2010-12-01 12:22:15 +00:00
|
|
|
/*
|
|
|
|
**
|
|
|
|
** Copyright 2010, 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.
|
|
|
|
*/
|
|
|
|
|
2010-12-06 08:38:58 +00:00
|
|
|
#include <assert.h>
|
2010-12-01 12:22:15 +00:00
|
|
|
#include <fcntl.h>
|
2010-12-06 12:28:24 +00:00
|
|
|
#include <stdio.h>
|
2010-12-01 12:22:15 +00:00
|
|
|
#include <string.h>
|
|
|
|
|
2010-12-02 05:53:24 +00:00
|
|
|
#define LOG_TAG "LatinIME: unigram_dictionary.cpp"
|
2010-12-01 12:22:15 +00:00
|
|
|
|
|
|
|
#include "basechars.h"
|
|
|
|
#include "char_utils.h"
|
2010-12-02 05:53:24 +00:00
|
|
|
#include "dictionary.h"
|
|
|
|
#include "unigram_dictionary.h"
|
2010-12-01 12:22:15 +00:00
|
|
|
|
|
|
|
namespace latinime {
|
|
|
|
|
2011-02-25 08:56:53 +00:00
|
|
|
const UnigramDictionary::digraph_t UnigramDictionary::GERMAN_UMLAUT_DIGRAPHS[] =
|
|
|
|
{ { 'a', 'e' },
|
|
|
|
{ 'o', 'e' },
|
|
|
|
{ 'u', 'e' } };
|
|
|
|
|
2010-12-02 05:53:24 +00:00
|
|
|
UnigramDictionary::UnigramDictionary(const unsigned char *dict, int typedLetterMultiplier,
|
2010-12-08 08:05:39 +00:00
|
|
|
int fullWordMultiplier, int maxWordLength, int maxWords, int maxProximityChars,
|
2010-12-02 09:11:54 +00:00
|
|
|
const bool isLatestDictVersion)
|
2011-02-10 11:53:58 +00:00
|
|
|
: DICT(dict), MAX_WORD_LENGTH(maxWordLength), MAX_WORDS(maxWords),
|
2010-12-08 08:05:39 +00:00
|
|
|
MAX_PROXIMITY_CHARS(maxProximityChars), IS_LATEST_DICT_VERSION(isLatestDictVersion),
|
|
|
|
TYPED_LETTER_MULTIPLIER(typedLetterMultiplier), FULL_WORD_MULTIPLIER(fullWordMultiplier),
|
2011-02-25 08:56:53 +00:00
|
|
|
ROOT_POS(isLatestDictVersion ? DICTIONARY_HEADER_SIZE : 0),
|
2011-03-04 03:17:48 +00:00
|
|
|
BYTES_IN_ONE_CHAR(MAX_PROXIMITY_CHARS * sizeof(*mInputCodes)),
|
|
|
|
MAX_UMLAUT_SEARCH_DEPTH(DEFAULT_MAX_UMLAUT_SEARCH_DEPTH) {
|
2010-12-09 13:08:33 +00:00
|
|
|
if (DEBUG_DICT) LOGI("UnigramDictionary - constructor");
|
2010-12-01 12:22:15 +00:00
|
|
|
}
|
|
|
|
|
2010-12-02 09:11:54 +00:00
|
|
|
UnigramDictionary::~UnigramDictionary() {}
|
2010-12-01 12:22:15 +00:00
|
|
|
|
2011-02-25 08:56:53 +00:00
|
|
|
static inline unsigned int getCodesBufferSize(const int* codes, const int codesSize,
|
|
|
|
const int MAX_PROXIMITY_CHARS) {
|
|
|
|
return sizeof(*codes) * MAX_PROXIMITY_CHARS * codesSize;
|
|
|
|
}
|
|
|
|
|
|
|
|
bool UnigramDictionary::isDigraph(const int* codes, const int i, const int codesSize) const {
|
|
|
|
|
|
|
|
// There can't be a digraph if we don't have at least 2 characters to examine
|
|
|
|
if (i + 2 > codesSize) return false;
|
|
|
|
|
|
|
|
// Search for the first char of some digraph
|
|
|
|
int lastDigraphIndex = -1;
|
|
|
|
const int thisChar = codes[i * MAX_PROXIMITY_CHARS];
|
|
|
|
for (lastDigraphIndex = sizeof(GERMAN_UMLAUT_DIGRAPHS) / sizeof(GERMAN_UMLAUT_DIGRAPHS[0]) - 1;
|
|
|
|
lastDigraphIndex >= 0; --lastDigraphIndex) {
|
|
|
|
if (thisChar == GERMAN_UMLAUT_DIGRAPHS[lastDigraphIndex].first) break;
|
|
|
|
}
|
|
|
|
// No match: return early
|
|
|
|
if (lastDigraphIndex < 0) return false;
|
|
|
|
|
|
|
|
// It's an interesting digraph if the second char matches too.
|
|
|
|
return GERMAN_UMLAUT_DIGRAPHS[lastDigraphIndex].second == codes[(i + 1) * MAX_PROXIMITY_CHARS];
|
|
|
|
}
|
|
|
|
|
|
|
|
// Mostly the same arguments as the non-recursive version, except:
|
|
|
|
// codes is the original value. It points to the start of the work buffer, and gets passed as is.
|
|
|
|
// codesSize is the size of the user input (thus, it is the size of codesSrc).
|
|
|
|
// codesDest is the current point in the work buffer.
|
|
|
|
// codesSrc is the current point in the user-input, original, content-unmodified buffer.
|
|
|
|
// codesRemain is the remaining size in codesSrc.
|
|
|
|
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,
|
2011-03-05 06:50:19 +00:00
|
|
|
const int currentDepth, int* codesDest, unsigned short* outWords, int* frequencies) {
|
2011-03-04 03:17:48 +00:00
|
|
|
|
|
|
|
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;
|
|
|
|
}
|
2011-02-25 08:56:53 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
// If we come here, we hit the end of the word: let's check it against the dictionary.
|
|
|
|
// In our example, we'll come here once for "prufen" and then once for "pruefen".
|
|
|
|
// If the word contains several digraphs, we'll come it for the product of them.
|
|
|
|
// eg. if the word is "ueberpruefen" we'll test, in order, against
|
|
|
|
// "uberprufen", "uberpruefen", "ueberprufen", "ueberpruefen".
|
|
|
|
const unsigned int remainingBytes = BYTES_IN_ONE_CHAR * codesRemain;
|
|
|
|
if (0 != remainingBytes)
|
|
|
|
memcpy(codesDest, codesSrc, remainingBytes);
|
|
|
|
|
|
|
|
getWordSuggestions(proximityInfo, xcoordinates, ycoordinates, codesBuffer,
|
|
|
|
(codesDest - codesBuffer) / MAX_PROXIMITY_CHARS + codesRemain, outWords, frequencies);
|
|
|
|
}
|
|
|
|
|
|
|
|
int UnigramDictionary::getSuggestions(const ProximityInfo *proximityInfo, const int *xcoordinates,
|
|
|
|
const int *ycoordinates, const int *codes, const int codesSize, const int flags,
|
|
|
|
unsigned short *outWords, int *frequencies) {
|
|
|
|
|
|
|
|
if (REQUIRES_GERMAN_UMLAUT_PROCESSING & flags)
|
|
|
|
{ // Incrementally tune the word and try all possibilities
|
|
|
|
int codesBuffer[getCodesBufferSize(codes, codesSize, MAX_PROXIMITY_CHARS)];
|
|
|
|
getWordWithDigraphSuggestionsRec(proximityInfo, xcoordinates, ycoordinates, codesBuffer,
|
2011-03-04 03:17:48 +00:00
|
|
|
codesSize, flags, codes, codesSize, 0, codesBuffer, outWords, frequencies);
|
2011-02-25 08:56:53 +00:00
|
|
|
} else { // Normal processing
|
|
|
|
getWordSuggestions(proximityInfo, xcoordinates, ycoordinates, codes, codesSize,
|
|
|
|
outWords, frequencies);
|
|
|
|
}
|
|
|
|
|
2011-03-04 14:06:45 +00:00
|
|
|
PROF_START(20);
|
2011-02-25 08:56:53 +00:00
|
|
|
// Get the word count
|
|
|
|
int suggestedWordsCount = 0;
|
|
|
|
while (suggestedWordsCount < MAX_WORDS && mFrequencies[suggestedWordsCount] > 0) {
|
|
|
|
suggestedWordsCount++;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (DEBUG_DICT) {
|
|
|
|
LOGI("Returning %d words", suggestedWordsCount);
|
|
|
|
LOGI("Next letters: ");
|
|
|
|
for (int k = 0; k < NEXT_LETTERS_SIZE; k++) {
|
|
|
|
if (mNextLettersFrequency[k] > 0) {
|
|
|
|
LOGI("%c = %d,", k, mNextLettersFrequency[k]);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
2011-03-04 14:06:45 +00:00
|
|
|
PROF_END(20);
|
2011-02-25 08:56:53 +00:00
|
|
|
PROF_CLOSE;
|
|
|
|
return suggestedWordsCount;
|
|
|
|
}
|
|
|
|
|
|
|
|
void UnigramDictionary::getWordSuggestions(const ProximityInfo *proximityInfo,
|
|
|
|
const int *xcoordinates, const int *ycoordinates, const int *codes, const int codesSize,
|
|
|
|
unsigned short *outWords, int *frequencies) {
|
|
|
|
|
2011-01-05 05:13:07 +00:00
|
|
|
PROF_OPEN;
|
|
|
|
PROF_START(0);
|
2010-12-01 12:22:15 +00:00
|
|
|
initSuggestions(codes, codesSize, outWords, frequencies);
|
2010-12-13 05:42:35 +00:00
|
|
|
if (DEBUG_DICT) assert(codesSize == mInputLength);
|
|
|
|
|
2010-12-09 13:08:33 +00:00
|
|
|
const int MAX_DEPTH = min(mInputLength * MAX_DEPTH_MULTIPLIER, MAX_WORD_LENGTH);
|
2011-01-05 05:13:07 +00:00
|
|
|
PROF_END(0);
|
|
|
|
|
|
|
|
PROF_START(1);
|
2011-02-10 11:53:58 +00:00
|
|
|
getSuggestionCandidates(-1, -1, -1, mNextLettersFrequency, NEXT_LETTERS_SIZE, MAX_DEPTH);
|
2011-01-05 05:13:07 +00:00
|
|
|
PROF_END(1);
|
2010-12-01 12:22:15 +00:00
|
|
|
|
2011-01-05 05:13:07 +00:00
|
|
|
PROF_START(2);
|
2010-12-08 08:05:39 +00:00
|
|
|
// Suggestion with missing character
|
|
|
|
if (SUGGEST_WORDS_WITH_MISSING_CHARACTER) {
|
2010-12-01 12:22:15 +00:00
|
|
|
for (int i = 0; i < codesSize; ++i) {
|
2010-12-08 07:04:16 +00:00
|
|
|
if (DEBUG_DICT) LOGI("--- Suggest missing characters %d", i);
|
2010-12-13 05:42:35 +00:00
|
|
|
getSuggestionCandidates(i, -1, -1, NULL, 0, MAX_DEPTH);
|
2010-12-08 07:04:16 +00:00
|
|
|
}
|
|
|
|
}
|
2011-01-05 05:13:07 +00:00
|
|
|
PROF_END(2);
|
2010-12-08 07:04:16 +00:00
|
|
|
|
2011-01-05 05:13:07 +00:00
|
|
|
PROF_START(3);
|
2010-12-08 08:05:39 +00:00
|
|
|
// Suggestion with excessive character
|
2010-12-13 05:42:35 +00:00
|
|
|
if (SUGGEST_WORDS_WITH_EXCESSIVE_CHARACTER
|
|
|
|
&& mInputLength >= MIN_USER_TYPED_LENGTH_FOR_EXCESSIVE_CHARACTER_SUGGESTION) {
|
2010-12-08 07:04:16 +00:00
|
|
|
for (int i = 0; i < codesSize; ++i) {
|
2010-12-13 05:42:35 +00:00
|
|
|
if (DEBUG_DICT) LOGI("--- Suggest excessive characters %d", i);
|
|
|
|
getSuggestionCandidates(-1, i, -1, NULL, 0, MAX_DEPTH);
|
2010-12-01 12:22:15 +00:00
|
|
|
}
|
|
|
|
}
|
2011-01-05 05:13:07 +00:00
|
|
|
PROF_END(3);
|
2010-12-01 12:22:15 +00:00
|
|
|
|
2011-01-05 05:13:07 +00:00
|
|
|
PROF_START(4);
|
2010-12-09 13:08:33 +00:00
|
|
|
// Suggestion with transposed characters
|
|
|
|
// Only suggest words that length is mInputLength
|
|
|
|
if (SUGGEST_WORDS_WITH_TRANSPOSED_CHARACTERS) {
|
|
|
|
for (int i = 0; i < codesSize; ++i) {
|
|
|
|
if (DEBUG_DICT) LOGI("--- Suggest transposed characters %d", i);
|
2010-12-13 05:42:35 +00:00
|
|
|
getSuggestionCandidates(-1, -1, i, NULL, 0, mInputLength - 1);
|
2010-12-09 13:08:33 +00:00
|
|
|
}
|
|
|
|
}
|
2011-01-05 05:13:07 +00:00
|
|
|
PROF_END(4);
|
2010-12-09 13:08:33 +00:00
|
|
|
|
2011-01-05 05:13:07 +00:00
|
|
|
PROF_START(5);
|
2010-12-08 08:05:39 +00:00
|
|
|
// Suggestions with missing space
|
2010-12-13 05:42:35 +00:00
|
|
|
if (SUGGEST_WORDS_WITH_MISSING_SPACE_CHARACTER
|
|
|
|
&& mInputLength >= MIN_USER_TYPED_LENGTH_FOR_MISSING_SPACE_SUGGESTION) {
|
2010-12-08 08:05:39 +00:00
|
|
|
for (int i = 1; i < codesSize; ++i) {
|
|
|
|
if (DEBUG_DICT) LOGI("--- Suggest missing space characters %d", i);
|
|
|
|
getMissingSpaceWords(mInputLength, i);
|
|
|
|
}
|
|
|
|
}
|
2011-01-05 05:13:07 +00:00
|
|
|
PROF_END(5);
|
2011-03-04 14:06:45 +00:00
|
|
|
|
|
|
|
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);
|
2010-12-01 12:22:15 +00:00
|
|
|
}
|
|
|
|
|
2011-02-25 08:56:53 +00:00
|
|
|
void UnigramDictionary::initSuggestions(const int *codes, const int codesSize,
|
|
|
|
unsigned short *outWords, int *frequencies) {
|
2010-12-06 12:28:24 +00:00
|
|
|
if (DEBUG_DICT) LOGI("initSuggest");
|
2010-12-01 12:22:15 +00:00
|
|
|
mFrequencies = frequencies;
|
|
|
|
mOutputChars = outWords;
|
|
|
|
mInputCodes = codes;
|
|
|
|
mInputLength = codesSize;
|
|
|
|
mMaxEditDistance = mInputLength < 5 ? 2 : mInputLength / 2;
|
|
|
|
}
|
|
|
|
|
2010-12-02 11:19:59 +00:00
|
|
|
void UnigramDictionary::registerNextLetter(
|
|
|
|
unsigned short c, int *nextLetters, int nextLettersSize) {
|
2010-12-01 12:22:15 +00:00
|
|
|
if (c < nextLettersSize) {
|
|
|
|
nextLetters[c]++;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2010-12-08 08:05:39 +00:00
|
|
|
// TODO: We need to optimize addWord by using STL or something
|
2010-12-03 07:39:16 +00:00
|
|
|
bool UnigramDictionary::addWord(unsigned short *word, int length, int frequency) {
|
2010-12-01 12:22:15 +00:00
|
|
|
word[length] = 0;
|
2010-12-08 08:05:39 +00:00
|
|
|
if (DEBUG_DICT && DEBUG_SHOW_FOUND_WORD) {
|
2010-12-01 12:22:15 +00:00
|
|
|
char s[length + 1];
|
|
|
|
for (int i = 0; i <= length; i++) s[i] = word[i];
|
2010-12-08 08:05:39 +00:00
|
|
|
LOGI("Found word = %s, freq = %d", s, frequency);
|
2010-12-01 12:22:15 +00:00
|
|
|
}
|
2010-12-06 12:28:24 +00:00
|
|
|
if (length > MAX_WORD_LENGTH) {
|
|
|
|
if (DEBUG_DICT) LOGI("Exceeded max word length.");
|
|
|
|
return false;
|
|
|
|
}
|
2010-12-01 12:22:15 +00:00
|
|
|
|
|
|
|
// Find the right insertion point
|
|
|
|
int insertAt = 0;
|
|
|
|
while (insertAt < MAX_WORDS) {
|
2010-12-02 11:19:59 +00:00
|
|
|
if (frequency > mFrequencies[insertAt] || (mFrequencies[insertAt] == frequency
|
|
|
|
&& length < Dictionary::wideStrLen(mOutputChars + insertAt * MAX_WORD_LENGTH))) {
|
2010-12-01 12:22:15 +00:00
|
|
|
break;
|
|
|
|
}
|
|
|
|
insertAt++;
|
|
|
|
}
|
|
|
|
if (insertAt < MAX_WORDS) {
|
2010-12-08 07:04:16 +00:00
|
|
|
if (DEBUG_DICT) {
|
|
|
|
char s[length + 1];
|
|
|
|
for (int i = 0; i <= length; i++) s[i] = word[i];
|
2010-12-08 08:05:39 +00:00
|
|
|
LOGI("Added word = %s, freq = %d", s, frequency);
|
2010-12-08 07:04:16 +00:00
|
|
|
}
|
2010-12-01 12:22:15 +00:00
|
|
|
memmove((char*) mFrequencies + (insertAt + 1) * sizeof(mFrequencies[0]),
|
|
|
|
(char*) mFrequencies + insertAt * sizeof(mFrequencies[0]),
|
|
|
|
(MAX_WORDS - insertAt - 1) * sizeof(mFrequencies[0]));
|
|
|
|
mFrequencies[insertAt] = frequency;
|
|
|
|
memmove((char*) mOutputChars + (insertAt + 1) * MAX_WORD_LENGTH * sizeof(short),
|
2010-12-02 11:19:59 +00:00
|
|
|
(char*) mOutputChars + insertAt * MAX_WORD_LENGTH * sizeof(short),
|
2010-12-01 12:22:15 +00:00
|
|
|
(MAX_WORDS - insertAt - 1) * sizeof(short) * MAX_WORD_LENGTH);
|
2010-12-02 11:19:59 +00:00
|
|
|
unsigned short *dest = mOutputChars + insertAt * MAX_WORD_LENGTH;
|
2010-12-01 12:22:15 +00:00
|
|
|
while (length--) {
|
|
|
|
*dest++ = *word++;
|
|
|
|
}
|
|
|
|
*dest = 0; // NULL terminate
|
2010-12-08 08:05:39 +00:00
|
|
|
if (DEBUG_DICT) LOGI("Added word at %d", insertAt);
|
2010-12-01 12:22:15 +00:00
|
|
|
return true;
|
|
|
|
}
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
|
2011-02-22 06:12:46 +00:00
|
|
|
unsigned short UnigramDictionary::toBaseLowerCase(unsigned short c) {
|
2010-12-01 12:22:15 +00:00
|
|
|
if (c < sizeof(BASE_CHARS) / sizeof(BASE_CHARS[0])) {
|
|
|
|
c = BASE_CHARS[c];
|
|
|
|
}
|
|
|
|
if (c >='A' && c <= 'Z') {
|
|
|
|
c |= 32;
|
|
|
|
} else if (c > 127) {
|
|
|
|
c = latin_tolower(c);
|
|
|
|
}
|
|
|
|
return c;
|
|
|
|
}
|
|
|
|
|
2010-12-03 07:39:16 +00:00
|
|
|
bool UnigramDictionary::sameAsTyped(unsigned short *word, int length) {
|
2010-12-01 12:22:15 +00:00
|
|
|
if (length != mInputLength) {
|
|
|
|
return false;
|
|
|
|
}
|
2011-02-25 08:56:53 +00:00
|
|
|
const int *inputCodes = mInputCodes;
|
2010-12-01 12:22:15 +00:00
|
|
|
while (length--) {
|
|
|
|
if ((unsigned int) *inputCodes != (unsigned int) *word) {
|
|
|
|
return false;
|
|
|
|
}
|
2010-12-08 08:05:39 +00:00
|
|
|
inputCodes += MAX_PROXIMITY_CHARS;
|
2010-12-01 12:22:15 +00:00
|
|
|
word++;
|
|
|
|
}
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
|
2010-12-02 11:19:59 +00:00
|
|
|
static const char QUOTE = '\'';
|
2010-12-08 08:05:39 +00:00
|
|
|
static const char SPACE = ' ';
|
2010-12-01 12:22:15 +00:00
|
|
|
|
2010-12-13 05:42:35 +00:00
|
|
|
void UnigramDictionary::getSuggestionCandidates(const int skipPos,
|
2010-12-09 13:08:33 +00:00
|
|
|
const int excessivePos, const int transposedPos, int *nextLetters,
|
|
|
|
const int nextLettersSize, const int maxDepth) {
|
2010-12-13 05:42:35 +00:00
|
|
|
if (DEBUG_DICT) {
|
|
|
|
LOGI("getSuggestionCandidates %d", maxDepth);
|
|
|
|
assert(transposedPos + 1 < mInputLength);
|
|
|
|
assert(excessivePos < mInputLength);
|
|
|
|
assert(missingPos < mInputLength);
|
|
|
|
}
|
2010-12-08 08:05:39 +00:00
|
|
|
int rootPosition = ROOT_POS;
|
2010-12-07 04:08:39 +00:00
|
|
|
// Get the number of child of root, then increment the position
|
|
|
|
int childCount = Dictionary::getCount(DICT, &rootPosition);
|
|
|
|
int depth = 0;
|
|
|
|
|
|
|
|
mStackChildCount[0] = childCount;
|
|
|
|
mStackTraverseAll[0] = (mInputLength <= 0);
|
|
|
|
mStackNodeFreq[0] = 1;
|
|
|
|
mStackInputIndex[0] = 0;
|
|
|
|
mStackDiffs[0] = 0;
|
|
|
|
mStackSiblingPos[0] = rootPosition;
|
|
|
|
|
2010-12-08 08:05:39 +00:00
|
|
|
// Depth first search
|
2010-12-07 04:08:39 +00:00
|
|
|
while (depth >= 0) {
|
|
|
|
if (mStackChildCount[depth] > 0) {
|
|
|
|
--mStackChildCount[depth];
|
|
|
|
bool traverseAllNodes = mStackTraverseAll[depth];
|
2011-02-22 06:12:46 +00:00
|
|
|
int matchWeight = mStackNodeFreq[depth];
|
2010-12-07 04:08:39 +00:00
|
|
|
int inputIndex = mStackInputIndex[depth];
|
|
|
|
int diffs = mStackDiffs[depth];
|
|
|
|
int siblingPos = mStackSiblingPos[depth];
|
|
|
|
int firstChildPos;
|
2010-12-09 13:08:33 +00:00
|
|
|
// depth will never be greater than maxDepth because in that case,
|
2010-12-07 04:08:39 +00:00
|
|
|
// needsToTraverseChildrenNodes should be false
|
|
|
|
const bool needsToTraverseChildrenNodes = processCurrentNode(siblingPos, depth,
|
2011-02-22 06:12:46 +00:00
|
|
|
maxDepth, traverseAllNodes, matchWeight, inputIndex, diffs, skipPos,
|
|
|
|
excessivePos, transposedPos, nextLetters, nextLettersSize, &childCount,
|
|
|
|
&firstChildPos, &traverseAllNodes, &matchWeight, &inputIndex, &diffs,
|
|
|
|
&siblingPos);
|
2010-12-08 08:05:39 +00:00
|
|
|
// Update next sibling pos
|
2010-12-07 04:08:39 +00:00
|
|
|
mStackSiblingPos[depth] = siblingPos;
|
|
|
|
if (needsToTraverseChildrenNodes) {
|
|
|
|
// Goes to child node
|
|
|
|
++depth;
|
|
|
|
mStackChildCount[depth] = childCount;
|
|
|
|
mStackTraverseAll[depth] = traverseAllNodes;
|
2011-02-22 06:12:46 +00:00
|
|
|
mStackNodeFreq[depth] = matchWeight;
|
2010-12-07 04:08:39 +00:00
|
|
|
mStackInputIndex[depth] = inputIndex;
|
|
|
|
mStackDiffs[depth] = diffs;
|
|
|
|
mStackSiblingPos[depth] = firstChildPos;
|
|
|
|
}
|
|
|
|
} else {
|
2010-12-08 07:04:16 +00:00
|
|
|
// Goes to parent sibling node
|
2010-12-07 04:08:39 +00:00
|
|
|
--depth;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2011-01-05 07:37:53 +00:00
|
|
|
inline static void multiplyRate(const int rate, int *freq) {
|
|
|
|
if (rate > 1000000) {
|
|
|
|
*freq = (*freq / 100) * rate;
|
|
|
|
} else {
|
|
|
|
*freq = *freq * rate / 100;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2011-03-04 14:06:45 +00:00
|
|
|
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
|
2011-03-05 06:50:19 +00:00
|
|
|
|| firstWordStartPos < 0 || secondWordStartPos + secondWordLength > inputLength)
|
2011-03-04 14:06:45 +00:00
|
|
|
return false;
|
|
|
|
const int newWordLength = firstWordLength + secondWordLength + 1;
|
2010-12-08 08:05:39 +00:00
|
|
|
// Allocating variable length array on stack
|
|
|
|
unsigned short word[newWordLength];
|
2011-03-04 14:06:45 +00:00
|
|
|
const int firstFreq = getBestWordFreq(firstWordStartPos, firstWordLength, mWord);
|
2010-12-09 10:21:51 +00:00
|
|
|
if (DEBUG_DICT) LOGI("First freq: %d", firstFreq);
|
|
|
|
if (firstFreq <= 0) return false;
|
|
|
|
|
2011-03-04 14:06:45 +00:00
|
|
|
for (int i = 0; i < firstWordLength; ++i) {
|
2010-12-09 10:21:51 +00:00
|
|
|
word[i] = mWord[i];
|
2010-12-08 08:05:39 +00:00
|
|
|
}
|
2010-12-09 10:21:51 +00:00
|
|
|
|
2011-03-04 14:06:45 +00:00
|
|
|
const int secondFreq = getBestWordFreq(secondWordStartPos, secondWordLength, mWord);
|
2010-12-09 13:08:33 +00:00
|
|
|
if (DEBUG_DICT) LOGI("Second freq: %d", secondFreq);
|
2010-12-09 10:21:51 +00:00
|
|
|
if (secondFreq <= 0) return false;
|
|
|
|
|
2011-03-04 14:06:45 +00:00
|
|
|
word[firstWordLength] = SPACE;
|
|
|
|
for (int i = (firstWordLength + 1); i < newWordLength; ++i) {
|
|
|
|
word[i] = mWord[i - firstWordLength - 1];
|
2010-12-08 08:05:39 +00:00
|
|
|
}
|
2010-12-09 10:21:51 +00:00
|
|
|
|
|
|
|
int pairFreq = ((firstFreq + secondFreq) / 2);
|
|
|
|
for (int i = 0; i < inputLength; ++i) pairFreq *= TYPED_LETTER_MULTIPLIER;
|
2011-01-05 07:37:53 +00:00
|
|
|
multiplyRate(WORDS_WITH_MISSING_SPACE_CHARACTER_DEMOTION_RATE, &pairFreq);
|
2010-12-08 08:05:39 +00:00
|
|
|
addWord(word, newWordLength, pairFreq);
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
|
2011-03-04 14:06:45 +00:00
|
|
|
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);
|
|
|
|
}
|
|
|
|
|
2010-12-08 08:05:39 +00:00
|
|
|
// Keep this for comparing spec to new getWords
|
|
|
|
void UnigramDictionary::getWordsOld(const int initialPos, const int inputLength, const int skipPos,
|
2010-12-09 13:08:33 +00:00
|
|
|
const int excessivePos, const int transposedPos,int *nextLetters,
|
|
|
|
const int nextLettersSize) {
|
2010-12-08 08:05:39 +00:00
|
|
|
int initialPosition = initialPos;
|
|
|
|
const int count = Dictionary::getCount(DICT, &initialPosition);
|
|
|
|
getWordsRec(count, initialPosition, 0,
|
|
|
|
min(inputLength * MAX_DEPTH_MULTIPLIER, MAX_WORD_LENGTH),
|
2010-12-09 13:08:33 +00:00
|
|
|
mInputLength <= 0, 1, 0, 0, skipPos, excessivePos, transposedPos, nextLetters,
|
|
|
|
nextLettersSize);
|
2010-12-08 08:05:39 +00:00
|
|
|
}
|
|
|
|
|
2010-12-03 10:38:08 +00:00
|
|
|
void UnigramDictionary::getWordsRec(const int childrenCount, const int pos, const int depth,
|
2011-02-22 06:12:46 +00:00
|
|
|
const int maxDepth, const bool traverseAllNodes, const int matchWeight,
|
|
|
|
const int inputIndex, const int diffs, const int skipPos, const int excessivePos,
|
|
|
|
const int transposedPos, int *nextLetters, const int nextLettersSize) {
|
2010-12-06 08:38:58 +00:00
|
|
|
int siblingPos = pos;
|
2010-12-03 10:38:08 +00:00
|
|
|
for (int i = 0; i < childrenCount; ++i) {
|
2010-12-06 08:38:58 +00:00
|
|
|
int newCount;
|
|
|
|
int newChildPosition;
|
2010-12-07 04:08:39 +00:00
|
|
|
const int newDepth = depth + 1;
|
2010-12-06 08:38:58 +00:00
|
|
|
bool newTraverseAllNodes;
|
2011-02-22 06:12:46 +00:00
|
|
|
int newMatchRate;
|
2010-12-06 08:38:58 +00:00
|
|
|
int newInputIndex;
|
|
|
|
int newDiffs;
|
|
|
|
int newSiblingPos;
|
|
|
|
const bool needsToTraverseChildrenNodes = processCurrentNode(siblingPos, depth, maxDepth,
|
2011-02-22 06:12:46 +00:00
|
|
|
traverseAllNodes, matchWeight, inputIndex, diffs,
|
|
|
|
skipPos, excessivePos, transposedPos,
|
2010-12-09 13:08:33 +00:00
|
|
|
nextLetters, nextLettersSize,
|
2011-02-22 06:12:46 +00:00
|
|
|
&newCount, &newChildPosition, &newTraverseAllNodes, &newMatchRate,
|
2010-12-06 08:38:58 +00:00
|
|
|
&newInputIndex, &newDiffs, &newSiblingPos);
|
|
|
|
siblingPos = newSiblingPos;
|
|
|
|
|
|
|
|
if (needsToTraverseChildrenNodes) {
|
|
|
|
getWordsRec(newCount, newChildPosition, newDepth, maxDepth, newTraverseAllNodes,
|
2011-02-22 06:12:46 +00:00
|
|
|
newMatchRate, newInputIndex, newDiffs, skipPos, excessivePos, transposedPos,
|
2010-12-09 13:08:33 +00:00
|
|
|
nextLetters, nextLettersSize);
|
2010-12-01 12:22:15 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2011-03-05 06:50:19 +00:00
|
|
|
static const int TWO_31ST_DIV_255 = S_INT_MAX / 255;
|
2011-02-18 08:50:58 +00:00
|
|
|
static inline int capped255MultForFullMatchAccentsOrCapitalizationDifference(const int num) {
|
|
|
|
return (num < TWO_31ST_DIV_255 ? 255 * num : S_INT_MAX);
|
|
|
|
}
|
2011-01-26 18:23:39 +00:00
|
|
|
inline int UnigramDictionary::calculateFinalFreq(const int inputIndex, const int depth,
|
2011-02-22 06:12:46 +00:00
|
|
|
const int matchWeight, const int skipPos, const int excessivePos, const int transposedPos,
|
2011-03-03 01:22:10 +00:00
|
|
|
const int freq, const bool sameLength) const {
|
2010-12-09 13:08:33 +00:00
|
|
|
// TODO: Demote by edit distance
|
2011-02-22 06:12:46 +00:00
|
|
|
int finalFreq = freq * matchWeight;
|
2011-03-03 01:22:10 +00:00
|
|
|
if (skipPos >= 0) {
|
|
|
|
if (mInputLength >= 3) {
|
|
|
|
multiplyRate(WORDS_WITH_MISSING_CHARACTER_DEMOTION_RATE *
|
|
|
|
(mInputLength - 2) / (mInputLength - 1), &finalFreq);
|
|
|
|
} else {
|
|
|
|
finalFreq = 0;
|
|
|
|
}
|
|
|
|
}
|
2011-01-05 07:37:53 +00:00
|
|
|
if (transposedPos >= 0) multiplyRate(
|
|
|
|
WORDS_WITH_TRANSPOSED_CHARACTERS_DEMOTION_RATE, &finalFreq);
|
2010-12-13 05:42:35 +00:00
|
|
|
if (excessivePos >= 0) {
|
2011-01-05 07:37:53 +00:00
|
|
|
multiplyRate(WORDS_WITH_EXCESSIVE_CHARACTER_DEMOTION_RATE, &finalFreq);
|
2010-12-13 05:42:35 +00:00
|
|
|
if (!existsAdjacentProximityChars(inputIndex, mInputLength)) {
|
2011-01-05 07:37:53 +00:00
|
|
|
multiplyRate(WORDS_WITH_EXCESSIVE_CHARACTER_OUT_OF_PROXIMITY_DEMOTION_RATE, &finalFreq);
|
2010-12-13 05:42:35 +00:00
|
|
|
}
|
|
|
|
}
|
2011-01-26 18:23:39 +00:00
|
|
|
int lengthFreq = TYPED_LETTER_MULTIPLIER;
|
|
|
|
for (int i = 0; i < depth; ++i) lengthFreq *= TYPED_LETTER_MULTIPLIER;
|
2011-02-22 06:12:46 +00:00
|
|
|
if (lengthFreq == matchWeight) {
|
2011-01-27 05:20:22 +00:00
|
|
|
if (depth > 1) {
|
|
|
|
if (DEBUG_DICT) LOGI("Found full matched word.");
|
|
|
|
multiplyRate(FULL_MATCHED_WORDS_PROMOTION_RATE, &finalFreq);
|
|
|
|
}
|
|
|
|
if (sameLength && transposedPos < 0 && skipPos < 0 && excessivePos < 0) {
|
2011-02-18 08:50:58 +00:00
|
|
|
finalFreq = capped255MultForFullMatchAccentsOrCapitalizationDifference(finalFreq);
|
2011-01-27 05:20:22 +00:00
|
|
|
}
|
2011-01-26 18:23:39 +00:00
|
|
|
}
|
2010-12-13 05:42:35 +00:00
|
|
|
if (sameLength && skipPos < 0) finalFreq *= FULL_WORD_MULTIPLIER;
|
|
|
|
return finalFreq;
|
|
|
|
}
|
2010-12-09 13:08:33 +00:00
|
|
|
|
2010-12-13 05:42:35 +00:00
|
|
|
inline void UnigramDictionary::onTerminalWhenUserTypedLengthIsGreaterThanInputLength(
|
2011-02-22 06:12:46 +00:00
|
|
|
unsigned short *word, const int inputIndex, const int depth, const int matchWeight,
|
2010-12-13 05:42:35 +00:00
|
|
|
int *nextLetters, const int nextLettersSize, const int skipPos, const int excessivePos,
|
|
|
|
const int transposedPos, const int freq) {
|
2011-02-22 06:12:46 +00:00
|
|
|
const int finalFreq = calculateFinalFreq(inputIndex, depth, matchWeight, skipPos, excessivePos,
|
2011-01-26 18:23:39 +00:00
|
|
|
transposedPos, freq, false);
|
2010-12-09 13:08:33 +00:00
|
|
|
if (depth >= MIN_SUGGEST_DEPTH) addWord(word, depth + 1, finalFreq);
|
2010-12-13 05:42:35 +00:00
|
|
|
if (depth >= mInputLength && skipPos < 0) {
|
2010-12-02 11:19:59 +00:00
|
|
|
registerNextLetter(mWord[mInputLength], nextLetters, nextLettersSize);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
inline void UnigramDictionary::onTerminalWhenUserTypedLengthIsSameAsInputLength(
|
2011-02-22 06:12:46 +00:00
|
|
|
unsigned short *word, const int inputIndex, const int depth, const int matchWeight,
|
2011-01-27 05:20:22 +00:00
|
|
|
const int skipPos, const int excessivePos, const int transposedPos, const int freq) {
|
2010-12-13 05:42:35 +00:00
|
|
|
if (sameAsTyped(word, depth + 1)) return;
|
2011-02-22 06:12:46 +00:00
|
|
|
const int finalFreq = calculateFinalFreq(inputIndex, depth, matchWeight, skipPos,
|
2010-12-13 05:42:35 +00:00
|
|
|
excessivePos, transposedPos, freq, true);
|
|
|
|
// Proximity collection will promote a word of the same length as what user typed.
|
|
|
|
if (depth >= MIN_SUGGEST_DEPTH) addWord(word, depth + 1, finalFreq);
|
2010-12-02 11:19:59 +00:00
|
|
|
}
|
2010-12-03 07:39:16 +00:00
|
|
|
|
|
|
|
inline bool UnigramDictionary::needsToSkipCurrentNode(const unsigned short c,
|
2010-12-03 10:38:08 +00:00
|
|
|
const int inputIndex, const int skipPos, const int depth) {
|
2011-02-22 08:28:55 +00:00
|
|
|
const unsigned short userTypedChar = getInputCharsAt(inputIndex)[0];
|
2010-12-03 07:39:16 +00:00
|
|
|
// Skip the ' or other letter and continue deeper
|
|
|
|
return (c == QUOTE && userTypedChar != QUOTE) || skipPos == depth;
|
|
|
|
}
|
|
|
|
|
2010-12-09 12:55:40 +00:00
|
|
|
inline bool UnigramDictionary::existsAdjacentProximityChars(const int inputIndex,
|
2011-03-03 01:22:10 +00:00
|
|
|
const int inputLength) const {
|
2010-12-09 12:55:40 +00:00
|
|
|
if (inputIndex < 0 || inputIndex >= inputLength) return false;
|
|
|
|
const int currentChar = *getInputCharsAt(inputIndex);
|
|
|
|
const int leftIndex = inputIndex - 1;
|
|
|
|
if (leftIndex >= 0) {
|
2011-02-25 08:56:53 +00:00
|
|
|
const int *leftChars = getInputCharsAt(leftIndex);
|
2010-12-09 12:55:40 +00:00
|
|
|
int i = 0;
|
|
|
|
while (leftChars[i] > 0 && i < MAX_PROXIMITY_CHARS) {
|
|
|
|
if (leftChars[i++] == currentChar) return true;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
const int rightIndex = inputIndex + 1;
|
|
|
|
if (rightIndex < inputLength) {
|
2011-02-25 08:56:53 +00:00
|
|
|
const int *rightChars = getInputCharsAt(rightIndex);
|
2010-12-09 12:55:40 +00:00
|
|
|
int i = 0;
|
|
|
|
while (rightChars[i] > 0 && i < MAX_PROXIMITY_CHARS) {
|
|
|
|
if (rightChars[i++] == currentChar) return true;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
|
2011-02-18 08:50:58 +00:00
|
|
|
|
|
|
|
// In the following function, c is the current character of the dictionary word
|
|
|
|
// currently examined.
|
|
|
|
// currentChars is an array containing the keys close to the character the
|
|
|
|
// user actually typed at the same position. We want to see if c is in it: if so,
|
|
|
|
// then the word contains at that position a character close to what the user
|
|
|
|
// typed.
|
|
|
|
// What the user typed is actually the first character of the array.
|
|
|
|
// Notice : accented characters do not have a proximity list, so they are alone
|
|
|
|
// in their list. The non-accented version of the character should be considered
|
|
|
|
// "close", but not the other keys close to the non-accented version.
|
2011-01-27 05:20:22 +00:00
|
|
|
inline UnigramDictionary::ProximityType UnigramDictionary::getMatchedProximityId(
|
|
|
|
const int *currentChars, const unsigned short c, const int skipPos,
|
|
|
|
const int excessivePos, const int transposedPos) {
|
2011-02-22 06:12:46 +00:00
|
|
|
const unsigned short baseLowerC = toBaseLowerCase(c);
|
2011-02-18 08:50:58 +00:00
|
|
|
|
|
|
|
// The first char in the array is what user typed. If it matches right away,
|
|
|
|
// that means the user typed that same char for this pos.
|
2011-02-22 06:12:46 +00:00
|
|
|
if (currentChars[0] == baseLowerC || currentChars[0] == c)
|
2011-02-18 08:50:58 +00:00
|
|
|
return SAME_OR_ACCENTED_OR_CAPITALIZED_CHAR;
|
|
|
|
|
|
|
|
// If one of those is true, we should not check for close characters at all.
|
|
|
|
if (skipPos >= 0 || excessivePos >= 0 || transposedPos >= 0)
|
|
|
|
return UNRELATED_CHAR;
|
|
|
|
|
|
|
|
// If the non-accented, lowercased version of that first character matches c,
|
|
|
|
// then we have a non-accented version of the accented character the user
|
|
|
|
// typed. Treat it as a close char.
|
2011-02-22 06:12:46 +00:00
|
|
|
if (toBaseLowerCase(currentChars[0]) == baseLowerC)
|
2011-02-18 08:50:58 +00:00
|
|
|
return NEAR_PROXIMITY_CHAR;
|
|
|
|
|
|
|
|
// Not an exact nor an accent-alike match: search the list of close keys
|
|
|
|
int j = 1;
|
2010-12-09 12:55:40 +00:00
|
|
|
while (currentChars[j] > 0 && j < MAX_PROXIMITY_CHARS) {
|
2011-02-22 06:12:46 +00:00
|
|
|
const bool matched = (currentChars[j] == baseLowerC || currentChars[j] == c);
|
2011-02-18 08:50:58 +00:00
|
|
|
if (matched) return NEAR_PROXIMITY_CHAR;
|
2010-12-03 07:39:16 +00:00
|
|
|
++j;
|
|
|
|
}
|
2011-02-18 08:50:58 +00:00
|
|
|
|
|
|
|
// Was not included, signal this as an unrelated character.
|
2011-01-27 05:20:22 +00:00
|
|
|
return UNRELATED_CHAR;
|
2010-12-03 07:39:16 +00:00
|
|
|
}
|
|
|
|
|
2010-12-06 08:38:58 +00:00
|
|
|
inline bool UnigramDictionary::processCurrentNode(const int pos, const int depth,
|
2011-02-22 06:12:46 +00:00
|
|
|
const int maxDepth, const bool traverseAllNodes, int matchWeight, int inputIndex,
|
2010-12-09 13:08:33 +00:00
|
|
|
const int diffs, const int skipPos, const int excessivePos, const int transposedPos,
|
|
|
|
int *nextLetters, const int nextLettersSize, int *newCount, int *newChildPosition,
|
2011-02-22 06:12:46 +00:00
|
|
|
bool *newTraverseAllNodes, int *newMatchRate, int *newInputIndex, int *newDiffs,
|
2010-12-09 13:08:33 +00:00
|
|
|
int *nextSiblingPosition) {
|
|
|
|
if (DEBUG_DICT) {
|
|
|
|
int inputCount = 0;
|
|
|
|
if (skipPos >= 0) ++inputCount;
|
|
|
|
if (excessivePos >= 0) ++inputCount;
|
|
|
|
if (transposedPos >= 0) ++inputCount;
|
|
|
|
assert(inputCount <= 1);
|
|
|
|
}
|
2010-12-06 08:38:58 +00:00
|
|
|
unsigned short c;
|
|
|
|
int childPosition;
|
|
|
|
bool terminal;
|
|
|
|
int freq;
|
2011-01-27 07:25:16 +00:00
|
|
|
bool isSameAsUserTypedLength = false;
|
2010-12-08 07:04:16 +00:00
|
|
|
|
2011-01-27 07:25:16 +00:00
|
|
|
if (excessivePos == depth && inputIndex < mInputLength - 1) ++inputIndex;
|
2010-12-08 07:04:16 +00:00
|
|
|
|
2010-12-06 08:38:58 +00:00
|
|
|
*nextSiblingPosition = Dictionary::setDictionaryValues(DICT, IS_LATEST_DICT_VERSION, pos, &c,
|
|
|
|
&childPosition, &terminal, &freq);
|
|
|
|
|
|
|
|
const bool needsToTraverseChildrenNodes = childPosition != 0;
|
|
|
|
|
|
|
|
// If we are only doing traverseAllNodes, no need to look at the typed characters.
|
|
|
|
if (traverseAllNodes || needsToSkipCurrentNode(c, inputIndex, skipPos, depth)) {
|
|
|
|
mWord[depth] = c;
|
|
|
|
if (traverseAllNodes && terminal) {
|
2010-12-13 05:42:35 +00:00
|
|
|
onTerminalWhenUserTypedLengthIsGreaterThanInputLength(mWord, inputIndex, depth,
|
2011-02-22 06:12:46 +00:00
|
|
|
matchWeight, nextLetters, nextLettersSize, skipPos, excessivePos, transposedPos,
|
|
|
|
freq);
|
2010-12-06 08:38:58 +00:00
|
|
|
}
|
|
|
|
if (!needsToTraverseChildrenNodes) return false;
|
|
|
|
*newTraverseAllNodes = traverseAllNodes;
|
2011-02-22 06:12:46 +00:00
|
|
|
*newMatchRate = matchWeight;
|
2010-12-06 08:38:58 +00:00
|
|
|
*newDiffs = diffs;
|
|
|
|
*newInputIndex = inputIndex;
|
|
|
|
} else {
|
2011-02-25 08:56:53 +00:00
|
|
|
const int *currentChars = getInputCharsAt(inputIndex);
|
2010-12-09 13:08:33 +00:00
|
|
|
|
|
|
|
if (transposedPos >= 0) {
|
|
|
|
if (inputIndex == transposedPos) currentChars += MAX_PROXIMITY_CHARS;
|
|
|
|
if (inputIndex == (transposedPos + 1)) currentChars -= MAX_PROXIMITY_CHARS;
|
|
|
|
}
|
|
|
|
|
|
|
|
int matchedProximityCharId = getMatchedProximityId(currentChars, c, skipPos, excessivePos,
|
|
|
|
transposedPos);
|
2011-01-27 05:20:22 +00:00
|
|
|
if (UNRELATED_CHAR == matchedProximityCharId) return false;
|
2010-12-06 08:38:58 +00:00
|
|
|
mWord[depth] = c;
|
|
|
|
// If inputIndex is greater than mInputLength, that means there is no
|
|
|
|
// proximity chars. So, we don't need to check proximity.
|
2011-01-27 05:20:22 +00:00
|
|
|
if (SAME_OR_ACCENTED_OR_CAPITALIZED_CHAR == matchedProximityCharId) {
|
2011-02-22 06:12:46 +00:00
|
|
|
matchWeight = matchWeight * TYPED_LETTER_MULTIPLIER;
|
2011-01-27 05:20:22 +00:00
|
|
|
}
|
2011-01-27 07:25:16 +00:00
|
|
|
bool isSameAsUserTypedLength = mInputLength == inputIndex + 1
|
|
|
|
|| (excessivePos == mInputLength - 1 && inputIndex == mInputLength - 2);
|
2010-12-06 08:38:58 +00:00
|
|
|
if (isSameAsUserTypedLength && terminal) {
|
2011-02-22 06:12:46 +00:00
|
|
|
onTerminalWhenUserTypedLengthIsSameAsInputLength(mWord, inputIndex, depth, matchWeight,
|
2011-01-27 05:20:22 +00:00
|
|
|
skipPos, excessivePos, transposedPos, freq);
|
2010-12-06 08:38:58 +00:00
|
|
|
}
|
|
|
|
if (!needsToTraverseChildrenNodes) return false;
|
|
|
|
// Start traversing all nodes after the index exceeds the user typed length
|
|
|
|
*newTraverseAllNodes = isSameAsUserTypedLength;
|
2011-02-22 06:12:46 +00:00
|
|
|
*newMatchRate = matchWeight;
|
2011-01-27 05:20:22 +00:00
|
|
|
*newDiffs = diffs + ((NEAR_PROXIMITY_CHAR == matchedProximityCharId) ? 1 : 0);
|
2010-12-06 08:38:58 +00:00
|
|
|
*newInputIndex = inputIndex + 1;
|
|
|
|
}
|
|
|
|
// Optimization: Prune out words that are too long compared to how much was typed.
|
2010-12-07 04:08:39 +00:00
|
|
|
if (depth >= maxDepth || *newDiffs > mMaxEditDistance) {
|
2010-12-06 08:38:58 +00:00
|
|
|
return false;
|
|
|
|
}
|
|
|
|
|
|
|
|
// If inputIndex is greater than mInputLength, that means there are no proximity chars.
|
2011-01-27 07:25:16 +00:00
|
|
|
// TODO: Check if this can be isSameAsUserTypedLength only.
|
|
|
|
if (isSameAsUserTypedLength || mInputLength <= *newInputIndex) {
|
2010-12-06 08:38:58 +00:00
|
|
|
*newTraverseAllNodes = true;
|
|
|
|
}
|
|
|
|
// get the count of nodes and increment childAddress.
|
|
|
|
*newCount = Dictionary::getCount(DICT, &childPosition);
|
|
|
|
*newChildPosition = childPosition;
|
|
|
|
if (DEBUG_DICT) assert(needsToTraverseChildrenNodes);
|
|
|
|
return needsToTraverseChildrenNodes;
|
|
|
|
}
|
|
|
|
|
2010-12-09 10:21:51 +00:00
|
|
|
inline int UnigramDictionary::getBestWordFreq(const int startInputIndex, const int inputLength,
|
|
|
|
unsigned short *word) {
|
2010-12-08 08:05:39 +00:00
|
|
|
int pos = ROOT_POS;
|
|
|
|
int count = Dictionary::getCount(DICT, &pos);
|
2010-12-09 10:21:51 +00:00
|
|
|
int maxFreq = 0;
|
|
|
|
int depth = 0;
|
|
|
|
unsigned short newWord[MAX_WORD_LENGTH_INTERNAL];
|
2010-12-08 08:05:39 +00:00
|
|
|
bool terminal = false;
|
|
|
|
|
2010-12-09 10:21:51 +00:00
|
|
|
mStackChildCount[0] = count;
|
|
|
|
mStackSiblingPos[0] = pos;
|
|
|
|
|
|
|
|
while (depth >= 0) {
|
|
|
|
if (mStackChildCount[depth] > 0) {
|
|
|
|
--mStackChildCount[depth];
|
|
|
|
int firstChildPos;
|
|
|
|
int newFreq;
|
|
|
|
int siblingPos = mStackSiblingPos[depth];
|
|
|
|
const bool needsToTraverseChildrenNodes = processCurrentNodeForExactMatch(siblingPos,
|
|
|
|
startInputIndex, depth, newWord, &firstChildPos, &count, &terminal, &newFreq,
|
|
|
|
&siblingPos);
|
|
|
|
mStackSiblingPos[depth] = siblingPos;
|
|
|
|
if (depth == (inputLength - 1)) {
|
|
|
|
// Traverse sibling node
|
|
|
|
if (terminal) {
|
|
|
|
if (newFreq > maxFreq) {
|
|
|
|
for (int i = 0; i < inputLength; ++i) word[i] = newWord[i];
|
|
|
|
if (DEBUG_DICT && DEBUG_NODE) {
|
|
|
|
char s[inputLength + 1];
|
|
|
|
for (int i = 0; i < inputLength; ++i) s[i] = word[i];
|
|
|
|
s[inputLength] = 0;
|
|
|
|
LOGI("New missing space word found: %d > %d (%s), %d, %d",
|
|
|
|
newFreq, maxFreq, s, inputLength, depth);
|
|
|
|
}
|
|
|
|
maxFreq = newFreq;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
} else if (needsToTraverseChildrenNodes) {
|
|
|
|
// Traverse children nodes
|
|
|
|
++depth;
|
|
|
|
mStackChildCount[depth] = count;
|
|
|
|
mStackSiblingPos[depth] = firstChildPos;
|
|
|
|
}
|
|
|
|
} else {
|
|
|
|
// Traverse parent node
|
|
|
|
--depth;
|
2010-12-08 08:05:39 +00:00
|
|
|
}
|
|
|
|
}
|
2010-12-09 10:21:51 +00:00
|
|
|
|
|
|
|
word[inputLength] = 0;
|
|
|
|
return maxFreq;
|
2010-12-08 08:05:39 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
inline bool UnigramDictionary::processCurrentNodeForExactMatch(const int firstChildPos,
|
2010-12-09 10:21:51 +00:00
|
|
|
const int startInputIndex, const int depth, unsigned short *word, int *newChildPosition,
|
|
|
|
int *newCount, bool *newTerminal, int *newFreq, int *siblingPos) {
|
|
|
|
const int inputIndex = startInputIndex + depth;
|
2011-02-22 08:28:55 +00:00
|
|
|
const int *currentChars = getInputCharsAt(inputIndex);
|
2010-12-08 08:05:39 +00:00
|
|
|
unsigned short c;
|
2010-12-09 10:21:51 +00:00
|
|
|
*siblingPos = Dictionary::setDictionaryValues(DICT, IS_LATEST_DICT_VERSION, firstChildPos, &c,
|
|
|
|
newChildPosition, newTerminal, newFreq);
|
|
|
|
const unsigned int inputC = currentChars[0];
|
|
|
|
if (DEBUG_DICT) assert(inputC <= U_SHORT_MAX);
|
2011-02-22 06:12:46 +00:00
|
|
|
const unsigned short baseLowerC = toBaseLowerCase(c);
|
|
|
|
const bool matched = (inputC == baseLowerC || inputC == c);
|
2010-12-09 10:21:51 +00:00
|
|
|
const bool hasChild = *newChildPosition != 0;
|
|
|
|
if (matched) {
|
|
|
|
word[depth] = c;
|
|
|
|
if (DEBUG_DICT && DEBUG_NODE) {
|
|
|
|
LOGI("Node(%c, %c)<%d>, %d, %d", inputC, c, matched, hasChild, *newFreq);
|
|
|
|
if (*newTerminal) LOGI("Terminal %d", *newFreq);
|
2010-12-08 08:05:39 +00:00
|
|
|
}
|
2010-12-09 10:21:51 +00:00
|
|
|
if (hasChild) {
|
|
|
|
*newCount = Dictionary::getCount(DICT, newChildPosition);
|
|
|
|
return true;
|
|
|
|
} else {
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
} else {
|
|
|
|
// If this node is not user typed character, this method treats this word as unmatched.
|
|
|
|
// Thus newTerminal shouldn't be true.
|
|
|
|
*newTerminal = false;
|
|
|
|
return false;
|
2010-12-08 08:05:39 +00:00
|
|
|
}
|
|
|
|
}
|
2010-12-01 12:22:15 +00:00
|
|
|
} // namespace latinime
|