Merge remote-tracking branch 'goog/master' into mergescript

Conflicts:
	java/src/com/android/inputmethod/keyboard/Keyboard.java
	java/src/com/android/inputmethod/latin/LatinIME.java
	java/src/com/android/inputmethod/latin/Suggest.java

Change-Id: I73c1cec1f4bca5139372392e93d55081d814fa6c

9 files changed, 77 insertions, 61 deletions

diff --git a/java/src/com/android/inputmethod/keyboard/Keyboard.java b/java/src/com/android/inputmethod/keyboard/Keyboard.java
index 6fc630d05..6d85fea27 100644
--- a/java/src/com/android/inputmethod/keyboard/Keyboard.java
+++ b/java/src/com/android/inputmethod/keyboard/Keyboard.java
@@ -184,19 +184,21 @@ public class Keyboard {
         if (code == CODE_UNSPECIFIED) {
             return null;
         }
-        final Integer keyCode = code;
-        if (mKeyCache.containsKey(keyCode)) {
-            return mKeyCache.get(keyCode);
-        }
+        synchronized (mKeyCache) {
+            final int index = mKeyCache.indexOfKey(code);
+            if (index >= 0) {
+                return mKeyCache.valueAt(index);
+            }
 
-        for (final Key key : mKeys) {
-            if (key.mCode == code) {
-                mKeyCache.put(keyCode, key);
-                return key;
+            for (final Key key : mKeys) {
+                if (key.mCode == code) {
+                    mKeyCache.put(code, key);
+                    return key;
+                }
             }
+            mKeyCache.put(code, null);
+            return null;
         }
-        mKeyCache.put(keyCode, null);
-        return null;
     }
 
     public boolean hasKey(Key aKey) {
diff --git a/java/src/com/android/inputmethod/latin/LatinIME.java b/java/src/com/android/inputmethod/latin/LatinIME.java
index 6e3540485..5cbf40400 100644
--- a/java/src/com/android/inputmethod/latin/LatinIME.java
+++ b/java/src/com/android/inputmethod/latin/LatinIME.java
@@ -208,13 +208,13 @@ public class LatinIME extends InputMethodService implements KeyboardActionListen
             final KeyboardSwitcher switcher = latinIme.mKeyboardSwitcher;
             switch (msg.what) {
             case MSG_UPDATE_SUGGESTIONS:
-                latinIme.updateSuggestions();
+                latinIme.updateSuggestionsOrPredictions(false /* isPredictions */);
                 break;
             case MSG_UPDATE_SHIFT_STATE:
                 switcher.updateShiftState();
                 break;
             case MSG_SET_BIGRAM_PREDICTIONS:
-                latinIme.updateBigramPredictions();
+                latinIme.updateSuggestionsOrPredictions(true /* isPredictions */);
                 break;
             }
         }
@@ -1006,7 +1006,7 @@ public class LatinIME extends InputMethodService implements KeyboardActionListen
     // the composing word, reset the last composed word, tell the inputconnection about it.
     private void resetEntireInputState() {
         resetComposingState(true /* alsoResetLastComposedWord */);
-        updateSuggestions();
+        updateSuggestionsOrPredictions(false /* isPredictions */);
         mConnection.finishComposingText();
     }
 
@@ -1029,7 +1029,7 @@ public class LatinIME extends InputMethodService implements KeyboardActionListen
                     LastComposedWord.COMMIT_TYPE_USER_TYPED_WORD, typedWord.toString(),
                     separatorCode, prevWord);
         }
-        updateSuggestions();
+        updateSuggestionsOrPredictions(false /* isPredictions */);
     }
 
     public int getCurrentAutoCapsState() {
@@ -1697,7 +1697,18 @@ public class LatinIME extends InputMethodService implements KeyboardActionListen
         }
     }
 
-    public void updateSuggestions() {
+    public void updateSuggestionsOrPredictions(final boolean isPredictions) {
+        if (isPredictions) {
+            updateBigramPredictions();
+        } else {
+            updateSuggestions();
+        }
+    }
+
+    private void updateSuggestions() {
+        mHandler.cancelUpdateSuggestions();
+        mHandler.cancelUpdateBigramPredictions();
+
         // Check if we have a suggestion engine attached.
         if ((mSuggest == null || !mCurrentSettings.isSuggestionsRequested(mDisplayOrientation))) {
             if (mWordComposer.isComposingWord()) {
@@ -1711,7 +1722,11 @@ public class LatinIME extends InputMethodService implements KeyboardActionListen
         mHandler.cancelUpdateBigramPredictions();
 
         if (!mWordComposer.isComposingWord()) {
-            setPunctuationSuggestions();
+            // We are never called with an empty word composer, but if because of a bug
+            // we are, what we should do here is just call updateBigramsPredictions. This will
+            // update the predictions if the "predict next word" option is on, or display
+            // punctuation signs if it's off.
+            updateBigramPredictions();
             return;
         }
 
@@ -1775,7 +1790,7 @@ public class LatinIME extends InputMethodService implements KeyboardActionListen
         // Complete any pending suggestions query first
         if (mHandler.hasPendingUpdateSuggestions()) {
             mHandler.cancelUpdateSuggestions();
-            updateSuggestions();
+            updateSuggestionsOrPredictions(false /* isPredictions */);
         }
         final CharSequence autoCorrection = mWordComposer.getAutoCorrectionOrNull();
         if (autoCorrection != null) {
@@ -1885,7 +1900,7 @@ public class LatinIME extends InputMethodService implements KeyboardActionListen
         if (!showingAddToDictionaryHint) {
             // If we're not showing the "Touch again to save", then show corrections again.
             // In case the cursor position doesn't change, make sure we show the suggestions again.
-            updateBigramPredictions();
+            updateSuggestionsOrPredictions(true /* isPredictions */);
             // Updating the predictions right away may be slow and feel unresponsive on slower
             // terminals. On the other hand if we just postUpdateBigramPredictions() it will
             // take a noticeable delay to update them which may feel uneasy.
diff --git a/java/src/com/android/inputmethod/latin/Suggest.java b/java/src/com/android/inputmethod/latin/Suggest.java
index 892245402..5aecc1371 100644
--- a/java/src/com/android/inputmethod/latin/Suggest.java
+++ b/java/src/com/android/inputmethod/latin/Suggest.java
@@ -272,31 +272,9 @@ public class Suggest implements Dictionary.WordCallback {
             mBigramSuggestions = new ArrayList<SuggestedWordInfo>(PREF_MAX_BIGRAMS);
 
             if (!TextUtils.isEmpty(prevWordForBigram)) {
-                getAllBigrams(prevWordForBigram, wordComposer);
-                if (TextUtils.isEmpty(consideredWord)) {
-                    // Nothing entered: return all bigrams for the previous word
-                    int insertCount = Math.min(mBigramSuggestions.size(), MAX_SUGGESTIONS);
-                    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 = consideredWord.charAt(0);
-                    // TODO: Must pay attention to locale when changing case.
-                    // TODO: Use codepoint instead of char
-                    final char currentCharUpper = Character.toUpperCase(currentChar);
-                    int count = 0;
-                    final int bigramSuggestionSize = mBigramSuggestions.size();
-                    for (int i = 0; i < bigramSuggestionSize; i++) {
-                        final SuggestedWordInfo bigramSuggestion = mBigramSuggestions.get(i);
-                        final char bigramSuggestionFirstChar =
-                                (char)bigramSuggestion.codePointAt(0);
-                        if (bigramSuggestionFirstChar == currentChar
-                                || bigramSuggestionFirstChar == currentCharUpper) {
-                            addBigramToSuggestions(bigramSuggestion);
-                            if (++count > MAX_SUGGESTIONS) break;
-                        }
-                    }
+                for (final String key : mDictionaries.keySet()) {
+                    final Dictionary dictionary = mDictionaries.get(key);
+                    suggestionsSet.addAll(dictionary.getBigrams(wordComposer, prevWordForBigram));
                 }
             }
 
diff --git a/java/src/com/android/inputmethod/latin/UserHistoryDictionaryBigramList.java b/java/src/com/android/inputmethod/latin/UserHistoryDictionaryBigramList.java
index 28847745e..610652ac1 100644
--- a/java/src/com/android/inputmethod/latin/UserHistoryDictionaryBigramList.java
+++ b/java/src/com/android/inputmethod/latin/UserHistoryDictionaryBigramList.java
@@ -98,11 +98,11 @@ public class UserHistoryDictionaryBigramList {
     }
 
     public HashMap<String, Byte> getBigrams(String word1) {
-        if (!mBigramMap.containsKey(word1)) {
-            return EMPTY_BIGRAM_MAP;
-        } else {
-            return mBigramMap.get(word1);
-        }
+        if (mBigramMap.containsKey(word1)) return mBigramMap.get(word1);
+        // TODO: lower case according to locale
+        final String lowerWord1 = word1.toLowerCase();
+        if (mBigramMap.containsKey(lowerWord1)) return mBigramMap.get(lowerWord1);
+        return EMPTY_BIGRAM_MAP;
     }
 
     public boolean removeBigram(String word1, String word2) {
diff --git a/native/jni/src/bigram_dictionary.cpp b/native/jni/src/bigram_dictionary.cpp
index 144336981..3bfbfad25 100644
--- a/native/jni/src/bigram_dictionary.cpp
+++ b/native/jni/src/bigram_dictionary.cpp
@@ -105,8 +105,15 @@ int BigramDictionary::getBigrams(const int32_t *prevWord, int prevWordLength, in
     // TODO: have "in" arguments before "out" ones, and make out args explicit in the name
 
     const uint8_t* const root = DICT;
-    int pos = getBigramListPositionForWord(prevWord, prevWordLength);
+    int pos = getBigramListPositionForWord(prevWord, prevWordLength,
+            false /* forceLowerCaseSearch */);
     // getBigramListPositionForWord returns 0 if this word isn't in the dictionary or has no bigrams
+    if (0 == pos) {
+        // If no bigrams for this exact word, search again in lower case.
+        pos = getBigramListPositionForWord(prevWord, prevWordLength,
+                true /* forceLowerCaseSearch */);
+    }
+    // If still no bigrams, we really don't have them!
     if (0 == pos) return 0;
     int bigramFlags;
     int bigramCount = 0;
@@ -141,10 +148,11 @@ int BigramDictionary::getBigrams(const int32_t *prevWord, int prevWordLength, in
 // Returns a pointer to the start of the bigram list.
 // If the word is not found or has no bigrams, this function returns 0.
 int BigramDictionary::getBigramListPositionForWord(const int32_t *prevWord,
-        const int prevWordLength) const {
+        const int prevWordLength, const bool forceLowerCaseSearch) const {
     if (0 >= prevWordLength) return 0;
     const uint8_t* const root = DICT;
-    int pos = BinaryFormat::getTerminalPosition(root, prevWord, prevWordLength);
+    int pos = BinaryFormat::getTerminalPosition(root, prevWord, prevWordLength,
+            forceLowerCaseSearch);
 
     if (NOT_VALID_WORD == pos) return 0;
     const int flags = BinaryFormat::getFlagsAndForwardPointer(root, &pos);
@@ -164,7 +172,13 @@ void BigramDictionary::fillBigramAddressToFrequencyMapAndFilter(const int32_t *p
         const int prevWordLength, std::map<int, int> *map, uint8_t *filter) const {
     memset(filter, 0, BIGRAM_FILTER_BYTE_SIZE);
     const uint8_t* const root = DICT;
-    int pos = getBigramListPositionForWord(prevWord, prevWordLength);
+    int pos = getBigramListPositionForWord(prevWord, prevWordLength,
+            false /* forceLowerCaseSearch */);
+    if (0 == pos) {
+        // If no bigrams for this exact string, search again in lower case.
+        pos = getBigramListPositionForWord(prevWord, prevWordLength,
+                true /* forceLowerCaseSearch */);
+    }
     if (0 == pos) return;
 
     int bigramFlags;
@@ -197,10 +211,11 @@ bool BigramDictionary::checkFirstCharacter(unsigned short *word, int *inputCodes
 bool BigramDictionary::isValidBigram(const int32_t *word1, int length1, const int32_t *word2,
         int length2) const {
     const uint8_t* const root = DICT;
-    int pos = getBigramListPositionForWord(word1, length1);
+    int pos = getBigramListPositionForWord(word1, length1, false /* forceLowerCaseSearch */);
     // getBigramListPositionForWord returns 0 if this word isn't in the dictionary or has no bigrams
     if (0 == pos) return false;
-    int nextWordPos = BinaryFormat::getTerminalPosition(root, word2, length2);
+    int nextWordPos = BinaryFormat::getTerminalPosition(root, word2, length2,
+            false /* forceLowerCaseSearch */);
     if (NOT_VALID_WORD == nextWordPos) return false;
     int bigramFlags;
     do {
diff --git a/native/jni/src/bigram_dictionary.h b/native/jni/src/bigram_dictionary.h
index 1ff1b2ec6..5372276cd 100644
--- a/native/jni/src/bigram_dictionary.h
+++ b/native/jni/src/bigram_dictionary.h
@@ -30,7 +30,8 @@ class BigramDictionary {
     BigramDictionary(const unsigned char *dict, int maxWordLength);
     int getBigrams(const int32_t *word, int length, int *inputCodes, int codesSize,
             unsigned short *outWords, int *frequencies, int maxWordLength, int maxBigrams) const;
-    int getBigramListPositionForWord(const int32_t *prevWord, const int prevWordLength) const;
+    int getBigramListPositionForWord(const int32_t *prevWord, const int prevWordLength,
+            const bool forceLowerCaseSearch) const;
     void fillBigramAddressToFrequencyMapAndFilter(const int32_t *prevWord, const int prevWordLength,
             std::map<int, int> *map, uint8_t *filter) const;
     bool isValidBigram(const int32_t *word1, int length1, const int32_t *word2, int length2) const;
diff --git a/native/jni/src/binary_format.h b/native/jni/src/binary_format.h
index 214ecfa8d..474c854fe 100644
--- a/native/jni/src/binary_format.h
+++ b/native/jni/src/binary_format.h
@@ -19,6 +19,7 @@
 
 #include <limits>
 #include "bloom_filter.h"
+#include "char_utils.h"
 #include "unigram_dictionary.h"
 
 namespace latinime {
@@ -65,7 +66,7 @@ class BinaryFormat {
     static int getAttributeAddressAndForwardPointer(const uint8_t* const dict, const uint8_t flags,
             int *pos);
     static int getTerminalPosition(const uint8_t* const root, const int32_t* const inWord,
-            const int length);
+            const int length, const bool forceLowerCaseSearch);
     static int getWordAtAddress(const uint8_t* const root, const int address, const int maxDepth,
             uint16_t* outWord, int* outUnigramFrequency);
     static int computeFrequencyForBigram(const int unigramFreq, const int bigramFreq);
@@ -309,7 +310,7 @@ inline int BinaryFormat::getAttributeAddressAndForwardPointer(const uint8_t* con
 // This function gets the byte position of the last chargroup of the exact matching word in the
 // dictionary. If no match is found, it returns NOT_VALID_WORD.
 inline int BinaryFormat::getTerminalPosition(const uint8_t* const root,
-        const int32_t* const inWord, const int length) {
+        const int32_t* const inWord, const int length, const bool forceLowerCaseSearch) {
     int pos = 0;
     int wordPos = 0;
 
@@ -318,7 +319,7 @@ inline int BinaryFormat::getTerminalPosition(const uint8_t* const root,
         // there was no match (or we would have found it).
         if (wordPos > length) return NOT_VALID_WORD;
         int charGroupCount = BinaryFormat::getGroupCountAndForwardPointer(root, &pos);
-        const int32_t wChar = inWord[wordPos];
+        const int32_t wChar = forceLowerCaseSearch ? toLowerCase(inWord[wordPos]) : inWord[wordPos];
         while (true) {
             // If there are no more character groups in this node, it means we could not
             // find a matching character for this depth, therefore there is no match.
diff --git a/native/jni/src/char_utils.h b/native/jni/src/char_utils.h
index 607dc5195..21dca9a0a 100644
--- a/native/jni/src/char_utils.h
+++ b/native/jni/src/char_utils.h
@@ -50,8 +50,7 @@ inline static unsigned short toBaseChar(unsigned short c) {
     return c;
 }
 
-inline static unsigned short toBaseLowerCase(unsigned short c) {
-    c = toBaseChar(c);
+inline static unsigned short toLowerCase(const unsigned short c) {
     if (isAsciiUpper(c)) {
         return toAsciiLower(c);
     } else if (isAscii(c)) {
@@ -60,6 +59,10 @@ inline static unsigned short toBaseLowerCase(unsigned short c) {
     return latin_tolower(c);
 }
 
+inline static unsigned short toBaseLowerCase(const unsigned short c) {
+    return toLowerCase(toBaseChar(c));
+}
+
 } // namespace latinime
 
 #endif // LATINIME_CHAR_UTILS_H
diff --git a/native/jni/src/unigram_dictionary.cpp b/native/jni/src/unigram_dictionary.cpp
index 3417d2ba7..22f1657ef 100644
--- a/native/jni/src/unigram_dictionary.cpp
+++ b/native/jni/src/unigram_dictionary.cpp
@@ -817,7 +817,8 @@ int UnigramDictionary::getMostFrequentWordLikeInner(const uint16_t * const inWor
 
 int UnigramDictionary::getFrequency(const int32_t* const inWord, const int length) const {
     const uint8_t* const root = DICT_ROOT;
-    int pos = BinaryFormat::getTerminalPosition(root, inWord, length);
+    int pos = BinaryFormat::getTerminalPosition(root, inWord, length,
+            false /* forceLowerCaseSearch */);
     if (NOT_VALID_WORD == pos) {
         return NOT_A_PROBABILITY;
     }