private static ArrayList<SuggestedWordInfo> getSuggestionsInfoListWithDebugInfo(
final String typedWord, final ArrayList<SuggestedWordInfo> suggestions) {
final SuggestedWordInfo typedWordInfo = suggestions.get(0);
typedWordInfo.setDebugString("+");
final int suggestionsSize = suggestions.size();
final ArrayList<SuggestedWordInfo> suggestionsList =
CollectionUtils.newArrayList(suggestionsSize);
suggestionsList.add(typedWordInfo);
// Note: i here is the index in mScores[], but the index in mSuggestions is one more
// than i because we added the typed word to mSuggestions without touching mScores.
for (int i = 0; i < suggestionsSize - 1; ++i) {
final SuggestedWordInfo cur = suggestions.get(i + 1);
final float normalizedScore =
BinaryDictionary.calcNormalizedScore(typedWord, cur.toString(), cur.mScore);
final String scoreInfoString;
if (normalizedScore > 0) {
scoreInfoString = String.format(Locale.ROOT, "%d (%4.2f)", cur.mScore, normalizedScore);
} else {
scoreInfoString = Integer.toString(cur.mScore);
}
cur.setDebugString(scoreInfoString);
suggestionsList.add(cur);
}
return suggestionsList;
}
private ArrayList<WeightedString> readShortcuts(final int terminalId) {
if (mShortcutAddressTable.get(0, terminalId) == SparseTable.NOT_EXIST) return null;
final ArrayList<WeightedString> ret = CollectionUtils.newArrayList();
final int posOfShortcuts =
mShortcutAddressTable.get(FormatSpec.SHORTCUT_CONTENT_INDEX, terminalId);
mShortcutBuffer.position(posOfShortcuts);
while (true) {
final int flags = mShortcutBuffer.readUnsignedByte();
final String word = CharEncoding.readString(mShortcutBuffer);
ret.add(new WeightedString(word, flags & FormatSpec.FLAG_BIGRAM_SHORTCUT_ATTR_FREQUENCY));
if (0 == (flags & FormatSpec.FLAG_BIGRAM_SHORTCUT_ATTR_HAS_NEXT)) break;
}
return ret;
}
// Retrieves suggestions for the batch input
// and calls the callback function with the suggestions.
private void getSuggestedWordsForBatchInput(
final WordComposer wordComposer,
final String prevWordForBigram,
final ProximityInfo proximityInfo,
final boolean blockOffensiveWords,
final int[] additionalFeaturesOptions,
final int sessionId,
final int sequenceNumber,
final OnGetSuggestedWordsCallback callback) {
final BoundedTreeSet suggestionsSet =
new BoundedTreeSet(sSuggestedWordInfoComparator, MAX_SUGGESTIONS);
// At second character typed, search the unigrams (scores being affected by bigrams)
for (final String key : mDictionaries.keySet()) {
final Dictionary dictionary = mDictionaries.get(key);
suggestionsSet.addAll(
dictionary.getSuggestionsWithSessionId(
wordComposer,
prevWordForBigram,
proximityInfo,
blockOffensiveWords,
additionalFeaturesOptions,
sessionId));
}
for (SuggestedWordInfo wordInfo : suggestionsSet) {
LatinImeLogger.onAddSuggestedWord(wordInfo.mWord, wordInfo.mSourceDict.mDictType);
}
final ArrayList<SuggestedWordInfo> suggestionsContainer =
CollectionUtils.newArrayList(suggestionsSet);
final int suggestionsCount = suggestionsContainer.size();
final boolean isFirstCharCapitalized = wordComposer.wasShiftedNoLock();
final boolean isAllUpperCase = wordComposer.isAllUpperCase();
if (isFirstCharCapitalized || isAllUpperCase) {
for (int i = 0; i < suggestionsCount; ++i) {
final SuggestedWordInfo wordInfo = suggestionsContainer.get(i);
final SuggestedWordInfo transformedWordInfo =
getTransformedSuggestedWordInfo(
wordInfo,
mLocale,
isAllUpperCase,
isFirstCharCapitalized,
0 /* trailingSingleQuotesCount */);
suggestionsContainer.set(i, transformedWordInfo);
}
}
if (suggestionsContainer.size() > 1
&& TextUtils.equals(
suggestionsContainer.get(0).mWord, wordComposer.getRejectedBatchModeSuggestion())) {
final SuggestedWordInfo rejected = suggestionsContainer.remove(0);
suggestionsContainer.add(1, rejected);
}
SuggestedWordInfo.removeDups(suggestionsContainer);
// For some reason some suggestions with MIN_VALUE are making their way here.
// TODO: Find a more robust way to detect distractors.
for (int i = suggestionsContainer.size() - 1; i >= 0; --i) {
if (suggestionsContainer.get(i).mScore < SUPPRESS_SUGGEST_THRESHOLD) {
suggestionsContainer.remove(i);
}
}
// In the batch input mode, the most relevant suggested word should act as a "typed word"
// (typedWordValid=true), not as an "auto correct word" (willAutoCorrect=false).
callback.onGetSuggestedWords(
new SuggestedWords(
suggestionsContainer,
true /* typedWordValid */,
false /* willAutoCorrect */,
false /* isPunctuationSuggestions */,
false /* isObsoleteSuggestions */,
false /* isPrediction */,
sequenceNumber));
}
// Retrieves suggestions for the typing input
// and calls the callback function with the suggestions.
private void getSuggestedWordsForTypingInput(
final WordComposer wordComposer,
final String prevWordForBigram,
final ProximityInfo proximityInfo,
final boolean blockOffensiveWords,
final boolean isCorrectionEnabled,
final int[] additionalFeaturesOptions,
final int sequenceNumber,
final OnGetSuggestedWordsCallback callback) {
final int trailingSingleQuotesCount = wordComposer.trailingSingleQuotesCount();
final BoundedTreeSet suggestionsSet =
new BoundedTreeSet(sSuggestedWordInfoComparator, MAX_SUGGESTIONS);
final String typedWord = wordComposer.getTypedWord();
final String consideredWord =
trailingSingleQuotesCount > 0
? typedWord.substring(0, typedWord.length() - trailingSingleQuotesCount)
: typedWord;
LatinImeLogger.onAddSuggestedWord(typedWord, Dictionary.TYPE_USER_TYPED);
final WordComposer wordComposerForLookup;
if (trailingSingleQuotesCount > 0) {
wordComposerForLookup = new WordComposer(wordComposer);
for (int i = trailingSingleQuotesCount - 1; i >= 0; --i) {
wordComposerForLookup.deleteLast();
}
} else {
wordComposerForLookup = wordComposer;
}
for (final String key : mDictionaries.keySet()) {
final Dictionary dictionary = mDictionaries.get(key);
suggestionsSet.addAll(
dictionary.getSuggestions(
wordComposerForLookup,
prevWordForBigram,
proximityInfo,
blockOffensiveWords,
additionalFeaturesOptions));
}
final String whitelistedWord;
if (suggestionsSet.isEmpty()) {
whitelistedWord = null;
} else if (SuggestedWordInfo.KIND_WHITELIST != suggestionsSet.first().mKind) {
whitelistedWord = null;
} else {
whitelistedWord = suggestionsSet.first().mWord;
}
// The word can be auto-corrected if it has a whitelist entry that is not itself,
// or if it's a 2+ characters non-word (i.e. it's not in the dictionary).
final boolean allowsToBeAutoCorrected =
(null != whitelistedWord && !whitelistedWord.equals(consideredWord))
|| (consideredWord.length() > 1
&& !AutoCorrectionUtils.isValidWord(
this, consideredWord, wordComposer.isFirstCharCapitalized()));
final boolean hasAutoCorrection;
// TODO: using isCorrectionEnabled here is not very good. It's probably useless, because
// any attempt to do auto-correction is already shielded with a test for this flag; at the
// same time, it feels wrong that the SuggestedWord object includes information about
// the current settings. It may also be useful to know, when the setting is off, whether
// the word *would* have been auto-corrected.
if (!isCorrectionEnabled
|| !allowsToBeAutoCorrected
|| !wordComposer.isComposingWord()
|| suggestionsSet.isEmpty()
|| wordComposer.hasDigits()
|| wordComposer.isMostlyCaps()
|| wordComposer.isResumed()
|| !hasMainDictionary()
|| SuggestedWordInfo.KIND_SHORTCUT == suggestionsSet.first().mKind) {
// If we don't have a main dictionary, we never want to auto-correct. The reason for
// this is, the user may have a contact whose name happens to match a valid word in
// their language, and it will unexpectedly auto-correct. For example, if the user
// types in English with no dictionary and has a "Will" in their contact list, "will"
// would always auto-correct to "Will" which is unwanted. Hence, no main dict => no
// auto-correct.
// Also, shortcuts should never auto-correct unless they are whitelist entries.
// TODO: we may want to have shortcut-only entries auto-correct in the future.
hasAutoCorrection = false;
} else {
hasAutoCorrection =
AutoCorrectionUtils.suggestionExceedsAutoCorrectionThreshold(
suggestionsSet.first(), consideredWord, mAutoCorrectionThreshold);
}
final ArrayList<SuggestedWordInfo> suggestionsContainer =
CollectionUtils.newArrayList(suggestionsSet);
final int suggestionsCount = suggestionsContainer.size();
final boolean isFirstCharCapitalized = wordComposer.isFirstCharCapitalized();
final boolean isAllUpperCase = wordComposer.isAllUpperCase();
if (isFirstCharCapitalized || isAllUpperCase || 0 != trailingSingleQuotesCount) {
for (int i = 0; i < suggestionsCount; ++i) {
final SuggestedWordInfo wordInfo = suggestionsContainer.get(i);
final SuggestedWordInfo transformedWordInfo =
getTransformedSuggestedWordInfo(
wordInfo,
mLocale,
isAllUpperCase,
isFirstCharCapitalized,
trailingSingleQuotesCount);
suggestionsContainer.set(i, transformedWordInfo);
}
}
for (int i = 0; i < suggestionsCount; ++i) {
final SuggestedWordInfo wordInfo = suggestionsContainer.get(i);
LatinImeLogger.onAddSuggestedWord(wordInfo.mWord.toString(), wordInfo.mSourceDict.mDictType);
}
if (!TextUtils.isEmpty(typedWord)) {
suggestionsContainer.add(
0,
new SuggestedWordInfo(
typedWord,
SuggestedWordInfo.MAX_SCORE,
SuggestedWordInfo.KIND_TYPED,
Dictionary.DICTIONARY_USER_TYPED,
SuggestedWordInfo.NOT_AN_INDEX /* indexOfTouchPointOfSecondWord */,
SuggestedWordInfo.NOT_A_CONFIDENCE /* autoCommitFirstWordConfidence */));
}
SuggestedWordInfo.removeDups(suggestionsContainer);
final ArrayList<SuggestedWordInfo> suggestionsList;
if (DBG && !suggestionsContainer.isEmpty()) {
suggestionsList = getSuggestionsInfoListWithDebugInfo(typedWord, suggestionsContainer);
} else {
suggestionsList = suggestionsContainer;
}
callback.onGetSuggestedWords(
new SuggestedWords(
suggestionsList,
// TODO: this first argument is lying. If this is a whitelisted word which is an
// actual word, it says typedWordValid = false, which looks wrong. We should either
// rename the attribute or change the value.
!allowsToBeAutoCorrected /* typedWordValid */,
hasAutoCorrection, /* willAutoCorrect */
false /* isPunctuationSuggestions */,
false /* isObsoleteSuggestions */,
!wordComposer.isComposingWord() /* isPrediction */,
sequenceNumber));
}