diff options
Diffstat (limited to 'native/jni/src')
59 files changed, 770 insertions, 356 deletions
diff --git a/native/jni/src/defines.h b/native/jni/src/defines.h index 742e388e4..fbcd612b7 100644 --- a/native/jni/src/defines.h +++ b/native/jni/src/defines.h @@ -392,24 +392,4 @@ typedef enum { // Create new word with space substitution CT_NEW_WORD_SPACE_SUBSTITUTION, } CorrectionType; - -// ErrorType is mainly decided by CorrectionType but it is also depending on if -// the correction has really been performed or not. -typedef enum { - // Substitution, omission and transposition - ET_EDIT_CORRECTION, - // Proximity error - ET_PROXIMITY_CORRECTION, - // Completion - ET_COMPLETION, - // New word - // TODO: Remove. - // A new word error should be an edit correction error or a proximity correction error. - ET_NEW_WORD, - // Treat error as an intentional omission when the CorrectionType is omission and the node can - // be intentional omission. - ET_INTENTIONAL_OMISSION, - // Not treated as an error. Tracked for checking exact match - ET_NOT_AN_ERROR -} ErrorType; #endif // LATINIME_DEFINES_H diff --git a/native/jni/src/suggest/core/dicnode/dic_node.h b/native/jni/src/suggest/core/dicnode/dic_node.h index 49cfdecac..0b2b4a9e8 100644 --- a/native/jni/src/suggest/core/dicnode/dic_node.h +++ b/native/jni/src/suggest/core/dicnode/dic_node.h @@ -99,7 +99,7 @@ class DicNode { virtual ~DicNode() {} // Init for copy - void initByCopy(const DicNode *dicNode) { + void initByCopy(const DicNode *const dicNode) { mIsUsed = true; mIsCachedForNextSuggestion = dicNode->mIsCachedForNextSuggestion; mDicNodeProperties.init(&dicNode->mDicNodeProperties); @@ -107,25 +107,25 @@ class DicNode { PROF_NODE_COPY(&dicNode->mProfiler, mProfiler); } - // Init for root with prevWordNodePos which is used for bigram - void initAsRoot(const int rootGroupPos, const int prevWordNodePos) { + // Init for root with prevWordPtNodePos which is used for bigram + void initAsRoot(const int rootPtNodeArrayPos, const int prevWordPtNodePos) { mIsUsed = true; mIsCachedForNextSuggestion = false; mDicNodeProperties.init( - NOT_A_DICT_POS /* pos */, rootGroupPos, NOT_A_CODE_POINT /* nodeCodePoint */, + NOT_A_DICT_POS /* pos */, rootPtNodeArrayPos, NOT_A_CODE_POINT /* nodeCodePoint */, NOT_A_PROBABILITY /* probability */, false /* isTerminal */, true /* hasChildren */, false /* isBlacklistedOrNotAWord */, 0 /* depth */, 0 /* terminalDepth */); - mDicNodeState.init(prevWordNodePos); + mDicNodeState.init(prevWordPtNodePos); PROF_NODE_RESET(mProfiler); } // Init for root with previous word - void initAsRootWithPreviousWord(DicNode *dicNode, const int rootGroupPos) { + void initAsRootWithPreviousWord(const DicNode *const dicNode, const int rootPtNodeArrayPos) { mIsUsed = true; mIsCachedForNextSuggestion = dicNode->mIsCachedForNextSuggestion; mDicNodeProperties.init( - NOT_A_DICT_POS /* pos */, rootGroupPos, NOT_A_CODE_POINT /* nodeCodePoint */, + NOT_A_DICT_POS /* pos */, rootPtNodeArrayPos, NOT_A_CODE_POINT /* nodeCodePoint */, NOT_A_PROBABILITY /* probability */, false /* isTerminal */, true /* hasChildren */, false /* isBlacklistedOrNotAWord */, 0 /* depth */, 0 /* terminalDepth */); @@ -138,7 +138,7 @@ class DicNode { mDicNodeState.mDicNodeStatePrevWord.init( dicNode->mDicNodeState.mDicNodeStatePrevWord.getPrevWordCount() + 1, dicNode->mDicNodeProperties.getProbability(), - dicNode->mDicNodeProperties.getPos(), + dicNode->mDicNodeProperties.getPtNodePos(), dicNode->mDicNodeState.mDicNodeStatePrevWord.mPrevWord, dicNode->mDicNodeState.mDicNodeStatePrevWord.getPrevWordLength(), dicNode->getOutputWordBuf(), @@ -148,26 +148,27 @@ class DicNode { PROF_NODE_COPY(&dicNode->mProfiler, mProfiler); } - void initAsPassingChild(DicNode *parentNode) { + void initAsPassingChild(DicNode *parentDicNode) { mIsUsed = true; - mIsCachedForNextSuggestion = parentNode->mIsCachedForNextSuggestion; - const int c = parentNode->getNodeTypedCodePoint(); - mDicNodeProperties.init(&parentNode->mDicNodeProperties, c); - mDicNodeState.init(&parentNode->mDicNodeState); - PROF_NODE_COPY(&parentNode->mProfiler, mProfiler); + mIsCachedForNextSuggestion = parentDicNode->mIsCachedForNextSuggestion; + const int parentCodePoint = parentDicNode->getNodeTypedCodePoint(); + mDicNodeProperties.init(&parentDicNode->mDicNodeProperties, parentCodePoint); + mDicNodeState.init(&parentDicNode->mDicNodeState); + PROF_NODE_COPY(&parentDicNode->mProfiler, mProfiler); } - void initAsChild(const DicNode *const dicNode, const int pos, const int childrenPos, - const int probability, const bool isTerminal, const bool hasChildren, - const bool isBlacklistedOrNotAWord, const uint16_t mergedNodeCodePointCount, - const int *const mergedNodeCodePoints) { + void initAsChild(const DicNode *const dicNode, const int ptNodePos, + const int childrenPtNodeArrayPos, const int probability, const bool isTerminal, + const bool hasChildren, const bool isBlacklistedOrNotAWord, + const uint16_t mergedNodeCodePointCount, const int *const mergedNodeCodePoints) { mIsUsed = true; uint16_t newDepth = static_cast<uint16_t>(dicNode->getNodeCodePointCount() + 1); mIsCachedForNextSuggestion = dicNode->mIsCachedForNextSuggestion; const uint16_t newLeavingDepth = static_cast<uint16_t>( dicNode->mDicNodeProperties.getLeavingDepth() + mergedNodeCodePointCount); - mDicNodeProperties.init(pos, childrenPos, mergedNodeCodePoints[0], probability, - isTerminal, hasChildren, isBlacklistedOrNotAWord, newDepth, newLeavingDepth); + mDicNodeProperties.init(ptNodePos, childrenPtNodeArrayPos, mergedNodeCodePoints[0], + probability, isTerminal, hasChildren, isBlacklistedOrNotAWord, newDepth, + newLeavingDepth); mDicNodeState.init(&dicNode->mDicNodeState, mergedNodeCodePointCount, mergedNodeCodePoints); PROF_NODE_COPY(&dicNode->mProfiler, mProfiler); @@ -234,7 +235,7 @@ class DicNode { } bool isFirstWord() const { - return mDicNodeState.mDicNodeStatePrevWord.getPrevWordNodePos() == NOT_A_DICT_POS; + return mDicNodeState.mDicNodeStatePrevWord.getPrevWordPtNodePos() == NOT_A_DICT_POS; } bool isCompletion(const int inputSize) const { @@ -246,29 +247,30 @@ class DicNode { } // Used to get bigram probability in DicNodeUtils - int getPos() const { - return mDicNodeProperties.getPos(); + int getPtNodePos() const { + return mDicNodeProperties.getPtNodePos(); } // Used to get bigram probability in DicNodeUtils - int getPrevWordPos() const { - return mDicNodeState.mDicNodeStatePrevWord.getPrevWordNodePos(); + int getPrevWordTerminalPtNodePos() const { + return mDicNodeState.mDicNodeStatePrevWord.getPrevWordPtNodePos(); } // Used in DicNodeUtils - int getChildrenPos() const { - return mDicNodeProperties.getChildrenPos(); + int getChildrenPtNodeArrayPos() const { + return mDicNodeProperties.getChildrenPtNodeArrayPos(); } int getProbability() const { return mDicNodeProperties.getProbability(); } - AK_FORCE_INLINE bool isTerminalWordNode() const { - const bool isTerminalNodes = mDicNodeProperties.isTerminal(); - const int currentNodeDepth = getNodeCodePointCount(); - const int terminalNodeDepth = mDicNodeProperties.getLeavingDepth(); - return isTerminalNodes && currentNodeDepth > 0 && currentNodeDepth == terminalNodeDepth; + AK_FORCE_INLINE bool isTerminalDicNode() const { + const bool isTerminalPtNode = mDicNodeProperties.isTerminal(); + const int currentDicNodeDepth = getNodeCodePointCount(); + const int terminalDicNodeDepth = mDicNodeProperties.getLeavingDepth(); + return isTerminalPtNode && currentDicNodeDepth > 0 + && currentDicNodeDepth == terminalDicNodeDepth; } bool shouldBeFilteredBySafetyNetForBigram() const { @@ -374,8 +376,8 @@ class DicNode { } // Used to commit input partially - int getPrevWordNodePos() const { - return mDicNodeState.mDicNodeStatePrevWord.getPrevWordNodePos(); + int getPrevWordPtNodePos() const { + return mDicNodeState.mDicNodeStatePrevWord.getPrevWordPtNodePos(); } AK_FORCE_INLINE const int *getOutputWordBuf() const { @@ -410,7 +412,7 @@ class DicNode { // TODO: Remove once touch path is merged into ProximityInfoState // Note: Returned codepoint may be a digraph codepoint if the node is in a composite glyph. int getNodeCodePoint() const { - const int codePoint = mDicNodeProperties.getNodeCodePoint(); + const int codePoint = mDicNodeProperties.getDicNodeCodePoint(); const DigraphUtils::DigraphCodePointIndex digraphIndex = mDicNodeState.mDicNodeStateScoring.getDigraphIndex(); if (digraphIndex == DigraphUtils::NOT_A_DIGRAPH_INDEX) { @@ -423,8 +425,8 @@ class DicNode { // Utils for cost calculation // //////////////////////////////// AK_FORCE_INLINE bool isSameNodeCodePoint(const DicNode *const dicNode) const { - return mDicNodeProperties.getNodeCodePoint() - == dicNode->mDicNodeProperties.getNodeCodePoint(); + return mDicNodeProperties.getDicNodeCodePoint() + == dicNode->mDicNodeProperties.getDicNodeCodePoint(); } // TODO: remove @@ -574,7 +576,8 @@ class DicNode { // Caveat: Must not be called outside Weighting // This restriction is guaranteed by "friend" AK_FORCE_INLINE void addCost(const float spatialCost, const float languageCost, - const bool doNormalization, const int inputSize, const ErrorType errorType) { + const bool doNormalization, const int inputSize, + const ErrorTypeUtils::ErrorType errorType) { if (DEBUG_GEO_FULL) { LOGI_SHOW_ADD_COST_PROP; } diff --git a/native/jni/src/suggest/core/dicnode/dic_node_utils.cpp b/native/jni/src/suggest/core/dicnode/dic_node_utils.cpp index ec65114c7..5540b6df5 100644 --- a/native/jni/src/suggest/core/dicnode/dic_node_utils.cpp +++ b/native/jni/src/suggest/core/dicnode/dic_node_utils.cpp @@ -22,7 +22,6 @@ #include "suggest/core/dicnode/dic_node_vector.h" #include "suggest/core/dictionary/multi_bigram_map.h" #include "suggest/core/policy/dictionary_structure_with_buffer_policy.h" -#include "utils/char_utils.h" namespace latinime { @@ -32,19 +31,20 @@ namespace latinime { /* static */ void DicNodeUtils::initAsRoot( const DictionaryStructureWithBufferPolicy *const dictionaryStructurePolicy, - const int prevWordNodePos, DicNode *const newRootNode) { - newRootNode->initAsRoot(dictionaryStructurePolicy->getRootPosition(), prevWordNodePos); + const int prevWordPtNodePos, DicNode *const newRootDicNode) { + newRootDicNode->initAsRoot(dictionaryStructurePolicy->getRootPosition(), prevWordPtNodePos); } /*static */ void DicNodeUtils::initAsRootWithPreviousWord( const DictionaryStructureWithBufferPolicy *const dictionaryStructurePolicy, - DicNode *const prevWordLastNode, DicNode *const newRootNode) { - newRootNode->initAsRootWithPreviousWord( - prevWordLastNode, dictionaryStructurePolicy->getRootPosition()); + const DicNode *const prevWordLastDicNode, DicNode *const newRootDicNode) { + newRootDicNode->initAsRootWithPreviousWord( + prevWordLastDicNode, dictionaryStructurePolicy->getRootPosition()); } -/* static */ void DicNodeUtils::initByCopy(DicNode *srcNode, DicNode *destNode) { - destNode->initByCopy(srcNode); +/* static */ void DicNodeUtils::initByCopy(const DicNode *const srcDicNode, + DicNode *const destDicNode) { + destDicNode->initByCopy(srcDicNode); } /////////////////////////////////// @@ -52,14 +52,14 @@ namespace latinime { /////////////////////////////////// /* static */ void DicNodeUtils::getAllChildDicNodes(DicNode *dicNode, const DictionaryStructureWithBufferPolicy *const dictionaryStructurePolicy, - DicNodeVector *childDicNodes) { + DicNodeVector *const childDicNodes) { if (dicNode->isTotalInputSizeExceedingLimit()) { return; } if (!dicNode->isLeavingNode()) { childDicNodes->pushPassingChild(dicNode); } else { - dictionaryStructurePolicy->createAndGetAllChildNodes(dicNode, childDicNodes); + dictionaryStructurePolicy->createAndGetAllChildDicNodes(dicNode, childDicNodes); } } @@ -71,11 +71,11 @@ namespace latinime { */ /* static */ float DicNodeUtils::getBigramNodeImprobability( const DictionaryStructureWithBufferPolicy *const dictionaryStructurePolicy, - const DicNode *const node, MultiBigramMap *multiBigramMap) { - if (node->hasMultipleWords() && !node->isValidMultipleWordSuggestion()) { + const DicNode *const dicNode, MultiBigramMap *const multiBigramMap) { + if (dicNode->hasMultipleWords() && !dicNode->isValidMultipleWordSuggestion()) { return static_cast<float>(MAX_VALUE_FOR_WEIGHTING); } - const int probability = getBigramNodeProbability(dictionaryStructurePolicy, node, + const int probability = getBigramNodeProbability(dictionaryStructurePolicy, dicNode, multiBigramMap); // TODO: This equation to calculate the improbability looks unreasonable. Investigate this. const float cost = static_cast<float>(MAX_PROBABILITY - probability) @@ -85,19 +85,19 @@ namespace latinime { /* static */ int DicNodeUtils::getBigramNodeProbability( const DictionaryStructureWithBufferPolicy *const dictionaryStructurePolicy, - const DicNode *const node, MultiBigramMap *multiBigramMap) { - const int unigramProbability = node->getProbability(); - const int wordPos = node->getPos(); - const int prevWordPos = node->getPrevWordPos(); - if (NOT_A_DICT_POS == wordPos || NOT_A_DICT_POS == prevWordPos) { + const DicNode *const dicNode, MultiBigramMap *const multiBigramMap) { + const int unigramProbability = dicNode->getProbability(); + const int ptNodePos = dicNode->getPtNodePos(); + const int prevWordTerminalPtNodePos = dicNode->getPrevWordTerminalPtNodePos(); + if (NOT_A_DICT_POS == ptNodePos || NOT_A_DICT_POS == prevWordTerminalPtNodePos) { // Note: Normally wordPos comes from the dictionary and should never equal // NOT_A_VALID_WORD_POS. return dictionaryStructurePolicy->getProbability(unigramProbability, NOT_A_PROBABILITY); } if (multiBigramMap) { - return multiBigramMap->getBigramProbability(dictionaryStructurePolicy, prevWordPos, - wordPos, unigramProbability); + return multiBigramMap->getBigramProbability(dictionaryStructurePolicy, + prevWordTerminalPtNodePos, ptNodePos, unigramProbability); } return dictionaryStructurePolicy->getProbability(unigramProbability, NOT_A_PROBABILITY); @@ -109,7 +109,7 @@ namespace latinime { // TODO: Move to char_utils? /* static */ int DicNodeUtils::appendTwoWords(const int *const src0, const int16_t length0, - const int *const src1, const int16_t length1, int *dest) { + const int *const src1, const int16_t length1, int *const dest) { int actualLength0 = 0; for (int i = 0; i < length0; ++i) { if (src0[i] == 0) { diff --git a/native/jni/src/suggest/core/dicnode/dic_node_utils.h b/native/jni/src/suggest/core/dicnode/dic_node_utils.h index 3fb351a61..3f1514a52 100644 --- a/native/jni/src/suggest/core/dicnode/dic_node_utils.h +++ b/native/jni/src/suggest/core/dicnode/dic_node_utils.h @@ -31,20 +31,20 @@ class MultiBigramMap; class DicNodeUtils { public: static int appendTwoWords(const int *src0, const int16_t length0, const int *src1, - const int16_t length1, int *dest); + const int16_t length1, int *const dest); static void initAsRoot( const DictionaryStructureWithBufferPolicy *const dictionaryStructurePolicy, - const int prevWordNodePos, DicNode *newRootNode); + const int prevWordPtNodePos, DicNode *const newRootDicNode); static void initAsRootWithPreviousWord( const DictionaryStructureWithBufferPolicy *const dictionaryStructurePolicy, - DicNode *prevWordLastNode, DicNode *newRootNode); - static void initByCopy(DicNode *srcNode, DicNode *destNode); + const DicNode *const prevWordLastDicNode, DicNode *const newRootDicNode); + static void initByCopy(const DicNode *const srcDicNode, DicNode *const destDicNode); static void getAllChildDicNodes(DicNode *dicNode, const DictionaryStructureWithBufferPolicy *const dictionaryStructurePolicy, DicNodeVector *childDicNodes); static float getBigramNodeImprobability( const DictionaryStructureWithBufferPolicy *const dictionaryStructurePolicy, - const DicNode *const node, MultiBigramMap *const multiBigramMap); + const DicNode *const dicNode, MultiBigramMap *const multiBigramMap); private: DISALLOW_IMPLICIT_CONSTRUCTORS(DicNodeUtils); @@ -53,7 +53,7 @@ class DicNodeUtils { static int getBigramNodeProbability( const DictionaryStructureWithBufferPolicy *const dictionaryStructurePolicy, - const DicNode *const node, MultiBigramMap *multiBigramMap); + const DicNode *const dicNode, MultiBigramMap *const multiBigramMap); }; } // namespace latinime #endif // LATINIME_DIC_NODE_UTILS_H diff --git a/native/jni/src/suggest/core/dicnode/dic_node_vector.h b/native/jni/src/suggest/core/dicnode/dic_node_vector.h index 42addae8d..9364e7751 100644 --- a/native/jni/src/suggest/core/dicnode/dic_node_vector.h +++ b/native/jni/src/suggest/core/dicnode/dic_node_vector.h @@ -62,14 +62,14 @@ class DicNodeVector { mDicNodes.back().initAsPassingChild(dicNode); } - void pushLeavingChild(const DicNode *const dicNode, const int pos, const int childrenPos, - const int probability, const bool isTerminal, const bool hasChildren, - const bool isBlacklistedOrNotAWord, const uint16_t mergedNodeCodePointCount, - const int *const mergedNodeCodePoints) { + void pushLeavingChild(const DicNode *const dicNode, const int ptNodePos, + const int childrenPtNodeArrayPos, const int probability, const bool isTerminal, + const bool hasChildren, const bool isBlacklistedOrNotAWord, + const uint16_t mergedNodeCodePointCount, const int *const mergedNodeCodePoints) { ASSERT(!mLock); mDicNodes.push_back(mEmptyNode); - mDicNodes.back().initAsChild(dicNode, pos, childrenPos, probability, isTerminal, - hasChildren, isBlacklistedOrNotAWord, mergedNodeCodePointCount, + mDicNodes.back().initAsChild(dicNode, ptNodePos, childrenPtNodeArrayPos, probability, + isTerminal, hasChildren, isBlacklistedOrNotAWord, mergedNodeCodePointCount, mergedNodeCodePoints); } diff --git a/native/jni/src/suggest/core/dicnode/internal/dic_node_properties.h b/native/jni/src/suggest/core/dicnode/internal/dic_node_properties.h index 9e0f62ceb..c41a7243a 100644 --- a/native/jni/src/suggest/core/dicnode/internal/dic_node_properties.h +++ b/native/jni/src/suggest/core/dicnode/internal/dic_node_properties.h @@ -24,15 +24,14 @@ namespace latinime { /** - * Node for traversing the lexicon trie. + * PtNode information related to the DicNode from the lexicon trie. */ -// TODO: Introduce a dictionary node class which has attribute members required to understand the -// dictionary structure. class DicNodeProperties { public: AK_FORCE_INLINE DicNodeProperties() - : mPos(0), mChildrenPos(0), mProbability(0), mNodeCodePoint(0), mIsTerminal(false), - mHasChildren(false), mIsBlacklistedOrNotAWord(false), mDepth(0), mLeavingDepth(0) {} + : mPtNodePos(0), mChildrenPtNodeArrayPos(0), mProbability(0), mDicNodeCodePoint(0), + mIsTerminal(false), mHasChildrenPtNodes(false), mIsBlacklistedOrNotAWord(false), + mDepth(0), mLeavingDepth(0) {} virtual ~DicNodeProperties() {} @@ -40,57 +39,57 @@ class DicNodeProperties { void init(const int pos, const int childrenPos, const int nodeCodePoint, const int probability, const bool isTerminal, const bool hasChildren, const bool isBlacklistedOrNotAWord, const uint16_t depth, const uint16_t leavingDepth) { - mPos = pos; - mChildrenPos = childrenPos; - mNodeCodePoint = nodeCodePoint; + mPtNodePos = pos; + mChildrenPtNodeArrayPos = childrenPos; + mDicNodeCodePoint = nodeCodePoint; mProbability = probability; mIsTerminal = isTerminal; - mHasChildren = hasChildren; + mHasChildrenPtNodes = hasChildren; mIsBlacklistedOrNotAWord = isBlacklistedOrNotAWord; mDepth = depth; mLeavingDepth = leavingDepth; } // Init for copy - void init(const DicNodeProperties *const nodeProp) { - mPos = nodeProp->mPos; - mChildrenPos = nodeProp->mChildrenPos; - mNodeCodePoint = nodeProp->mNodeCodePoint; - mProbability = nodeProp->mProbability; - mIsTerminal = nodeProp->mIsTerminal; - mHasChildren = nodeProp->mHasChildren; - mIsBlacklistedOrNotAWord = nodeProp->mIsBlacklistedOrNotAWord; - mDepth = nodeProp->mDepth; - mLeavingDepth = nodeProp->mLeavingDepth; + void init(const DicNodeProperties *const dicNodeProp) { + mPtNodePos = dicNodeProp->mPtNodePos; + mChildrenPtNodeArrayPos = dicNodeProp->mChildrenPtNodeArrayPos; + mDicNodeCodePoint = dicNodeProp->mDicNodeCodePoint; + mProbability = dicNodeProp->mProbability; + mIsTerminal = dicNodeProp->mIsTerminal; + mHasChildrenPtNodes = dicNodeProp->mHasChildrenPtNodes; + mIsBlacklistedOrNotAWord = dicNodeProp->mIsBlacklistedOrNotAWord; + mDepth = dicNodeProp->mDepth; + mLeavingDepth = dicNodeProp->mLeavingDepth; } // Init as passing child - void init(const DicNodeProperties *const nodeProp, const int codePoint) { - mPos = nodeProp->mPos; - mChildrenPos = nodeProp->mChildrenPos; - mNodeCodePoint = codePoint; // Overwrite the node char of a passing child - mProbability = nodeProp->mProbability; - mIsTerminal = nodeProp->mIsTerminal; - mHasChildren = nodeProp->mHasChildren; - mIsBlacklistedOrNotAWord = nodeProp->mIsBlacklistedOrNotAWord; - mDepth = nodeProp->mDepth + 1; // Increment the depth of a passing child - mLeavingDepth = nodeProp->mLeavingDepth; + void init(const DicNodeProperties *const dicNodeProp, const int codePoint) { + mPtNodePos = dicNodeProp->mPtNodePos; + mChildrenPtNodeArrayPos = dicNodeProp->mChildrenPtNodeArrayPos; + mDicNodeCodePoint = codePoint; // Overwrite the node char of a passing child + mProbability = dicNodeProp->mProbability; + mIsTerminal = dicNodeProp->mIsTerminal; + mHasChildrenPtNodes = dicNodeProp->mHasChildrenPtNodes; + mIsBlacklistedOrNotAWord = dicNodeProp->mIsBlacklistedOrNotAWord; + mDepth = dicNodeProp->mDepth + 1; // Increment the depth of a passing child + mLeavingDepth = dicNodeProp->mLeavingDepth; } - int getPos() const { - return mPos; + int getPtNodePos() const { + return mPtNodePos; } - int getChildrenPos() const { - return mChildrenPos; + int getChildrenPtNodeArrayPos() const { + return mChildrenPtNodeArrayPos; } int getProbability() const { return mProbability; } - int getNodeCodePoint() const { - return mNodeCodePoint; + int getDicNodeCodePoint() const { + return mDicNodeCodePoint; } uint16_t getDepth() const { @@ -107,7 +106,7 @@ class DicNodeProperties { } bool hasChildren() const { - return mHasChildren || mDepth != mLeavingDepth; + return mHasChildrenPtNodes || mDepth != mLeavingDepth; } bool isBlacklistedOrNotAWord() const { @@ -118,12 +117,12 @@ class DicNodeProperties { // Caution!!! // Use a default copy constructor and an assign operator because shallow copies are ok // for this class - int mPos; - int mChildrenPos; + int mPtNodePos; + int mChildrenPtNodeArrayPos; int mProbability; - int mNodeCodePoint; + int mDicNodeCodePoint; bool mIsTerminal; - bool mHasChildren; + bool mHasChildrenPtNodes; bool mIsBlacklistedOrNotAWord; uint16_t mDepth; uint16_t mLeavingDepth; diff --git a/native/jni/src/suggest/core/dicnode/internal/dic_node_state_prevword.h b/native/jni/src/suggest/core/dicnode/internal/dic_node_state_prevword.h index b8986203d..dba57056b 100644 --- a/native/jni/src/suggest/core/dicnode/internal/dic_node_state_prevword.h +++ b/native/jni/src/suggest/core/dicnode/internal/dic_node_state_prevword.h @@ -30,7 +30,7 @@ class DicNodeStatePrevWord { public: AK_FORCE_INLINE DicNodeStatePrevWord() : mPrevWordCount(0), mPrevWordLength(0), mPrevWordStart(0), mPrevWordProbability(0), - mPrevWordNodePos(NOT_A_DICT_POS), mSecondWordFirstInputIndex(NOT_AN_INDEX) { + mPrevWordPtNodePos(NOT_A_DICT_POS), mSecondWordFirstInputIndex(NOT_AN_INDEX) { memset(mPrevWord, 0, sizeof(mPrevWord)); } @@ -41,7 +41,7 @@ class DicNodeStatePrevWord { mPrevWordCount = 0; mPrevWordStart = 0; mPrevWordProbability = -1; - mPrevWordNodePos = NOT_A_DICT_POS; + mPrevWordPtNodePos = NOT_A_DICT_POS; mSecondWordFirstInputIndex = NOT_AN_INDEX; } @@ -50,7 +50,7 @@ class DicNodeStatePrevWord { mPrevWordCount = 0; mPrevWordStart = 0; mPrevWordProbability = -1; - mPrevWordNodePos = prevWordNodePos; + mPrevWordPtNodePos = prevWordNodePos; mSecondWordFirstInputIndex = NOT_AN_INDEX; } @@ -60,7 +60,7 @@ class DicNodeStatePrevWord { mPrevWordCount = prevWord->mPrevWordCount; mPrevWordStart = prevWord->mPrevWordStart; mPrevWordProbability = prevWord->mPrevWordProbability; - mPrevWordNodePos = prevWord->mPrevWordNodePos; + mPrevWordPtNodePos = prevWord->mPrevWordPtNodePos; mSecondWordFirstInputIndex = prevWord->mSecondWordFirstInputIndex; memcpy(mPrevWord, prevWord->mPrevWord, prevWord->mPrevWordLength * sizeof(mPrevWord[0])); } @@ -71,7 +71,7 @@ class DicNodeStatePrevWord { const int prevWordSecondWordFirstInputIndex, const int lastInputIndex) { mPrevWordCount = min(prevWordCount, static_cast<int16_t>(MAX_RESULTS)); mPrevWordProbability = prevWordProbability; - mPrevWordNodePos = prevWordNodePos; + mPrevWordPtNodePos = prevWordNodePos; int twoWordsLen = DicNodeUtils::appendTwoWords(src0, length0, src1, length1, mPrevWord); if (twoWordsLen >= MAX_WORD_LENGTH) { @@ -116,8 +116,8 @@ class DicNodeStatePrevWord { return mPrevWordStart; } - int getPrevWordNodePos() const { - return mPrevWordNodePos; + int getPrevWordPtNodePos() const { + return mPrevWordPtNodePos; } int getPrevWordCodePointAt(const int id) const { @@ -147,7 +147,7 @@ class DicNodeStatePrevWord { int16_t mPrevWordLength; int16_t mPrevWordStart; int16_t mPrevWordProbability; - int mPrevWordNodePos; + int mPrevWordPtNodePos; int mSecondWordFirstInputIndex; }; } // namespace latinime diff --git a/native/jni/src/suggest/core/dicnode/internal/dic_node_state_scoring.h b/native/jni/src/suggest/core/dicnode/internal/dic_node_state_scoring.h index 3c85d0e9d..74f9eee92 100644 --- a/native/jni/src/suggest/core/dicnode/internal/dic_node_state_scoring.h +++ b/native/jni/src/suggest/core/dicnode/internal/dic_node_state_scoring.h @@ -21,6 +21,7 @@ #include "defines.h" #include "suggest/core/dictionary/digraph_utils.h" +#include "suggest/core/dictionary/error_type_utils.h" namespace latinime { @@ -31,7 +32,7 @@ class DicNodeStateScoring { mDigraphIndex(DigraphUtils::NOT_A_DIGRAPH_INDEX), mEditCorrectionCount(0), mProximityCorrectionCount(0), mNormalizedCompoundDistance(0.0f), mSpatialDistance(0.0f), mLanguageDistance(0.0f), - mRawLength(0.0f), mExactMatch(true), + mRawLength(0.0f), mContainingErrorTypes(ErrorTypeUtils::NOT_AN_ERROR), mNormalizedCompoundDistanceAfterFirstWord(MAX_VALUE_FOR_WEIGHTING) { } @@ -47,7 +48,7 @@ class DicNodeStateScoring { mDoubleLetterLevel = NOT_A_DOUBLE_LETTER; mDigraphIndex = DigraphUtils::NOT_A_DIGRAPH_INDEX; mNormalizedCompoundDistanceAfterFirstWord = MAX_VALUE_FOR_WEIGHTING; - mExactMatch = true; + mContainingErrorTypes = ErrorTypeUtils::NOT_AN_ERROR; } AK_FORCE_INLINE void init(const DicNodeStateScoring *const scoring) { @@ -59,34 +60,21 @@ class DicNodeStateScoring { mRawLength = scoring->mRawLength; mDoubleLetterLevel = scoring->mDoubleLetterLevel; mDigraphIndex = scoring->mDigraphIndex; - mExactMatch = scoring->mExactMatch; + mContainingErrorTypes = scoring->mContainingErrorTypes; mNormalizedCompoundDistanceAfterFirstWord = scoring->mNormalizedCompoundDistanceAfterFirstWord; } void addCost(const float spatialCost, const float languageCost, const bool doNormalization, - const int inputSize, const int totalInputIndex, const ErrorType errorType) { + const int inputSize, const int totalInputIndex, + const ErrorTypeUtils::ErrorType errorType) { addDistance(spatialCost, languageCost, doNormalization, inputSize, totalInputIndex); - switch (errorType) { - case ET_EDIT_CORRECTION: - ++mEditCorrectionCount; - mExactMatch = false; - break; - case ET_PROXIMITY_CORRECTION: - ++mProximityCorrectionCount; - mExactMatch = false; - break; - case ET_COMPLETION: - mExactMatch = false; - break; - case ET_NEW_WORD: - mExactMatch = false; - break; - case ET_INTENTIONAL_OMISSION: - mExactMatch = false; - break; - case ET_NOT_AN_ERROR: - break; + mContainingErrorTypes = mContainingErrorTypes | errorType; + if (ErrorTypeUtils::isEditCorrectionError(errorType)) { + ++mEditCorrectionCount; + } + if (ErrorTypeUtils::isProximityCorrectionError(errorType)) { + ++mProximityCorrectionCount; } } @@ -182,7 +170,7 @@ class DicNodeStateScoring { } bool isExactMatch() const { - return mExactMatch; + return ErrorTypeUtils::isExactMatch(mContainingErrorTypes); } private: @@ -199,7 +187,8 @@ class DicNodeStateScoring { float mSpatialDistance; float mLanguageDistance; float mRawLength; - bool mExactMatch; + // All accumulated error types so far + ErrorTypeUtils::ErrorType mContainingErrorTypes; float mNormalizedCompoundDistanceAfterFirstWord; AK_FORCE_INLINE void addDistance(float spatialDistance, float languageDistance, diff --git a/native/jni/src/suggest/core/dictionary/bigram_dictionary.cpp b/native/jni/src/suggest/core/dictionary/bigram_dictionary.cpp index 71f4ef6ea..c2a15a312 100644 --- a/native/jni/src/suggest/core/dictionary/bigram_dictionary.cpp +++ b/native/jni/src/suggest/core/dictionary/bigram_dictionary.cpp @@ -144,7 +144,7 @@ int BigramDictionary::getPredictions(const int *prevWord, const int prevWordLeng int BigramDictionary::getBigramListPositionForWord(const int *prevWord, const int prevWordLength, const bool forceLowerCaseSearch) const { if (0 >= prevWordLength) return NOT_A_DICT_POS; - int pos = mDictionaryStructurePolicy->getTerminalNodePositionOfWord(prevWord, prevWordLength, + int pos = mDictionaryStructurePolicy->getTerminalPtNodePositionOfWord(prevWord, prevWordLength, forceLowerCaseSearch); if (NOT_A_DICT_POS == pos) return NOT_A_DICT_POS; return mDictionaryStructurePolicy->getBigramsPositionOfPtNode(pos); @@ -155,7 +155,7 @@ int BigramDictionary::getBigramProbability(const int *word0, int length0, const int pos = getBigramListPositionForWord(word0, length0, false /* forceLowerCaseSearch */); // getBigramListPositionForWord returns 0 if this word isn't in the dictionary or has no bigrams if (NOT_A_DICT_POS == pos) return NOT_A_PROBABILITY; - int nextWordPos = mDictionaryStructurePolicy->getTerminalNodePositionOfWord(word1, length1, + int nextWordPos = mDictionaryStructurePolicy->getTerminalPtNodePositionOfWord(word1, length1, false /* forceLowerCaseSearch */); if (NOT_A_DICT_POS == nextWordPos) return NOT_A_PROBABILITY; diff --git a/native/jni/src/suggest/core/dictionary/dictionary.cpp b/native/jni/src/suggest/core/dictionary/dictionary.cpp index 59ead1894..264b46056 100644 --- a/native/jni/src/suggest/core/dictionary/dictionary.cpp +++ b/native/jni/src/suggest/core/dictionary/dictionary.cpp @@ -88,7 +88,7 @@ int Dictionary::getBigrams(const int *word, int length, int *outWords, int *freq } int Dictionary::getProbability(const int *word, int length) const { - int pos = getDictionaryStructurePolicy()->getTerminalNodePositionOfWord(word, length, + int pos = getDictionaryStructurePolicy()->getTerminalPtNodePositionOfWord(word, length, false /* forceLowerCaseSearch */); if (NOT_A_DICT_POS == pos) { return NOT_A_PROBABILITY; diff --git a/native/jni/src/suggest/core/dictionary/error_type_utils.cpp b/native/jni/src/suggest/core/dictionary/error_type_utils.cpp new file mode 100644 index 000000000..0635fef7e --- /dev/null +++ b/native/jni/src/suggest/core/dictionary/error_type_utils.cpp @@ -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. + */ + +#include "suggest/core/dictionary/error_type_utils.h" + +namespace latinime { + +const ErrorTypeUtils::ErrorType ErrorTypeUtils::NOT_AN_ERROR = 0x0; +const ErrorTypeUtils::ErrorType ErrorTypeUtils::MATCH_WITH_CASE_ERROR = 0x1; +const ErrorTypeUtils::ErrorType ErrorTypeUtils::MATCH_WITH_ACCENT_ERROR = 0x2; +const ErrorTypeUtils::ErrorType ErrorTypeUtils::MATCH_WITH_DIGRAPH = 0x4; +const ErrorTypeUtils::ErrorType ErrorTypeUtils::INTENTIONAL_OMISSION = 0x8; +const ErrorTypeUtils::ErrorType ErrorTypeUtils::EDIT_CORRECTION = 0x10; +const ErrorTypeUtils::ErrorType ErrorTypeUtils::PROXIMITY_CORRECTION = 0x20; +const ErrorTypeUtils::ErrorType ErrorTypeUtils::COMPLETION = 0x40; +const ErrorTypeUtils::ErrorType ErrorTypeUtils::NEW_WORD = 0x80; + +const ErrorTypeUtils::ErrorType ErrorTypeUtils::ERRORS_TREATED_AS_AN_EXACT_MATCH = + NOT_AN_ERROR | MATCH_WITH_CASE_ERROR | MATCH_WITH_ACCENT_ERROR | MATCH_WITH_DIGRAPH; + +} // namespace latinime diff --git a/native/jni/src/suggest/core/dictionary/error_type_utils.h b/native/jni/src/suggest/core/dictionary/error_type_utils.h new file mode 100644 index 000000000..ab4a65e48 --- /dev/null +++ b/native/jni/src/suggest/core/dictionary/error_type_utils.h @@ -0,0 +1,69 @@ +/* + * 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. + */ + +#ifndef LATINIME_ERROR_TYPE_UTILS_H +#define LATINIME_ERROR_TYPE_UTILS_H + +#include <stdint.h> + +#include "defines.h" + +namespace latinime { + +class ErrorTypeUtils { + public: + // ErrorType is mainly decided by CorrectionType but it is also depending on if + // the correction has really been performed or not. + typedef uint32_t ErrorType; + + static const ErrorType NOT_AN_ERROR; + static const ErrorType MATCH_WITH_CASE_ERROR; + static const ErrorType MATCH_WITH_ACCENT_ERROR; + static const ErrorType MATCH_WITH_DIGRAPH; + // Treat error as an intentional omission when the CorrectionType is omission and the node can + // be intentional omission. + static const ErrorType INTENTIONAL_OMISSION; + // Substitution, omission and transposition + static const ErrorType EDIT_CORRECTION; + // Proximity error + static const ErrorType PROXIMITY_CORRECTION; + // Completion + static const ErrorType COMPLETION; + // New word + // TODO: Remove. + // A new word error should be an edit correction error or a proximity correction error. + static const ErrorType NEW_WORD; + + // TODO: Differentiate errors. + static bool isExactMatch(const ErrorType containingErrors) { + return (containingErrors & ~ERRORS_TREATED_AS_AN_EXACT_MATCH) == 0; + } + + static bool isEditCorrectionError(const ErrorType errorType) { + return (errorType & EDIT_CORRECTION) != 0; + } + + static bool isProximityCorrectionError(const ErrorType errorType) { + return (errorType & PROXIMITY_CORRECTION) != 0; + } + + private: + DISALLOW_IMPLICIT_CONSTRUCTORS(ErrorTypeUtils); + + static const ErrorType ERRORS_TREATED_AS_AN_EXACT_MATCH; +}; +} // namespace latinime +#endif // LATINIME_ERROR_TYPE_UTILS_H diff --git a/native/jni/src/suggest/core/layout/proximity_info.cpp b/native/jni/src/suggest/core/layout/proximity_info.cpp index e64476d82..ee8e59ef9 100644 --- a/native/jni/src/suggest/core/layout/proximity_info.cpp +++ b/native/jni/src/suggest/core/layout/proximity_info.cpp @@ -71,7 +71,7 @@ ProximityInfo::ProximityInfo(JNIEnv *env, const jstring localeJStr, && sweetSpotCenterYs && sweetSpotRadii), mProximityCharsArray(new int[GRID_WIDTH * GRID_HEIGHT * MAX_PROXIMITY_CHARS_SIZE /* proximityCharsLength */]), - mCodeToKeyMap() { + mLowerCodePointToKeyMap() { /* Let's check the input array length here to make sure */ const jsize proximityCharsLength = env->GetArrayLength(proximityChars); if (proximityCharsLength != GRID_WIDTH * GRID_HEIGHT * MAX_PROXIMITY_CHARS_SIZE) { @@ -147,7 +147,14 @@ int ProximityInfo::getCodePointOf(const int keyIndex) const { if (keyIndex < 0 || keyIndex >= KEY_COUNT) { return NOT_A_CODE_POINT; } - return mKeyIndexToCodePointG[keyIndex]; + return mKeyIndexToLowerCodePointG[keyIndex]; +} + +int ProximityInfo::getOriginalCodePointOf(const int keyIndex) const { + if (keyIndex < 0 || keyIndex >= KEY_COUNT) { + return NOT_A_CODE_POINT; + } + return mKeyIndexToOriginalCodePoint[keyIndex]; } void ProximityInfo::initializeG() { @@ -164,8 +171,9 @@ void ProximityInfo::initializeG() { const float gapY = sweetSpotCenterY - mCenterYsG[i]; mSweetSpotCenterYsG[i] = static_cast<int>(mCenterYsG[i] + gapY * verticalScale); } - mCodeToKeyMap[lowerCode] = i; - mKeyIndexToCodePointG[i] = lowerCode; + mLowerCodePointToKeyMap[lowerCode] = i; + mKeyIndexToOriginalCodePoint[i] = code; + mKeyIndexToLowerCodePointG[i] = lowerCode; } for (int i = 0; i < KEY_COUNT; i++) { mKeyKeyDistancesG[i][i] = 0; diff --git a/native/jni/src/suggest/core/layout/proximity_info.h b/native/jni/src/suggest/core/layout/proximity_info.h index f25949001..a91b9d674 100644 --- a/native/jni/src/suggest/core/layout/proximity_info.h +++ b/native/jni/src/suggest/core/layout/proximity_info.h @@ -39,6 +39,7 @@ class ProximityInfo { float getNormalizedSquaredDistanceFromCenterFloatG( const int keyId, const int x, const int y, const bool isGeometric) const; int getCodePointOf(const int keyIndex) const; + int getOriginalCodePointOf(const int keyIndex) const; bool hasSweetSpotData(const int keyIndex) const { // When there are no calibration data for a key, // the radius of the key is assigned to zero. @@ -76,11 +77,11 @@ class ProximityInfo { ProximityInfoUtils::initializeProximities(inputCodes, inputXCoordinates, inputYCoordinates, inputSize, mKeyXCoordinates, mKeyYCoordinates, mKeyWidths, mKeyHeights, mProximityCharsArray, CELL_HEIGHT, CELL_WIDTH, GRID_WIDTH, MOST_COMMON_KEY_WIDTH, - KEY_COUNT, mLocaleStr, &mCodeToKeyMap, allInputCodes); + KEY_COUNT, mLocaleStr, &mLowerCodePointToKeyMap, allInputCodes); } AK_FORCE_INLINE int getKeyIndexOf(const int c) const { - return ProximityInfoUtils::getKeyIndexOf(KEY_COUNT, c, &mCodeToKeyMap); + return ProximityInfoUtils::getKeyIndexOf(KEY_COUNT, c, &mLowerCodePointToKeyMap); } AK_FORCE_INLINE bool isCodePointOnKeyboard(const int codePoint) const { @@ -117,9 +118,9 @@ class ProximityInfo { // Sweet spots for geometric input. Note that we have extra sweet spots only for Y coordinates. float mSweetSpotCenterYsG[MAX_KEY_COUNT_IN_A_KEYBOARD]; float mSweetSpotRadii[MAX_KEY_COUNT_IN_A_KEYBOARD]; - hash_map_compat<int, int> mCodeToKeyMap; - - int mKeyIndexToCodePointG[MAX_KEY_COUNT_IN_A_KEYBOARD]; + hash_map_compat<int, int> mLowerCodePointToKeyMap; + int mKeyIndexToOriginalCodePoint[MAX_KEY_COUNT_IN_A_KEYBOARD]; + int mKeyIndexToLowerCodePointG[MAX_KEY_COUNT_IN_A_KEYBOARD]; int mCenterXsG[MAX_KEY_COUNT_IN_A_KEYBOARD]; int mCenterYsG[MAX_KEY_COUNT_IN_A_KEYBOARD]; int mKeyKeyDistancesG[MAX_KEY_COUNT_IN_A_KEYBOARD][MAX_KEY_COUNT_IN_A_KEYBOARD]; diff --git a/native/jni/src/suggest/core/layout/proximity_info_state.cpp b/native/jni/src/suggest/core/layout/proximity_info_state.cpp index fbabd92f2..bb4b41714 100644 --- a/native/jni/src/suggest/core/layout/proximity_info_state.cpp +++ b/native/jni/src/suggest/core/layout/proximity_info_state.cpp @@ -30,6 +30,12 @@ namespace latinime { +int ProximityInfoState::getPrimaryOriginalCodePointAt(const int index) const { + const int primaryCodePoint = getPrimaryCodePointAt(index); + const int keyIndex = mProximityInfo->getKeyIndexOf(primaryCodePoint); + return mProximityInfo->getOriginalCodePointOf(keyIndex); +} + // TODO: Remove the dependency of "isGeometric" void ProximityInfoState::initInputParams(const int pointerId, const float maxPointToKeyLength, const ProximityInfo *proximityInfo, const int *const inputCodes, const int inputSize, diff --git a/native/jni/src/suggest/core/layout/proximity_info_state.h b/native/jni/src/suggest/core/layout/proximity_info_state.h index c94060fa9..e941e43d8 100644 --- a/native/jni/src/suggest/core/layout/proximity_info_state.h +++ b/native/jni/src/suggest/core/layout/proximity_info_state.h @@ -65,6 +65,8 @@ class ProximityInfoState { return getProximityCodePointsAt(index)[0]; } + int getPrimaryOriginalCodePointAt(const int index) const; + inline bool sameAsTyped(const int *word, int length) const { if (length != mSampledInputSize) { return false; diff --git a/native/jni/src/suggest/core/policy/dictionary_structure_with_buffer_policy.h b/native/jni/src/suggest/core/policy/dictionary_structure_with_buffer_policy.h index 41f82049f..610de48ab 100644 --- a/native/jni/src/suggest/core/policy/dictionary_structure_with_buffer_policy.h +++ b/native/jni/src/suggest/core/policy/dictionary_structure_with_buffer_policy.h @@ -37,14 +37,14 @@ class DictionaryStructureWithBufferPolicy { virtual int getRootPosition() const = 0; - virtual void createAndGetAllChildNodes(const DicNode *const dicNode, + virtual void createAndGetAllChildDicNodes(const DicNode *const dicNode, DicNodeVector *const childDicNodes) const = 0; virtual int getCodePointsAndProbabilityAndReturnCodePointCount( const int nodePos, const int maxCodePointCount, int *const outCodePoints, int *const outUnigramProbability) const = 0; - virtual int getTerminalNodePositionOfWord(const int *const inWord, + virtual int getTerminalPtNodePositionOfWord(const int *const inWord, const int length, const bool forceLowerCaseSearch) const = 0; virtual int getProbability(const int unigramProbability, diff --git a/native/jni/src/suggest/core/policy/weighting.cpp b/native/jni/src/suggest/core/policy/weighting.cpp index 0c4016893..c202b81fe 100644 --- a/native/jni/src/suggest/core/policy/weighting.cpp +++ b/native/jni/src/suggest/core/policy/weighting.cpp @@ -20,6 +20,7 @@ #include "suggest/core/dicnode/dic_node.h" #include "suggest/core/dicnode/dic_node_profiler.h" #include "suggest/core/dicnode/dic_node_utils.h" +#include "suggest/core/dictionary/error_type_utils.h" #include "suggest/core/session/dic_traverse_session.h" namespace latinime { @@ -82,8 +83,8 @@ static inline void profile(const CorrectionType correctionType, DicNode *const n traverseSession, parentDicNode, dicNode, &inputStateG); const float languageCost = Weighting::getLanguageCost(weighting, correctionType, traverseSession, parentDicNode, dicNode, multiBigramMap); - const ErrorType errorType = weighting->getErrorType(correctionType, traverseSession, - parentDicNode, dicNode); + const ErrorTypeUtils::ErrorType errorType = weighting->getErrorType(correctionType, + traverseSession, parentDicNode, dicNode); profile(correctionType, dicNode); if (inputStateG.mNeedsToUpdateInputStateG) { dicNode->updateInputIndexG(&inputStateG); diff --git a/native/jni/src/suggest/core/policy/weighting.h b/native/jni/src/suggest/core/policy/weighting.h index 2d49e98a6..bd6b3cf41 100644 --- a/native/jni/src/suggest/core/policy/weighting.h +++ b/native/jni/src/suggest/core/policy/weighting.h @@ -18,6 +18,7 @@ #define LATINIME_WEIGHTING_H #include "defines.h" +#include "suggest/core/dictionary/error_type_utils.h" namespace latinime { @@ -84,7 +85,7 @@ class Weighting { virtual float getSpaceSubstitutionCost(const DicTraverseSession *const traverseSession, const DicNode *const dicNode) const = 0; - virtual ErrorType getErrorType(const CorrectionType correctionType, + virtual ErrorTypeUtils::ErrorType getErrorType(const CorrectionType correctionType, const DicTraverseSession *const traverseSession, const DicNode *const parentDicNode, const DicNode *const dicNode) const = 0; diff --git a/native/jni/src/suggest/core/session/dic_traverse_session.cpp b/native/jni/src/suggest/core/session/dic_traverse_session.cpp index 50f2bbd8d..5070491f4 100644 --- a/native/jni/src/suggest/core/session/dic_traverse_session.cpp +++ b/native/jni/src/suggest/core/session/dic_traverse_session.cpp @@ -35,16 +35,16 @@ void DicTraverseSession::init(const Dictionary *const dictionary, const int *pre ->getMultiWordCostMultiplier(); mSuggestOptions = suggestOptions; if (!prevWord) { - mPrevWordPos = NOT_A_DICT_POS; + mPrevWordPtNodePos = NOT_A_DICT_POS; return; } // TODO: merge following similar calls to getTerminalPosition into one case-insensitive call. - mPrevWordPos = getDictionaryStructurePolicy()->getTerminalNodePositionOfWord( + mPrevWordPtNodePos = getDictionaryStructurePolicy()->getTerminalPtNodePositionOfWord( prevWord, prevWordLength, false /* forceLowerCaseSearch */); - if (mPrevWordPos == NOT_A_DICT_POS) { + if (mPrevWordPtNodePos == NOT_A_DICT_POS) { // Check bigrams for lower-cased previous word if original was not found. Useful for // auto-capitalized words like "The [current_word]". - mPrevWordPos = getDictionaryStructurePolicy()->getTerminalNodePositionOfWord( + mPrevWordPtNodePos = getDictionaryStructurePolicy()->getTerminalPtNodePositionOfWord( prevWord, prevWordLength, true /* forceLowerCaseSearch */); } } diff --git a/native/jni/src/suggest/core/session/dic_traverse_session.h b/native/jni/src/suggest/core/session/dic_traverse_session.h index e0b1c67d9..6e4dda44d 100644 --- a/native/jni/src/suggest/core/session/dic_traverse_session.h +++ b/native/jni/src/suggest/core/session/dic_traverse_session.h @@ -59,7 +59,7 @@ class DicTraverseSession { } AK_FORCE_INLINE DicTraverseSession(JNIEnv *env, jstring localeStr, bool usesLargeCache) - : mPrevWordPos(NOT_A_DICT_POS), mProximityInfo(0), + : mPrevWordPtNodePos(NOT_A_DICT_POS), mProximityInfo(0), mDictionary(0), mSuggestOptions(0), mDicNodesCache(usesLargeCache), mMultiBigramMap(), mInputSize(0), mPartiallyCommited(false), mMaxPointerCount(1), mMultiWordCostMultiplier(1.0f) { @@ -86,11 +86,9 @@ class DicTraverseSession { //-------------------- const ProximityInfo *getProximityInfo() const { return mProximityInfo; } const SuggestOptions *getSuggestOptions() const { return mSuggestOptions; } - int getPrevWordPos() const { return mPrevWordPos; } + int getPrevWordPtNodePos() const { return mPrevWordPtNodePos; } // TODO: REMOVE - void setPrevWordPos(int pos) { mPrevWordPos = pos; } - // TODO: Use proper parameter when changed - int getDicRootPos() const { return 0; } + void setPrevWordPtNodePos(const int ptNodePos) { mPrevWordPtNodePos = ptNodePos; } DicNodesCache *getDicTraverseCache() { return &mDicNodesCache; } MultiBigramMap *getMultiBigramMap() { return &mMultiBigramMap; } const ProximityInfoState *getProximityInfoState(int id) const { @@ -119,26 +117,13 @@ class DicTraverseSession { return true; } - void getSearchKeys(const DicNode *node, std::vector<int> *const outputSearchKeyVector) const { - for (int i = 0; i < MAX_POINTER_COUNT_G; ++i) { - if (!mProximityInfoStates[i].isUsed()) { - continue; - } - const int pointerId = node->getInputIndex(i); - const std::vector<int> *const searchKeyVector = - mProximityInfoStates[i].getSearchKeyVector(pointerId); - outputSearchKeyVector->insert(outputSearchKeyVector->end(), searchKeyVector->begin(), - searchKeyVector->end()); - } - } - - ProximityType getProximityTypeG(const DicNode *const node, const int childCodePoint) const { + ProximityType getProximityTypeG(const DicNode *const dicNode, const int childCodePoint) const { ProximityType proximityType = UNRELATED_CHAR; for (int i = 0; i < MAX_POINTER_COUNT_G; ++i) { if (!mProximityInfoStates[i].isUsed()) { continue; } - const int pointerId = node->getInputIndex(i); + const int pointerId = dicNode->getInputIndex(i); proximityType = mProximityInfoStates[i].getProximityTypeG(pointerId, childCodePoint); ASSERT(proximityType == UNRELATED_CHAR || proximityType == MATCH_CHAR); // TODO: Make this more generic @@ -192,7 +177,7 @@ class DicTraverseSession { const int *const inputYs, const int *const times, const int *const pointerIds, const int inputSize, const float maxSpatialDistance, const int maxPointerCount); - int mPrevWordPos; + int mPrevWordPtNodePos; const ProximityInfo *mProximityInfo; const Dictionary *mDictionary; const SuggestOptions *mSuggestOptions; diff --git a/native/jni/src/suggest/core/suggest.cpp b/native/jni/src/suggest/core/suggest.cpp index 73ccebc88..2eda414f4 100644 --- a/native/jni/src/suggest/core/suggest.cpp +++ b/native/jni/src/suggest/core/suggest.cpp @@ -98,7 +98,7 @@ void Suggest::initializeSearch(DicTraverseSession *traverseSession, int commitPo // Continue suggestion after partial commit. DicNode *topDicNode = traverseSession->getDicTraverseCache()->setCommitPoint(commitPoint); - traverseSession->setPrevWordPos(topDicNode->getPrevWordNodePos()); + traverseSession->setPrevWordPtNodePos(topDicNode->getPrevWordPtNodePos()); traverseSession->getDicTraverseCache()->continueSearch(); traverseSession->setPartiallyCommited(); } @@ -109,7 +109,7 @@ void Suggest::initializeSearch(DicTraverseSession *traverseSession, int commitPo // Create a new dic node here DicNode rootNode; DicNodeUtils::initAsRoot(traverseSession->getDictionaryStructurePolicy(), - traverseSession->getPrevWordPos(), &rootNode); + traverseSession->getPrevWordPtNodePos(), &rootNode); traverseSession->getDicTraverseCache()->copyPushActive(&rootNode); } } @@ -231,7 +231,7 @@ int Suggest::outputSuggestions(DicTraverseSession *traverseSession, int *frequen BinaryDictionaryShortcutIterator shortcutIt( traverseSession->getDictionaryStructurePolicy()->getShortcutsStructurePolicy(), traverseSession->getDictionaryStructurePolicy() - ->getShortcutPositionOfPtNode(terminalDicNode->getPos())); + ->getShortcutPositionOfPtNode(terminalDicNode->getPtNodePos())); // Shortcut is not supported for multiple words suggestions. // TODO: Check shortcuts during traversal for multiple words suggestions. const bool sameAsTyped = TRAVERSAL->sameAsTyped(traverseSession, terminalDicNode); @@ -421,15 +421,15 @@ void Suggest::expandCurrentDicNodes(DicTraverseSession *traverseSession) const { } break; case UNRELATED_CHAR: - // Just drop this node and do nothing. + // Just drop this dicNode and do nothing. break; default: - // Just drop this node and do nothing. + // Just drop this dicNode and do nothing. break; } } - // Push the node for look-ahead correction + // Push the dicNode for look-ahead correction if (allowsErrorCorrections && canDoLookAheadCorrection) { traverseSession->getDicTraverseCache()->copyPushNextActive(&dicNode); } @@ -442,7 +442,7 @@ void Suggest::processTerminalDicNode( if (dicNode->getCompoundDistance() >= static_cast<float>(MAX_VALUE_FOR_WEIGHTING)) { return; } - if (!dicNode->isTerminalWordNode()) { + if (!dicNode->isTerminalDicNode()) { return; } if (dicNode->shouldBeFilteredBySafetyNetForBigram()) { @@ -463,7 +463,7 @@ void Suggest::processTerminalDicNode( /** * Adds the expanded dicNode to the next search priority queue. Also creates an additional next word - * (by the space omission error correction) search path if input dicNode is on a terminal node. + * (by the space omission error correction) search path if input dicNode is on a terminal. */ void Suggest::processExpandedDicNode( DicTraverseSession *traverseSession, DicNode *dicNode) const { @@ -505,7 +505,7 @@ void Suggest::processDicNodeAsSubstitution(DicTraverseSession *traverseSession, processExpandedDicNode(traverseSession, childDicNode); } -// Process the node codepoint as a digraph. This means that composite glyphs like the German +// Process the DicNode codepoint as a digraph. This means that composite glyphs like the German // u-umlaut is expanded to the transliteration "ue". Note that this happens in parallel with // the normal non-digraph traversal, so both "uber" and "ueber" can be corrected to "[u-umlaut]ber". void Suggest::processDicNodeAsDigraph(DicTraverseSession *traverseSession, @@ -518,7 +518,7 @@ void Suggest::processDicNodeAsDigraph(DicTraverseSession *traverseSession, /** * Handle the dicNode as an omission error (e.g., ths => this). Skip the current letter and consider * matches for all possible next letters. Note that just skipping the current letter without any - * other conditions tends to flood the search dic nodes cache with omission nodes. Instead, check + * other conditions tends to flood the search DicNodes cache with omission DicNodes. Instead, check * the possible *next* letters after the omission to better limit search to plausible omissions. * Note that apostrophes are handled as omissions. */ @@ -605,7 +605,7 @@ void Suggest::processDicNodeAsTransposition(DicTraverseSession *traverseSession, } /** - * Weight child node by aligning it to the key + * Weight child dicNode by aligning it to the key */ void Suggest::weightChildNode(DicTraverseSession *traverseSession, DicNode *dicNode) const { const int inputSize = traverseSession->getInputSize(); diff --git a/native/jni/src/suggest/policyimpl/dictionary/bigram/bigram_list_read_write_utils.cpp b/native/jni/src/suggest/policyimpl/dictionary/bigram/bigram_list_read_write_utils.cpp index 1926b9831..de9fc9bbc 100644 --- a/native/jni/src/suggest/policyimpl/dictionary/bigram/bigram_list_read_write_utils.cpp +++ b/native/jni/src/suggest/policyimpl/dictionary/bigram/bigram_list_read_write_utils.cpp @@ -16,7 +16,7 @@ #include "suggest/policyimpl/dictionary/bigram/bigram_list_read_write_utils.h" -#include "suggest/policyimpl/dictionary/dynamic_patricia_trie_reading_utils.h" +#include "suggest/policyimpl/dictionary/structure/v3/dynamic_patricia_trie_reading_utils.h" #include "suggest/policyimpl/dictionary/utils/byte_array_utils.h" #include "suggest/policyimpl/dictionary/utils/buffer_with_extendable_buffer.h" diff --git a/native/jni/src/suggest/policyimpl/dictionary/bigram/dynamic_bigram_list_policy.cpp b/native/jni/src/suggest/policyimpl/dictionary/bigram/dynamic_bigram_list_policy.cpp index b1170e251..d97501265 100644 --- a/native/jni/src/suggest/policyimpl/dictionary/bigram/dynamic_bigram_list_policy.cpp +++ b/native/jni/src/suggest/policyimpl/dictionary/bigram/dynamic_bigram_list_policy.cpp @@ -17,8 +17,8 @@ #include "suggest/policyimpl/dictionary/bigram/dynamic_bigram_list_policy.h" #include "suggest/core/policy/dictionary_shortcuts_structure_policy.h" -#include "suggest/policyimpl/dictionary/dynamic_patricia_trie_node_reader.h" -#include "suggest/policyimpl/dictionary/dynamic_patricia_trie_writing_helper.h" +#include "suggest/policyimpl/dictionary/structure/v3/dynamic_patricia_trie_node_reader.h" +#include "suggest/policyimpl/dictionary/structure/v3/dynamic_patricia_trie_writing_helper.h" #include "suggest/policyimpl/dictionary/utils/buffer_with_extendable_buffer.h" #include "suggest/policyimpl/dictionary/utils/forgetting_curve_utils.h" diff --git a/native/jni/src/suggest/policyimpl/dictionary/bigram/dynamic_bigram_list_policy.h b/native/jni/src/suggest/policyimpl/dictionary/bigram/dynamic_bigram_list_policy.h index 0504b59d5..5de456656 100644 --- a/native/jni/src/suggest/policyimpl/dictionary/bigram/dynamic_bigram_list_policy.h +++ b/native/jni/src/suggest/policyimpl/dictionary/bigram/dynamic_bigram_list_policy.h @@ -22,7 +22,7 @@ #include "defines.h" #include "suggest/core/policy/dictionary_bigrams_structure_policy.h" #include "suggest/policyimpl/dictionary/bigram/bigram_list_read_write_utils.h" -#include "suggest/policyimpl/dictionary/dynamic_patricia_trie_writing_helper.h" +#include "suggest/policyimpl/dictionary/structure/v3/dynamic_patricia_trie_writing_helper.h" namespace latinime { diff --git a/native/jni/src/suggest/policyimpl/dictionary/header/header_policy.h b/native/jni/src/suggest/policyimpl/dictionary/header/header_policy.h index a9c7805a8..7c06a7117 100644 --- a/native/jni/src/suggest/policyimpl/dictionary/header/header_policy.h +++ b/native/jni/src/suggest/policyimpl/dictionary/header/header_policy.h @@ -30,8 +30,8 @@ namespace latinime { class HeaderPolicy : public DictionaryHeaderStructurePolicy { public: // Reads information from existing dictionary buffer. - HeaderPolicy(const uint8_t *const dictBuf, const int dictSize) - : mDictFormatVersion(FormatUtils::detectFormatVersion(dictBuf, dictSize)), + HeaderPolicy(const uint8_t *const dictBuf, const FormatUtils::FORMAT_VERSION formatVersion) + : mDictFormatVersion(formatVersion), mDictionaryFlags(HeaderReadWriteUtils::getFlags(dictBuf)), mSize(HeaderReadWriteUtils::getHeaderSize(dictBuf)), mAttributeMap(createAttributeMapAndReadAllAttributes(dictBuf)), diff --git a/native/jni/src/suggest/policyimpl/dictionary/header/header_read_write_utils.cpp b/native/jni/src/suggest/policyimpl/dictionary/header/header_read_write_utils.cpp index 5ded8f6a1..5ef8e50b4 100644 --- a/native/jni/src/suggest/policyimpl/dictionary/header/header_read_write_utils.cpp +++ b/native/jni/src/suggest/policyimpl/dictionary/header/header_read_write_utils.cpp @@ -118,6 +118,9 @@ const char *const HeaderReadWriteUtils::REQUIRES_FRENCH_LIGATURE_PROCESSING_KEY case FormatUtils::VERSION_3: return buffer->writeUintAndAdvancePosition(3 /* data */, HEADER_DICTIONARY_VERSION_SIZE, writingPos); + case FormatUtils::VERSION_4: + return buffer->writeUintAndAdvancePosition(4 /* data */, + HEADER_DICTIONARY_VERSION_SIZE, writingPos); default: return false; } diff --git a/native/jni/src/suggest/policyimpl/dictionary/dictionary_structure_with_buffer_policy_factory.cpp b/native/jni/src/suggest/policyimpl/dictionary/structure/dictionary_structure_with_buffer_policy_factory.cpp index ff80dd2f6..baa9cf142 100644 --- a/native/jni/src/suggest/policyimpl/dictionary/dictionary_structure_with_buffer_policy_factory.cpp +++ b/native/jni/src/suggest/policyimpl/dictionary/structure/dictionary_structure_with_buffer_policy_factory.cpp @@ -14,13 +14,13 @@ * limitations under the License. */ -#include "suggest/policyimpl/dictionary/dictionary_structure_with_buffer_policy_factory.h" +#include "suggest/policyimpl/dictionary/structure/dictionary_structure_with_buffer_policy_factory.h" #include <stdint.h> #include "defines.h" -#include "suggest/policyimpl/dictionary/dynamic_patricia_trie_policy.h" -#include "suggest/policyimpl/dictionary/patricia_trie_policy.h" +#include "suggest/policyimpl/dictionary/structure/v2/patricia_trie_policy.h" +#include "suggest/policyimpl/dictionary/structure/v3/dynamic_patricia_trie_policy.h" #include "suggest/policyimpl/dictionary/utils/format_utils.h" #include "suggest/policyimpl/dictionary/utils/mmapped_buffer.h" @@ -42,6 +42,9 @@ namespace latinime { return new PatriciaTriePolicy(mmapedBuffer); case FormatUtils::VERSION_3: return new DynamicPatriciaTriePolicy(mmapedBuffer); + case FormatUtils::VERSION_4: + // TODO: Support version 4 dictionary format. + // Fall through. default: AKLOGE("DICT: dictionary format is unknown, bad magic number"); delete mmapedBuffer; diff --git a/native/jni/src/suggest/policyimpl/dictionary/dictionary_structure_with_buffer_policy_factory.h b/native/jni/src/suggest/policyimpl/dictionary/structure/dictionary_structure_with_buffer_policy_factory.h index 8cebc3b16..8cebc3b16 100644 --- a/native/jni/src/suggest/policyimpl/dictionary/dictionary_structure_with_buffer_policy_factory.h +++ b/native/jni/src/suggest/policyimpl/dictionary/structure/dictionary_structure_with_buffer_policy_factory.h diff --git a/native/jni/src/suggest/policyimpl/dictionary/patricia_trie_policy.cpp b/native/jni/src/suggest/policyimpl/dictionary/structure/v2/patricia_trie_policy.cpp index 8a84bd261..960c1b936 100644 --- a/native/jni/src/suggest/policyimpl/dictionary/patricia_trie_policy.cpp +++ b/native/jni/src/suggest/policyimpl/dictionary/structure/v2/patricia_trie_policy.cpp @@ -15,22 +15,22 @@ */ -#include "suggest/policyimpl/dictionary/patricia_trie_policy.h" +#include "suggest/policyimpl/dictionary/structure/v2/patricia_trie_policy.h" #include "defines.h" #include "suggest/core/dicnode/dic_node.h" #include "suggest/core/dicnode/dic_node_vector.h" -#include "suggest/policyimpl/dictionary/patricia_trie_reading_utils.h" +#include "suggest/policyimpl/dictionary/structure/v2/patricia_trie_reading_utils.h" #include "suggest/policyimpl/dictionary/utils/probability_utils.h" namespace latinime { -void PatriciaTriePolicy::createAndGetAllChildNodes(const DicNode *const dicNode, +void PatriciaTriePolicy::createAndGetAllChildDicNodes(const DicNode *const dicNode, DicNodeVector *const childDicNodes) const { if (!dicNode->hasChildren()) { return; } - int nextPos = dicNode->getChildrenPos(); + int nextPos = dicNode->getChildrenPtNodeArrayPos(); if (nextPos < 0 || nextPos >= mDictBufferSize) { AKLOGE("Children PtNode array position is invalid. pos: %d, dict size: %d", nextPos, mDictBufferSize); @@ -52,14 +52,14 @@ void PatriciaTriePolicy::createAndGetAllChildNodes(const DicNode *const dicNode, // This retrieves code points and the probability of the word by its terminal position. // Due to the fact that words are ordered in the dictionary in a strict breadth-first order, -// it is possible to check for this with advantageous complexity. For each node, we search +// it is possible to check for this with advantageous complexity. For each PtNode array, we search // for PtNodes with children and compare the children position with the position we look for. // When we shoot the position we look for, it means the word we look for is in the children // of the previous PtNode. The only tricky part is the fact that if we arrive at the end of a // PtNode array with the last PtNode's children position still less than what we are searching for, // we must descend the last PtNode's children (for example, if the word we are searching for starts // with a z, it's the last PtNode of the root array, so all children addresses will be smaller -// than the position we look for, and we have to descend the z node). +// than the position we look for, and we have to descend the z PtNode). /* Parameters : * ptNodePos: the byte position of the terminal PtNode of the word we are searching for (this is * what is stored as the "bigram position" in each bigram) @@ -74,9 +74,9 @@ int PatriciaTriePolicy::getCodePointsAndProbabilityAndReturnCodePointCount( int pos = getRootPosition(); int wordPos = 0; // One iteration of the outer loop iterates through PtNode arrays. As stated above, we will - // only traverse nodes that are actually a part of the terminal we are searching, so each time - // we enter this loop we are one depth level further than last time. - // The only reason we count nodes is because we want to reduce the probability of infinite + // only traverse PtNodes that are actually a part of the terminal we are searching, so each + // time we enter this loop we are one depth level further than last time. + // The only reason we count PtNodes is because we want to reduce the probability of infinite // looping in case there is a bug. Since we know there is an upper bound to the depth we are // supposed to traverse, it does not hurt to count iterations. for (int loopCount = maxCodePointCount; loopCount > 0; --loopCount) { @@ -140,8 +140,9 @@ int PatriciaTriePolicy::getCodePointsAndProbabilityAndReturnCodePointCount( found = true; } else if (1 >= ptNodeCount) { // However if we are on the LAST PtNode of this array, and we have NOT shot the - // position we should descend THIS node. So we trick the lastCandidatePtNodePos - // so that we will descend this PtNode, not the previous one. + // position we should descend THIS PtNode. So we trick the + // lastCandidatePtNodePos so that we will descend this PtNode, not the previous + // one. lastCandidatePtNodePos = startPos; found = true; } else { @@ -149,7 +150,7 @@ int PatriciaTriePolicy::getCodePointsAndProbabilityAndReturnCodePointCount( found = false; } } else { - // Even if we don't have children here, we could still be on the last PtNode of / + // Even if we don't have children here, we could still be on the last PtNode of // this array. If this is the case, we should descend the last PtNode that had // children, and their position is already in lastCandidatePtNodePos. found = (1 >= ptNodeCount); @@ -230,9 +231,9 @@ int PatriciaTriePolicy::getCodePointsAndProbabilityAndReturnCodePointCount( return 0; } -// This function gets the position of the terminal node of the exact matching word in the +// This function gets the position of the terminal PtNode of the exact matching word in the // dictionary. If no match is found, it returns NOT_A_DICT_POS. -int PatriciaTriePolicy::getTerminalNodePositionOfWord(const int *const inWord, +int PatriciaTriePolicy::getTerminalPtNodePositionOfWord(const int *const inWord, const int length, const bool forceLowerCaseSearch) const { int pos = getRootPosition(); int wordPos = 0; diff --git a/native/jni/src/suggest/policyimpl/dictionary/patricia_trie_policy.h b/native/jni/src/suggest/policyimpl/dictionary/structure/v2/patricia_trie_policy.h index 0f8662aea..31e6186b7 100644 --- a/native/jni/src/suggest/policyimpl/dictionary/patricia_trie_policy.h +++ b/native/jni/src/suggest/policyimpl/dictionary/structure/v2/patricia_trie_policy.h @@ -24,6 +24,7 @@ #include "suggest/policyimpl/dictionary/bigram/bigram_list_policy.h" #include "suggest/policyimpl/dictionary/header/header_policy.h" #include "suggest/policyimpl/dictionary/shortcut/shortcut_list_policy.h" +#include "suggest/policyimpl/dictionary/utils/format_utils.h" #include "suggest/policyimpl/dictionary/utils/mmapped_buffer.h" namespace latinime { @@ -34,7 +35,7 @@ class DicNodeVector; class PatriciaTriePolicy : public DictionaryStructureWithBufferPolicy { public: PatriciaTriePolicy(const MmappedBuffer *const buffer) - : mBuffer(buffer), mHeaderPolicy(mBuffer->getBuffer(), buffer->getBufferSize()), + : mBuffer(buffer), mHeaderPolicy(mBuffer->getBuffer(), FormatUtils::VERSION_2), mDictRoot(mBuffer->getBuffer() + mHeaderPolicy.getSize()), mDictBufferSize(mBuffer->getBufferSize() - mHeaderPolicy.getSize()), mBigramListPolicy(mDictRoot), mShortcutListPolicy(mDictRoot) {} @@ -47,14 +48,14 @@ class PatriciaTriePolicy : public DictionaryStructureWithBufferPolicy { return 0; } - void createAndGetAllChildNodes(const DicNode *const dicNode, + void createAndGetAllChildDicNodes(const DicNode *const dicNode, DicNodeVector *const childDicNodes) const; int getCodePointsAndProbabilityAndReturnCodePointCount( const int terminalNodePos, const int maxCodePointCount, int *const outCodePoints, int *const outUnigramProbability) const; - int getTerminalNodePositionOfWord(const int *const inWord, + int getTerminalPtNodePositionOfWord(const int *const inWord, const int length, const bool forceLowerCaseSearch) const; int getProbability(const int unigramProbability, const int bigramProbability) const; diff --git a/native/jni/src/suggest/policyimpl/dictionary/patricia_trie_reading_utils.cpp b/native/jni/src/suggest/policyimpl/dictionary/structure/v2/patricia_trie_reading_utils.cpp index 7df55815f..82b3593c8 100644 --- a/native/jni/src/suggest/policyimpl/dictionary/patricia_trie_reading_utils.cpp +++ b/native/jni/src/suggest/policyimpl/dictionary/structure/v2/patricia_trie_reading_utils.cpp @@ -14,7 +14,7 @@ * limitations under the License. */ -#include "suggest/policyimpl/dictionary/patricia_trie_reading_utils.h" +#include "suggest/policyimpl/dictionary/structure/v2/patricia_trie_reading_utils.h" #include "defines.h" #include "suggest/policyimpl/dictionary/utils/byte_array_utils.h" diff --git a/native/jni/src/suggest/policyimpl/dictionary/patricia_trie_reading_utils.h b/native/jni/src/suggest/policyimpl/dictionary/structure/v2/patricia_trie_reading_utils.h index 8420ee95a..8420ee95a 100644 --- a/native/jni/src/suggest/policyimpl/dictionary/patricia_trie_reading_utils.h +++ b/native/jni/src/suggest/policyimpl/dictionary/structure/v2/patricia_trie_reading_utils.h diff --git a/native/jni/src/suggest/policyimpl/dictionary/dynamic_patricia_trie_gc_event_listeners.cpp b/native/jni/src/suggest/policyimpl/dictionary/structure/v3/dynamic_patricia_trie_gc_event_listeners.cpp index 5724c5d88..126b7681e 100644 --- a/native/jni/src/suggest/policyimpl/dictionary/dynamic_patricia_trie_gc_event_listeners.cpp +++ b/native/jni/src/suggest/policyimpl/dictionary/structure/v3/dynamic_patricia_trie_gc_event_listeners.cpp @@ -14,7 +14,7 @@ * limitations under the License. */ -#include "suggest/policyimpl/dictionary/dynamic_patricia_trie_gc_event_listeners.h" +#include "suggest/policyimpl/dictionary/structure/v3/dynamic_patricia_trie_gc_event_listeners.h" #include "suggest/core/policy/dictionary_header_structure_policy.h" #include "suggest/policyimpl/dictionary/utils/forgetting_curve_utils.h" diff --git a/native/jni/src/suggest/policyimpl/dictionary/dynamic_patricia_trie_gc_event_listeners.h b/native/jni/src/suggest/policyimpl/dictionary/structure/v3/dynamic_patricia_trie_gc_event_listeners.h index 9755120b0..ab59cc645 100644 --- a/native/jni/src/suggest/policyimpl/dictionary/dynamic_patricia_trie_gc_event_listeners.h +++ b/native/jni/src/suggest/policyimpl/dictionary/structure/v3/dynamic_patricia_trie_gc_event_listeners.h @@ -21,9 +21,9 @@ #include "defines.h" #include "suggest/policyimpl/dictionary/bigram/dynamic_bigram_list_policy.h" -#include "suggest/policyimpl/dictionary/dynamic_patricia_trie_reading_helper.h" -#include "suggest/policyimpl/dictionary/dynamic_patricia_trie_writing_helper.h" -#include "suggest/policyimpl/dictionary/dynamic_patricia_trie_writing_utils.h" +#include "suggest/policyimpl/dictionary/structure/v3/dynamic_patricia_trie_reading_helper.h" +#include "suggest/policyimpl/dictionary/structure/v3/dynamic_patricia_trie_writing_helper.h" +#include "suggest/policyimpl/dictionary/structure/v3/dynamic_patricia_trie_writing_utils.h" #include "suggest/policyimpl/dictionary/utils/buffer_with_extendable_buffer.h" #include "utils/hash_map_compat.h" diff --git a/native/jni/src/suggest/policyimpl/dictionary/dynamic_patricia_trie_node_reader.cpp b/native/jni/src/suggest/policyimpl/dictionary/structure/v3/dynamic_patricia_trie_node_reader.cpp index 2fa3111d3..4fd2484e1 100644 --- a/native/jni/src/suggest/policyimpl/dictionary/dynamic_patricia_trie_node_reader.cpp +++ b/native/jni/src/suggest/policyimpl/dictionary/structure/v3/dynamic_patricia_trie_node_reader.cpp @@ -14,11 +14,11 @@ * limitations under the License. */ -#include "suggest/policyimpl/dictionary/dynamic_patricia_trie_node_reader.h" +#include "suggest/policyimpl/dictionary/structure/v3/dynamic_patricia_trie_node_reader.h" #include "suggest/core/policy/dictionary_bigrams_structure_policy.h" #include "suggest/core/policy/dictionary_shortcuts_structure_policy.h" -#include "suggest/policyimpl/dictionary/dynamic_patricia_trie_reading_utils.h" +#include "suggest/policyimpl/dictionary/structure/v3/dynamic_patricia_trie_reading_utils.h" #include "suggest/policyimpl/dictionary/utils/buffer_with_extendable_buffer.h" namespace latinime { diff --git a/native/jni/src/suggest/policyimpl/dictionary/dynamic_patricia_trie_node_reader.h b/native/jni/src/suggest/policyimpl/dictionary/structure/v3/dynamic_patricia_trie_node_reader.h index 3b36d425f..fac078d0a 100644 --- a/native/jni/src/suggest/policyimpl/dictionary/dynamic_patricia_trie_node_reader.h +++ b/native/jni/src/suggest/policyimpl/dictionary/structure/v3/dynamic_patricia_trie_node_reader.h @@ -20,8 +20,8 @@ #include <stdint.h> #include "defines.h" -#include "suggest/policyimpl/dictionary/dynamic_patricia_trie_reading_utils.h" -#include "suggest/policyimpl/dictionary/patricia_trie_reading_utils.h" +#include "suggest/policyimpl/dictionary/structure/v2/patricia_trie_reading_utils.h" +#include "suggest/policyimpl/dictionary/structure/v3/dynamic_patricia_trie_reading_utils.h" namespace latinime { diff --git a/native/jni/src/suggest/policyimpl/dictionary/dynamic_patricia_trie_policy.cpp b/native/jni/src/suggest/policyimpl/dictionary/structure/v3/dynamic_patricia_trie_policy.cpp index 495b146c2..246458d09 100644 --- a/native/jni/src/suggest/policyimpl/dictionary/dynamic_patricia_trie_policy.cpp +++ b/native/jni/src/suggest/policyimpl/dictionary/structure/v3/dynamic_patricia_trie_policy.cpp @@ -14,7 +14,7 @@ * limitations under the License. */ -#include "suggest/policyimpl/dictionary/dynamic_patricia_trie_policy.h" +#include "suggest/policyimpl/dictionary/structure/v3/dynamic_patricia_trie_policy.h" #include <cstdio> #include <cstring> @@ -23,11 +23,11 @@ #include "defines.h" #include "suggest/core/dicnode/dic_node.h" #include "suggest/core/dicnode/dic_node_vector.h" -#include "suggest/policyimpl/dictionary/dynamic_patricia_trie_node_reader.h" -#include "suggest/policyimpl/dictionary/dynamic_patricia_trie_reading_helper.h" -#include "suggest/policyimpl/dictionary/dynamic_patricia_trie_reading_utils.h" -#include "suggest/policyimpl/dictionary/dynamic_patricia_trie_writing_helper.h" -#include "suggest/policyimpl/dictionary/patricia_trie_reading_utils.h" +#include "suggest/policyimpl/dictionary/structure/v2/patricia_trie_reading_utils.h" +#include "suggest/policyimpl/dictionary/structure/v3/dynamic_patricia_trie_node_reader.h" +#include "suggest/policyimpl/dictionary/structure/v3/dynamic_patricia_trie_reading_helper.h" +#include "suggest/policyimpl/dictionary/structure/v3/dynamic_patricia_trie_reading_utils.h" +#include "suggest/policyimpl/dictionary/structure/v3/dynamic_patricia_trie_writing_helper.h" #include "suggest/policyimpl/dictionary/utils/forgetting_curve_utils.h" #include "suggest/policyimpl/dictionary/utils/probability_utils.h" @@ -45,14 +45,14 @@ const int DynamicPatriciaTriePolicy::MAX_DICT_EXTENDED_REGION_SIZE = 1024 * 1024 const int DynamicPatriciaTriePolicy::MIN_DICT_SIZE_TO_REFUSE_DYNAMIC_OPERATIONS = DynamicPatriciaTrieWritingHelper::MAX_DICTIONARY_SIZE - 1024; -void DynamicPatriciaTriePolicy::createAndGetAllChildNodes(const DicNode *const dicNode, +void DynamicPatriciaTriePolicy::createAndGetAllChildDicNodes(const DicNode *const dicNode, DicNodeVector *const childDicNodes) const { if (!dicNode->hasChildren()) { return; } DynamicPatriciaTrieReadingHelper readingHelper(&mBufferWithExtendableBuffer, getBigramsStructurePolicy(), getShortcutsStructurePolicy()); - readingHelper.initWithPtNodeArrayPos(dicNode->getChildrenPos()); + readingHelper.initWithPtNodeArrayPos(dicNode->getChildrenPtNodeArrayPos()); const DynamicPatriciaTrieNodeReader *const nodeReader = readingHelper.getNodeReader(); while (!readingHelper.isEnd()) { bool isTerminal = nodeReader->isTerminal() && !nodeReader->isDeleted(); @@ -114,7 +114,7 @@ int DynamicPatriciaTriePolicy::getCodePointsAndProbabilityAndReturnCodePointCoun return codePointCount; } -int DynamicPatriciaTriePolicy::getTerminalNodePositionOfWord(const int *const inWord, +int DynamicPatriciaTriePolicy::getTerminalPtNodePositionOfWord(const int *const inWord, const int length, const bool forceLowerCaseSearch) const { int searchCodePoints[length]; for (int i = 0; i < length; ++i) { @@ -253,12 +253,12 @@ bool DynamicPatriciaTriePolicy::addBigramWords(const int *const word0, const int AKLOGE("The dictionary is too large to dynamically update."); return false; } - const int word0Pos = getTerminalNodePositionOfWord(word0, length0, + const int word0Pos = getTerminalPtNodePositionOfWord(word0, length0, false /* forceLowerCaseSearch */); if (word0Pos == NOT_A_DICT_POS) { return false; } - const int word1Pos = getTerminalNodePositionOfWord(word1, length1, + const int word1Pos = getTerminalPtNodePositionOfWord(word1, length1, false /* forceLowerCaseSearch */); if (word1Pos == NOT_A_DICT_POS) { return false; @@ -287,12 +287,12 @@ bool DynamicPatriciaTriePolicy::removeBigramWords(const int *const word0, const AKLOGE("The dictionary is too large to dynamically update."); return false; } - const int word0Pos = getTerminalNodePositionOfWord(word0, length0, + const int word0Pos = getTerminalPtNodePositionOfWord(word0, length0, false /* forceLowerCaseSearch */); if (word0Pos == NOT_A_DICT_POS) { return false; } - const int word1Pos = getTerminalNodePositionOfWord(word1, length1, + const int word1Pos = getTerminalPtNodePositionOfWord(word1, length1, false /* forceLowerCaseSearch */); if (word1Pos == NOT_A_DICT_POS) { return false; diff --git a/native/jni/src/suggest/policyimpl/dictionary/dynamic_patricia_trie_policy.h b/native/jni/src/suggest/policyimpl/dictionary/structure/v3/dynamic_patricia_trie_policy.h index be97ee1a5..1a924c177 100644 --- a/native/jni/src/suggest/policyimpl/dictionary/dynamic_patricia_trie_policy.h +++ b/native/jni/src/suggest/policyimpl/dictionary/structure/v3/dynamic_patricia_trie_policy.h @@ -23,6 +23,7 @@ #include "suggest/policyimpl/dictionary/header/header_policy.h" #include "suggest/policyimpl/dictionary/shortcut/dynamic_shortcut_list_policy.h" #include "suggest/policyimpl/dictionary/utils/buffer_with_extendable_buffer.h" +#include "suggest/policyimpl/dictionary/utils/format_utils.h" #include "suggest/policyimpl/dictionary/utils/mmapped_buffer.h" namespace latinime { @@ -33,9 +34,10 @@ class DicNodeVector; class DynamicPatriciaTriePolicy : public DictionaryStructureWithBufferPolicy { public: DynamicPatriciaTriePolicy(const MmappedBuffer *const buffer) - : mBuffer(buffer), mHeaderPolicy(mBuffer->getBuffer(), buffer->getBufferSize()), + : mBuffer(buffer), mHeaderPolicy(mBuffer->getBuffer(), FormatUtils::VERSION_3), mBufferWithExtendableBuffer(mBuffer->getBuffer() + mHeaderPolicy.getSize(), - mBuffer->getBufferSize() - mHeaderPolicy.getSize()), + mBuffer->getBufferSize() - mHeaderPolicy.getSize(), + BufferWithExtendableBuffer::DEFAULT_MAX_ADDITIONAL_BUFFER_SIZE), mShortcutListPolicy(&mBufferWithExtendableBuffer), mBigramListPolicy(&mHeaderPolicy, &mBufferWithExtendableBuffer, &mShortcutListPolicy, mHeaderPolicy.isDecayingDict()), @@ -50,14 +52,14 @@ class DynamicPatriciaTriePolicy : public DictionaryStructureWithBufferPolicy { return 0; } - void createAndGetAllChildNodes(const DicNode *const dicNode, + void createAndGetAllChildDicNodes(const DicNode *const dicNode, DicNodeVector *const childDicNodes) const; int getCodePointsAndProbabilityAndReturnCodePointCount( const int terminalPtNodePos, const int maxCodePointCount, int *const outCodePoints, int *const outUnigramProbability) const; - int getTerminalNodePositionOfWord(const int *const inWord, + int getTerminalPtNodePositionOfWord(const int *const inWord, const int length, const bool forceLowerCaseSearch) const; int getProbability(const int unigramProbability, const int bigramProbability) const; diff --git a/native/jni/src/suggest/policyimpl/dictionary/dynamic_patricia_trie_reading_helper.cpp b/native/jni/src/suggest/policyimpl/dictionary/structure/v3/dynamic_patricia_trie_reading_helper.cpp index f108c219f..f3410affc 100644 --- a/native/jni/src/suggest/policyimpl/dictionary/dynamic_patricia_trie_reading_helper.cpp +++ b/native/jni/src/suggest/policyimpl/dictionary/structure/v3/dynamic_patricia_trie_reading_helper.cpp @@ -14,7 +14,7 @@ * limitations under the License. */ -#include "suggest/policyimpl/dictionary/dynamic_patricia_trie_reading_helper.h" +#include "suggest/policyimpl/dictionary/structure/v3/dynamic_patricia_trie_reading_helper.h" #include "suggest/policyimpl/dictionary/utils/buffer_with_extendable_buffer.h" @@ -22,7 +22,7 @@ namespace latinime { // To avoid infinite loop caused by invalid or malicious forward links. const int DynamicPatriciaTrieReadingHelper::MAX_CHILD_COUNT_TO_AVOID_INFINITE_LOOP = 100000; -const int DynamicPatriciaTrieReadingHelper::MAX_NODE_ARRAY_COUNT_TO_AVOID_INFINITE_LOOP = 100000; +const int DynamicPatriciaTrieReadingHelper::MAX_PT_NODE_ARRAY_COUNT_TO_AVOID_INFINITE_LOOP = 100000; const size_t DynamicPatriciaTrieReadingHelper::MAX_READING_STATE_STACK_SIZE = MAX_WORD_LENGTH; // Visits all PtNodes in post-order depth first manner. @@ -170,35 +170,41 @@ void DynamicPatriciaTrieReadingHelper::nextPtNodeArray() { mReadingState.mPos = NOT_A_DICT_POS; return; } - mReadingState.mPosOfLastPtNodeArrayHead = mReadingState.mPos; + mReadingState.mPosOfThisPtNodeArrayHead = mReadingState.mPos; const bool usesAdditionalBuffer = mBuffer->isInAdditionalBuffer(mReadingState.mPos); const uint8_t *const dictBuf = mBuffer->getBuffer(usesAdditionalBuffer); if (usesAdditionalBuffer) { mReadingState.mPos -= mBuffer->getOriginalBufferSize(); } - mReadingState.mNodeCount = PatriciaTrieReadingUtils::getPtNodeArraySizeAndAdvancePosition( - dictBuf, &mReadingState.mPos); + mReadingState.mRemainingPtNodeCountInThisArray = + PatriciaTrieReadingUtils::getPtNodeArraySizeAndAdvancePosition(dictBuf, + &mReadingState.mPos); if (usesAdditionalBuffer) { mReadingState.mPos += mBuffer->getOriginalBufferSize(); } // Count up nodes and node arrays to avoid infinite loop. - mReadingState.mTotalNodeCount += mReadingState.mNodeCount; - mReadingState.mNodeArrayCount++; - if (mReadingState.mNodeCount < 0 - || mReadingState.mTotalNodeCount > MAX_CHILD_COUNT_TO_AVOID_INFINITE_LOOP - || mReadingState.mNodeArrayCount > MAX_NODE_ARRAY_COUNT_TO_AVOID_INFINITE_LOOP) { + mReadingState.mTotalPtNodeIndexInThisArrayChain += + mReadingState.mRemainingPtNodeCountInThisArray; + mReadingState.mPtNodeArrayIndexInThisArrayChain++; + if (mReadingState.mRemainingPtNodeCountInThisArray < 0 + || mReadingState.mTotalPtNodeIndexInThisArrayChain + > MAX_CHILD_COUNT_TO_AVOID_INFINITE_LOOP + || mReadingState.mPtNodeArrayIndexInThisArrayChain + > MAX_PT_NODE_ARRAY_COUNT_TO_AVOID_INFINITE_LOOP) { // Invalid dictionary. AKLOGI("Invalid dictionary. nodeCount: %d, totalNodeCount: %d, MAX_CHILD_COUNT: %d" "nodeArrayCount: %d, MAX_NODE_ARRAY_COUNT: %d", - mReadingState.mNodeCount, mReadingState.mTotalNodeCount, - MAX_CHILD_COUNT_TO_AVOID_INFINITE_LOOP, mReadingState.mNodeArrayCount, - MAX_NODE_ARRAY_COUNT_TO_AVOID_INFINITE_LOOP); + mReadingState.mRemainingPtNodeCountInThisArray, + mReadingState.mTotalPtNodeIndexInThisArrayChain, + MAX_CHILD_COUNT_TO_AVOID_INFINITE_LOOP, + mReadingState.mPtNodeArrayIndexInThisArrayChain, + MAX_PT_NODE_ARRAY_COUNT_TO_AVOID_INFINITE_LOOP); ASSERT(false); mIsError = true; mReadingState.mPos = NOT_A_DICT_POS; return; } - if (mReadingState.mNodeCount == 0) { + if (mReadingState.mRemainingPtNodeCountInThisArray == 0) { // Empty node array. Try following forward link. followForwardLink(); } diff --git a/native/jni/src/suggest/policyimpl/dictionary/dynamic_patricia_trie_reading_helper.h b/native/jni/src/suggest/policyimpl/dictionary/structure/v3/dynamic_patricia_trie_reading_helper.h index a71c06971..f8d32c9cd 100644 --- a/native/jni/src/suggest/policyimpl/dictionary/dynamic_patricia_trie_reading_helper.h +++ b/native/jni/src/suggest/policyimpl/dictionary/structure/v3/dynamic_patricia_trie_reading_helper.h @@ -21,9 +21,9 @@ #include <vector> #include "defines.h" -#include "suggest/policyimpl/dictionary/dynamic_patricia_trie_node_reader.h" -#include "suggest/policyimpl/dictionary/dynamic_patricia_trie_reading_utils.h" -#include "suggest/policyimpl/dictionary/patricia_trie_reading_utils.h" +#include "suggest/policyimpl/dictionary/structure/v2/patricia_trie_reading_utils.h" +#include "suggest/policyimpl/dictionary/structure/v3/dynamic_patricia_trie_node_reader.h" +#include "suggest/policyimpl/dictionary/structure/v3/dynamic_patricia_trie_reading_utils.h" namespace latinime { @@ -84,9 +84,9 @@ class DynamicPatriciaTrieReadingHelper { } else { mIsError = false; mReadingState.mPos = ptNodeArrayPos; - mReadingState.mPrevTotalCodePointCount = 0; - mReadingState.mTotalNodeCount = 0; - mReadingState.mNodeArrayCount = 0; + mReadingState.mTotalCodePointCountSinceInitialization = 0; + mReadingState.mTotalPtNodeIndexInThisArrayChain = 0; + mReadingState.mPtNodeArrayIndexInThisArrayChain = 0; mReadingState.mPosOfLastForwardLinkField = NOT_A_DICT_POS; mReadingStateStack.clear(); nextPtNodeArray(); @@ -103,12 +103,12 @@ class DynamicPatriciaTrieReadingHelper { } else { mIsError = false; mReadingState.mPos = ptNodePos; - mReadingState.mNodeCount = 1; - mReadingState.mPrevTotalCodePointCount = 0; - mReadingState.mTotalNodeCount = 1; - mReadingState.mNodeArrayCount = 1; + mReadingState.mRemainingPtNodeCountInThisArray = 1; + mReadingState.mTotalCodePointCountSinceInitialization = 0; + mReadingState.mTotalPtNodeIndexInThisArrayChain = 1; + mReadingState.mPtNodeArrayIndexInThisArrayChain = 1; mReadingState.mPosOfLastForwardLinkField = NOT_A_DICT_POS; - mReadingState.mPosOfLastPtNodeArrayHead = NOT_A_DICT_POS; + mReadingState.mPosOfThisPtNodeArrayHead = NOT_A_DICT_POS; mReadingStateStack.clear(); fetchPtNodeInfo(); } @@ -128,12 +128,13 @@ class DynamicPatriciaTrieReadingHelper { // Return code point count exclude the last read node's code points. AK_FORCE_INLINE int getPrevTotalCodePointCount() const { - return mReadingState.mPrevTotalCodePointCount; + return mReadingState.mTotalCodePointCountSinceInitialization; } // Return code point count include the last read node's code points. AK_FORCE_INLINE int getTotalCodePointCount() const { - return mReadingState.mPrevTotalCodePointCount + mNodeReader.getCodePointCount(); + return mReadingState.mTotalCodePointCountSinceInitialization + + mNodeReader.getCodePointCount(); } AK_FORCE_INLINE void fetchMergedNodeCodePointsInReverseOrder( @@ -149,9 +150,9 @@ class DynamicPatriciaTrieReadingHelper { } AK_FORCE_INLINE void readNextSiblingNode() { - mReadingState.mNodeCount -= 1; + mReadingState.mRemainingPtNodeCountInThisArray -= 1; mReadingState.mPos = mNodeReader.getSiblingNodePos(); - if (mReadingState.mNodeCount <= 0) { + if (mReadingState.mRemainingPtNodeCountInThisArray <= 0) { // All nodes in the current node array have been read. followForwardLink(); if (!isEnd()) { @@ -165,9 +166,10 @@ class DynamicPatriciaTrieReadingHelper { // Read the first child node of the current node. AK_FORCE_INLINE void readChildNode() { if (mNodeReader.hasChildren()) { - mReadingState.mPrevTotalCodePointCount += mNodeReader.getCodePointCount(); - mReadingState.mTotalNodeCount = 0; - mReadingState.mNodeArrayCount = 0; + mReadingState.mTotalCodePointCountSinceInitialization += + mNodeReader.getCodePointCount(); + mReadingState.mTotalPtNodeIndexInThisArrayChain = 0; + mReadingState.mPtNodeArrayIndexInThisArrayChain = 0; mReadingState.mPos = mNodeReader.getChildrenPos(); mReadingState.mPosOfLastForwardLinkField = NOT_A_DICT_POS; // Read children node array. @@ -183,13 +185,14 @@ class DynamicPatriciaTrieReadingHelper { // Read the parent node of the current node. AK_FORCE_INLINE void readParentNode() { if (mNodeReader.getParentPos() != NOT_A_DICT_POS) { - mReadingState.mPrevTotalCodePointCount += mNodeReader.getCodePointCount(); - mReadingState.mTotalNodeCount = 1; - mReadingState.mNodeArrayCount = 1; - mReadingState.mNodeCount = 1; + mReadingState.mTotalCodePointCountSinceInitialization += + mNodeReader.getCodePointCount(); + mReadingState.mTotalPtNodeIndexInThisArrayChain = 1; + mReadingState.mPtNodeArrayIndexInThisArrayChain = 1; + mReadingState.mRemainingPtNodeCountInThisArray = 1; mReadingState.mPos = mNodeReader.getParentPos(); mReadingState.mPosOfLastForwardLinkField = NOT_A_DICT_POS; - mReadingState.mPosOfLastPtNodeArrayHead = NOT_A_DICT_POS; + mReadingState.mPosOfThisPtNodeArrayHead = NOT_A_DICT_POS; fetchPtNodeInfo(); } else { mReadingState.mPos = NOT_A_DICT_POS; @@ -201,7 +204,7 @@ class DynamicPatriciaTrieReadingHelper { } AK_FORCE_INLINE int getPosOfLastPtNodeArrayHead() const { - return mReadingState.mPosOfLastPtNodeArrayHead; + return mReadingState.mPosOfThisPtNodeArrayHead; } AK_FORCE_INLINE void reloadCurrentPtNodeInfo() { @@ -218,35 +221,41 @@ class DynamicPatriciaTrieReadingHelper { private: DISALLOW_COPY_AND_ASSIGN(DynamicPatriciaTrieReadingHelper); - class ReadingState { + // This class encapsulates the reading state of a position in the dictionary. It points at a + // specific PtNode in the dictionary. + class PtNodeReadingState { public: // Note that copy constructor and assignment operator are used for this class to use // std::vector. - ReadingState() : mPos(NOT_A_DICT_POS), mNodeCount(0), mPrevTotalCodePointCount(0), - mTotalNodeCount(0), mNodeArrayCount(0), mPosOfLastForwardLinkField(NOT_A_DICT_POS), - mPosOfLastPtNodeArrayHead(NOT_A_DICT_POS) {} + PtNodeReadingState() : mPos(NOT_A_DICT_POS), mRemainingPtNodeCountInThisArray(0), + mTotalCodePointCountSinceInitialization(0), mTotalPtNodeIndexInThisArrayChain(0), + mPtNodeArrayIndexInThisArrayChain(0), mPosOfLastForwardLinkField(NOT_A_DICT_POS), + mPosOfThisPtNodeArrayHead(NOT_A_DICT_POS) {} int mPos; - // Node count of a node array. - int mNodeCount; - int mPrevTotalCodePointCount; - int mTotalNodeCount; - int mNodeArrayCount; + // Remaining node count in the current array. + int mRemainingPtNodeCountInThisArray; + int mTotalCodePointCountSinceInitialization; + // Counter of PtNodes used to avoid infinite loops caused by broken or malicious links. + int mTotalPtNodeIndexInThisArrayChain; + // Counter of PtNode arrays used to avoid infinite loops caused by cyclic links of empty + // PtNode arrays. + int mPtNodeArrayIndexInThisArrayChain; int mPosOfLastForwardLinkField; - int mPosOfLastPtNodeArrayHead; + int mPosOfThisPtNodeArrayHead; }; static const int MAX_CHILD_COUNT_TO_AVOID_INFINITE_LOOP; - static const int MAX_NODE_ARRAY_COUNT_TO_AVOID_INFINITE_LOOP; + static const int MAX_PT_NODE_ARRAY_COUNT_TO_AVOID_INFINITE_LOOP; static const size_t MAX_READING_STATE_STACK_SIZE; // TODO: Introduce error code to track what caused the error. bool mIsError; - ReadingState mReadingState; + PtNodeReadingState mReadingState; const BufferWithExtendableBuffer *const mBuffer; DynamicPatriciaTrieNodeReader mNodeReader; int mMergedNodeCodePoints[MAX_WORD_LENGTH]; - std::vector<ReadingState> mReadingStateStack; + std::vector<PtNodeReadingState> mReadingStateStack; void nextPtNodeArray(); diff --git a/native/jni/src/suggest/policyimpl/dictionary/dynamic_patricia_trie_reading_utils.cpp b/native/jni/src/suggest/policyimpl/dictionary/structure/v3/dynamic_patricia_trie_reading_utils.cpp index d68446db6..e94925365 100644 --- a/native/jni/src/suggest/policyimpl/dictionary/dynamic_patricia_trie_reading_utils.cpp +++ b/native/jni/src/suggest/policyimpl/dictionary/structure/v3/dynamic_patricia_trie_reading_utils.cpp @@ -14,7 +14,7 @@ * limitations under the License. */ -#include "suggest/policyimpl/dictionary/dynamic_patricia_trie_reading_utils.h" +#include "suggest/policyimpl/dictionary/structure/v3/dynamic_patricia_trie_reading_utils.h" #include "defines.h" #include "suggest/policyimpl/dictionary/utils/byte_array_utils.h" diff --git a/native/jni/src/suggest/policyimpl/dictionary/dynamic_patricia_trie_reading_utils.h b/native/jni/src/suggest/policyimpl/dictionary/structure/v3/dynamic_patricia_trie_reading_utils.h index 67c3cc57e..67c3cc57e 100644 --- a/native/jni/src/suggest/policyimpl/dictionary/dynamic_patricia_trie_reading_utils.h +++ b/native/jni/src/suggest/policyimpl/dictionary/structure/v3/dynamic_patricia_trie_reading_utils.h diff --git a/native/jni/src/suggest/policyimpl/dictionary/dynamic_patricia_trie_writing_helper.cpp b/native/jni/src/suggest/policyimpl/dictionary/structure/v3/dynamic_patricia_trie_writing_helper.cpp index 052558bfc..d856c50f4 100644 --- a/native/jni/src/suggest/policyimpl/dictionary/dynamic_patricia_trie_writing_helper.cpp +++ b/native/jni/src/suggest/policyimpl/dictionary/structure/v3/dynamic_patricia_trie_writing_helper.cpp @@ -14,16 +14,16 @@ * limitations under the License. */ -#include "suggest/policyimpl/dictionary/dynamic_patricia_trie_writing_helper.h" +#include "suggest/policyimpl/dictionary/structure/v3/dynamic_patricia_trie_writing_helper.h" #include "suggest/policyimpl/dictionary/bigram/dynamic_bigram_list_policy.h" -#include "suggest/policyimpl/dictionary/dynamic_patricia_trie_gc_event_listeners.h" -#include "suggest/policyimpl/dictionary/dynamic_patricia_trie_node_reader.h" -#include "suggest/policyimpl/dictionary/dynamic_patricia_trie_reading_helper.h" -#include "suggest/policyimpl/dictionary/dynamic_patricia_trie_reading_utils.h" -#include "suggest/policyimpl/dictionary/dynamic_patricia_trie_writing_utils.h" +#include "suggest/policyimpl/dictionary/structure/v2/patricia_trie_reading_utils.h" +#include "suggest/policyimpl/dictionary/structure/v3/dynamic_patricia_trie_gc_event_listeners.h" +#include "suggest/policyimpl/dictionary/structure/v3/dynamic_patricia_trie_node_reader.h" +#include "suggest/policyimpl/dictionary/structure/v3/dynamic_patricia_trie_reading_helper.h" +#include "suggest/policyimpl/dictionary/structure/v3/dynamic_patricia_trie_reading_utils.h" +#include "suggest/policyimpl/dictionary/structure/v3/dynamic_patricia_trie_writing_utils.h" #include "suggest/policyimpl/dictionary/header/header_policy.h" -#include "suggest/policyimpl/dictionary/patricia_trie_reading_utils.h" #include "suggest/policyimpl/dictionary/shortcut/dynamic_shortcut_list_policy.h" #include "suggest/policyimpl/dictionary/utils/dict_file_writing_utils.h" #include "suggest/policyimpl/dictionary/utils/forgetting_curve_utils.h" @@ -149,7 +149,8 @@ bool DynamicPatriciaTrieWritingHelper::removeBigramWords(const int word0Pos, con void DynamicPatriciaTrieWritingHelper::writeToDictFile(const char *const fileName, const HeaderPolicy *const headerPolicy, const int unigramCount, const int bigramCount) { - BufferWithExtendableBuffer headerBuffer(0 /* originalBuffer */, 0 /* originalBufferSize */); + BufferWithExtendableBuffer headerBuffer( + BufferWithExtendableBuffer::DEFAULT_MAX_ADDITIONAL_BUFFER_SIZE); const int extendedRegionSize = headerPolicy->getExtendedRegionSize() + mBuffer->getUsedAdditionalBufferSize(); if (!headerPolicy->writeHeaderToBuffer(&headerBuffer, false /* updatesLastUpdatedTime */, @@ -161,8 +162,7 @@ void DynamicPatriciaTrieWritingHelper::writeToDictFile(const char *const fileNam void DynamicPatriciaTrieWritingHelper::writeToDictFileWithGC(const int rootPtNodeArrayPos, const char *const fileName, const HeaderPolicy *const headerPolicy) { - BufferWithExtendableBuffer newDictBuffer(0 /* originalBuffer */, 0 /* originalBufferSize */, - MAX_DICTIONARY_SIZE); + BufferWithExtendableBuffer newDictBuffer(MAX_DICTIONARY_SIZE); int unigramCount = 0; int bigramCount = 0; if (mNeedsToDecay) { @@ -171,7 +171,8 @@ void DynamicPatriciaTrieWritingHelper::writeToDictFileWithGC(const int rootPtNod if (!runGC(rootPtNodeArrayPos, headerPolicy, &newDictBuffer, &unigramCount, &bigramCount)) { return; } - BufferWithExtendableBuffer headerBuffer(0 /* originalBuffer */, 0 /* originalBufferSize */); + BufferWithExtendableBuffer headerBuffer( + BufferWithExtendableBuffer::DEFAULT_MAX_ADDITIONAL_BUFFER_SIZE); if (!headerPolicy->writeHeaderToBuffer(&headerBuffer, true /* updatesLastUpdatedTime */, mNeedsToDecay, unigramCount, bigramCount, 0 /* extendedRegionSize */)) { return; diff --git a/native/jni/src/suggest/policyimpl/dictionary/dynamic_patricia_trie_writing_helper.h b/native/jni/src/suggest/policyimpl/dictionary/structure/v3/dynamic_patricia_trie_writing_helper.h index ca8664729..ca8664729 100644 --- a/native/jni/src/suggest/policyimpl/dictionary/dynamic_patricia_trie_writing_helper.h +++ b/native/jni/src/suggest/policyimpl/dictionary/structure/v3/dynamic_patricia_trie_writing_helper.h diff --git a/native/jni/src/suggest/policyimpl/dictionary/dynamic_patricia_trie_writing_utils.cpp b/native/jni/src/suggest/policyimpl/dictionary/structure/v3/dynamic_patricia_trie_writing_utils.cpp index 30ff10cd6..67733660b 100644 --- a/native/jni/src/suggest/policyimpl/dictionary/dynamic_patricia_trie_writing_utils.cpp +++ b/native/jni/src/suggest/policyimpl/dictionary/structure/v3/dynamic_patricia_trie_writing_utils.cpp @@ -14,7 +14,7 @@ * limitations under the License. */ -#include "suggest/policyimpl/dictionary/dynamic_patricia_trie_writing_utils.h" +#include "suggest/policyimpl/dictionary/structure/v3/dynamic_patricia_trie_writing_utils.h" #include <cstddef> #include <cstdlib> diff --git a/native/jni/src/suggest/policyimpl/dictionary/dynamic_patricia_trie_writing_utils.h b/native/jni/src/suggest/policyimpl/dictionary/structure/v3/dynamic_patricia_trie_writing_utils.h index af76bc6b5..5654105ee 100644 --- a/native/jni/src/suggest/policyimpl/dictionary/dynamic_patricia_trie_writing_utils.h +++ b/native/jni/src/suggest/policyimpl/dictionary/structure/v3/dynamic_patricia_trie_writing_utils.h @@ -20,7 +20,7 @@ #include <cstddef> #include "defines.h" -#include "suggest/policyimpl/dictionary/dynamic_patricia_trie_reading_utils.h" +#include "suggest/policyimpl/dictionary/structure/v3/dynamic_patricia_trie_reading_utils.h" namespace latinime { diff --git a/native/jni/src/suggest/policyimpl/dictionary/structure/v4/ver4_patricia_trie_policy.cpp b/native/jni/src/suggest/policyimpl/dictionary/structure/v4/ver4_patricia_trie_policy.cpp new file mode 100644 index 000000000..b9ee4891c --- /dev/null +++ b/native/jni/src/suggest/policyimpl/dictionary/structure/v4/ver4_patricia_trie_policy.cpp @@ -0,0 +1,96 @@ +/* + * 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. + */ + +#include "suggest/policyimpl/dictionary/structure/v4/ver4_patricia_trie_policy.h" + +namespace latinime { + +void Ver4PatriciaTriePolicy::createAndGetAllChildDicNodes(const DicNode *const dicNode, + DicNodeVector *const childDicNodes) const { + // TODO: Implement. +} + +int Ver4PatriciaTriePolicy::getCodePointsAndProbabilityAndReturnCodePointCount( + const int ptNodePos, const int maxCodePointCount, int *const outCodePoints, + int *const outUnigramProbability) const { + // TODO: Implement. + return 0; +} + +int Ver4PatriciaTriePolicy::getTerminalPtNodePositionOfWord(const int *const inWord, + const int length, const bool forceLowerCaseSearch) const { + // TODO: Implement. + return NOT_A_DICT_POS; +} + +int Ver4PatriciaTriePolicy::getProbability(const int unigramProbability, + const int bigramProbability) const { + // TODO: Implement. + return NOT_A_PROBABILITY; +} + +int Ver4PatriciaTriePolicy::getUnigramProbabilityOfPtNode(const int ptNodePos) const { + // TODO: Implement. + return NOT_A_PROBABILITY; +} + +int Ver4PatriciaTriePolicy::getShortcutPositionOfPtNode(const int ptNodePos) const { + // TODO: Implement. + return NOT_A_DICT_POS; +} + +int Ver4PatriciaTriePolicy::getBigramsPositionOfPtNode(const int ptNodePos) const { + // TODO: Implement. + return NOT_A_DICT_POS; +} + +bool Ver4PatriciaTriePolicy::addUnigramWord(const int *const word, const int length, + const int probability) { + // TODO: Implement. + return false; +} + +bool Ver4PatriciaTriePolicy::addBigramWords(const int *const word0, const int length0, + const int *const word1, const int length1, const int probability) { + // TODO: Implement. + return false; +} + +bool Ver4PatriciaTriePolicy::removeBigramWords(const int *const word0, const int length0, + const int *const word1, const int length1) { + // TODO: Implement. + return false; +} + +void Ver4PatriciaTriePolicy::flush(const char *const filePath) { + // TODO: Implement. +} + +void Ver4PatriciaTriePolicy::flushWithGC(const char *const filePath) { + // TODO: Implement. +} + +bool Ver4PatriciaTriePolicy::needsToRunGC(const bool mindsBlockByGC) const { + // TODO: Implement. + return false; +} + +void Ver4PatriciaTriePolicy::getProperty(const char *const query, char *const outResult, + const int maxResultLength) { + // TODO: Implement. +} + +} // namespace latinime diff --git a/native/jni/src/suggest/policyimpl/dictionary/structure/v4/ver4_patricia_trie_policy.h b/native/jni/src/suggest/policyimpl/dictionary/structure/v4/ver4_patricia_trie_policy.h new file mode 100644 index 000000000..86a4f8bce --- /dev/null +++ b/native/jni/src/suggest/policyimpl/dictionary/structure/v4/ver4_patricia_trie_policy.h @@ -0,0 +1,88 @@ +/* + * 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. + */ + +#ifndef LATINIME_VER4_PATRICIA_TRIE_POLICY_H +#define LATINIME_VER4_PATRICIA_TRIE_POLICY_H + +#include "defines.h" +#include "suggest/core/policy/dictionary_structure_with_buffer_policy.h" + +namespace latinime { + +class DicNode; +class DicNodeVector; + +// TODO: Implement. +class Ver4PatriciaTriePolicy : public DictionaryStructureWithBufferPolicy { + public: + ~Ver4PatriciaTriePolicy() {} + + AK_FORCE_INLINE int getRootPosition() const { + return 0; + } + + void createAndGetAllChildDicNodes(const DicNode *const dicNode, + DicNodeVector *const childDicNodes) const; + + int getCodePointsAndProbabilityAndReturnCodePointCount( + const int terminalPtNodePos, const int maxCodePointCount, int *const outCodePoints, + int *const outUnigramProbability) const; + + int getTerminalPtNodePositionOfWord(const int *const inWord, + const int length, const bool forceLowerCaseSearch) const; + + int getProbability(const int unigramProbability, const int bigramProbability) const; + + int getUnigramProbabilityOfPtNode(const int ptNodePos) const; + + int getShortcutPositionOfPtNode(const int ptNodePos) const; + + int getBigramsPositionOfPtNode(const int ptNodePos) const; + + const DictionaryHeaderStructurePolicy *getHeaderStructurePolicy() const { + return 0; + } + + const DictionaryBigramsStructurePolicy *getBigramsStructurePolicy() const { + return 0; + } + + const DictionaryShortcutsStructurePolicy *getShortcutsStructurePolicy() const { + return 0; + } + + bool addUnigramWord(const int *const word, const int length, const int probability); + + bool addBigramWords(const int *const word0, const int length0, const int *const word1, + const int length1, const int probability); + + bool removeBigramWords(const int *const word0, const int length0, const int *const word1, + const int length1); + + void flush(const char *const filePath); + + void flushWithGC(const char *const filePath); + + bool needsToRunGC(const bool mindsBlockByGC) const; + + void getProperty(const char *const query, char *const outResult, + const int maxResultLength); + + private: + DISALLOW_IMPLICIT_CONSTRUCTORS(Ver4PatriciaTriePolicy); +}; +} // namespace latinime +#endif // LATINIME_VER4_PATRICIA_TRIE_POLICY_H diff --git a/native/jni/src/suggest/policyimpl/dictionary/utils/buffer_with_extendable_buffer.cpp b/native/jni/src/suggest/policyimpl/dictionary/utils/buffer_with_extendable_buffer.cpp index f692882f2..5032131ab 100644 --- a/native/jni/src/suggest/policyimpl/dictionary/utils/buffer_with_extendable_buffer.cpp +++ b/native/jni/src/suggest/policyimpl/dictionary/utils/buffer_with_extendable_buffer.cpp @@ -18,7 +18,7 @@ namespace latinime { -const size_t BufferWithExtendableBuffer::MAX_ADDITIONAL_BUFFER_SIZE = 1024 * 1024; +const size_t BufferWithExtendableBuffer::DEFAULT_MAX_ADDITIONAL_BUFFER_SIZE = 1024 * 1024; const int BufferWithExtendableBuffer::NEAR_BUFFER_LIMIT_THRESHOLD_PERCENTILE = 90; // TODO: Needs to allocate larger memory corresponding to the current vector size. const size_t BufferWithExtendableBuffer::EXTEND_ADDITIONAL_BUFFER_SIZE_STEP = 128 * 1024; diff --git a/native/jni/src/suggest/policyimpl/dictionary/utils/buffer_with_extendable_buffer.h b/native/jni/src/suggest/policyimpl/dictionary/utils/buffer_with_extendable_buffer.h index 9dc34823c..1e27a1bec 100644 --- a/native/jni/src/suggest/policyimpl/dictionary/utils/buffer_with_extendable_buffer.h +++ b/native/jni/src/suggest/policyimpl/dictionary/utils/buffer_with_extendable_buffer.h @@ -32,12 +32,20 @@ namespace latinime { // raw pointer but provides several methods that handle boundary checking for writing data. class BufferWithExtendableBuffer { public: + static const size_t DEFAULT_MAX_ADDITIONAL_BUFFER_SIZE; + BufferWithExtendableBuffer(uint8_t *const originalBuffer, const int originalBufferSize, - const int maxAdditionalBufferSize = MAX_ADDITIONAL_BUFFER_SIZE) + const int maxAdditionalBufferSize) : mOriginalBuffer(originalBuffer), mOriginalBufferSize(originalBufferSize), mAdditionalBuffer(EXTEND_ADDITIONAL_BUFFER_SIZE_STEP), mUsedAdditionalBufferSize(0), mMaxAdditionalBufferSize(maxAdditionalBufferSize) {} + // Without original buffer. + BufferWithExtendableBuffer(const int maxAdditionalBufferSize) + : mOriginalBuffer(0), mOriginalBufferSize(0), + mAdditionalBuffer(EXTEND_ADDITIONAL_BUFFER_SIZE_STEP), mUsedAdditionalBufferSize(0), + mMaxAdditionalBufferSize(maxAdditionalBufferSize) {} + AK_FORCE_INLINE int getTailPosition() const { return mOriginalBufferSize + mUsedAdditionalBufferSize; } @@ -86,7 +94,6 @@ class BufferWithExtendableBuffer { private: DISALLOW_COPY_AND_ASSIGN(BufferWithExtendableBuffer); - static const size_t MAX_ADDITIONAL_BUFFER_SIZE; static const int NEAR_BUFFER_LIMIT_THRESHOLD_PERCENTILE; static const size_t EXTEND_ADDITIONAL_BUFFER_SIZE_STEP; diff --git a/native/jni/src/suggest/policyimpl/dictionary/utils/dict_file_writing_utils.cpp b/native/jni/src/suggest/policyimpl/dictionary/utils/dict_file_writing_utils.cpp index 994826fa8..b48e5b005 100644 --- a/native/jni/src/suggest/policyimpl/dictionary/utils/dict_file_writing_utils.cpp +++ b/native/jni/src/suggest/policyimpl/dictionary/utils/dict_file_writing_utils.cpp @@ -20,7 +20,7 @@ #include <cstring> #include "suggest/policyimpl/dictionary/header/header_policy.h" -#include "suggest/policyimpl/dictionary/dynamic_patricia_trie_writing_utils.h" +#include "suggest/policyimpl/dictionary/structure/v3/dynamic_patricia_trie_writing_utils.h" #include "suggest/policyimpl/dictionary/utils/buffer_with_extendable_buffer.h" #include "suggest/policyimpl/dictionary/utils/format_utils.h" @@ -33,6 +33,9 @@ const char *const DictFileWritingUtils::TEMP_FILE_SUFFIX_FOR_WRITING_DICT_FILE = switch (dictVersion) { case 3: return createEmptyV3DictFile(filePath, attributeMap); + case 4: + // TODO: Support version 4 dictionary format. + return false; default: // Only version 3 dictionary is supported for now. return false; @@ -41,12 +44,14 @@ const char *const DictFileWritingUtils::TEMP_FILE_SUFFIX_FOR_WRITING_DICT_FILE = /* static */ bool DictFileWritingUtils::createEmptyV3DictFile(const char *const filePath, const HeaderReadWriteUtils::AttributeMap *const attributeMap) { - BufferWithExtendableBuffer headerBuffer(0 /* originalBuffer */, 0 /* originalBufferSize */); + BufferWithExtendableBuffer headerBuffer( + BufferWithExtendableBuffer::DEFAULT_MAX_ADDITIONAL_BUFFER_SIZE); HeaderPolicy headerPolicy(FormatUtils::VERSION_3, attributeMap); headerPolicy.writeHeaderToBuffer(&headerBuffer, true /* updatesLastUpdatedTime */, true /* updatesLastDecayedTime */, 0 /* unigramCount */, 0 /* bigramCount */, 0 /* extendedRegionSize */); - BufferWithExtendableBuffer bodyBuffer(0 /* originalBuffer */, 0 /* originalBufferSize */); + BufferWithExtendableBuffer bodyBuffer( + BufferWithExtendableBuffer::DEFAULT_MAX_ADDITIONAL_BUFFER_SIZE); if (!DynamicPatriciaTrieWritingUtils::writeEmptyDictionary(&bodyBuffer, 0 /* rootPos */)) { return false; } diff --git a/native/jni/src/suggest/policyimpl/dictionary/utils/file_utils.h b/native/jni/src/suggest/policyimpl/dictionary/utils/file_utils.h new file mode 100644 index 000000000..59b894fa6 --- /dev/null +++ b/native/jni/src/suggest/policyimpl/dictionary/utils/file_utils.h @@ -0,0 +1,50 @@ +/* + * 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. + */ + +#ifndef LATINIME_FILE_UTILS_H +#define LATINIME_FILE_UTILS_H + +#include <sys/types.h> +#include <sys/stat.h> +#include <fcntl.h> +#include <unistd.h> + +#include "defines.h" + +namespace latinime { + +class FileUtils { + public: + // Returns -1 on error. + static int getFileSize(const char *const filePath) { + const int fd = open(filePath, O_RDONLY); + if (fd == -1) { + return -1; + } + struct stat statBuf; + if (fstat(fd, &statBuf) != 0) { + close(fd); + return -1; + } + close(fd); + return static_cast<int>(statBuf.st_size); + } + + private: + DISALLOW_IMPLICIT_CONSTRUCTORS(FileUtils); +}; +} // namespace latinime +#endif /* LATINIME_FILE_UTILS_H */ diff --git a/native/jni/src/suggest/policyimpl/dictionary/utils/format_utils.cpp b/native/jni/src/suggest/policyimpl/dictionary/utils/format_utils.cpp index 1d77d5c27..4843650ad 100644 --- a/native/jni/src/suggest/policyimpl/dictionary/utils/format_utils.cpp +++ b/native/jni/src/suggest/policyimpl/dictionary/utils/format_utils.cpp @@ -45,6 +45,8 @@ const int FormatUtils::DICTIONARY_MINIMUM_SIZE = 12; return VERSION_2; } else if (ByteArrayUtils::readUint16(dict, 4) == 3) { return VERSION_3; + } else if (ByteArrayUtils::readUint16(dict, 4) == 4) { + return VERSION_4; } else { return UNKNOWN_VERSION; } diff --git a/native/jni/src/suggest/policyimpl/dictionary/utils/format_utils.h b/native/jni/src/suggest/policyimpl/dictionary/utils/format_utils.h index 79ed0de29..b90393a53 100644 --- a/native/jni/src/suggest/policyimpl/dictionary/utils/format_utils.h +++ b/native/jni/src/suggest/policyimpl/dictionary/utils/format_utils.h @@ -31,6 +31,7 @@ class FormatUtils { enum FORMAT_VERSION { VERSION_2, VERSION_3, + VERSION_4, UNKNOWN_VERSION }; diff --git a/native/jni/src/suggest/policyimpl/dictionary/utils/mmapped_buffer.h b/native/jni/src/suggest/policyimpl/dictionary/utils/mmapped_buffer.h index 6b69116eb..82138355d 100644 --- a/native/jni/src/suggest/policyimpl/dictionary/utils/mmapped_buffer.h +++ b/native/jni/src/suggest/policyimpl/dictionary/utils/mmapped_buffer.h @@ -18,18 +18,21 @@ #define LATINIME_MMAPPED_BUFFER_H #include <cerrno> +#include <climits> +#include <cstdio> #include <fcntl.h> #include <stdint.h> #include <sys/mman.h> #include <unistd.h> #include "defines.h" +#include "suggest/policyimpl/dictionary/utils/file_utils.h" namespace latinime { class MmappedBuffer { public: - static MmappedBuffer* openBuffer(const char *const path, const int bufferOffset, + static MmappedBuffer *openBuffer(const char *const path, const int bufferOffset, const int bufferSize, const bool isUpdatable) { const int openMode = isUpdatable ? O_RDWR : O_RDONLY; const int mmapFd = open(path, openMode); @@ -59,7 +62,34 @@ class MmappedBuffer { isUpdatable); } + // Mmap entire file. + static MmappedBuffer *openBuffer(const char *const path, const bool isUpdatable) { + const int fileSize = FileUtils::getFileSize(path); + if (fileSize == -1) { + return 0; + } else if (fileSize == 0) { + return new MmappedBuffer(isUpdatable); + } else { + return openBuffer(path, 0 /* bufferOffset */, fileSize, isUpdatable); + } + } + + static MmappedBuffer *openBuffer(const char *const dirPath, const char *const fileName, + const bool isUpdatable) { + const int filePathBufferSize = PATH_MAX + 1 /* terminator */; + char filePath[filePathBufferSize]; + const int filePathLength = snprintf(filePath, filePathBufferSize, "%s%s", dirPath, + fileName); + if (filePathLength >= filePathBufferSize) { + return 0; + } + return openBuffer(filePath, isUpdatable); + } + ~MmappedBuffer() { + if (mAlignedSize == 0) { + return; + } int ret = munmap(mMmappedBuffer, mAlignedSize); if (ret != 0) { AKLOGE("DICT: Failure in munmap. ret=%d errno=%d", ret, errno); @@ -89,6 +119,11 @@ class MmappedBuffer { : mBuffer(buffer), mBufferSize(bufferSize), mMmappedBuffer(mmappedBuffer), mAlignedSize(alignedSize), mMmapFd(mmapFd), mIsUpdatable(isUpdatable) {} + // Empty file. We have to handle an empty file as a valid part of a dictionary. + AK_FORCE_INLINE MmappedBuffer(const bool isUpdatable) + : mBuffer(0), mBufferSize(0), mMmappedBuffer(0), mAlignedSize(0), mMmapFd(0), + mIsUpdatable(isUpdatable) {} + DISALLOW_IMPLICIT_CONSTRUCTORS(MmappedBuffer); uint8_t *const mBuffer; diff --git a/native/jni/src/suggest/policyimpl/typing/typing_traversal.h b/native/jni/src/suggest/policyimpl/typing/typing_traversal.h index 007c19e0a..fd0ac9eb6 100644 --- a/native/jni/src/suggest/policyimpl/typing/typing_traversal.h +++ b/native/jni/src/suggest/policyimpl/typing/typing_traversal.h @@ -81,7 +81,7 @@ class TypingTraversal : public Traversal { return false; } const int point0Index = dicNode->getInputIndex(0); - return dicNode->isTerminalWordNode() + return dicNode->isTerminalDicNode() && traverseSession->getProximityInfoState(0)-> hasSpaceProximity(point0Index); } @@ -96,7 +96,7 @@ class TypingTraversal : public Traversal { if (dicNode->isCompletion(inputSize)) { return false; } - if (!dicNode->isTerminalWordNode()) { + if (!dicNode->isTerminalDicNode()) { return false; } const int16_t pointIndex = dicNode->getInputIndex(0); diff --git a/native/jni/src/suggest/policyimpl/typing/typing_weighting.cpp b/native/jni/src/suggest/policyimpl/typing/typing_weighting.cpp index 5b6b5e874..54f65c786 100644 --- a/native/jni/src/suggest/policyimpl/typing/typing_weighting.cpp +++ b/native/jni/src/suggest/policyimpl/typing/typing_weighting.cpp @@ -23,39 +23,64 @@ namespace latinime { const TypingWeighting TypingWeighting::sInstance; -ErrorType TypingWeighting::getErrorType(const CorrectionType correctionType, +ErrorTypeUtils::ErrorType TypingWeighting::getErrorType(const CorrectionType correctionType, const DicTraverseSession *const traverseSession, const DicNode *const parentDicNode, const DicNode *const dicNode) const { switch (correctionType) { case CT_MATCH: if (isProximityDicNode(traverseSession, dicNode)) { - return ET_PROXIMITY_CORRECTION; + return ErrorTypeUtils::PROXIMITY_CORRECTION; + } else if (dicNode->isInDigraph()) { + return ErrorTypeUtils::MATCH_WITH_DIGRAPH; } else { - return ET_NOT_AN_ERROR; + // Compare the node code point with original primary code point on the keyboard. + const ProximityInfoState *const pInfoState = + traverseSession->getProximityInfoState(0); + const int primaryOriginalCodePoint = pInfoState->getPrimaryOriginalCodePointAt( + dicNode->getInputIndex(0)); + const int nodeCodePoint = dicNode->getNodeCodePoint(); + if (primaryOriginalCodePoint == nodeCodePoint) { + // Node code point is same as original code point on the keyboard. + return ErrorTypeUtils::NOT_AN_ERROR; + } else if (CharUtils::toLowerCase(primaryOriginalCodePoint) == + CharUtils::toLowerCase(nodeCodePoint)) { + // Only cases of the code points are different. + return ErrorTypeUtils::MATCH_WITH_CASE_ERROR; + } else if (CharUtils::toBaseCodePoint(primaryOriginalCodePoint) == + CharUtils::toBaseCodePoint(nodeCodePoint)) { + // Node code point is a variant of original code point. + return ErrorTypeUtils::MATCH_WITH_ACCENT_ERROR; + } else { + // Node code point is a variant of original code point and the cases are also + // different. + return ErrorTypeUtils::MATCH_WITH_ACCENT_ERROR + | ErrorTypeUtils::MATCH_WITH_CASE_ERROR; + } } + break; case CT_ADDITIONAL_PROXIMITY: - return ET_PROXIMITY_CORRECTION; + return ErrorTypeUtils::PROXIMITY_CORRECTION; case CT_OMISSION: if (parentDicNode->canBeIntentionalOmission()) { - return ET_INTENTIONAL_OMISSION; + return ErrorTypeUtils::INTENTIONAL_OMISSION; } else { - return ET_EDIT_CORRECTION; + return ErrorTypeUtils::EDIT_CORRECTION; } break; case CT_SUBSTITUTION: case CT_INSERTION: case CT_TERMINAL_INSERTION: case CT_TRANSPOSITION: - return ET_EDIT_CORRECTION; + return ErrorTypeUtils::EDIT_CORRECTION; case CT_NEW_WORD_SPACE_OMISSION: case CT_NEW_WORD_SPACE_SUBSTITUTION: - return ET_NEW_WORD; + return ErrorTypeUtils::NEW_WORD; case CT_TERMINAL: - return ET_NOT_AN_ERROR; + return ErrorTypeUtils::NOT_AN_ERROR; case CT_COMPLETION: - return ET_COMPLETION; + return ErrorTypeUtils::COMPLETION; default: - return ET_NOT_AN_ERROR; + return ErrorTypeUtils::NOT_AN_ERROR; } } } // namespace latinime diff --git a/native/jni/src/suggest/policyimpl/typing/typing_weighting.h b/native/jni/src/suggest/policyimpl/typing/typing_weighting.h index 9f0a331e3..41314ef52 100644 --- a/native/jni/src/suggest/policyimpl/typing/typing_weighting.h +++ b/native/jni/src/suggest/policyimpl/typing/typing_weighting.h @@ -19,6 +19,7 @@ #include "defines.h" #include "suggest/core/dicnode/dic_node_utils.h" +#include "suggest/core/dictionary/error_type_utils.h" #include "suggest/core/layout/touch_position_correction_utils.h" #include "suggest/core/policy/weighting.h" #include "suggest/core/session/dic_traverse_session.h" @@ -204,7 +205,7 @@ class TypingWeighting : public Weighting { return cost * traverseSession->getMultiWordCostMultiplier(); } - ErrorType getErrorType(const CorrectionType correctionType, + ErrorTypeUtils::ErrorType getErrorType(const CorrectionType correctionType, const DicTraverseSession *const traverseSession, const DicNode *const parentDicNode, const DicNode *const dicNode) const; |