeArray) { // StringBuilder s = new StringBuilder(); -// for (CharGroup g : node.data) { +// for (CharGroup g : nodeArray.data) { // s.append(g.frequency); // for (int ch : g.chars) { // s.append(Character.toChars(ch)); @@ -901,6 +902,6 @@ public final class FusionDictionary implements Iterable<Word> { */ @Override public Iterator<Word> iterator() { - return new DictionaryIterator(mRoot.mData); + return new DictionaryIterator(mRootNodeArray.mData); } } diff --git a/java/src/com/android/inputmethod/latin/makedict/decoder/HeaderReader.java b/java/src/com/android/inputmethod/latin/makedict/decoder/HeaderReader.java new file mode 100644 index 000000000..f2badb444 --- /dev/null +++ b/java/src/com/android/inputmethod/latin/makedict/decoder/HeaderReader.java @@ -0,0 +1,32 @@ +/* + * Copyright (C) 2013 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.makedict.decoder; + +import com.android.inputmethod.latin.makedict.UnsupportedFormatException; + +import java.io.IOException; +import java.util.HashMap; + +/** + * An interface to read a binary dictionary file header. + */ +public interface HeaderReader { + public int readVersion() throws IOException, UnsupportedFormatException; + public int readOptionFlags(); + public int readHeaderSize(); + public HashMap<String, String> readAttributes(final int headerSize); +} diff --git a/java/src/com/android/inputmethod/latin/personalization/DynamicPredictionDictionaryBase.java b/java/src/com/android/inputmethod/latin/personalization/DynamicPredictionDictionaryBase.java index bb6ec6b1c..916be4481 100644 --- a/java/src/com/android/inputmethod/latin/personalization/DynamicPredictionDictionaryBase.java +++ b/java/src/com/android/inputmethod/latin/personalization/DynamicPredictionDictionaryBase.java @@ -28,9 +28,9 @@ import com.android.inputmethod.latin.ExpandableDictionary; import com.android.inputmethod.latin.LatinImeLogger; import com.android.inputmethod.latin.SuggestedWords.SuggestedWordInfo; import com.android.inputmethod.latin.WordComposer; +import com.android.inputmethod.latin.makedict.BinaryDictDecoder; import com.android.inputmethod.latin.makedict.FormatSpec.FormatOptions; import com.android.inputmethod.latin.settings.Settings; -import com.android.inputmethod.latin.utils.ByteArrayWrapper; import com.android.inputmethod.latin.utils.CollectionUtils; import com.android.inputmethod.latin.utils.UserHistoryDictIOUtils; import com.android.inputmethod.latin.utils.UserHistoryDictIOUtils.BigramDictionaryInterface; @@ -39,11 +39,11 @@ import com.android.inputmethod.latin.utils.UserHistoryForgettingCurveUtils; import com.android.inputmethod.latin.utils.UserHistoryForgettingCurveUtils.ForgettingCurveParams; import java.io.File; -import java.io.FileInputStream; import java.io.FileNotFoundException; import java.io.FileOutputStream; import java.io.IOException; import java.util.ArrayList; +import java.util.concurrent.atomic.AtomicReference; import java.util.concurrent.locks.ReentrantLock; /** @@ -76,6 +76,8 @@ public abstract class DynamicPredictionDictionaryBase extends ExpandableDictiona private final ArrayList<PersonalizationDictionaryUpdateSession> mSessions = CollectionUtils.newArrayList(); + private final AtomicReference<AsyncTask<Void, Void, Void>> mWaitingTask; + // Should always be false except when we use this class for test @UsedForTesting boolean mIsTest = false; @@ -84,6 +86,7 @@ public abstract class DynamicPredictionDictionaryBase extends ExpandableDictiona super(context, dictionaryType); mLocale = locale; mPrefs = sp; + mWaitingTask = new AtomicReference<AsyncTask<Void, Void, Void>>(); if (mLocale != null && mLocale.length() > 1) { loadDictionary(); } @@ -175,7 +178,11 @@ public abstract class DynamicPredictionDictionaryBase extends ExpandableDictiona */ private void flushPendingWrites() { // Create a background thread to write the pending entries - new UpdateBinaryTask(mBigramList, mLocale, this, mPrefs, getContext()).execute(); + final AsyncTask<Void, Void, Void> old = mWaitingTask.getAndSet(new UpdateBinaryTask( + mBigramList, mLocale, this, mPrefs, getContext()).execute()); + if (old != null) { + old.cancel(false); + } } @Override @@ -234,27 +241,17 @@ public abstract class DynamicPredictionDictionaryBase extends ExpandableDictiona }; // Load the dictionary from binary file - FileInputStream inStream = null; + final BinaryDictDecoder reader = new BinaryDictDecoder( + new File(getContext().getFilesDir(), fileName)); try { - final File file = new File(getContext().getFilesDir(), fileName); - final byte[] buffer = new byte[(int)file.length()]; - inStream = new FileInputStream(file); - inStream.read(buffer); - UserHistoryDictIOUtils.readDictionaryBinary( - new ByteArrayWrapper(buffer), listener); + reader.openDictBuffer(new BinaryDictDecoder.DictionaryBufferFromByteArrayFactory()); + UserHistoryDictIOUtils.readDictionaryBinary(reader, listener); } catch (FileNotFoundException e) { // This is an expected condition: we don't have a user history dictionary for this // language yet. It will be created sometime later. } catch (IOException e) { Log.e(TAG, "IOException on opening a bytebuffer", e); } finally { - if (inStream != null) { - try { - inStream.close(); - } catch (IOException e) { - // do nothing - } - } if (PROFILE_SAVE_RESTORE) { final long diff = System.currentTimeMillis() - now; Log.d(TAG, "PROF: Load UserHistoryDictionary: " @@ -298,6 +295,7 @@ public abstract class DynamicPredictionDictionaryBase extends ExpandableDictiona @Override protected Void doInBackground(final Void... v) { + if (isCancelled()) return null; if (mDynamicPredictionDictionary.mIsTest) { // If mIsTest == true, wait until the lock is released. mDynamicPredictionDictionary.mBigramListLock.lock(); @@ -317,6 +315,9 @@ public abstract class DynamicPredictionDictionaryBase extends ExpandableDictiona } private void doWriteTaskLocked() { + if (isCancelled()) return; + mDynamicPredictionDictionary.mWaitingTask.compareAndSet(this, null); + if (DBG_STRESS_TEST) { try { Log.w(TAG, "Start stress in closing: " + mLocale); @@ -402,7 +403,12 @@ public abstract class DynamicPredictionDictionaryBase extends ExpandableDictiona } public void registerUpdateSession(PersonalizationDictionaryUpdateSession session) { - session.setDictionary(this); + session.setPredictionDictionary(mLocale, this); mSessions.add(session); + session.onDictionaryReady(); + } + + public void unRegisterUpdateSession(PersonalizationDictionaryUpdateSession session) { + mSessions.remove(session); } } diff --git a/java/src/com/android/inputmethod/latin/personalization/PersonalizationDictionary.java b/java/src/com/android/inputmethod/latin/personalization/PersonalizationDictionary.java index e38a235e9..275ce2fdc 100644 --- a/java/src/com/android/inputmethod/latin/personalization/PersonalizationDictionary.java +++ b/java/src/com/android/inputmethod/latin/personalization/PersonalizationDictionary.java @@ -36,7 +36,9 @@ public class PersonalizationDictionary extends ExpandableBinaryDictionary { // Singleton private PersonalizationDictionary(final Context context, final String locale) { - super(context, getFilenameWithLocale(NAME, locale), Dictionary.TYPE_PERSONALIZATION); + // TODO: Make isUpdatable true. + super(context, getFilenameWithLocale(NAME, locale), Dictionary.TYPE_PERSONALIZATION, + false /* isUpdatable */); mLocale = locale; } diff --git a/java/src/com/android/inputmethod/latin/personalization/PersonalizationDictionaryHelper.java b/java/src/com/android/inputmethod/latin/personalization/PersonalizationDictionaryHelper.java index da256f827..b4fd25024 100644 --- a/java/src/com/android/inputmethod/latin/personalization/PersonalizationDictionaryHelper.java +++ b/java/src/com/android/inputmethod/latin/personalization/PersonalizationDictionaryHelper.java @@ -59,9 +59,8 @@ public class PersonalizationDictionaryHelper { } } - public static void - registerPersonalizationDictionaryUpdateSession(final Context context, - final PersonalizationDictionaryUpdateSession session) { + public static void registerPersonalizationDictionaryUpdateSession(final Context context, + final PersonalizationDictionaryUpdateSession session) { final PersonalizationPredictionDictionary dictionary = getPersonalizationPredictionDictionary(context, context.getResources().getConfiguration().locale.toString(), diff --git a/java/src/com/android/inputmethod/latin/personalization/PersonalizationDictionarySessionRegister.java b/java/src/com/android/inputmethod/latin/personalization/PersonalizationDictionarySessionRegister.java new file mode 100644 index 000000000..da59333f5 --- /dev/null +++ b/java/src/com/android/inputmethod/latin/personalization/PersonalizationDictionarySessionRegister.java @@ -0,0 +1,34 @@ +/* + * Copyright (C) 2013 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.personalization; + +import android.content.Context; +import android.content.res.Configuration; + +public class PersonalizationDictionarySessionRegister { + public static void init(Context context) { + } + + public static void onConfigurationChanged(final Context context, final Configuration conf) { + } + + public static void onUpdateData(Context context, String type) { + } + + public static void onRemoveData(Context context, String type) { + } +} diff --git a/java/src/com/android/inputmethod/latin/personalization/PersonalizationDictionaryUpdateSession.java b/java/src/com/android/inputmethod/latin/personalization/PersonalizationDictionaryUpdateSession.java index d62aec19d..433c69c1c 100644 --- a/java/src/com/android/inputmethod/latin/personalization/PersonalizationDictionaryUpdateSession.java +++ b/java/src/com/android/inputmethod/latin/personalization/PersonalizationDictionaryUpdateSession.java @@ -16,6 +16,8 @@ package com.android.inputmethod.latin.personalization; +import android.content.Context; + import java.lang.ref.WeakReference; import java.util.ArrayList; @@ -43,18 +45,49 @@ public abstract class PersonalizationDictionaryUpdateSession { } // TODO: Use a dynamic binary dictionary instead - public WeakReference<DynamicPredictionDictionaryBase> mDictionary; + public WeakReference<DynamicPredictionDictionaryBase> mPredictionDictionary; + public String mLocale; public abstract void onDictionaryReady(); - public void setDictionary(DynamicPredictionDictionaryBase dictionary) { - mDictionary = new WeakReference<DynamicPredictionDictionaryBase>(dictionary); + public abstract void onDictionaryClosed(Context context); + + public void setPredictionDictionary(String locale, DynamicPredictionDictionaryBase dictionary) { + mPredictionDictionary = new WeakReference<DynamicPredictionDictionaryBase>(dictionary); + mLocale = locale; + } + + protected DynamicPredictionDictionaryBase getPredictionDictionary() { + return mPredictionDictionary == null ? null : mPredictionDictionary.get(); + } + + private void unsetPredictionDictionary() { + final DynamicPredictionDictionaryBase dictionary = getPredictionDictionary(); + if (dictionary == null) { + return; + } + dictionary.unRegisterUpdateSession(this); + } + + + public void closeSession(Context context) { + unsetPredictionDictionary(); + onDictionaryClosed(context); + } + + public void addBigramToPersonalizationDictionary(String word0, String word1, boolean isValid, + int frequency) { + final DynamicPredictionDictionaryBase dictionary = getPredictionDictionary(); + if (dictionary == null) { + return; + } + dictionary.addToPersonalizationPredictionDictionary(word0, word1, isValid); } - public void addToPersonalizationDictionary( + // Bulk import + public void addBigramsToPersonalizationDictionary( final ArrayList<PersonalizationLanguageModelParam> lmParams) { - final DynamicPredictionDictionaryBase dictionary = mDictionary == null - ? null : mDictionary.get(); + final DynamicPredictionDictionaryBase dictionary = getPredictionDictionary(); if (dictionary == null) { return; } diff --git a/java/src/com/android/inputmethod/latin/personalization/PersonalizationPredictionDictionary.java b/java/src/com/android/inputmethod/latin/personalization/PersonalizationPredictionDictionary.java index 955bd2762..a038d0ab2 100644 --- a/java/src/com/android/inputmethod/latin/personalization/PersonalizationPredictionDictionary.java +++ b/java/src/com/android/inputmethod/latin/personalization/PersonalizationPredictionDictionary.java @@ -17,6 +17,7 @@ package com.android.inputmethod.latin.personalization; import com.android.inputmethod.latin.Dictionary; +import com.android.inputmethod.latin.ExpandableBinaryDictionary; import android.content.Context; import android.content.SharedPreferences; @@ -31,6 +32,6 @@ public class PersonalizationPredictionDictionary extends DynamicPredictionDictio @Override protected String getDictionaryFileName() { - return NAME + "." + getLocale() + ".dict"; + return NAME + "." + getLocale() + ExpandableBinaryDictionary.DICT_FILE_EXTENSION; } } diff --git a/java/src/com/android/inputmethod/latin/personalization/UserHistoryDictionaryBigramList.java b/java/src/com/android/inputmethod/latin/personalization/UserHistoryDictionaryBigramList.java index f21db25a6..9f289e9ff 100644 --- a/java/src/com/android/inputmethod/latin/personalization/UserHistoryDictionaryBigramList.java +++ b/java/src/com/android/inputmethod/latin/personalization/UserHistoryDictionaryBigramList.java @@ -53,7 +53,7 @@ public final class UserHistoryDictionaryBigramList { * Called when loaded from the SQL DB. */ public void addBigram(String word1, String word2, byte fcValue) { - if (UserHistoryPredictionDictionary.DBG_SAVE_RESTORE) { + if (DynamicPredictionDictionaryBase.DBG_SAVE_RESTORE) { Log.d(TAG, "--- add bigram: " + word1 + ", " + word2 + ", " + fcValue); } final HashMap<String, Byte> map; @@ -73,7 +73,7 @@ public final class UserHistoryDictionaryBigramList { * Called when inserted to the SQL DB. */ public void updateBigram(String word1, String word2, byte fcValue) { - if (UserHistoryPredictionDictionary.DBG_SAVE_RESTORE) { + if (DynamicPredictionDictionaryBase.DBG_SAVE_RESTORE) { Log.d(TAG, "--- update bigram: " + word1 + ", " + word2 + ", " + fcValue); } final HashMap<String, Byte> map; diff --git a/java/src/com/android/inputmethod/latin/personalization/UserHistoryPredictionDictionary.java b/java/src/com/android/inputmethod/latin/personalization/UserHistoryPredictionDictionary.java index d11784454..76e48c744 100644 --- a/java/src/com/android/inputmethod/latin/personalization/UserHistoryPredictionDictionary.java +++ b/java/src/com/android/inputmethod/latin/personalization/UserHistoryPredictionDictionary.java @@ -17,6 +17,7 @@ package com.android.inputmethod.latin.personalization; import com.android.inputmethod.latin.Dictionary; +import com.android.inputmethod.latin.ExpandableBinaryDictionary; import android.content.Context; import android.content.SharedPreferences; @@ -26,7 +27,8 @@ import android.content.SharedPreferences; * cancellation or manual picks. This allows the keyboard to adapt to the typist over time. */ public class UserHistoryPredictionDictionary extends DynamicPredictionDictionaryBase { - private static final String NAME = UserHistoryPredictionDictionary.class.getSimpleName(); + /* package for tests */ static final String NAME = + UserHistoryPredictionDictionary.class.getSimpleName(); /* package */ UserHistoryPredictionDictionary(final Context context, final String locale, final SharedPreferences sp) { super(context, locale, sp, Dictionary.TYPE_USER_HISTORY); @@ -34,6 +36,6 @@ public class UserHistoryPredictionDictionary extends DynamicPredictionDictionary @Override protected String getDictionaryFileName() { - return NAME + "." + getLocale() + ".dict"; + return NAME + "." + getLocale() + ExpandableBinaryDictionary.DICT_FILE_EXTENSION; } } diff --git a/java/src/com/android/inputmethod/latin/settings/Settings.java b/java/src/com/android/inputmethod/latin/settings/Settings.java index d432087d3..20b675257 100644 --- a/java/src/com/android/inputmethod/latin/settings/Settings.java +++ b/java/src/com/android/inputmethod/latin/settings/Settings.java @@ -27,10 +27,10 @@ import com.android.inputmethod.latin.AudioAndHapticFeedbackManager; import com.android.inputmethod.latin.InputAttributes; import com.android.inputmethod.latin.R; import com.android.inputmethod.latin.utils.AdditionalSubtypeUtils; -import com.android.inputmethod.latin.utils.DebugLogUtils; import com.android.inputmethod.latin.utils.LocaleUtils; import com.android.inputmethod.latin.utils.ResourceUtils; import com.android.inputmethod.latin.utils.RunInLocale; +import com.android.inputmethod.latin.utils.StringUtils; import java.util.HashMap; import java.util.Locale; @@ -90,6 +90,8 @@ public final class Settings implements SharedPreferences.OnSharedPreferenceChang private static final String PREF_SUPPRESS_LANGUAGE_SWITCH_KEY = "pref_suppress_language_switch_key"; + private static final String PREF_LAST_USED_PERSONALIZATION_TOKEN = + "pref_last_used_personalization_token"; public static final String PREF_SEND_FEEDBACK = "send_feedback"; public static final String PREF_ABOUT_KEYBOARD = "about_keyboard"; @@ -343,4 +345,14 @@ public final class Settings implements SharedPreferences.OnSharedPreferenceChang return prefs.getBoolean( DebugSettings.PREF_USE_ONLY_PERSONALIZATION_DICTIONARY_FOR_DEBUG, false); } + + public void writeLastUsedPersonalizationToken(byte[] token) { + final String tokenStr = StringUtils.byteArrayToHexString(token); + mPrefs.edit().putString(PREF_LAST_USED_PERSONALIZATION_TOKEN, tokenStr).apply(); + } + + public byte[] readLastUsedPersonalizationToken() { + final String tokenStr = mPrefs.getString(PREF_LAST_USED_PERSONALIZATION_TOKEN, null); + return StringUtils.hexStringToByteArray(tokenStr); + } } diff --git a/java/src/com/android/inputmethod/latin/settings/SettingsFragment.java b/java/src/com/android/inputmethod/latin/settings/SettingsFragment.java index 446777704..1677e1828 100644 --- a/java/src/com/android/inputmethod/latin/settings/SettingsFragment.java +++ b/java/src/com/android/inputmethod/latin/settings/SettingsFragment.java @@ -229,10 +229,10 @@ public final class SettingsFragment extends InputMethodSettingsFragment if (!Settings.readFromBuildConfigIfGestureInputEnabled(res)) { removePreference(Settings.PREF_GESTURE_SETTINGS, getPreferenceScreen()); - } else { - AdditionalFeaturesSettingUtils.addAdditionalFeaturesPreferences(context, this); } + AdditionalFeaturesSettingUtils.addAdditionalFeaturesPreferences(context, this); + setupKeyLongpressTimeoutSettings(prefs, res); setupKeypressVibrationDurationSettings(prefs, res); setupKeypressSoundVolumeSettings(prefs, res); diff --git a/java/src/com/android/inputmethod/latin/settings/SettingsValues.java b/java/src/com/android/inputmethod/latin/settings/SettingsValues.java index a25cf620c..a0b744dd2 100644 --- a/java/src/com/android/inputmethod/latin/settings/SettingsValues.java +++ b/java/src/com/android/inputmethod/latin/settings/SettingsValues.java @@ -22,6 +22,7 @@ import android.content.res.Resources; import android.util.Log; import android.view.inputmethod.EditorInfo; +import com.android.inputmethod.annotations.UsedForTesting; import com.android.inputmethod.keyboard.internal.KeySpecParser; import com.android.inputmethod.latin.Dictionary; import com.android.inputmethod.latin.InputAttributes; @@ -56,6 +57,7 @@ public final class SettingsValues { public final SuggestedWords mSuggestPuncList; public final String mWordSeparators; public final CharSequence mHintToSaveText; + public final boolean mCurrentLanguageHasSpaces; // From preferences, in the same order as xml/prefs.xml: public final boolean mAutoCap; @@ -117,6 +119,7 @@ public final class SettingsValues { mSuggestPuncList = createSuggestPuncList(suggestPuncsSpec); mWordSeparators = res.getString(R.string.symbols_word_separators); mHintToSaveText = res.getText(R.string.hint_add_to_dictionary); + mCurrentLanguageHasSpaces = res.getBoolean(R.bool.current_language_has_spaces); // Store the input attributes if (null == inputAttributes) { @@ -170,6 +173,56 @@ public final class SettingsValues { mIsInternal = Settings.isInternal(prefs); } + // Only for tests + private SettingsValues(final Locale locale) { + // TODO: locale is saved, but not used yet. May have to change this if tests require. + mLocale = locale; + mDelayUpdateOldSuggestions = 0; + mSymbolsPrecededBySpace = new int[] { '(', '[', '{', '&' }; + Arrays.sort(mSymbolsPrecededBySpace); + mSymbolsFollowedBySpace = new int[] { '.', ',', ';', ':', '!', '?', ')', ']', '}', '&' }; + Arrays.sort(mSymbolsFollowedBySpace); + mWordConnectors = new int[] { '\'', '-' }; + Arrays.sort(mWordConnectors); + final String[] suggestPuncsSpec = new String[] { "!", "?", ",", ":", ";" }; + mSuggestPuncList = createSuggestPuncList(suggestPuncsSpec); + mWordSeparators = "&\t \n()[]{}*&<>+=|.,;:!?/_\""; + mHintToSaveText = "Touch again to save"; + mCurrentLanguageHasSpaces = true; + mInputAttributes = new InputAttributes(null, false /* isFullscreenMode */); + mAutoCap = true; + mVibrateOn = true; + mSoundOn = true; + mKeyPreviewPopupOn = true; + mSlidingKeyInputPreviewEnabled = true; + mVoiceMode = "0"; + mIncludesOtherImesInLanguageSwitchList = false; + mShowsLanguageSwitchKey = true; + mUseContactsDict = true; + mUseDoubleSpacePeriod = true; + mBlockPotentiallyOffensive = true; + mAutoCorrectEnabled = true; + mBigramPredictionEnabled = true; + mKeyLongpressTimeout = 300; + mKeypressVibrationDuration = 5; + mKeypressSoundVolume = 1; + mKeyPreviewPopupDismissDelay = 70; + mAutoCorrectionThreshold = 1; + mVoiceKeyEnabled = true; + mVoiceKeyOnMain = true; + mGestureInputEnabled = true; + mGestureTrailEnabled = true; + mGestureFloatingPreviewTextEnabled = true; + mCorrectionEnabled = mAutoCorrectEnabled && !mInputAttributes.mInputTypeNoAutoCorrect; + mSuggestionVisibility = 0; + mIsInternal = false; + } + + @UsedForTesting + public static SettingsValues makeDummySettingsValuesForTest(final Locale locale) { + return new SettingsValues(locale); + } + public boolean isApplicationSpecifiedCompletionsOn() { return mInputAttributes.mApplicationSpecifiedCompletionOn; } @@ -194,6 +247,10 @@ public final class SettingsValues { return Arrays.binarySearch(mWordConnectors, code) >= 0; } + public boolean isWordCodePoint(final int code) { + return Character.isLetter(code) || isWordConnector(code); + } + public boolean isUsuallyPrecededBySpace(final int code) { return Arrays.binarySearch(mSymbolsPrecededBySpace, code) >= 0; } diff --git a/java/src/com/android/inputmethod/latin/suggestions/MoreSuggestions.java b/java/src/com/android/inputmethod/latin/suggestions/MoreSuggestions.java index e97069dff..acd47450b 100644 --- a/java/src/com/android/inputmethod/latin/suggestions/MoreSuggestions.java +++ b/java/src/com/android/inputmethod/latin/suggestions/MoreSuggestions.java @@ -210,7 +210,8 @@ public final class MoreSuggestions extends Keyboard { final int indexInMoreSuggestions = index + SUGGESTION_CODE_BASE; final Key key = new Key( params, word, info, KeyboardIconsSet.ICON_UNDEFINED, indexInMoreSuggestions, - null, x, y, width, params.mDefaultRowHeight, 0); + null /* outputText */, x, y, width, params.mDefaultRowHeight, + 0 /* labelFlags */, Key.BACKGROUND_TYPE_NORMAL); params.markAsEdgeKey(key, index); params.onAddKey(key); final int columnNumber = params.getColumnNumber(index); diff --git a/java/src/com/android/inputmethod/latin/suggestions/SuggestionStripLayoutHelper.java b/java/src/com/android/inputmethod/latin/suggestions/SuggestionStripLayoutHelper.java index 1dd04fc4d..bcf64a8e8 100644 --- a/java/src/com/android/inputmethod/latin/suggestions/SuggestionStripLayoutHelper.java +++ b/java/src/com/android/inputmethod/latin/suggestions/SuggestionStripLayoutHelper.java @@ -177,20 +177,9 @@ final class SuggestionStripLayoutHelper { return mMaxMoreSuggestionsRow; } - private int getMoreSuggestionsHeight() { - return mMaxMoreSuggestionsRow * mMoreSuggestionsRowHeight + mMoreSuggestionsBottomGap; - } - - public int setMoreSuggestionsHeight(final int remainingHeight) { - final int currentHeight = getMoreSuggestionsHeight(); - if (currentHeight <= remainingHeight) { - return currentHeight; - } - + public void setMoreSuggestionsHeight(final int remainingHeight) { mMaxMoreSuggestionsRow = (remainingHeight - mMoreSuggestionsBottomGap) / mMoreSuggestionsRowHeight; - final int newHeight = getMoreSuggestionsHeight(); - return newHeight; } private static Drawable getMoreSuggestionsHint(final Resources res, final float textSize, diff --git a/java/src/com/android/inputmethod/latin/suggestions/SuggestionStripView.java b/java/src/com/android/inputmethod/latin/suggestions/SuggestionStripView.java index a8a14a825..badc942b9 100644 --- a/java/src/com/android/inputmethod/latin/suggestions/SuggestionStripView.java +++ b/java/src/com/android/inputmethod/latin/suggestions/SuggestionStripView.java @@ -135,8 +135,8 @@ public final class SuggestionStripView extends RelativeLayout implements OnClick } } - public int setMoreSuggestionsHeight(final int remainingHeight) { - return mLayoutHelper.setMoreSuggestionsHeight(remainingHeight); + public void setMoreSuggestionsHeight(final int remainingHeight) { + mLayoutHelper.setMoreSuggestionsHeight(remainingHeight); } public boolean isShowingAddToDictionaryHint() { @@ -198,7 +198,7 @@ public final class SuggestionStripView extends RelativeLayout implements OnClick @Override public boolean onLongClick(final View view) { - AudioAndHapticFeedbackManager.getInstance().hapticAndAudioFeedback( + AudioAndHapticFeedbackManager.getInstance().performHapticAndAudioFeedback( Constants.NOT_A_CODE, this); return showMoreSuggestions(); } diff --git a/java/src/com/android/inputmethod/latin/utils/ByteArrayWrapper.java b/java/src/com/android/inputmethod/latin/utils/ByteArrayDictBuffer.java index 1bb27aa2b..2028298f2 100644 --- a/java/src/com/android/inputmethod/latin/utils/ByteArrayWrapper.java +++ b/java/src/com/android/inputmethod/latin/utils/ByteArrayDictBuffer.java @@ -16,17 +16,17 @@ package com.android.inputmethod.latin.utils; -import com.android.inputmethod.latin.makedict.BinaryDictInputOutput.FusionDictionaryBufferInterface; +import com.android.inputmethod.latin.makedict.BinaryDictDecoderUtils.DictBuffer; /** * This class provides an implementation for the FusionDictionary buffer interface that is backed * by a simpled byte array. It allows to create a binary dictionary in memory. */ -public final class ByteArrayWrapper implements FusionDictionaryBufferInterface { +public final class ByteArrayDictBuffer implements DictBuffer { private byte[] mBuffer; private int mPosition; - public ByteArrayWrapper(final byte[] buffer) { + public ByteArrayDictBuffer(final byte[] buffer) { mBuffer = buffer; mPosition = 0; } diff --git a/java/src/com/android/inputmethod/latin/utils/CollectionUtils.java b/java/src/com/android/inputmethod/latin/utils/CollectionUtils.java index 98f0d8b68..cc25102ce 100644 --- a/java/src/com/android/inputmethod/latin/utils/CollectionUtils.java +++ b/java/src/com/android/inputmethod/latin/utils/CollectionUtils.java @@ -18,6 +18,7 @@ package com.android.inputmethod.latin.utils; import android.util.SparseArray; +import java.util.ArrayDeque; import java.util.ArrayList; import java.util.Collection; import java.util.Collections; @@ -94,6 +95,10 @@ public final class CollectionUtils { return new CopyOnWriteArrayList<E>(array); } + public static <E> ArrayDeque<E> newArrayDeque() { + return new ArrayDeque<E>(); + } + public static <E> SparseArray<E> newSparseArray() { return new SparseArray<E>(); } diff --git a/java/src/com/android/inputmethod/latin/utils/StringUtils.java b/java/src/com/android/inputmethod/latin/utils/StringUtils.java index 7406d855a..be4184093 100644 --- a/java/src/com/android/inputmethod/latin/utils/StringUtils.java +++ b/java/src/com/android/inputmethod/latin/utils/StringUtils.java @@ -18,7 +18,9 @@ package com.android.inputmethod.latin.utils; import android.text.TextUtils; +import com.android.inputmethod.annotations.UsedForTesting; import com.android.inputmethod.latin.Constants; +import com.android.inputmethod.latin.settings.SettingsValues; import java.util.ArrayList; import java.util.Locale; @@ -193,27 +195,56 @@ public final class StringUtils { } public static boolean isIdenticalAfterUpcase(final String text) { - final int len = text.length(); - for (int i = 0; i < len; i = text.offsetByCodePoints(i, 1)) { + final int length = text.length(); + int i = 0; + while (i < length) { final int codePoint = text.codePointAt(i); if (Character.isLetter(codePoint) && !Character.isUpperCase(codePoint)) { return false; } + i += Character.charCount(codePoint); } return true; } public static boolean isIdenticalAfterDowncase(final String text) { - final int len = text.length(); - for (int i = 0; i < len; i = text.offsetByCodePoints(i, 1)) { + final int length = text.length(); + int i = 0; + while (i < length) { final int codePoint = text.codePointAt(i); if (Character.isLetter(codePoint) && !Character.isLowerCase(codePoint)) { return false; } + i += Character.charCount(codePoint); } return true; } + @UsedForTesting + public static boolean looksValidForDictionaryInsertion(final CharSequence text, + final SettingsValues settings) { + if (TextUtils.isEmpty(text)) return false; + final int length = text.length(); + int i = 0; + int digitCount = 0; + while (i < length) { + final int codePoint = Character.codePointAt(text, i); + final int charCount = Character.charCount(codePoint); + i += charCount; + if (Character.isDigit(codePoint)) { + // Count digits: see below + digitCount += charCount; + continue; + } + if (!settings.isWordCodePoint(codePoint)) return false; + } + // We reject strings entirely comprised of digits to avoid using PIN codes or credit + // card numbers. It would come in handy for word prediction though; a good example is + // when writing one's address where the street number is usually quite discriminative, + // as well as the postal code. + return digitCount < length; + } + public static boolean isIdenticalAfterCapitalizeEachWord(final String text, final String separators) { boolean needCapsNext = true; @@ -316,4 +347,47 @@ public final class StringUtils { // Otherwise, it doesn't look like an URL. return false; } + + public static boolean isEmptyStringOrWhiteSpaces(String s) { + final int N = codePointCount(s); + for (int i = 0; i < N; ++i) { + if (!Character.isWhitespace(s.codePointAt(i))) { + return false; + } + } + return true; + } + + @UsedForTesting + public static String byteArrayToHexString(byte[] bytes) { + if (bytes == null || bytes.length == 0) { + return ""; + } + final StringBuilder sb = new StringBuilder(); + for (byte b : bytes) { + sb.append(String.format("%02x", b & 0xff)); + } + return sb.toString(); + } + + /** + * Convert hex string to byte array. The string length must be an even number. + */ + @UsedForTesting + public static byte[] hexStringToByteArray(String hexString) { + if (TextUtils.isEmpty(hexString)) { + return null; + } + final int N = hexString.length(); + if (N % 2 != 0) { + throw new NumberFormatException("Input hex string length must be an even number." + + " Length = " + N); + } + final byte[] bytes = new byte[N / 2]; + for (int i = 0; i < N; i += 2) { + bytes[i / 2] = (byte) ((Character.digit(hexString.charAt(i), 16) << 4) + + Character.digit(hexString.charAt(i + 1), 16)); + } + return bytes; + } } diff --git a/java/src/com/android/inputmethod/latin/utils/UserHistoryDictIOUtils.java b/java/src/com/android/inputmethod/latin/utils/UserHistoryDictIOUtils.java index d02f7187e..dd7f534dc 100644 --- a/java/src/com/android/inputmethod/latin/utils/UserHistoryDictIOUtils.java +++ b/java/src/com/android/inputmethod/latin/utils/UserHistoryDictIOUtils.java @@ -19,12 +19,12 @@ package com.android.inputmethod.latin.utils; import android.util.Log; import com.android.inputmethod.annotations.UsedForTesting; +import com.android.inputmethod.latin.makedict.BinaryDictDecoder; +import com.android.inputmethod.latin.makedict.BinaryDictEncoder; import com.android.inputmethod.latin.makedict.BinaryDictIOUtils; -import com.android.inputmethod.latin.makedict.BinaryDictInputOutput; -import com.android.inputmethod.latin.makedict.BinaryDictInputOutput.FusionDictionaryBufferInterface; import com.android.inputmethod.latin.makedict.FormatSpec.FormatOptions; import com.android.inputmethod.latin.makedict.FusionDictionary; -import com.android.inputmethod.latin.makedict.FusionDictionary.Node; +import com.android.inputmethod.latin.makedict.FusionDictionary.PtNodeArray; import com.android.inputmethod.latin.makedict.PendingAttribute; import com.android.inputmethod.latin.makedict.UnsupportedFormatException; import com.android.inputmethod.latin.personalization.UserHistoryDictionaryBigramList; @@ -62,7 +62,7 @@ public final class UserHistoryDictIOUtils { final FormatOptions formatOptions) { final FusionDictionary fusionDict = constructFusionDictionary(dict, bigrams); try { - BinaryDictInputOutput.writeDictionaryBinary(destination, fusionDict, formatOptions); + BinaryDictEncoder.writeDictionaryBinary(destination, fusionDict, formatOptions); Log.d(TAG, "end writing"); } catch (IOException e) { Log.e(TAG, "IO exception while writing file", e); @@ -77,7 +77,7 @@ public final class UserHistoryDictIOUtils { @UsedForTesting static FusionDictionary constructFusionDictionary( final BigramDictionaryInterface dict, final UserHistoryDictionaryBigramList bigrams) { - final FusionDictionary fusionDict = new FusionDictionary(new Node(), + final FusionDictionary fusionDict = new FusionDictionary(new PtNodeArray(), new FusionDictionary.DictionaryOptions(new HashMap<String, String>(), false, false)); int profTotal = 0; @@ -101,7 +101,7 @@ public final class UserHistoryDictIOUtils { if (word1 == null) { // unigram fusionDict.add(word2, freq, null, false /* isNotAWord */); } else { // bigram - if (FusionDictionary.findWordInTree(fusionDict.mRoot, word1) == null) { + if (FusionDictionary.findWordInTree(fusionDict.mRootNodeArray, word1) == null) { fusionDict.add(word1, 2, null, false /* isNotAWord */); } fusionDict.setBigram(word1, word2, freq); @@ -118,13 +118,13 @@ public final class UserHistoryDictIOUtils { /** * Reads dictionary from file. */ - public static void readDictionaryBinary(final FusionDictionaryBufferInterface buffer, + public static void readDictionaryBinary(final BinaryDictDecoder dictDecoder, final OnAddWordListener dict) { final Map<Integer, String> unigrams = CollectionUtils.newTreeMap(); final Map<Integer, Integer> frequencies = CollectionUtils.newTreeMap(); final Map<Integer, ArrayList<PendingAttribute>> bigrams = CollectionUtils.newTreeMap(); try { - BinaryDictIOUtils.readUnigramsAndBigramsBinary(buffer, unigrams, frequencies, + BinaryDictIOUtils.readUnigramsAndBigramsBinary(dictDecoder, unigrams, frequencies, bigrams); } catch (IOException e) { Log.e(TAG, "IO exception while reading file", e); @@ -156,7 +156,7 @@ public final class UserHistoryDictIOUtils { continue; } to.setBigram(word1, word2, - BinaryDictInputOutput.reconstructBigramFrequency(unigramFrequency, + BinaryDictIOUtils.reconstructBigramFrequency(unigramFrequency, attr.mFrequency)); } } diff --git a/java/src/com/android/inputmethod/latin/utils/UserLogRingCharBuffer.java b/java/src/com/android/inputmethod/latin/utils/UserLogRingCharBuffer.java index 161386e2e..a75d353c9 100644 --- a/java/src/com/android/inputmethod/latin/utils/UserLogRingCharBuffer.java +++ b/java/src/com/android/inputmethod/latin/utils/UserLogRingCharBuffer.java @@ -19,6 +19,7 @@ package com.android.inputmethod.latin.utils; import android.inputmethodservice.InputMethodService; import com.android.inputmethod.annotations.UsedForTesting; +import com.android.inputmethod.latin.LatinImeLogger; import com.android.inputmethod.latin.settings.Settings; public final class UserLogRingCharBuffer { @@ -64,6 +65,9 @@ public final class UserLogRingCharBuffer { if (!mEnabled) { return; } + if (LatinImeLogger.sUsabilityStudy) { + UsabilityStudyLogUtils.getInstance().writeChar(c, x, y); + } mCharBuf[mEnd] = c; mXBuf[mEnd] = x; mYBuf[mEnd] = y; |