Merge "Refactor: Move utility functions and no suggestion functions from unigram_dictionary.cpp to dictionary.cpp"

main
satok 2010-12-01 23:18:39 -08:00 committed by Android (Google) Code Review
commit 9fee221248
6 changed files with 220 additions and 200 deletions

View File

@ -15,6 +15,8 @@
** limitations under the License.
*/
#define LOG_TAG "LatinIME: bigram_dictionary.cpp"
#include "bigram_dictionary.h"
namespace latinime {

53
native/src/defines.h Normal file
View File

@ -0,0 +1,53 @@
/*
**
** 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.
*/
#ifndef LATINIME_DEFINES_H
#define LATINIME_DEFINES_H
#ifdef FLAG_DBG
#include <cutils/log.h>
#ifndef LOG_TAG
#define LOG_TAG "LatinIME: "
#endif
#define DEBUG_DICT 1
#else // FLAG_DBG
#define LOGI
#define DEBUG_DICT 0
#endif // FLAG_DBG
// 22-bit address = ~4MB dictionary size limit, which on average would be about 200k-300k words
#define ADDRESS_MASK 0x3FFFFF
// The bit that decides if an address follows in the next 22 bits
#define FLAG_ADDRESS_MASK 0x40
// The bit that decides if this is a terminal node for a word. The node could still have children,
// if the word has other endings.
#define FLAG_TERMINAL_MASK 0x80
#define FLAG_BIGRAM_READ 0x80
#define FLAG_BIGRAM_CHILDEXIST 0x40
#define FLAG_BIGRAM_CONTINUED 0x80
#define FLAG_BIGRAM_FREQ 0x7F
#define DICTIONARY_VERSION_MIN 200
#define DICTIONARY_HEADER_SIZE 2
#define NOT_VALID_WORD -99
#define SUGGEST_MISSING_CHARACTERS true
#define SUGGEST_MISSING_CHARACTERS_THRESHOLD 5
#endif // LATINIME_DEFINES_H

View File

@ -17,15 +17,22 @@
#include <stdio.h>
#define LOG_TAG "LatinIME: dictionary.cpp"
#include "dictionary.h"
namespace latinime {
Dictionary::Dictionary(void *dict, int typedLetterMultiplier, int fullWordMultiplier,
int maxWordLength, int maxWords, int maxAlternatives)
: DICT((unsigned char*) dict),
// Checks whether it has the latest dictionary or the old dictionary
IS_LATEST_DICT_VERSION((((unsigned char*) dict)[0] & 0xFF) >= DICTIONARY_VERSION_MIN)
{
mUnigramDictionary = new UnigramDictionary(dict, typedLetterMultiplier, fullWordMultiplier,
maxWordLength, maxWords, maxAlternatives, this);
LOGI("IN NATIVE SUGGEST Version: %d \n", (DICT[0] & 0xFF));
mUnigramDictionary = new UnigramDictionary(DICT, typedLetterMultiplier, fullWordMultiplier,
maxWordLength, maxWords, maxAlternatives, IS_LATEST_DICT_VERSION,
hasBigram(), this);
mBigramDictionary = new BigramDictionary(dict, typedLetterMultiplier, fullWordMultiplier,
maxWordLength, maxWords, maxAlternatives, this);
}
@ -35,4 +42,51 @@ Dictionary::~Dictionary()
delete mUnigramDictionary;
delete mBigramDictionary;
}
bool Dictionary::hasBigram() {
return ((DICT[1] & 0xFF) == 1);
}
// TODO: use uint16_t instead of unsigned short
bool Dictionary::isValidWord(unsigned short *word, int length)
{
if (IS_LATEST_DICT_VERSION) {
return (isValidWordRec(DICTIONARY_HEADER_SIZE, word, 0, length) != NOT_VALID_WORD);
} else {
return (isValidWordRec(0, word, 0, length) != NOT_VALID_WORD);
}
}
int Dictionary::isValidWordRec(int pos, unsigned short *word, int offset, int length) {
// returns address of bigram data of that word
// return -99 if not found
int count = Dictionary::getCount(DICT, &pos);
unsigned short currentChar = (unsigned short) word[offset];
for (int j = 0; j < count; j++) {
unsigned short c = Dictionary::getChar(DICT, &pos);
int terminal = Dictionary::getTerminal(DICT, &pos);
int childPos = Dictionary::getAddress(DICT, &pos);
if (c == currentChar) {
if (offset == length - 1) {
if (terminal) {
return (pos+1);
}
} else {
if (childPos != 0) {
int t = isValidWordRec(childPos, word, offset + 1, length);
if (t > 0) {
return t;
}
}
}
}
if (terminal) {
Dictionary::getFreq(DICT, IS_LATEST_DICT_VERSION, &pos);
}
// There could be two instances of each alphabet - upper and lower case. So continue
// looking ...
}
return NOT_VALID_WORD;
}
} // namespace latinime

View File

@ -18,24 +18,11 @@
#define LATINIME_DICTIONARY_H
#include "bigram_dictionary.h"
#include "defines.h"
#include "unigram_dictionary.h"
namespace latinime {
// 22-bit address = ~4MB dictionary size limit, which on average would be about 200k-300k words
#define ADDRESS_MASK 0x3FFFFF
// The bit that decides if an address follows in the next 22 bits
#define FLAG_ADDRESS_MASK 0x40
// The bit that decides if this is a terminal node for a word. The node could still have children,
// if the word has other endings.
#define FLAG_TERMINAL_MASK 0x80
#define FLAG_BIGRAM_READ 0x80
#define FLAG_BIGRAM_CHILDEXIST 0x40
#define FLAG_BIGRAM_CONTINUED 0x80
#define FLAG_BIGRAM_FREQ 0x7F
class Dictionary {
public:
Dictionary(void *dict, int typedLetterMultipler, int fullWordMultiplier, int maxWordLength,
@ -53,21 +40,82 @@ public:
return mUnigramDictionary->getBigrams(word, length, codes, codesSize, outWords, frequencies,
maxWordLength, maxBigrams, maxAlternatives);
}
bool isValidWord(unsigned short *word, int length) {
return mUnigramDictionary->isValidWord(word, length);
}
bool isValidWord(unsigned short *word, int length);
int isValidWordRec(int pos, unsigned short *word, int offset, int length);
void setAsset(void *asset) { mAsset = asset; }
void *getAsset() { return mAsset; }
~Dictionary();
// public static utility methods
// static inline methods should be defined in the header file
static unsigned short getChar(const unsigned char *dict, int *pos);
static int getCount(const unsigned char *dict, int *pos);
static bool getTerminal(const unsigned char *dict, int *pos);
static int getAddress(const unsigned char *dict, int *pos);
static int getFreq(const unsigned char *dict, const bool isLatestDictVersion, int *pos);
private:
bool hasBigram();
const unsigned char *DICT;
const bool IS_LATEST_DICT_VERSION;
void *mAsset;
BigramDictionary *mBigramDictionary;
UnigramDictionary *mUnigramDictionary;
};
// ----------------------------------------------------------------------------
// public static utility methods
// static inline methods should be defined in the header file
inline unsigned short Dictionary::getChar(const unsigned char *dict, int *pos) {
unsigned short ch = (unsigned short) (dict[(*pos)++] & 0xFF);
// If the code is 255, then actual 16 bit code follows (in big endian)
if (ch == 0xFF) {
ch = ((dict[*pos] & 0xFF) << 8) | (dict[*pos + 1] & 0xFF);
(*pos) += 2;
}
return ch;
}
inline int Dictionary::getCount(const unsigned char *dict, int *pos) {
return dict[(*pos)++] & 0xFF;
}
inline bool Dictionary::getTerminal(const unsigned char *dict, int *pos) {
return (dict[*pos] & FLAG_TERMINAL_MASK) > 0;
}
inline int Dictionary::getAddress(const unsigned char *dict, int *pos) {
int address = 0;
if ((dict[*pos] & FLAG_ADDRESS_MASK) == 0) {
*pos += 1;
} else {
address += (dict[*pos] & (ADDRESS_MASK >> 16)) << 16;
address += (dict[*pos + 1] & 0xFF) << 8;
address += (dict[*pos + 2] & 0xFF);
*pos += 3;
}
return address;
}
inline int Dictionary::getFreq(const unsigned char *dict,
const bool isLatestDictVersion, int *pos) {
int freq = dict[(*pos)++] & 0xFF;
if (isLatestDictVersion) {
// skipping bigram
int bigramExist = (dict[*pos] & FLAG_BIGRAM_READ);
if (bigramExist > 0) {
int nextBigramExist = 1;
while (nextBigramExist > 0) {
(*pos) += 3;
nextBigramExist = (dict[(*pos)++] & FLAG_BIGRAM_CONTINUED);
}
} else {
(*pos)++;
}
}
return freq;
}
}; // namespace latinime
#endif // LATINIME_DICTIONARY_H

View File

@ -20,39 +20,26 @@
#include <sys/mman.h>
#include <string.h>
#ifdef FLAG_DBG
#define LOG_TAG "LatinIME: dictionary.cpp"
#include <cutils/log.h>
#define DEBUG_DICT 1
#else // FLAG_DBG
#define LOGI
#define DEBUG_DICT 0
#endif // FLAG_DBG
#define LOG_TAG "LatinIME: unigram_dictionary.cpp"
#include "unigram_dictionary.h"
#include "basechars.h"
#include "char_utils.h"
#define DICTIONARY_VERSION_MIN 200
#define DICTIONARY_HEADER_SIZE 2
#define NOT_VALID_WORD -99
#define SUGGEST_MISSING_CHARACTERS true
#define SUGGEST_MISSING_CHARACTERS_THRESHOLD 5
#include "dictionary.h"
#include "unigram_dictionary.h"
namespace latinime {
UnigramDictionary::UnigramDictionary(void *dict, int typedLetterMultiplier, int fullWordMultiplier,
int maxWordLength, int maxWords, int maxAlternatives, Dictionary *parentDictionary)
: MAX_WORD_LENGTH(maxWordLength),MAX_WORDS(maxWords), MAX_ALTERNATIVES(maxAlternatives)
UnigramDictionary::UnigramDictionary(const unsigned char *dict, int typedLetterMultiplier,
int fullWordMultiplier, int maxWordLength, int maxWords, int maxAlternatives,
const bool isLatestDictVersion, const bool hasBigram, Dictionary *parentDictionary)
: DICT(dict), MAX_WORD_LENGTH(maxWordLength),MAX_WORDS(maxWords),
MAX_ALTERNATIVES(maxAlternatives), IS_LATEST_DICT_VERSION(isLatestDictVersion),
HAS_BIGRAM(hasBigram), mParentDictionary(parentDictionary)
{
LOGI("UnigramDictionary - constructor");
mDict = (unsigned char*) dict;
LOGI("Has Bigram : %d \n", hasBigram);
mTypedLetterMultiplier = typedLetterMultiplier;
mFullWordMultiplier = fullWordMultiplier;
mParentDictionary = parentDictionary;
getVersionNumber();
}
UnigramDictionary::~UnigramDictionary()
@ -106,7 +93,7 @@ void UnigramDictionary::initSuggestions(int *codes, int codesSize, unsigned shor
int UnigramDictionary::getSuggestionCandidates(int inputLength, int skipPos,
int *nextLetters, int nextLettersSize) {
if (checkIfDictVersionIsLatest()) {
if (IS_LATEST_DICT_VERSION) {
getWordsRec(DICTIONARY_HEADER_SIZE, 0, inputLength * 3, false, 1, 0, 0, skipPos,
nextLetters, nextLettersSize);
} else {
@ -127,72 +114,6 @@ void UnigramDictionary::registerNextLetter(unsigned short c, int *nextLetters, i
}
}
// TODO: Should be const static variable calculate in the constructor
void
UnigramDictionary::getVersionNumber()
{
mVersion = (mDict[0] & 0xFF);
mBigram = (mDict[1] & 0xFF);
LOGI("IN NATIVE SUGGEST Version: %d Bigram : %d \n", mVersion, mBigram);
}
// TODO: Should be const static variable calculate in the constructor
// Checks whether it has the latest dictionary or the old dictionary
bool
UnigramDictionary::checkIfDictVersionIsLatest()
{
return (mVersion >= DICTIONARY_VERSION_MIN) && (mBigram == 1 || mBigram == 0);
}
unsigned short
UnigramDictionary::getChar(int *pos)
{
unsigned short ch = (unsigned short) (mDict[(*pos)++] & 0xFF);
// If the code is 255, then actual 16 bit code follows (in big endian)
if (ch == 0xFF) {
ch = ((mDict[*pos] & 0xFF) << 8) | (mDict[*pos + 1] & 0xFF);
(*pos) += 2;
}
return ch;
}
int
UnigramDictionary::getAddress(int *pos)
{
int address = 0;
if ((mDict[*pos] & FLAG_ADDRESS_MASK) == 0) {
*pos += 1;
} else {
address += (mDict[*pos] & (ADDRESS_MASK >> 16)) << 16;
address += (mDict[*pos + 1] & 0xFF) << 8;
address += (mDict[*pos + 2] & 0xFF);
*pos += 3;
}
return address;
}
int
UnigramDictionary::getFreq(int *pos)
{
int freq = mDict[(*pos)++] & 0xFF;
if (checkIfDictVersionIsLatest()) {
// skipping bigram
int bigramExist = (mDict[*pos] & FLAG_BIGRAM_READ);
if (bigramExist > 0) {
int nextBigramExist = 1;
while (nextBigramExist > 0) {
(*pos) += 3;
nextBigramExist = (mDict[(*pos)++] & FLAG_BIGRAM_CONTINUED);
}
} else {
(*pos)++;
}
}
return freq;
}
int
UnigramDictionary::wideStrLen(unsigned short *str)
{
@ -325,7 +246,7 @@ UnigramDictionary::getWordsRec(int pos, int depth, int maxDepth, bool completion
if (diffs > mMaxEditDistance) {
return;
}
int count = getCount(&pos);
int count = Dictionary::getCount(DICT, &pos);
int *currentChars = NULL;
if (mInputLength <= inputIndex) {
completion = true;
@ -335,14 +256,14 @@ UnigramDictionary::getWordsRec(int pos, int depth, int maxDepth, bool completion
for (int i = 0; i < count; i++) {
// -- at char
unsigned short c = getChar(&pos);
unsigned short c = Dictionary::getChar(DICT, &pos);
// -- at flag/add
unsigned short lowerC = toLowerCase(c);
bool terminal = getTerminal(&pos);
int childrenAddress = getAddress(&pos);
bool terminal = Dictionary::getTerminal(DICT, &pos);
int childrenAddress = Dictionary::getAddress(DICT, &pos);
// -- after address or flag
int freq = 1;
if (terminal) freq = getFreq(&pos);
if (terminal) freq = Dictionary::getFreq(DICT, IS_LATEST_DICT_VERSION, &pos);
// -- after add or freq
// If we are only doing completions, no need to look at the typed characters.
@ -403,9 +324,9 @@ UnigramDictionary::getBigramAddress(int *pos, bool advance)
{
int address = 0;
address += (mDict[*pos] & 0x3F) << 16;
address += (mDict[*pos + 1] & 0xFF) << 8;
address += (mDict[*pos + 2] & 0xFF);
address += (DICT[*pos] & 0x3F) << 16;
address += (DICT[*pos + 1] & 0xFF) << 8;
address += (DICT[*pos + 2] & 0xFF);
if (advance) {
*pos += 3;
@ -417,7 +338,7 @@ UnigramDictionary::getBigramAddress(int *pos, bool advance)
int
UnigramDictionary::getBigramFreq(int *pos)
{
int freq = mDict[(*pos)++] & FLAG_BIGRAM_FREQ;
int freq = DICT[(*pos)++] & FLAG_BIGRAM_FREQ;
return freq;
}
@ -434,8 +355,8 @@ UnigramDictionary::getBigrams(unsigned short *prevWord, int prevWordLength, int
mInputLength = codesSize;
mMaxBigrams = maxBigrams;
if (mBigram == 1 && checkIfDictVersionIsLatest()) {
int pos = isValidWordRec(
if (HAS_BIGRAM && IS_LATEST_DICT_VERSION) {
int pos = mParentDictionary->isValidWordRec(
DICTIONARY_HEADER_SIZE, prevWord, 0, prevWordLength);
LOGI("Pos -> %d\n", pos);
if (pos < 0) {
@ -443,15 +364,15 @@ UnigramDictionary::getBigrams(unsigned short *prevWord, int prevWordLength, int
}
int bigramCount = 0;
int bigramExist = (mDict[pos] & FLAG_BIGRAM_READ);
int bigramExist = (DICT[pos] & FLAG_BIGRAM_READ);
if (bigramExist > 0) {
int nextBigramExist = 1;
while (nextBigramExist > 0 && bigramCount < maxBigrams) {
int bigramAddress = getBigramAddress(&pos, true);
int frequency = (FLAG_BIGRAM_FREQ & mDict[pos]);
int frequency = (FLAG_BIGRAM_FREQ & DICT[pos]);
// search for all bigrams and store them
searchForTerminalNode(bigramAddress, frequency);
nextBigramExist = (mDict[pos++] & FLAG_BIGRAM_CONTINUED);
nextBigramExist = (DICT[pos++] & FLAG_BIGRAM_CONTINUED);
bigramCount++;
}
}
@ -482,7 +403,7 @@ UnigramDictionary::searchForTerminalNode(int addressLookingFor, int frequency)
word[depth] = (unsigned short) followingChar;
}
pos = followDownBranchAddress; // pos start at count
int count = mDict[pos] & 0xFF;
int count = DICT[pos] & 0xFF;
LOGI("count - %d\n",count);
pos++;
for (int i = 0; i < count; i++) {
@ -502,7 +423,7 @@ UnigramDictionary::searchForTerminalNode(int addressLookingFor, int frequency)
}
} else {
followDownBranchAddress = addr;
followingChar = (char)(0xFF & mDict[pos-1]);
followingChar = (char)(0xFF & DICT[pos-1]);
if (firstAddress) {
firstAddress = false;
haveToSearchAll = false;
@ -513,7 +434,7 @@ UnigramDictionary::searchForTerminalNode(int addressLookingFor, int frequency)
} else if (getFirstBitOfByte(&pos)) { // terminal
if (addressLookingFor == (pos-1)) { // found !!
depth++;
word[depth] = (0xFF & mDict[pos-1]);
word[depth] = (0xFF & DICT[pos-1]);
found = true;
break;
}
@ -530,7 +451,7 @@ UnigramDictionary::searchForTerminalNode(int addressLookingFor, int frequency)
}
} else {
followDownBranchAddress = addr;
followingChar = (char)(0xFF & mDict[pos-1]);
followingChar = (char)(0xFF & DICT[pos-1]);
if (firstAddress) {
firstAddress = false;
haveToSearchAll = true;
@ -543,12 +464,12 @@ UnigramDictionary::searchForTerminalNode(int addressLookingFor, int frequency)
}
// skipping bigram
int bigramExist = (mDict[pos] & FLAG_BIGRAM_READ);
int bigramExist = (DICT[pos] & FLAG_BIGRAM_READ);
if (bigramExist > 0) {
int nextBigramExist = 1;
while (nextBigramExist > 0) {
pos += 3;
nextBigramExist = (mDict[pos++] & FLAG_BIGRAM_CONTINUED);
nextBigramExist = (DICT[pos++] & FLAG_BIGRAM_CONTINUED);
}
} else {
pos++;
@ -584,48 +505,4 @@ UnigramDictionary::checkFirstCharacter(unsigned short *word)
return false;
}
// TODO: Move to parent dictionary
bool
UnigramDictionary::isValidWord(unsigned short *word, int length)
{
if (checkIfDictVersionIsLatest()) {
return (isValidWordRec(DICTIONARY_HEADER_SIZE, word, 0, length) != NOT_VALID_WORD);
} else {
return (isValidWordRec(0, word, 0, length) != NOT_VALID_WORD);
}
}
int
UnigramDictionary::isValidWordRec(int pos, unsigned short *word, int offset, int length) {
// returns address of bigram data of that word
// return -99 if not found
int count = getCount(&pos);
unsigned short currentChar = (unsigned short) word[offset];
for (int j = 0; j < count; j++) {
unsigned short c = getChar(&pos);
int terminal = getTerminal(&pos);
int childPos = getAddress(&pos);
if (c == currentChar) {
if (offset == length - 1) {
if (terminal) {
return (pos+1);
}
} else {
if (childPos != 0) {
int t = isValidWordRec(childPos, word, offset + 1, length);
if (t > 0) {
return t;
}
}
}
}
if (terminal) {
getFreq(&pos);
}
// There could be two instances of each alphabet - upper and lower case. So continue
// looking ...
}
return NOT_VALID_WORD;
}
} // namespace latinime

View File

@ -17,33 +17,20 @@
#ifndef LATINIME_UNIGRAM_DICTIONARY_H
#define LATINIME_UNIGRAM_DICTIONARY_H
#include "defines.h"
namespace latinime {
// 22-bit address = ~4MB dictionary size limit, which on average would be about 200k-300k words
#define ADDRESS_MASK 0x3FFFFF
// The bit that decides if an address follows in the next 22 bits
#define FLAG_ADDRESS_MASK 0x40
// The bit that decides if this is a terminal node for a word. The node could still have children,
// if the word has other endings.
#define FLAG_TERMINAL_MASK 0x80
#define FLAG_BIGRAM_READ 0x80
#define FLAG_BIGRAM_CHILDEXIST 0x40
#define FLAG_BIGRAM_CONTINUED 0x80
#define FLAG_BIGRAM_FREQ 0x7F
class Dictionary;
class UnigramDictionary {
public:
UnigramDictionary(void *dict, int typedLetterMultipler, int fullWordMultiplier, int maxWordLength,
int maxWords, int maxAlternatives, Dictionary *parentDictionary);
UnigramDictionary(const unsigned char *dict, int typedLetterMultipler, int fullWordMultiplier,
int maxWordLength, int maxWords, int maxAlternatives, const bool isLatestDictVersion,
const bool hasBigram, Dictionary *parentDictionary);
int getSuggestions(int *codes, int codesSize, unsigned short *outWords, int *frequencies,
int *nextLetters, int nextLettersSize);
int getBigrams(unsigned short *word, int length, int *codes, int codesSize,
unsigned short *outWords, int *frequencies, int maxWordLength, int maxBigrams,
int maxAlternatives);
bool isValidWord(unsigned short *word, int length);
~UnigramDictionary();
private:
@ -57,11 +44,9 @@ private:
int getBigramFreq(int *pos);
void searchForTerminalNode(int address, int frequency);
bool getFirstBitOfByte(int *pos) { return (mDict[*pos] & 0x80) > 0; }
bool getSecondBitOfByte(int *pos) { return (mDict[*pos] & 0x40) > 0; }
bool getTerminal(int *pos) { return (mDict[*pos] & FLAG_TERMINAL_MASK) > 0; }
int getCount(int *pos) { return mDict[(*pos)++] & 0xFF; }
unsigned short getChar(int *pos);
bool getFirstBitOfByte(int *pos) { return (DICT[*pos] & 0x80) > 0; }
bool getSecondBitOfByte(int *pos) { return (DICT[*pos] & 0x40) > 0; }
bool getTerminal(int *pos) { return (DICT[*pos] & FLAG_TERMINAL_MASK) > 0; }
int wideStrLen(unsigned short *str);
bool sameAsTyped(unsigned short *word, int length);
@ -72,15 +57,16 @@ private:
void getWordsRec(int pos, int depth, int maxDepth, bool completion, int frequency,
int inputIndex, int diffs, int skipPos, int *nextLetters, int nextLettersSize);
void registerNextLetter(unsigned short c, int *nextLetters, int nextLettersSize);
int isValidWordRec(int pos, unsigned short *word, int offset, int length);
unsigned char *mDict;
Dictionary *mParentDictionary;
const unsigned char *DICT;
const int MAX_WORDS;
const int MAX_WORD_LENGTH;
const int MAX_ALTERNATIVES;
const bool IS_LATEST_DICT_VERSION;
const bool HAS_BIGRAM;
Dictionary *mParentDictionary;
int *mFrequencies;
int *mBigramFreq;
int mMaxBigrams;