Merge "Read shortcuts as strings in the dictionary."

native/jni/src/bigram_dictionary.cpp

@@ -123,6 +123,7 @@ int BigramDictionary::getBigrams(unsigned short *prevWord, int prevWordLength, i
     }
     pos = BinaryFormat::skipChildrenPosition(flags, pos);
     pos = BinaryFormat::skipFrequency(flags, pos);
+    pos = BinaryFormat::skipShortcuts(root, flags, pos);
     int bigramFlags;
     int bigramCount = 0;
     do {

native/jni/src/binary_format.h

@@ -40,6 +40,9 @@ class BinaryFormat {
     // implementations. On this occasion, we made the magic number 32 bits long.
     const static uint32_t FORMAT_VERSION_2_MAGIC_NUMBER = 0x9BC13AFE;
 
+    const static int CHARACTER_ARRAY_TERMINATOR_SIZE = 1;
+    const static int SHORTCUT_LIST_SIZE_SIZE = 2;
+
     static int detectFormat(const uint8_t* const dict);
     static unsigned int getHeaderSize(const uint8_t* const dict);
     static int getGroupCountAndForwardPointer(const uint8_t* const dict, int* pos);
@@ -47,9 +50,10 @@ class BinaryFormat {
     static int32_t getCharCodeAndForwardPointer(const uint8_t* const dict, int* pos);
     static int readFrequencyWithoutMovingPointer(const uint8_t* const dict, const int pos);
     static int skipOtherCharacters(const uint8_t* const dict, const int pos);
-    static int skipAttributes(const uint8_t* const dict, const int pos);
     static int skipChildrenPosition(const uint8_t flags, const int pos);
     static int skipFrequency(const uint8_t flags, const int pos);
+    static int skipShortcuts(const uint8_t* const dict, const uint8_t flags, const int pos);
+    static int skipBigrams(const uint8_t* const dict, const uint8_t flags, const int pos);
     static int skipAllAttributes(const uint8_t* const dict, const uint8_t flags, const int pos);
     static int skipChildrenPosAndAttributes(const uint8_t* const dict, const uint8_t flags,
             const int pos);
@@ -157,12 +161,12 @@ static inline int attributeAddressSize(const uint8_t flags) {
     */
 }
 
-inline int BinaryFormat::skipAttributes(const uint8_t* const dict, const int pos) {
+static inline int skipExistingBigrams(const uint8_t* const dict, const int pos) {
     int currentPos = pos;
-    uint8_t flags = getFlagsAndForwardPointer(dict, &currentPos);
+    uint8_t flags = BinaryFormat::getFlagsAndForwardPointer(dict, &currentPos);
     while (flags & UnigramDictionary::FLAG_ATTRIBUTE_HAS_NEXT) {
         currentPos += attributeAddressSize(flags);
-        flags = getFlagsAndForwardPointer(dict, &currentPos);
+        flags = BinaryFormat::getFlagsAndForwardPointer(dict, &currentPos);
     }
     currentPos += attributeAddressSize(flags);
     return currentPos;
@@ -174,6 +178,10 @@ static inline int childrenAddressSize(const uint8_t flags) {
     /* See the note in attributeAddressSize. The same applies here */
 }
 
+static inline int shortcutByteSize(const uint8_t* const dict, const int pos) {
+    return ((int)(dict[pos] << 8)) + (dict[pos + 1]);
+}
+
 inline int BinaryFormat::skipChildrenPosition(const uint8_t flags, const int pos) {
     return pos + childrenAddressSize(flags);
 }
@@ -182,16 +190,30 @@ inline int BinaryFormat::skipFrequency(const uint8_t flags, const int pos) {
     return UnigramDictionary::FLAG_IS_TERMINAL & flags ? pos + 1 : pos;
 }
 
+inline int BinaryFormat::skipShortcuts(const uint8_t* const dict, const uint8_t flags,
+        const int pos) {
+    if (UnigramDictionary::FLAG_HAS_SHORTCUT_TARGETS & flags) {
+        return pos + shortcutByteSize(dict, pos);
+    } else {
+        return pos;
+    }
+}
+
+inline int BinaryFormat::skipBigrams(const uint8_t* const dict, const uint8_t flags,
+        const int pos) {
+    if (UnigramDictionary::FLAG_HAS_BIGRAMS & flags) {
+        return skipExistingBigrams(dict, pos);
+    } else {
+        return pos;
+    }
+}
+
 inline int BinaryFormat::skipAllAttributes(const uint8_t* const dict, const uint8_t flags,
         const int pos) {
     // This function skips all attributes: shortcuts and bigrams.
     int newPos = pos;
-    if (UnigramDictionary::FLAG_HAS_SHORTCUT_TARGETS & flags) {
+    newPos = skipShortcuts(dict, flags, newPos);
-        newPos = skipAttributes(dict, newPos);
+    newPos = skipBigrams(dict, flags, newPos);
-    }
-    if (UnigramDictionary::FLAG_HAS_BIGRAMS & flags) {
-        newPos = skipAttributes(dict, newPos);
-    }
     return newPos;
 }
 

native/jni/src/terminal_attributes.h

@@ -45,13 +45,19 @@ class TerminalAttributes {
 
         // Gets the shortcut target itself as a uint16_t string. For parameters and return value
         // see BinaryFormat::getWordAtAddress.
+        // TODO: make the output an uint32_t* to handle the whole unicode range.
         inline int getNextShortcutTarget(const int maxDepth, uint16_t* outWord) {
             const int shortcutFlags = BinaryFormat::getFlagsAndForwardPointer(mDict, &mPos);
             mHasNextShortcutTarget =
                     0 != (shortcutFlags & UnigramDictionary::FLAG_ATTRIBUTE_HAS_NEXT);
-            int shortcutAddress =
+            unsigned int i;
-                    BinaryFormat::getAttributeAddressAndForwardPointer(mDict, shortcutFlags, &mPos);
+            for (i = 0; i < MAX_WORD_LENGTH_INTERNAL; ++i) {
-            return BinaryFormat::getWordAtAddress(mDict, shortcutAddress, maxDepth, outWord);
+                const int charCode = BinaryFormat::getCharCodeAndForwardPointer(mDict, &mPos);
+                if (NOT_A_CHARACTER == charCode) break;
+                outWord[i] = (uint16_t)charCode;
+            }
+            mPos += BinaryFormat::CHARACTER_ARRAY_TERMINATOR_SIZE;
+            return i;
         }
     };
 
@@ -65,12 +71,10 @@ class TerminalAttributes {
             mDict(dict), mFlags(flags), mStartPos(pos) {
     }
 
-    inline bool isShortcutOnly() const {
-        return 0 != (mFlags & UnigramDictionary::FLAG_IS_SHORTCUT_ONLY);
-    }
-
     inline ShortcutIterator getShortcutIterator() const {
-        return ShortcutIterator(mDict, mStartPos, mFlags);
+        // The size of the shortcuts is stored here so that the whole shortcut chunk can be
+        // skipped quickly, so we ignore it.
+        return ShortcutIterator(mDict, mStartPos + BinaryFormat::SHORTCUT_LIST_SIZE_SIZE, mFlags);
     }
 };
 } // namespace latinime

native/jni/src/unigram_dictionary.cpp

@@ -366,10 +366,9 @@ inline void UnigramDictionary::onTerminal(const int freq,
         WordsPriorityQueue *masterQueue = queuePool->getMasterQueue();
         const int finalFreq = correction->getFinalFreq(freq, &wordPointer, &wordLength);
         if (finalFreq != NOT_A_FREQUENCY) {
-            if (!terminalAttributes.isShortcutOnly()) {
+            addWord(wordPointer, wordLength, finalFreq, masterQueue);
-                addWord(wordPointer, wordLength, finalFreq, masterQueue);
-            }
 
+            const int shortcutFreq = finalFreq > 0 ? finalFreq - 1 : 0;
             // Please note that the shortcut candidates will be added to the master queue only.
             TerminalAttributes::ShortcutIterator iterator =
                     terminalAttributes.getShortcutIterator();
@@ -379,11 +378,12 @@ inline void UnigramDictionary::onTerminal(const int freq,
                 // We need to either modulate the frequency of each shortcut according
                 // to its own shortcut frequency or to make the queue
                 // so that the insert order is protected inside the queue for words
-                // with the same score.
+                // with the same score. For the moment we use -1 to make sure the shortcut will
+                // never be in front of the word.
                 uint16_t shortcutTarget[MAX_WORD_LENGTH_INTERNAL];
                 const int shortcutTargetStringLength = iterator.getNextShortcutTarget(
                         MAX_WORD_LENGTH_INTERNAL, shortcutTarget);
-                addWord(shortcutTarget, shortcutTargetStringLength, finalFreq, masterQueue);
+                addWord(shortcutTarget, shortcutTargetStringLength, shortcutFreq, masterQueue);
             }
         }
     }

native/jni/src/unigram_dictionary.h

@@ -49,10 +49,6 @@ class UnigramDictionary {
     static const int FLAG_HAS_SHORTCUT_TARGETS = 0x08;
     // Flag for bigram presence
     static const int FLAG_HAS_BIGRAMS = 0x04;
-    // Flag for shortcut-only words. Some words are shortcut-only, which means they match when
-    // the user types them but they don't pop in the suggestion strip, only the words they are
-    // shortcuts for do.
-    static const int FLAG_IS_SHORTCUT_ONLY = 0x02;
 
     // Attribute (bigram/shortcut) related flags:
     // Flag for presence of more attributes