diff options
Diffstat (limited to 'java/src/com/android/inputmethod/latin/Suggest.java')
| -rw-r--r-- | java/src/com/android/inputmethod/latin/Suggest.java | 184 |
1 files changed, 98 insertions, 86 deletions
diff --git a/java/src/com/android/inputmethod/latin/Suggest.java b/java/src/com/android/inputmethod/latin/Suggest.java index 0de474e59..ca75866c0 100644 --- a/java/src/com/android/inputmethod/latin/Suggest.java +++ b/java/src/com/android/inputmethod/latin/Suggest.java @@ -27,6 +27,7 @@ import java.util.ArrayList; import java.util.Arrays; import java.util.HashMap; import java.util.HashSet; +import java.util.Locale; import java.util.Map; import java.util.Set; @@ -47,7 +48,7 @@ public class Suggest implements Dictionary.WordCallback { /** * Words that appear in both bigram and unigram data gets multiplier ranging from - * BIGRAM_MULTIPLIER_MIN to BIGRAM_MULTIPLIER_MAX depending on the frequency score from + * BIGRAM_MULTIPLIER_MIN to BIGRAM_MULTIPLIER_MAX depending on the score from * bigram data. */ public static final double BIGRAM_MULTIPLIER_MIN = 1.2; @@ -55,7 +56,7 @@ public class Suggest implements Dictionary.WordCallback { /** * Maximum possible bigram frequency. Will depend on how many bits are being used in data - * structure. Maximum bigram freqeuncy will get the BIGRAM_MULTIPLIER_MAX as the multiplier. + * structure. Maximum bigram frequency will get the BIGRAM_MULTIPLIER_MAX as the multiplier. */ public static final int MAXIMUM_BIGRAM_FREQUENCY = 127; @@ -74,13 +75,11 @@ public class Suggest implements Dictionary.WordCallback { public static final String DICT_KEY_USER_BIGRAM = "user_bigram"; public static final String DICT_KEY_WHITELIST ="whitelist"; - static final int LARGE_DICTIONARY_THRESHOLD = 200 * 1000; - private static final boolean DBG = LatinImeLogger.sDBG; private AutoCorrection mAutoCorrection; - private BinaryDictionary mMainDict; + private Dictionary mMainDict; private WhitelistDictionary mWhiteListDictionary; private final Map<String, Dictionary> mUnigramDictionaries = new HashMap<String, Dictionary>(); private final Map<String, Dictionary> mBigramDictionaries = new HashMap<String, Dictionary>(); @@ -92,13 +91,13 @@ public class Suggest implements Dictionary.WordCallback { private boolean mQuickFixesEnabled; private double mAutoCorrectionThreshold; - private int[] mPriorities = new int[mPrefMaxSuggestions]; - private int[] mBigramPriorities = new int[PREF_MAX_BIGRAMS]; + private int[] mScores = new int[mPrefMaxSuggestions]; + private int[] mBigramScores = new int[PREF_MAX_BIGRAMS]; private ArrayList<CharSequence> mSuggestions = new ArrayList<CharSequence>(); ArrayList<CharSequence> mBigramSuggestions = new ArrayList<CharSequence>(); private ArrayList<CharSequence> mStringPool = new ArrayList<CharSequence>(); - private String mLowerOriginalWord; + private CharSequence mTypedWord; // TODO: Remove these member variables by passing more context to addWord() callback method private boolean mIsFirstCharCapitalized; @@ -106,15 +105,18 @@ public class Suggest implements Dictionary.WordCallback { private int mCorrectionMode = CORRECTION_BASIC; - public Suggest(Context context, int dictionaryResId) { - init(context, BinaryDictionary.initDictionary(context, dictionaryResId, DIC_MAIN)); + public Suggest(Context context, int dictionaryResId, Locale locale) { + init(context, DictionaryFactory.createDictionaryFromManager(context, locale, + dictionaryResId)); } - /* package for test */ Suggest(File dictionary, long startOffset, long length) { - init(null, BinaryDictionary.initDictionary(dictionary, startOffset, length, DIC_MAIN)); + /* package for test */ Suggest(Context context, File dictionary, long startOffset, long length, + Flag[] flagArray) { + init(null, DictionaryFactory.createDictionaryForTest(context, dictionary, startOffset, + length, flagArray)); } - private void init(Context context, BinaryDictionary mainDict) { + private void init(Context context, Dictionary mainDict) { if (mainDict != null) { mMainDict = mainDict; mUnigramDictionaries.put(DICT_KEY_MAIN, mainDict); @@ -128,6 +130,19 @@ public class Suggest implements Dictionary.WordCallback { initPool(); } + public void resetMainDict(Context context, int dictionaryResId, Locale locale) { + final Dictionary newMainDict = DictionaryFactory.createDictionaryFromManager( + context, locale, dictionaryResId); + mMainDict = newMainDict; + if (null == newMainDict) { + mUnigramDictionaries.remove(DICT_KEY_MAIN); + mBigramDictionaries.remove(DICT_KEY_MAIN); + } else { + mUnigramDictionaries.put(DICT_KEY_MAIN, newMainDict); + mBigramDictionaries.put(DICT_KEY_MAIN, newMainDict); + } + } + private void initPool() { for (int i = 0; i < mPrefMaxSuggestions; i++) { StringBuilder sb = new StringBuilder(getApproxMaxWordLength()); @@ -148,7 +163,7 @@ public class Suggest implements Dictionary.WordCallback { } public boolean hasMainDictionary() { - return mMainDict != null && mMainDict.getSize() > LARGE_DICTIONARY_THRESHOLD; + return mMainDict != null; } public Map<String, Dictionary> getUnigramDictionaries() { @@ -207,8 +222,8 @@ public class Suggest implements Dictionary.WordCallback { throw new IllegalArgumentException("maxSuggestions must be between 1 and 100"); } mPrefMaxSuggestions = maxSuggestions; - mPriorities = new int[mPrefMaxSuggestions]; - mBigramPriorities = new int[PREF_MAX_BIGRAMS]; + mScores = new int[mPrefMaxSuggestions]; + mBigramScores = new int[PREF_MAX_BIGRAMS]; collectGarbage(mSuggestions, mPrefMaxSuggestions); while (mStringPool.size() < mPrefMaxSuggestions) { StringBuilder sb = new StringBuilder(getApproxMaxWordLength()); @@ -248,6 +263,16 @@ public class Suggest implements Dictionary.WordCallback { return sb; } + protected void addBigramToSuggestions(CharSequence bigram) { + final int poolSize = mStringPool.size(); + final StringBuilder sb = poolSize > 0 ? + (StringBuilder) mStringPool.remove(poolSize - 1) + : new StringBuilder(getApproxMaxWordLength()); + sb.setLength(0); + sb.append(bigram); + mSuggestions.add(sb); + } + // TODO: cleanup dictionaries looking up and suggestions building with SuggestedWords.Builder public SuggestedWords.Builder getSuggestedWordBuilder(View view, WordComposer wordComposer, CharSequence prevWordForBigram) { @@ -256,25 +281,23 @@ public class Suggest implements Dictionary.WordCallback { mIsFirstCharCapitalized = wordComposer.isFirstCharCapitalized(); mIsAllUpperCase = wordComposer.isAllUpperCase(); collectGarbage(mSuggestions, mPrefMaxSuggestions); - Arrays.fill(mPriorities, 0); + Arrays.fill(mScores, 0); // Save a lowercase version of the original word CharSequence typedWord = wordComposer.getTypedWord(); if (typedWord != null) { final String typedWordString = typedWord.toString(); typedWord = typedWordString; - mLowerOriginalWord = typedWordString.toLowerCase(); // Treating USER_TYPED as UNIGRAM suggestion for logging now. LatinImeLogger.onAddSuggestedWord(typedWordString, Suggest.DIC_USER_TYPED, Dictionary.DataType.UNIGRAM); - } else { - mLowerOriginalWord = ""; } + mTypedWord = typedWord; - if (wordComposer.size() == 1 && (mCorrectionMode == CORRECTION_FULL_BIGRAM + if (wordComposer.size() <= 1 && (mCorrectionMode == CORRECTION_FULL_BIGRAM || mCorrectionMode == CORRECTION_BASIC)) { // At first character typed, search only the bigrams - Arrays.fill(mBigramPriorities, 0); + Arrays.fill(mBigramScores, 0); collectGarbage(mBigramSuggestions, PREF_MAX_BIGRAMS); if (!TextUtils.isEmpty(prevWordForBigram)) { @@ -285,21 +308,26 @@ public class Suggest implements Dictionary.WordCallback { for (final Dictionary dictionary : mBigramDictionaries.values()) { dictionary.getBigrams(wordComposer, prevWordForBigram, this); } - char currentChar = wordComposer.getTypedWord().charAt(0); - char currentCharUpper = Character.toUpperCase(currentChar); - int count = 0; - int bigramSuggestionSize = mBigramSuggestions.size(); - for (int i = 0; i < bigramSuggestionSize; i++) { - if (mBigramSuggestions.get(i).charAt(0) == currentChar - || mBigramSuggestions.get(i).charAt(0) == currentCharUpper) { - int poolSize = mStringPool.size(); - StringBuilder sb = poolSize > 0 ? - (StringBuilder) mStringPool.remove(poolSize - 1) - : new StringBuilder(getApproxMaxWordLength()); - sb.setLength(0); - sb.append(mBigramSuggestions.get(i)); - mSuggestions.add(count++, sb); - if (count > mPrefMaxSuggestions) break; + if (TextUtils.isEmpty(typedWord)) { + // Nothing entered: return all bigrams for the previous word + int insertCount = Math.min(mBigramSuggestions.size(), mPrefMaxSuggestions); + for (int i = 0; i < insertCount; ++i) { + addBigramToSuggestions(mBigramSuggestions.get(i)); + } + } else { + // Word entered: return only bigrams that match the first char of the typed word + final char currentChar = typedWord.charAt(0); + final char currentCharUpper = Character.toUpperCase(currentChar); + int count = 0; + final int bigramSuggestionSize = mBigramSuggestions.size(); + for (int i = 0; i < bigramSuggestionSize; i++) { + final CharSequence bigramSuggestion = mBigramSuggestions.get(i); + final char bigramSuggestionFirstChar = bigramSuggestion.charAt(0); + if (bigramSuggestionFirstChar == currentChar + || bigramSuggestionFirstChar == currentCharUpper) { + addBigramToSuggestions(bigramSuggestion); + if (++count > mPrefMaxSuggestions) break; + } } } } @@ -346,7 +374,7 @@ public class Suggest implements Dictionary.WordCallback { mWhiteListDictionary.getWhiteListedWord(typedWordString)); mAutoCorrection.updateAutoCorrectionStatus(mUnigramDictionaries, wordComposer, - mSuggestions, mPriorities, typedWord, mAutoCorrectionThreshold, mCorrectionMode, + mSuggestions, mScores, typedWord, mAutoCorrectionThreshold, mCorrectionMode, autoText, whitelistedWord); if (autoText != null) { @@ -364,26 +392,25 @@ public class Suggest implements Dictionary.WordCallback { if (DBG) { double normalizedScore = mAutoCorrection.getNormalizedScore(); - ArrayList<SuggestedWords.SuggestedWordInfo> frequencyInfoList = + ArrayList<SuggestedWords.SuggestedWordInfo> scoreInfoList = new ArrayList<SuggestedWords.SuggestedWordInfo>(); - frequencyInfoList.add(new SuggestedWords.SuggestedWordInfo("+", false)); - final int priorityLength = mPriorities.length; - for (int i = 0; i < priorityLength; ++i) { + scoreInfoList.add(new SuggestedWords.SuggestedWordInfo("+", false)); + for (int i = 0; i < mScores.length; ++i) { if (normalizedScore > 0) { - final String priorityThreshold = Integer.toString(mPriorities[i]) + " (" + - normalizedScore + ")"; - frequencyInfoList.add( - new SuggestedWords.SuggestedWordInfo(priorityThreshold, false)); + final String scoreThreshold = String.format("%d (%4.2f)", mScores[i], + normalizedScore); + scoreInfoList.add( + new SuggestedWords.SuggestedWordInfo(scoreThreshold, false)); normalizedScore = 0.0; } else { - final String priority = Integer.toString(mPriorities[i]); - frequencyInfoList.add(new SuggestedWords.SuggestedWordInfo(priority, false)); + final String score = Integer.toString(mScores[i]); + scoreInfoList.add(new SuggestedWords.SuggestedWordInfo(score, false)); } } - for (int i = priorityLength; i < mSuggestions.size(); ++i) { - frequencyInfoList.add(new SuggestedWords.SuggestedWordInfo("--", false)); + for (int i = mScores.length; i < mSuggestions.size(); ++i) { + scoreInfoList.add(new SuggestedWords.SuggestedWordInfo("--", false)); } - return new SuggestedWords.Builder().addWords(mSuggestions, frequencyInfoList); + return new SuggestedWords.Builder().addWords(mSuggestions, scoreInfoList); } return new SuggestedWords.Builder().addWords(mSuggestions, null); } @@ -419,52 +446,37 @@ public class Suggest implements Dictionary.WordCallback { return mAutoCorrection.hasAutoCorrection(); } - private static boolean compareCaseInsensitive(final String lowerOriginalWord, - final char[] word, final int offset, final int length) { - final int originalLength = lowerOriginalWord.length(); - if (originalLength == length && Character.isUpperCase(word[offset])) { - for (int i = 0; i < originalLength; i++) { - if (lowerOriginalWord.charAt(i) != Character.toLowerCase(word[offset+i])) { - return false; - } - } - return true; - } - return false; - } - @Override - public boolean addWord(final char[] word, final int offset, final int length, int freq, + public boolean addWord(final char[] word, final int offset, final int length, int score, final int dicTypeId, final Dictionary.DataType dataType) { Dictionary.DataType dataTypeForLog = dataType; - ArrayList<CharSequence> suggestions; - int[] priorities; - int prefMaxSuggestions; + final ArrayList<CharSequence> suggestions; + final int[] sortedScores; + final int prefMaxSuggestions; if(dataType == Dictionary.DataType.BIGRAM) { suggestions = mBigramSuggestions; - priorities = mBigramPriorities; + sortedScores = mBigramScores; prefMaxSuggestions = PREF_MAX_BIGRAMS; } else { suggestions = mSuggestions; - priorities = mPriorities; + sortedScores = mScores; prefMaxSuggestions = mPrefMaxSuggestions; } int pos = 0; // Check if it's the same word, only caps are different - if (compareCaseInsensitive(mLowerOriginalWord, word, offset, length)) { + if (Utils.equalsIgnoreCase(mTypedWord, word, offset, length)) { // TODO: remove this surrounding if clause and move this logic to // getSuggestedWordBuilder. if (suggestions.size() > 0) { - final String currentHighestWordLowerCase = - suggestions.get(0).toString().toLowerCase(); + final String currentHighestWord = suggestions.get(0).toString(); // If the current highest word is also equal to typed word, we need to compare // frequency to determine the insertion position. This does not ensure strictly // correct ordering, but ensures the top score is on top which is enough for // removing duplicates correctly. - if (compareCaseInsensitive(currentHighestWordLowerCase, word, offset, length) - && freq <= priorities[0]) { + if (Utils.equalsIgnoreCase(currentHighestWord, word, offset, length) + && score <= sortedScores[0]) { pos = 1; } } @@ -475,24 +487,24 @@ public class Suggest implements Dictionary.WordCallback { if(bigramSuggestion >= 0) { dataTypeForLog = Dictionary.DataType.BIGRAM; // turn freq from bigram into multiplier specified above - double multiplier = (((double) mBigramPriorities[bigramSuggestion]) + double multiplier = (((double) mBigramScores[bigramSuggestion]) / MAXIMUM_BIGRAM_FREQUENCY) * (BIGRAM_MULTIPLIER_MAX - BIGRAM_MULTIPLIER_MIN) + BIGRAM_MULTIPLIER_MIN; /* Log.d(TAG,"bigram num: " + bigramSuggestion + " wordB: " + mBigramSuggestions.get(bigramSuggestion).toString() - + " currentPriority: " + freq + " bigramPriority: " - + mBigramPriorities[bigramSuggestion] + + " currentScore: " + score + " bigramScore: " + + mBigramScores[bigramSuggestion] + " multiplier: " + multiplier); */ - freq = (int)Math.round((freq * multiplier)); + score = (int)Math.round((score * multiplier)); } } - // Check the last one's priority and bail - if (priorities[prefMaxSuggestions - 1] >= freq) return true; + // Check the last one's score and bail + if (sortedScores[prefMaxSuggestions - 1] >= score) return true; while (pos < prefMaxSuggestions) { - if (priorities[pos] < freq - || (priorities[pos] == freq && length < suggestions.get(pos).length())) { + if (sortedScores[pos] < score + || (sortedScores[pos] == score && length < suggestions.get(pos).length())) { break; } pos++; @@ -502,8 +514,8 @@ public class Suggest implements Dictionary.WordCallback { return true; } - System.arraycopy(priorities, pos, priorities, pos + 1, prefMaxSuggestions - pos - 1); - priorities[pos] = freq; + System.arraycopy(sortedScores, pos, sortedScores, pos + 1, prefMaxSuggestions - pos - 1); + sortedScores[pos] = score; int poolSize = mStringPool.size(); StringBuilder sb = poolSize > 0 ? (StringBuilder) mStringPool.remove(poolSize - 1) : new StringBuilder(getApproxMaxWordLength()); |