diff options
Diffstat (limited to 'native/src/unigram_dictionary.cpp')
| -rw-r--r-- | native/src/unigram_dictionary.cpp | 124 |
1 files changed, 80 insertions, 44 deletions
diff --git a/native/src/unigram_dictionary.cpp b/native/src/unigram_dictionary.cpp index 17a87a708..3487d4f11 100644 --- a/native/src/unigram_dictionary.cpp +++ b/native/src/unigram_dictionary.cpp @@ -41,7 +41,8 @@ UnigramDictionary::UnigramDictionary(const unsigned char *dict, int typedLetterM MAX_PROXIMITY_CHARS(maxProximityChars), IS_LATEST_DICT_VERSION(isLatestDictVersion), TYPED_LETTER_MULTIPLIER(typedLetterMultiplier), FULL_WORD_MULTIPLIER(fullWordMultiplier), ROOT_POS(isLatestDictVersion ? DICTIONARY_HEADER_SIZE : 0), - BYTES_IN_ONE_CHAR(MAX_PROXIMITY_CHARS * sizeof(*mInputCodes)) { + BYTES_IN_ONE_CHAR(MAX_PROXIMITY_CHARS * sizeof(*mInputCodes)), + MAX_UMLAUT_SEARCH_DEPTH(DEFAULT_MAX_UMLAUT_SEARCH_DEPTH) { if (DEBUG_DICT) LOGI("UnigramDictionary - constructor"); } @@ -80,30 +81,37 @@ bool UnigramDictionary::isDigraph(const int* codes, const int i, const int codes void UnigramDictionary::getWordWithDigraphSuggestionsRec(const ProximityInfo *proximityInfo, const int *xcoordinates, const int* ycoordinates, const int *codesBuffer, const int codesBufferSize, const int flags, const int* codesSrc, const int codesRemain, - int* codesDest, unsigned short* outWords, int* frequencies) { - - for (int i = 0; i < codesRemain; ++i) { - if (isDigraph(codesSrc, i, codesRemain)) { - // Found a digraph. We will try both spellings. eg. the word is "pruefen" - - // Copy the word up to the first char of the digraph, then continue processing - // on the remaining part of the word, skipping the second char of the digraph. - // In our example, copy "pru" and continue running on "fen" - memcpy(codesDest, codesSrc, i * BYTES_IN_ONE_CHAR); - getWordWithDigraphSuggestionsRec(proximityInfo, xcoordinates, ycoordinates, codesBuffer, - codesBufferSize, flags, codesSrc + (i + 1) * MAX_PROXIMITY_CHARS, - codesRemain - i - 1, codesDest + i * MAX_PROXIMITY_CHARS, - outWords, frequencies); - - // Copy the second char of the digraph in place, then continue processing on - // the remaining part of the word. - // In our example, after "pru" in the buffer copy the "e", and continue running on "fen" - memcpy(codesDest + i * MAX_PROXIMITY_CHARS, codesSrc + i * MAX_PROXIMITY_CHARS, - BYTES_IN_ONE_CHAR); - getWordWithDigraphSuggestionsRec(proximityInfo, xcoordinates, ycoordinates, codesBuffer, - codesBufferSize, flags, codesSrc + i * MAX_PROXIMITY_CHARS, codesRemain - i, - codesDest + i * MAX_PROXIMITY_CHARS, outWords, frequencies); - return; + int currentDepth, int* codesDest, unsigned short* outWords, int* frequencies) { + + if (currentDepth < MAX_UMLAUT_SEARCH_DEPTH) { + for (int i = 0; i < codesRemain; ++i) { + if (isDigraph(codesSrc, i, codesRemain)) { + // Found a digraph. We will try both spellings. eg. the word is "pruefen" + + // Copy the word up to the first char of the digraph, then continue processing + // on the remaining part of the word, skipping the second char of the digraph. + // In our example, copy "pru" and continue running on "fen" + // Make i the index of the second char of the digraph for simplicity. Forgetting + // to do that results in an infinite recursion so take care! + ++i; + memcpy(codesDest, codesSrc, i * BYTES_IN_ONE_CHAR); + getWordWithDigraphSuggestionsRec(proximityInfo, xcoordinates, ycoordinates, + codesBuffer, codesBufferSize, flags, + codesSrc + (i + 1) * MAX_PROXIMITY_CHARS, codesRemain - i - 1, + currentDepth + 1, codesDest + i * MAX_PROXIMITY_CHARS, outWords, + frequencies); + + // Copy the second char of the digraph in place, then continue processing on + // the remaining part of the word. + // In our example, after "pru" in the buffer copy the "e", and continue on "fen" + memcpy(codesDest + i * MAX_PROXIMITY_CHARS, codesSrc + i * MAX_PROXIMITY_CHARS, + BYTES_IN_ONE_CHAR); + getWordWithDigraphSuggestionsRec(proximityInfo, xcoordinates, ycoordinates, + codesBuffer, codesBufferSize, flags, codesSrc + i * MAX_PROXIMITY_CHARS, + codesRemain - i, currentDepth + 1, codesDest + i * MAX_PROXIMITY_CHARS, + outWords, frequencies); + return; + } } } @@ -128,13 +136,13 @@ int UnigramDictionary::getSuggestions(const ProximityInfo *proximityInfo, const { // Incrementally tune the word and try all possibilities int codesBuffer[getCodesBufferSize(codes, codesSize, MAX_PROXIMITY_CHARS)]; getWordWithDigraphSuggestionsRec(proximityInfo, xcoordinates, ycoordinates, codesBuffer, - codesSize, flags, codes, codesSize, codesBuffer, outWords, frequencies); + codesSize, flags, codes, codesSize, 0, codesBuffer, outWords, frequencies); } else { // Normal processing getWordSuggestions(proximityInfo, xcoordinates, ycoordinates, codes, codesSize, outWords, frequencies); } - PROF_START(6); + PROF_START(20); // Get the word count int suggestedWordsCount = 0; while (suggestedWordsCount < MAX_WORDS && mFrequencies[suggestedWordsCount] > 0) { @@ -150,7 +158,7 @@ int UnigramDictionary::getSuggestions(const ProximityInfo *proximityInfo, const } } } - PROF_END(6); + PROF_END(20); PROF_CLOSE; return suggestedWordsCount; } @@ -164,12 +172,6 @@ void UnigramDictionary::getWordSuggestions(const ProximityInfo *proximityInfo, initSuggestions(codes, codesSize, outWords, frequencies); if (DEBUG_DICT) assert(codesSize == mInputLength); - if (DEBUG_PROXIMITY_INFO) { - for (int i = 0; i < codesSize; ++i) { - LOGI("Input[%d] x = %d, y = %d", i, xcoordinates[i], ycoordinates[i]); - } - } - const int MAX_DEPTH = min(mInputLength * MAX_DEPTH_MULTIPLIER, MAX_WORD_LENGTH); PROF_END(0); @@ -219,6 +221,25 @@ void UnigramDictionary::getWordSuggestions(const ProximityInfo *proximityInfo, } } PROF_END(5); + + PROF_START(6); + if (SUGGEST_WORDS_WITH_SPACE_PROXIMITY) { + // The first and last "mistyped spaces" are taken care of by excessive character handling + for (int i = 1; i < codesSize - 1; ++i) { + if (DEBUG_DICT) LOGI("--- Suggest words with proximity space %d", i); + const int x = xcoordinates[i]; + const int y = ycoordinates[i]; + if (DEBUG_PROXIMITY_INFO) + LOGI("Input[%d] x = %d, y = %d, has space proximity = %d", + i, x, y, proximityInfo->hasSpaceProximity(x, y)); + + if (proximityInfo->hasSpaceProximity(x, y)) { + getMistypedSpaceWords(mInputLength, i); + } + + } + } + PROF_END(6); } void UnigramDictionary::initSuggestions(const int *codes, const int codesSize, @@ -379,27 +400,31 @@ inline static void multiplyRate(const int rate, int *freq) { } } -bool UnigramDictionary::getMissingSpaceWords(const int inputLength, const int missingSpacePos) { - if (missingSpacePos <= 0 || missingSpacePos >= inputLength - || inputLength >= MAX_WORD_LENGTH) return false; - const int newWordLength = inputLength + 1; +bool UnigramDictionary::getSplitTwoWordsSuggestion(const int inputLength, + const int firstWordStartPos, const int firstWordLength, const int secondWordStartPos, + const int secondWordLength) { + if (inputLength >= MAX_WORD_LENGTH) return false; + if (0 >= firstWordLength || 0 >= secondWordLength || firstWordStartPos >= secondWordStartPos + || firstWordStartPos < 0 || secondWordStartPos >= inputLength) + return false; + const int newWordLength = firstWordLength + secondWordLength + 1; // Allocating variable length array on stack unsigned short word[newWordLength]; - const int firstFreq = getBestWordFreq(0, missingSpacePos, mWord); + const int firstFreq = getBestWordFreq(firstWordStartPos, firstWordLength, mWord); if (DEBUG_DICT) LOGI("First freq: %d", firstFreq); if (firstFreq <= 0) return false; - for (int i = 0; i < missingSpacePos; ++i) { + for (int i = 0; i < firstWordLength; ++i) { word[i] = mWord[i]; } - const int secondFreq = getBestWordFreq(missingSpacePos, inputLength - missingSpacePos, mWord); + const int secondFreq = getBestWordFreq(secondWordStartPos, secondWordLength, mWord); if (DEBUG_DICT) LOGI("Second freq: %d", secondFreq); if (secondFreq <= 0) return false; - word[missingSpacePos] = SPACE; - for (int i = (missingSpacePos + 1); i < newWordLength; ++i) { - word[i] = mWord[i - missingSpacePos - 1]; + word[firstWordLength] = SPACE; + for (int i = (firstWordLength + 1); i < newWordLength; ++i) { + word[i] = mWord[i - firstWordLength - 1]; } int pairFreq = ((firstFreq + secondFreq) / 2); @@ -409,6 +434,17 @@ bool UnigramDictionary::getMissingSpaceWords(const int inputLength, const int mi return true; } +bool UnigramDictionary::getMissingSpaceWords(const int inputLength, const int missingSpacePos) { + return getSplitTwoWordsSuggestion( + inputLength, 0, missingSpacePos, missingSpacePos, inputLength - missingSpacePos); +} + +bool UnigramDictionary::getMistypedSpaceWords(const int inputLength, const int spaceProximityPos) { + return getSplitTwoWordsSuggestion( + inputLength, 0, spaceProximityPos, spaceProximityPos + 1, + inputLength - spaceProximityPos - 1); +} + // Keep this for comparing spec to new getWords void UnigramDictionary::getWordsOld(const int initialPos, const int inputLength, const int skipPos, const int excessivePos, const int transposedPos,int *nextLetters, |