nyancrimew
/
LatinIME
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

(Step2)Move functions related to proximity to proximity_info.cpp 

Change-Id: Iae0eb2a5cd758bda820fa42b4bc3eb3d2665bf96
main
satok 2011-07-14 15:43:42 +09:00
parent f7f2e82e8b
commit d24df43eaf
6 changed files with 150 additions and 119 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

20
native/src/basechars.h
View File

@ -1,3 +1,22 @@
/*
* Copyright (C) 2009 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_BASECHARS_H
#define LATINIME_BASECHARS_H
/**
* Table mapping most combined Latin, Greek, and Cyrillic characters
* to their base characters. If c is in range, BASE_CHARS[c] == c
@ -170,3 +189,4 @@ static unsigned short BASE_CHARS[] = {
// generated with:
// cat UnicodeData.txt | perl -e 'while (<>) { @foo = split(/;/); $foo[5] =~ s/<.*> //; $base[hex($foo[0])] = hex($foo[5]);} for ($i = 0; $i < 0x500; $i += 8) { for ($j = $i; $j < $i + 8; $j++) { printf("0x%04x, ", $base[$j] ? $base[$j] : $j)}; print "\n"; }'
#endif // LATINIME_BASECHARS_H

17
native/src/dictionary.h
View File

@ -17,7 +17,9 @@
#ifndef LATINIME_DICTIONARY_H
#define LATINIME_DICTIONARY_H
#include "basechars.h"
#include "bigram_dictionary.h"
#include "char_utils.h"
#include "defines.h"
#include "proximity_info.h"
#include "unigram_dictionary.h"
@ -61,7 +63,7 @@ public:
static int setDictionaryValues(const unsigned char *dict, const bool isLatestDictVersion,
const int pos, unsigned short *c, int *childrenPosition,
bool *terminal, int *freq);
static inline unsigned short toBaseLowerCase(unsigned short c);
// TODO: delete this
int getBigramPosition(unsigned short *word, int length);
@ -156,6 +158,19 @@ inline int Dictionary::setDictionaryValues(const unsigned char *dict,
return position;
}
inline unsigned short Dictionary::toBaseLowerCase(unsigned short c) {
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;
}
} // namespace latinime
#endif // LATINIME_DICTIONARY_H

75
native/src/proximity_info.cpp
View File

@ -19,6 +19,7 @@
#define LOG_TAG "LatinIME: proximity_info.cpp"
#include "dictionary.h"
#include "proximity_info.h"
namespace latinime {
@ -69,10 +70,82 @@ void ProximityInfo::setInputParams(const int* inputCodes, const int inputLength)
mInputLength = inputLength;
}
const int* ProximityInfo::getProximityCharsAt(const int index) const {
inline const int* ProximityInfo::getProximityCharsAt(const int index) const {
return mInputCodes + (index * MAX_PROXIMITY_CHARS_SIZE);
}
unsigned short ProximityInfo::getPrimaryCharAt(const int index) const {
return getProximityCharsAt(index)[0];
}
bool ProximityInfo::existsCharInProximityAt(const int index, const int c) const {
const int *chars = getProximityCharsAt(index);
int i = 0;
while (chars[i] > 0 && i < MAX_PROXIMITY_CHARS_SIZE) {
if (chars[i++] == c) {
return true;
}
}
return false;
}
bool ProximityInfo::existsAdjacentProximityChars(const int index) const {
if (index < 0 || index >= mInputLength) return false;
const int currentChar = getPrimaryCharAt(index);
const int leftIndex = index - 1;
if (leftIndex >= 0 && existsCharInProximityAt(leftIndex, currentChar)) {
return true;
}
const int rightIndex = index + 1;
if (rightIndex < mInputLength && existsCharInProximityAt(rightIndex, currentChar)) {
return true;
}
return false;
}
// 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.
ProximityInfo::ProximityType ProximityInfo::getMatchedProximityId(
const int index, const unsigned short c, const int skipPos,
const int excessivePos, const int transposedPos) const {
const int *currentChars = getProximityCharsAt(index);
const unsigned short baseLowerC = Dictionary::toBaseLowerCase(c);
// 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.
if (currentChars[0] == baseLowerC || currentChars[0] == c)
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.
if (Dictionary::toBaseLowerCase(currentChars[0]) == baseLowerC)
return NEAR_PROXIMITY_CHAR;
// Not an exact nor an accent-alike match: search the list of close keys
int j = 1;
while (currentChars[j] > 0 && j < MAX_PROXIMITY_CHARS_SIZE) {
const bool matched = (currentChars[j] == baseLowerC || currentChars[j] == c);
if (matched) return NEAR_PROXIMITY_CHAR;
++j;
}
// Was not included, signal this as an unrelated character.
return UNRELATED_CHAR;
}
bool ProximityInfo::sameAsTyped(const unsigned short *word, int length) const {
if (length != mInputLength) {
return false;

12
native/src/proximity_info.h
View File

@ -25,6 +25,12 @@ namespace latinime {
class ProximityInfo {
public:
typedef enum { // Used as a return value for character comparison
SAME_OR_ACCENTED_OR_CAPITALIZED_CHAR, // Same char, possibly with different case or accent
NEAR_PROXIMITY_CHAR, // It is a char located nearby on the keyboard
UNRELATED_CHAR // It is an unrelated char
} ProximityType;
ProximityInfo(const int maxProximityCharsSize, const int keyboardWidth,
const int keybaordHeight, const int gridWidth, const int gridHeight,
const uint32_t *proximityCharsArray);
@ -32,6 +38,12 @@ public:
bool hasSpaceProximity(const int x, const int y) const;
void setInputParams(const int* inputCodes, const int inputLength);
const int* getProximityCharsAt(const int index) const;
unsigned short getPrimaryCharAt(const int index) const;
bool existsCharInProximityAt(const int index, const int c) const;
bool existsAdjacentProximityChars(const int index) const;
ProximityType getMatchedProximityId(
const int index, const unsigned short c, const int skipPos,
const int excessivePos, const int transposedPos) const;
bool sameAsTyped(const unsigned short *word, int length) const;
private:
int getStartIndexFromCoordinates(const int x, const int y) const;

132
native/src/unigram_dictionary.cpp
View File

@ -20,7 +20,6 @@
#define LOG_TAG "LatinIME: unigram_dictionary.cpp"
#include "basechars.h"
#include "char_utils.h"
#include "dictionary.h"
#include "unigram_dictionary.h"
@ -351,18 +350,6 @@ bool UnigramDictionary::addWord(unsigned short *word, int length, int frequency)
return false;
}
static inline unsigned short toBaseLowerCase(unsigned short c) {
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;
}
static const char QUOTE = '\'';
static const char SPACE = ' ';
@ -556,7 +543,7 @@ inline int UnigramDictionary::calculateFinalFreq(const int inputIndex, const int
WORDS_WITH_TRANSPOSED_CHARACTERS_DEMOTION_RATE, &finalFreq);
if (excessivePos >= 0) {
multiplyRate(WORDS_WITH_EXCESSIVE_CHARACTER_DEMOTION_RATE, &finalFreq);
if (!existsAdjacentProximityChars(inputIndex, mInputLength)) {
if (!mProximityInfo->existsAdjacentProximityChars(inputIndex)) {
// If an excessive character is not adjacent to the left char or the right char,
// we will demote this word.
multiplyRate(WORDS_WITH_EXCESSIVE_CHARACTER_OUT_OF_PROXIMITY_DEMOTION_RATE, &finalFreq);
@ -592,75 +579,11 @@ inline int UnigramDictionary::calculateFinalFreq(const int inputIndex, const int
inline bool UnigramDictionary::needsToSkipCurrentNode(const unsigned short c,
const int inputIndex, const int skipPos, const int depth) {
const unsigned short userTypedChar = getInputCharsAt(inputIndex)[0];
const unsigned short userTypedChar = mProximityInfo->getPrimaryCharAt(inputIndex);
// Skip the ' or other letter and continue deeper
return (c == QUOTE && userTypedChar != QUOTE) || skipPos == depth;
}
inline bool UnigramDictionary::existsAdjacentProximityChars(const int inputIndex,
const int inputLength) const {
if (inputIndex < 0 || inputIndex >= inputLength) return false;
const int currentChar = *getInputCharsAt(inputIndex);
const int leftIndex = inputIndex - 1;
if (leftIndex >= 0) {
const int *leftChars = getInputCharsAt(leftIndex);
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) {
const int *rightChars = getInputCharsAt(rightIndex);
int i = 0;
while (rightChars[i] > 0 && i < MAX_PROXIMITY_CHARS) {
if (rightChars[i++] == currentChar) return true;
}
}
return false;
}
// 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.
inline UnigramDictionary::ProximityType UnigramDictionary::getMatchedProximityId(
const int *currentChars, const unsigned short c, const int skipPos,
const int excessivePos, const int transposedPos) {
const unsigned short baseLowerC = toBaseLowerCase(c);
// 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.
if (currentChars[0] == baseLowerC || currentChars[0] == c)
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.
if (toBaseLowerCase(currentChars[0]) == baseLowerC)
return NEAR_PROXIMITY_CHAR;
// Not an exact nor an accent-alike match: search the list of close keys
int j = 1;
while (currentChars[j] > 0 && j < MAX_PROXIMITY_CHARS) {
const bool matched = (currentChars[j] == baseLowerC || currentChars[j] == c);
if (matched) return NEAR_PROXIMITY_CHAR;
++j;
}
// Was not included, signal this as an unrelated character.
return UNRELATED_CHAR;
}
inline void UnigramDictionary::onTerminal(unsigned short int* word, const int depth,
const uint8_t* const root, const uint8_t flags, const int pos,
@ -826,15 +749,14 @@ inline bool UnigramDictionary::processCurrentNodeForExactMatch(const int firstCh
const int startInputIndex, const int depth, unsigned short *word, int *newChildPosition,
int *newCount, bool *newTerminal, int *newFreq, int *siblingPos) {
const int inputIndex = startInputIndex + depth;
const int *currentChars = getInputCharsAt(inputIndex);
unsigned short c;
*siblingPos = Dictionary::setDictionaryValues(DICT_ROOT, IS_LATEST_DICT_VERSION, firstChildPos,
&c, newChildPosition, newTerminal, newFreq);
const unsigned int inputC = currentChars[0];
const unsigned int inputC = mProximityInfo->getPrimaryCharAt(inputIndex);
if (DEBUG_DICT) {
assert(inputC <= U_SHORT_MAX);
}
const unsigned short baseLowerC = toBaseLowerCase(c);
const unsigned short baseLowerC = Dictionary::toBaseLowerCase(c);
const bool matched = (inputC == baseLowerC || inputC == c);
const bool hasChild = *newChildPosition != 0;
if (matched) {
@ -952,20 +874,20 @@ inline bool UnigramDictionary::processCurrentNode(const int initialPos, const in
*newDiffs = diffs;
*newInputIndex = inputIndex;
} else {
const int *currentChars = getInputCharsAt(inputIndex);
int inputIndexForProximity = inputIndex;
if (transposedPos >= 0) {
if (inputIndex == transposedPos) currentChars += MAX_PROXIMITY_CHARS;
if (inputIndex == (transposedPos + 1)) currentChars -= MAX_PROXIMITY_CHARS;
if (inputIndex == transposedPos) ++inputIndexForProximity;
if (inputIndex == (transposedPos + 1)) --inputIndexForProximity;
}
int matchedProximityCharId = getMatchedProximityId(currentChars, c, skipPos, excessivePos,
transposedPos);
if (UNRELATED_CHAR == matchedProximityCharId) return false;
ProximityInfo::ProximityType matchedProximityCharId = mProximityInfo->getMatchedProximityId(
inputIndexForProximity, c, skipPos, excessivePos, transposedPos);
if (ProximityInfo::UNRELATED_CHAR == matchedProximityCharId) return false;
mWord[depth] = c;
// If inputIndex is greater than mInputLength, that means there is no
// proximity chars. So, we don't need to check proximity.
if (SAME_OR_ACCENTED_OR_CAPITALIZED_CHAR == matchedProximityCharId) {
if (ProximityInfo::SAME_OR_ACCENTED_OR_CAPITALIZED_CHAR == matchedProximityCharId) {
multiplyIntCapped(TYPED_LETTER_MULTIPLIER, &matchWeight);
}
bool isSameAsUserTypedLength = mInputLength == inputIndex + 1
@ -978,7 +900,8 @@ inline bool UnigramDictionary::processCurrentNode(const int initialPos, const in
// Start traversing all nodes after the index exceeds the user typed length
*newTraverseAllNodes = isSameAsUserTypedLength;
*newMatchRate = matchWeight;
*newDiffs = diffs + ((NEAR_PROXIMITY_CHAR == matchedProximityCharId) ? 1 : 0);
*newDiffs = diffs
+ ((ProximityInfo::NEAR_PROXIMITY_CHAR == matchedProximityCharId) ? 1 : 0);
*newInputIndex = inputIndex + 1;
}
// Optimization: Prune out words that are too long compared to how much was typed.
@ -1007,7 +930,7 @@ inline int UnigramDictionary::getMostFrequentWordLike(const int startInputIndex,
uint16_t inWord[inputLength];
for (int i = 0; i < inputLength; ++i) {
inWord[i] = *getInputCharsAt(startInputIndex + i);
inWord[i] = (uint16_t)mProximityInfo->getPrimaryCharAt(startInputIndex + i);
}
return getMostFrequentWordLikeInner(inWord, inputLength, word);
}
@ -1031,8 +954,8 @@ static inline bool testCharGroupForContinuedLikeness(const uint8_t flags,
const bool hasMultipleChars = (0 != (UnigramDictionary::FLAG_HAS_MULTIPLE_CHARS & flags));
int pos = startPos;
int32_t character = BinaryFormat::getCharCodeAndForwardPointer(root, &pos);
int32_t baseChar = toBaseLowerCase(character);
const uint16_t wChar = toBaseLowerCase(inWord[startInputIndex]);
int32_t baseChar = Dictionary::toBaseLowerCase(character);
const uint16_t wChar = Dictionary::toBaseLowerCase(inWord[startInputIndex]);
if (baseChar != wChar) {
*outPos = hasMultipleChars ? BinaryFormat::skipOtherCharacters(root, pos) : pos;
@ -1044,8 +967,8 @@ static inline bool testCharGroupForContinuedLikeness(const uint8_t flags,
if (hasMultipleChars) {
character = BinaryFormat::getCharCodeAndForwardPointer(root, &pos);
while (NOT_A_CHARACTER != character) {
baseChar = toBaseLowerCase(character);
if (toBaseLowerCase(inWord[++inputIndex]) != baseChar) {
baseChar = Dictionary::toBaseLowerCase(character);
if (Dictionary::toBaseLowerCase(inWord[++inputIndex]) != baseChar) {
*outPos = BinaryFormat::skipOtherCharacters(root, pos);
*outInputIndex = startInputIndex;
return false;
@ -1290,7 +1213,7 @@ inline bool UnigramDictionary::processCurrentNode(const int initialPos, const in
const bool hasChildren = (!isLastChar) || BinaryFormat::hasChildrenInFlags(flags);
// This has to be done for each virtual char (this forwards the "inputIndex" which
// is the index in the user-inputted chars, as read by getInputCharsAt.
// is the index in the user-inputted chars, as read by proximity chars.
if (excessivePos == depth && inputIndex < mInputLength - 1) ++inputIndex;
if (traverseAllNodes || needsToSkipCurrentNode(c, inputIndex, skipPos, depth)) {
mWord[depth] = c;
@ -1314,16 +1237,16 @@ inline bool UnigramDictionary::processCurrentNode(const int initialPos, const in
return false;
}
} else {
const int *currentChars = getInputCharsAt(inputIndex);
int inputIndexForProximity = inputIndex;
if (transposedPos >= 0) {
if (inputIndex == transposedPos) currentChars += MAX_PROXIMITY_CHARS;
if (inputIndex == (transposedPos + 1)) currentChars -= MAX_PROXIMITY_CHARS;
if (inputIndex == transposedPos) ++inputIndexForProximity;
if (inputIndex == (transposedPos + 1)) --inputIndexForProximity;
}
const int matchedProximityCharId = getMatchedProximityId(currentChars, c, skipPos,
excessivePos, transposedPos);
if (UNRELATED_CHAR == matchedProximityCharId) {
int matchedProximityCharId = mProximityInfo->getMatchedProximityId(
inputIndexForProximity, c, skipPos, excessivePos, transposedPos);
if (ProximityInfo::UNRELATED_CHAR == matchedProximityCharId) {
// We found that this is an unrelated character, so we should give up traversing
// this node and its children entirely.
// However we may not be on the last virtual node yet so we skip the remaining
@ -1342,7 +1265,7 @@ inline bool UnigramDictionary::processCurrentNode(const int initialPos, const in
mWord[depth] = c;
// If inputIndex is greater than mInputLength, that means there is no
// proximity chars. So, we don't need to check proximity.
if (SAME_OR_ACCENTED_OR_CAPITALIZED_CHAR == matchedProximityCharId) {
if (ProximityInfo::SAME_OR_ACCENTED_OR_CAPITALIZED_CHAR == matchedProximityCharId) {
multiplyIntCapped(TYPED_LETTER_MULTIPLIER, &matchWeight);
}
const bool isSameAsUserTypedLength = mInputLength == inputIndex + 1
@ -1366,7 +1289,8 @@ inline bool UnigramDictionary::processCurrentNode(const int initialPos, const in
}
// Start traversing all nodes after the index exceeds the user typed length
traverseAllNodes = isSameAsUserTypedLength;
diffs = diffs + ((NEAR_PROXIMITY_CHAR == matchedProximityCharId) ? 1 : 0);
diffs = diffs
+ ((ProximityInfo::NEAR_PROXIMITY_CHAR == matchedProximityCharId) ? 1 : 0);
// Finally, we are ready to go to the next character, the next "virtual node".
// We should advance the input index.
// We do this in this branch of the 'if traverseAllNodes' because we are still matching

13
native/src/unigram_dictionary.h
View File

@ -29,12 +29,6 @@ namespace latinime {
class UnigramDictionary {
typedef enum { // Used as a return value for character comparison
SAME_OR_ACCENTED_OR_CAPITALIZED_CHAR, // Same char, possibly with different case or accent
NEAR_PROXIMITY_CHAR, // It is a char located nearby on the keyboard
UNRELATED_CHAR // It is an unrelated char
} ProximityType;
public:
#ifdef NEW_DICTIONARY_FORMAT
@ -118,8 +112,6 @@ private:
int *nextLetters, const int nextLettersSize);
bool needsToSkipCurrentNode(const unsigned short c,
const int inputIndex, const int skipPos, const int depth);
ProximityType getMatchedProximityId(const int *currentChars, const unsigned short c,
const int skipPos, const int excessivePos, const int transposedPos);
// Process a node by considering proximity, missing and excessive character
bool processCurrentNode(const int initialPos, const int initialDepth,
const int maxDepth, const bool initialTraverseAllNodes, const int snr, int inputIndex,
@ -127,10 +119,6 @@ private:
const int transposedPos, int *nextLetters, const int nextLettersSize, int *newCount,
int *newChildPosition, bool *newTraverseAllNodes, int *newSnr, int*newInputIndex,
int *newDiffs, int *nextSiblingPosition, int *nextOutputIndex);
bool existsAdjacentProximityChars(const int inputIndex, const int inputLength) const;
inline const int* getInputCharsAt(const int index) const {
return mProximityInfo->getProximityCharsAt(index);
}
int getMostFrequentWordLike(const int startInputIndex, const int inputLength,
unsigned short *word);
#ifndef NEW_DICTIONARY_FORMAT
@ -189,7 +177,6 @@ private:
int mStackOutputIndex[MAX_WORD_LENGTH_INTERNAL];
int mNextLettersFrequency[NEXT_LETTERS_SIZE];
};
} // namespace latinime
#endif // LATINIME_UNIGRAM_DICTIONARY_H