aboutsummaryrefslogtreecommitdiffstats
path: root/native/src
diff options
context:
space:
mode:
Diffstat (limited to 'native/src')
-rw-r--r--native/src/bigram_dictionary.cpp254
-rw-r--r--native/src/bigram_dictionary.h55
-rw-r--r--native/src/defines.h57
-rw-r--r--native/src/dictionary.cpp558
-rw-r--r--native/src/dictionary.h176
-rw-r--r--native/src/unigram_dictionary.cpp321
-rw-r--r--native/src/unigram_dictionary.h84
7 files changed, 909 insertions, 596 deletions
diff --git a/native/src/bigram_dictionary.cpp b/native/src/bigram_dictionary.cpp
new file mode 100644
index 000000000..095b80531
--- /dev/null
+++ b/native/src/bigram_dictionary.cpp
@@ -0,0 +1,254 @@
+/*
+**
+** 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.
+*/
+
+#include <string.h>
+
+#define LOG_TAG "LatinIME: bigram_dictionary.cpp"
+
+#include "bigram_dictionary.h"
+#include "dictionary.h"
+
+namespace latinime {
+
+BigramDictionary::BigramDictionary(const unsigned char *dict, int maxWordLength,
+ int maxAlternatives, const bool isLatestDictVersion, const bool hasBigram,
+ Dictionary *parentDictionary)
+ : DICT(dict), MAX_WORD_LENGTH(maxWordLength),
+ MAX_ALTERNATIVES(maxAlternatives), IS_LATEST_DICT_VERSION(isLatestDictVersion),
+ HAS_BIGRAM(hasBigram), mParentDictionary(parentDictionary) {
+ LOGI("BigramDictionary - constructor");
+ LOGI("Has Bigram : %d \n", hasBigram);
+}
+
+BigramDictionary::~BigramDictionary() {
+}
+
+bool BigramDictionary::addWordBigram(unsigned short *word, int length, int frequency) {
+ word[length] = 0;
+ if (DEBUG_DICT) {
+ char s[length + 1];
+ for (int i = 0; i <= length; i++) s[i] = word[i];
+ LOGI("Bigram: Found word = %s, freq = %d : \n", s, frequency);
+ }
+
+ // Find the right insertion point
+ int insertAt = 0;
+ while (insertAt < mMaxBigrams) {
+ if (frequency > mBigramFreq[insertAt] || (mBigramFreq[insertAt] == frequency
+ && length < Dictionary::wideStrLen(mBigramChars + insertAt * MAX_WORD_LENGTH))) {
+ break;
+ }
+ insertAt++;
+ }
+ LOGI("Bigram: InsertAt -> %d maxBigrams: %d\n", insertAt, mMaxBigrams);
+ if (insertAt < mMaxBigrams) {
+ memmove((char*) mBigramFreq + (insertAt + 1) * sizeof(mBigramFreq[0]),
+ (char*) mBigramFreq + insertAt * sizeof(mBigramFreq[0]),
+ (mMaxBigrams - insertAt - 1) * sizeof(mBigramFreq[0]));
+ mBigramFreq[insertAt] = frequency;
+ memmove((char*) mBigramChars + (insertAt + 1) * MAX_WORD_LENGTH * sizeof(short),
+ (char*) mBigramChars + (insertAt ) * MAX_WORD_LENGTH * sizeof(short),
+ (mMaxBigrams - insertAt - 1) * sizeof(short) * MAX_WORD_LENGTH);
+ unsigned short *dest = mBigramChars + (insertAt ) * MAX_WORD_LENGTH;
+ while (length--) {
+ *dest++ = *word++;
+ }
+ *dest = 0; // NULL terminate
+ if (DEBUG_DICT) LOGI("Bigram: Added word at %d\n", insertAt);
+ return true;
+ }
+ return false;
+}
+
+int BigramDictionary::getBigramAddress(int *pos, bool advance) {
+ int address = 0;
+
+ address += (DICT[*pos] & 0x3F) << 16;
+ address += (DICT[*pos + 1] & 0xFF) << 8;
+ address += (DICT[*pos + 2] & 0xFF);
+
+ if (advance) {
+ *pos += 3;
+ }
+
+ return address;
+}
+
+int BigramDictionary::getBigramFreq(int *pos) {
+ int freq = DICT[(*pos)++] & FLAG_BIGRAM_FREQ;
+
+ return freq;
+}
+
+
+int BigramDictionary::getBigrams(unsigned short *prevWord, int prevWordLength, int *codes,
+ int codesSize, unsigned short *bigramChars, int *bigramFreq, int maxWordLength,
+ int maxBigrams, int maxAlternatives) {
+ mBigramFreq = bigramFreq;
+ mBigramChars = bigramChars;
+ mInputCodes = codes;
+ mInputLength = codesSize;
+ mMaxBigrams = maxBigrams;
+
+ 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) {
+ return 0;
+ }
+
+ int bigramCount = 0;
+ 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 & DICT[pos]);
+ // search for all bigrams and store them
+ searchForTerminalNode(bigramAddress, frequency);
+ nextBigramExist = (DICT[pos++] & FLAG_BIGRAM_CONTINUED);
+ bigramCount++;
+ }
+ }
+
+ return bigramCount;
+ }
+ return 0;
+}
+
+void BigramDictionary::searchForTerminalNode(int addressLookingFor, int frequency) {
+ // track word with such address and store it in an array
+ unsigned short word[MAX_WORD_LENGTH];
+
+ int pos;
+ int followDownBranchAddress = DICTIONARY_HEADER_SIZE;
+ bool found = false;
+ char followingChar = ' ';
+ int depth = -1;
+
+ while(!found) {
+ bool followDownAddressSearchStop = false;
+ bool firstAddress = true;
+ bool haveToSearchAll = true;
+
+ if (depth >= 0) {
+ word[depth] = (unsigned short) followingChar;
+ }
+ pos = followDownBranchAddress; // pos start at count
+ int count = DICT[pos] & 0xFF;
+ LOGI("count - %d\n",count);
+ pos++;
+ for (int i = 0; i < count; i++) {
+ // pos at data
+ pos++;
+ // pos now at flag
+ if (!getFirstBitOfByte(&pos)) { // non-terminal
+ if (!followDownAddressSearchStop) {
+ int addr = getBigramAddress(&pos, false);
+ if (addr > addressLookingFor) {
+ followDownAddressSearchStop = true;
+ if (firstAddress) {
+ firstAddress = false;
+ haveToSearchAll = true;
+ } else if (!haveToSearchAll) {
+ break;
+ }
+ } else {
+ followDownBranchAddress = addr;
+ followingChar = (char)(0xFF & DICT[pos-1]);
+ if (firstAddress) {
+ firstAddress = false;
+ haveToSearchAll = false;
+ }
+ }
+ }
+ pos += 3;
+ } else if (getFirstBitOfByte(&pos)) { // terminal
+ if (addressLookingFor == (pos-1)) { // found !!
+ depth++;
+ word[depth] = (0xFF & DICT[pos-1]);
+ found = true;
+ break;
+ }
+ if (getSecondBitOfByte(&pos)) { // address + freq (4 byte)
+ if (!followDownAddressSearchStop) {
+ int addr = getBigramAddress(&pos, false);
+ if (addr > addressLookingFor) {
+ followDownAddressSearchStop = true;
+ if (firstAddress) {
+ firstAddress = false;
+ haveToSearchAll = true;
+ } else if (!haveToSearchAll) {
+ break;
+ }
+ } else {
+ followDownBranchAddress = addr;
+ followingChar = (char)(0xFF & DICT[pos-1]);
+ if (firstAddress) {
+ firstAddress = false;
+ haveToSearchAll = true;
+ }
+ }
+ }
+ pos += 4;
+ } else { // freq only (2 byte)
+ pos += 2;
+ }
+
+ // 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++;
+ }
+ }
+ }
+ depth++;
+ if (followDownBranchAddress == 0) {
+ LOGI("ERROR!!! Cannot find bigram!!");
+ break;
+ }
+ }
+ if (checkFirstCharacter(word)) {
+ addWordBigram(word, depth, frequency);
+ }
+}
+
+bool BigramDictionary::checkFirstCharacter(unsigned short *word) {
+ // Checks whether this word starts with same character or neighboring characters of
+ // what user typed.
+
+ int *inputCodes = mInputCodes;
+ int maxAlt = MAX_ALTERNATIVES;
+ while (maxAlt > 0) {
+ if ((unsigned int) *inputCodes == (unsigned int) *word) {
+ return true;
+ }
+ inputCodes++;
+ maxAlt--;
+ }
+ return false;
+}
+
+// TODO: Move functions related to bigram to here
+} // namespace latinime
diff --git a/native/src/bigram_dictionary.h b/native/src/bigram_dictionary.h
new file mode 100644
index 000000000..d658b93e6
--- /dev/null
+++ b/native/src/bigram_dictionary.h
@@ -0,0 +1,55 @@
+/*
+ * Copyright (C) 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_BIGRAM_DICTIONARY_H
+#define LATINIME_BIGRAM_DICTIONARY_H
+
+namespace latinime {
+
+class Dictionary;
+class BigramDictionary {
+public:
+ BigramDictionary(const unsigned char *dict, int maxWordLength, int maxAlternatives,
+ const bool isLatestDictVersion, const bool hasBigram, Dictionary *parentDictionary);
+ int getBigrams(unsigned short *word, int length, int *codes, int codesSize,
+ unsigned short *outWords, int *frequencies, int maxWordLength, int maxBigrams,
+ int maxAlternatives);
+ ~BigramDictionary();
+private:
+ bool addWordBigram(unsigned short *word, int length, int frequency);
+ int getBigramAddress(int *pos, bool advance);
+ int getBigramFreq(int *pos);
+ void searchForTerminalNode(int addressLookingFor, int frequency);
+ bool getFirstBitOfByte(int *pos) { return (DICT[*pos] & 0x80) > 0; }
+ bool getSecondBitOfByte(int *pos) { return (DICT[*pos] & 0x40) > 0; }
+ bool checkFirstCharacter(unsigned short *word);
+
+ const unsigned char *DICT;
+ const int MAX_WORD_LENGTH;
+ const int MAX_ALTERNATIVES;
+ const bool IS_LATEST_DICT_VERSION;
+ const bool HAS_BIGRAM;
+
+ Dictionary *mParentDictionary;
+ int *mBigramFreq;
+ int mMaxBigrams;
+ unsigned short *mBigramChars;
+ int *mInputCodes;
+ int mInputLength;
+};
+// ----------------------------------------------------------------------------
+}; // namespace latinime
+#endif // LATINIME_BIGRAM_DICTIONARY_H
diff --git a/native/src/defines.h b/native/src/defines.h
new file mode 100644
index 000000000..953905fb2
--- /dev/null
+++ b/native/src/defines.h
@@ -0,0 +1,57 @@
+/*
+**
+** 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
+
+#define MAX_WORD_LENGTH_INTERNAL 64
+
+#define MAX_DEPTH_MULTIPLIER 3
+
+#endif // LATINIME_DEFINES_H
diff --git a/native/src/dictionary.cpp b/native/src/dictionary.cpp
index 1a39f585b4..6936dc928 100644
--- a/native/src/dictionary.cpp
+++ b/native/src/dictionary.cpp
@@ -16,559 +16,59 @@
*/
#include <stdio.h>
-#include <fcntl.h>
-#include <sys/mman.h>
-#include <string.h>
-//#define LOG_TAG "dictionary.cpp"
-//#include <cutils/log.h>
-#define LOGI
-#include "dictionary.h"
-#include "basechars.h"
-#include "char_utils.h"
+#define LOG_TAG "LatinIME: dictionary.cpp"
-#define DEBUG_DICT 0
-#define DICTIONARY_VERSION_MIN 200
-#define DICTIONARY_HEADER_SIZE 2
-#define NOT_VALID_WORD -99
+#include "dictionary.h"
namespace latinime {
-Dictionary::Dictionary(void *dict, int typedLetterMultiplier, int fullWordMultiplier)
-{
- mDict = (unsigned char*) dict;
- mTypedLetterMultiplier = typedLetterMultiplier;
- mFullWordMultiplier = fullWordMultiplier;
- getVersionNumber();
-}
-
-Dictionary::~Dictionary()
-{
-}
-
-int Dictionary::getSuggestions(int *codes, int codesSize, unsigned short *outWords, int *frequencies,
- int maxWordLength, int maxWords, int maxAlternatives, int skipPos,
- int *nextLetters, int nextLettersSize)
-{
- int suggWords;
- mFrequencies = frequencies;
- mOutputChars = outWords;
- mInputCodes = codes;
- mInputLength = codesSize;
- mMaxAlternatives = maxAlternatives;
- mMaxWordLength = maxWordLength;
- mMaxWords = maxWords;
- mSkipPos = skipPos;
- mMaxEditDistance = mInputLength < 5 ? 2 : mInputLength / 2;
- mNextLettersFrequencies = nextLetters;
- mNextLettersSize = nextLettersSize;
-
- if (checkIfDictVersionIsLatest()) {
- getWordsRec(DICTIONARY_HEADER_SIZE, 0, mInputLength * 3, false, 1, 0, 0);
- } else {
- getWordsRec(0, 0, mInputLength * 3, false, 1, 0, 0);
- }
-
- // Get the word count
- suggWords = 0;
- while (suggWords < mMaxWords && mFrequencies[suggWords] > 0) suggWords++;
- if (DEBUG_DICT) LOGI("Returning %d words", suggWords);
-
- if (DEBUG_DICT) {
- LOGI("Next letters: ");
- for (int k = 0; k < nextLettersSize; k++) {
- if (mNextLettersFrequencies[k] > 0) {
- LOGI("%c = %d,", k, mNextLettersFrequencies[k]);
- }
- }
- LOGI("\n");
- }
- return suggWords;
-}
-
-void
-Dictionary::registerNextLetter(unsigned short c)
-{
- if (c < mNextLettersSize) {
- mNextLettersFrequencies[c]++;
- }
-}
-
-void
-Dictionary::getVersionNumber()
-{
- mVersion = (mDict[0] & 0xFF);
- mBigram = (mDict[1] & 0xFF);
- LOGI("IN NATIVE SUGGEST Version: %d Bigram : %d \n", mVersion, mBigram);
-}
-
-// Checks whether it has the latest dictionary or the old dictionary
-bool
-Dictionary::checkIfDictVersionIsLatest()
-{
- return (mVersion >= DICTIONARY_VERSION_MIN) && (mBigram == 1 || mBigram == 0);
-}
-
-unsigned short
-Dictionary::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
-Dictionary::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
-Dictionary::getFreq(int *pos)
+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)
{
- 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)++;
- }
+ if (MAX_WORD_LENGTH_INTERNAL < maxWordLength) {
+ LOGI("Max word length (%d) is greater than %d", maxWordLength, MAX_WORD_LENGTH_INTERNAL);
}
-
- return freq;
-}
-
-int
-Dictionary::wideStrLen(unsigned short *str)
-{
- if (!str) return 0;
- unsigned short *end = str;
- while (*end)
- end++;
- return end - str;
+ LOGI("IN NATIVE SUGGEST Version: %d \n", (DICT[0] & 0xFF));
+ mUnigramDictionary = new UnigramDictionary(DICT, typedLetterMultiplier, fullWordMultiplier,
+ maxWordLength, maxWords, maxAlternatives, IS_LATEST_DICT_VERSION);
+ mBigramDictionary = new BigramDictionary(DICT, maxWordLength, maxAlternatives,
+ IS_LATEST_DICT_VERSION, hasBigram(), this);
}
-bool
-Dictionary::addWord(unsigned short *word, int length, int frequency)
-{
- word[length] = 0;
- if (DEBUG_DICT) {
- char s[length + 1];
- for (int i = 0; i <= length; i++) s[i] = word[i];
- LOGI("Found word = %s, freq = %d : \n", s, frequency);
- }
-
- // Find the right insertion point
- int insertAt = 0;
- while (insertAt < mMaxWords) {
- if (frequency > mFrequencies[insertAt]
- || (mFrequencies[insertAt] == frequency
- && length < wideStrLen(mOutputChars + insertAt * mMaxWordLength))) {
- break;
- }
- insertAt++;
- }
- if (insertAt < mMaxWords) {
- memmove((char*) mFrequencies + (insertAt + 1) * sizeof(mFrequencies[0]),
- (char*) mFrequencies + insertAt * sizeof(mFrequencies[0]),
- (mMaxWords - insertAt - 1) * sizeof(mFrequencies[0]));
- mFrequencies[insertAt] = frequency;
- memmove((char*) mOutputChars + (insertAt + 1) * mMaxWordLength * sizeof(short),
- (char*) mOutputChars + (insertAt ) * mMaxWordLength * sizeof(short),
- (mMaxWords - insertAt - 1) * sizeof(short) * mMaxWordLength);
- unsigned short *dest = mOutputChars + (insertAt ) * mMaxWordLength;
- while (length--) {
- *dest++ = *word++;
- }
- *dest = 0; // NULL terminate
- if (DEBUG_DICT) LOGI("Added word at %d\n", insertAt);
- return true;
- }
- return false;
-}
-
-bool
-Dictionary::addWordBigram(unsigned short *word, int length, int frequency)
-{
- word[length] = 0;
- if (DEBUG_DICT) {
- char s[length + 1];
- for (int i = 0; i <= length; i++) s[i] = word[i];
- LOGI("Bigram: Found word = %s, freq = %d : \n", s, frequency);
- }
-
- // Find the right insertion point
- int insertAt = 0;
- while (insertAt < mMaxBigrams) {
- if (frequency > mBigramFreq[insertAt]
- || (mBigramFreq[insertAt] == frequency
- && length < wideStrLen(mBigramChars + insertAt * mMaxWordLength))) {
- break;
- }
- insertAt++;
- }
- LOGI("Bigram: InsertAt -> %d maxBigrams: %d\n", insertAt, mMaxBigrams);
- if (insertAt < mMaxBigrams) {
- memmove((char*) mBigramFreq + (insertAt + 1) * sizeof(mBigramFreq[0]),
- (char*) mBigramFreq + insertAt * sizeof(mBigramFreq[0]),
- (mMaxBigrams - insertAt - 1) * sizeof(mBigramFreq[0]));
- mBigramFreq[insertAt] = frequency;
- memmove((char*) mBigramChars + (insertAt + 1) * mMaxWordLength * sizeof(short),
- (char*) mBigramChars + (insertAt ) * mMaxWordLength * sizeof(short),
- (mMaxBigrams - insertAt - 1) * sizeof(short) * mMaxWordLength);
- unsigned short *dest = mBigramChars + (insertAt ) * mMaxWordLength;
- while (length--) {
- *dest++ = *word++;
- }
- *dest = 0; // NULL terminate
- if (DEBUG_DICT) LOGI("Bigram: Added word at %d\n", insertAt);
- return true;
- }
- return false;
-}
-
-unsigned short
-Dictionary::toLowerCase(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;
-}
-
-bool
-Dictionary::sameAsTyped(unsigned short *word, int length)
-{
- if (length != mInputLength) {
- return false;
- }
- int *inputCodes = mInputCodes;
- while (length--) {
- if ((unsigned int) *inputCodes != (unsigned int) *word) {
- return false;
- }
- inputCodes += mMaxAlternatives;
- word++;
- }
- return true;
-}
-
-static char QUOTE = '\'';
-
-void
-Dictionary::getWordsRec(int pos, int depth, int maxDepth, bool completion, int snr, int inputIndex,
- int diffs)
-{
- // Optimization: Prune out words that are too long compared to how much was typed.
- if (depth > maxDepth) {
- return;
- }
- if (diffs > mMaxEditDistance) {
- return;
- }
- int count = getCount(&pos);
- int *currentChars = NULL;
- if (mInputLength <= inputIndex) {
- completion = true;
- } else {
- currentChars = mInputCodes + (inputIndex * mMaxAlternatives);
- }
-
- for (int i = 0; i < count; i++) {
- // -- at char
- unsigned short c = getChar(&pos);
- // -- at flag/add
- unsigned short lowerC = toLowerCase(c);
- bool terminal = getTerminal(&pos);
- int childrenAddress = getAddress(&pos);
- // -- after address or flag
- int freq = 1;
- if (terminal) freq = getFreq(&pos);
- // -- after add or freq
-
- // If we are only doing completions, no need to look at the typed characters.
- if (completion) {
- mWord[depth] = c;
- if (terminal) {
- addWord(mWord, depth + 1, freq * snr);
- if (depth >= mInputLength && mSkipPos < 0) {
- registerNextLetter(mWord[mInputLength]);
- }
- }
- if (childrenAddress != 0) {
- getWordsRec(childrenAddress, depth + 1, maxDepth,
- completion, snr, inputIndex, diffs);
- }
- } else if ((c == QUOTE && currentChars[0] != QUOTE) || mSkipPos == depth) {
- // Skip the ' or other letter and continue deeper
- mWord[depth] = c;
- if (childrenAddress != 0) {
- getWordsRec(childrenAddress, depth + 1, maxDepth, false, snr, inputIndex, diffs);
- }
- } else {
- int j = 0;
- while (currentChars[j] > 0) {
- if (currentChars[j] == lowerC || currentChars[j] == c) {
- int addedWeight = j == 0 ? mTypedLetterMultiplier : 1;
- mWord[depth] = c;
- if (mInputLength == inputIndex + 1) {
- if (terminal) {
- if (//INCLUDE_TYPED_WORD_IF_VALID ||
- !sameAsTyped(mWord, depth + 1)) {
- int finalFreq = freq * snr * addedWeight;
- if (mSkipPos < 0) finalFreq *= mFullWordMultiplier;
- addWord(mWord, depth + 1, finalFreq);
- }
- }
- if (childrenAddress != 0) {
- getWordsRec(childrenAddress, depth + 1,
- maxDepth, true, snr * addedWeight, inputIndex + 1,
- diffs + (j > 0));
- }
- } else if (childrenAddress != 0) {
- getWordsRec(childrenAddress, depth + 1, maxDepth,
- false, snr * addedWeight, inputIndex + 1, diffs + (j > 0));
- }
- }
- j++;
- if (mSkipPos >= 0) break;
- }
- }
- }
-}
-
-int
-Dictionary::getBigramAddress(int *pos, bool advance)
-{
- int address = 0;
-
- address += (mDict[*pos] & 0x3F) << 16;
- address += (mDict[*pos + 1] & 0xFF) << 8;
- address += (mDict[*pos + 2] & 0xFF);
-
- if (advance) {
- *pos += 3;
- }
-
- return address;
-}
-
-int
-Dictionary::getBigramFreq(int *pos)
-{
- int freq = mDict[(*pos)++] & FLAG_BIGRAM_FREQ;
-
- return freq;
-}
-
-
-int
-Dictionary::getBigrams(unsigned short *prevWord, int prevWordLength, int *codes, int codesSize,
- unsigned short *bigramChars, int *bigramFreq, int maxWordLength, int maxBigrams,
- int maxAlternatives)
-{
- mBigramFreq = bigramFreq;
- mBigramChars = bigramChars;
- mInputCodes = codes;
- mInputLength = codesSize;
- mMaxWordLength = maxWordLength;
- mMaxBigrams = maxBigrams;
- mMaxAlternatives = maxAlternatives;
-
- if (mBigram == 1 && checkIfDictVersionIsLatest()) {
- int pos = isValidWordRec(DICTIONARY_HEADER_SIZE, prevWord, 0, prevWordLength);
- LOGI("Pos -> %d\n", pos);
- if (pos < 0) {
- return 0;
- }
-
- int bigramCount = 0;
- int bigramExist = (mDict[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]);
- // search for all bigrams and store them
- searchForTerminalNode(bigramAddress, frequency);
- nextBigramExist = (mDict[pos++] & FLAG_BIGRAM_CONTINUED);
- bigramCount++;
- }
- }
-
- return bigramCount;
- }
- return 0;
-}
-
-void
-Dictionary::searchForTerminalNode(int addressLookingFor, int frequency)
+Dictionary::~Dictionary()
{
- // track word with such address and store it in an array
- unsigned short word[mMaxWordLength];
-
- int pos;
- int followDownBranchAddress = DICTIONARY_HEADER_SIZE;
- bool found = false;
- char followingChar = ' ';
- int depth = -1;
-
- while(!found) {
- bool followDownAddressSearchStop = false;
- bool firstAddress = true;
- bool haveToSearchAll = true;
-
- if (depth >= 0) {
- word[depth] = (unsigned short) followingChar;
- }
- pos = followDownBranchAddress; // pos start at count
- int count = mDict[pos] & 0xFF;
- LOGI("count - %d\n",count);
- pos++;
- for (int i = 0; i < count; i++) {
- // pos at data
- pos++;
- // pos now at flag
- if (!getFirstBitOfByte(&pos)) { // non-terminal
- if (!followDownAddressSearchStop) {
- int addr = getBigramAddress(&pos, false);
- if (addr > addressLookingFor) {
- followDownAddressSearchStop = true;
- if (firstAddress) {
- firstAddress = false;
- haveToSearchAll = true;
- } else if (!haveToSearchAll) {
- break;
- }
- } else {
- followDownBranchAddress = addr;
- followingChar = (char)(0xFF & mDict[pos-1]);
- if (firstAddress) {
- firstAddress = false;
- haveToSearchAll = false;
- }
- }
- }
- pos += 3;
- } else if (getFirstBitOfByte(&pos)) { // terminal
- if (addressLookingFor == (pos-1)) { // found !!
- depth++;
- word[depth] = (0xFF & mDict[pos-1]);
- found = true;
- break;
- }
- if (getSecondBitOfByte(&pos)) { // address + freq (4 byte)
- if (!followDownAddressSearchStop) {
- int addr = getBigramAddress(&pos, false);
- if (addr > addressLookingFor) {
- followDownAddressSearchStop = true;
- if (firstAddress) {
- firstAddress = false;
- haveToSearchAll = true;
- } else if (!haveToSearchAll) {
- break;
- }
- } else {
- followDownBranchAddress = addr;
- followingChar = (char)(0xFF & mDict[pos-1]);
- if (firstAddress) {
- firstAddress = false;
- haveToSearchAll = true;
- }
- }
- }
- pos += 4;
- } else { // freq only (2 byte)
- pos += 2;
- }
-
- // 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++;
- }
- }
- }
- depth++;
- if (followDownBranchAddress == 0) {
- LOGI("ERROR!!! Cannot find bigram!!");
- break;
- }
- }
- if (checkFirstCharacter(word)) {
- addWordBigram(word, depth, frequency);
- }
+ delete mUnigramDictionary;
+ delete mBigramDictionary;
}
-bool
-Dictionary::checkFirstCharacter(unsigned short *word)
-{
- // Checks whether this word starts with same character or neighboring characters of
- // what user typed.
-
- int *inputCodes = mInputCodes;
- int maxAlt = mMaxAlternatives;
- while (maxAlt > 0) {
- if ((unsigned int) *inputCodes == (unsigned int) *word) {
- return true;
- }
- inputCodes++;
- maxAlt--;
- }
- return false;
+bool Dictionary::hasBigram() {
+ return ((DICT[1] & 0xFF) == 1);
}
-bool
-Dictionary::isValidWord(unsigned short *word, int length)
+// TODO: use uint16_t instead of unsigned short
+bool Dictionary::isValidWord(unsigned short *word, int length)
{
- if (checkIfDictVersionIsLatest()) {
+ 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) {
+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 = getCount(&pos);
+ int count = Dictionary::getCount(DICT, &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);
+ 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) {
@@ -584,13 +84,11 @@ Dictionary::isValidWordRec(int pos, unsigned short *word, int offset, int length
}
}
if (terminal) {
- getFreq(&pos);
+ 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
diff --git a/native/src/dictionary.h b/native/src/dictionary.h
index d13496e01..da876242d 100644
--- a/native/src/dictionary.h
+++ b/native/src/dictionary.h
@@ -17,90 +17,134 @@
#ifndef LATINIME_DICTIONARY_H
#define LATINIME_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
+#include "bigram_dictionary.h"
+#include "defines.h"
+#include "unigram_dictionary.h"
-#define FLAG_BIGRAM_READ 0x80
-#define FLAG_BIGRAM_CHILDEXIST 0x40
-#define FLAG_BIGRAM_CONTINUED 0x80
-#define FLAG_BIGRAM_FREQ 0x7F
+namespace latinime {
class Dictionary {
public:
- Dictionary(void *dict, int typedLetterMultipler, int fullWordMultiplier);
+ Dictionary(void *dict, int typedLetterMultipler, int fullWordMultiplier, int maxWordLength,
+ int maxWords, int maxAlternatives);
int getSuggestions(int *codes, int codesSize, unsigned short *outWords, int *frequencies,
- int maxWordLength, int maxWords, int maxAlternatives, int skipPos,
- int *nextLetters, int nextLettersSize);
+ int *nextLetters, int nextLettersSize) {
+ return mUnigramDictionary->getSuggestions(codes, codesSize, outWords, frequencies,
+ nextLetters, nextLettersSize);
+ }
+
+ // TODO: Call mBigramDictionary instead of mUnigramDictionary
int getBigrams(unsigned short *word, int length, int *codes, int codesSize,
unsigned short *outWords, int *frequencies, int maxWordLength, int maxBigrams,
- int maxAlternatives);
+ int maxAlternatives) {
+ return mBigramDictionary->getBigrams(word, length, codes, codesSize, outWords, frequencies,
+ maxWordLength, maxBigrams, maxAlternatives);
+ }
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();
-private:
+ // 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);
+ static int wideStrLen(unsigned short *str);
+ // returns next sibling's position
+ static int setDictionaryValues(const unsigned char *dict, const bool isLatestDictVersion,
+ const int pos, unsigned short *c, int *childrenPosition,
+ bool *terminal, int *freq);
- void getVersionNumber();
- bool checkIfDictVersionIsLatest();
- int getAddress(int *pos);
- int getBigramAddress(int *pos, bool advance);
- int getFreq(int *pos);
- 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);
- int wideStrLen(unsigned short *str);
-
- bool sameAsTyped(unsigned short *word, int length);
- bool checkFirstCharacter(unsigned short *word);
- bool addWord(unsigned short *word, int length, int frequency);
- bool addWordBigram(unsigned short *word, int length, int frequency);
- unsigned short toLowerCase(unsigned short c);
- void getWordsRec(int pos, int depth, int maxDepth, bool completion, int frequency,
- int inputIndex, int diffs);
- int isValidWordRec(int pos, unsigned short *word, int offset, int length);
- void registerNextLetter(unsigned short c);
+private:
+ bool hasBigram();
- unsigned char *mDict;
+ const unsigned char *DICT;
+ const bool IS_LATEST_DICT_VERSION;
void *mAsset;
-
- int *mFrequencies;
- int *mBigramFreq;
- int mMaxWords;
- int mMaxBigrams;
- int mMaxWordLength;
- unsigned short *mOutputChars;
- unsigned short *mBigramChars;
- int *mInputCodes;
- int mInputLength;
- int mMaxAlternatives;
- unsigned short mWord[128];
- int mSkipPos;
- int mMaxEditDistance;
-
- int mFullWordMultiplier;
- int mTypedLetterMultiplier;
- int *mNextLettersFrequencies;
- int mNextLettersSize;
- int mVersion;
- int mBigram;
+ 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;
+}
+
+
+inline int Dictionary::wideStrLen(unsigned short *str) {
+ if (!str) return 0;
+ unsigned short *end = str;
+ while (*end)
+ end++;
+ return end - str;
+}
+
+inline int Dictionary::setDictionaryValues(const unsigned char *dict,
+ const bool isLatestDictVersion, const int pos, unsigned short *c,int *childrenPosition,
+ bool *terminal, int *freq) {
+ int position = pos;
+ // -- at char
+ *c = Dictionary::getChar(dict, &position);
+ // -- at flag/add
+ *terminal = Dictionary::getTerminal(dict, &position);
+ *childrenPosition = Dictionary::getAddress(dict, &position);
+ // -- after address or flag
+ *freq = (*terminal) ? Dictionary::getFreq(dict, isLatestDictVersion, &position) : 1;
+ // returns next sibling's position
+ return position;
+}
}; // namespace latinime
-
#endif // LATINIME_DICTIONARY_H
diff --git a/native/src/unigram_dictionary.cpp b/native/src/unigram_dictionary.cpp
new file mode 100644
index 000000000..fa4e29632
--- /dev/null
+++ b/native/src/unigram_dictionary.cpp
@@ -0,0 +1,321 @@
+/*
+**
+** 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.
+*/
+
+#include <assert.h>
+#include <stdio.h>
+#include <fcntl.h>
+#include <string.h>
+
+#define LOG_TAG "LatinIME: unigram_dictionary.cpp"
+
+#include "basechars.h"
+#include "char_utils.h"
+#include "dictionary.h"
+#include "unigram_dictionary.h"
+
+namespace latinime {
+
+UnigramDictionary::UnigramDictionary(const unsigned char *dict, int typedLetterMultiplier,
+ int fullWordMultiplier, int maxWordLength, int maxWords, int maxAlternatives,
+ const bool isLatestDictVersion)
+ : DICT(dict), MAX_WORD_LENGTH(maxWordLength),MAX_WORDS(maxWords),
+ MAX_ALTERNATIVES(maxAlternatives), IS_LATEST_DICT_VERSION(isLatestDictVersion),
+ TYPED_LETTER_MULTIPLIER(typedLetterMultiplier), FULL_WORD_MULTIPLIER(fullWordMultiplier) {
+ LOGI("UnigramDictionary - constructor");
+}
+
+UnigramDictionary::~UnigramDictionary() {}
+
+int UnigramDictionary::getSuggestions(int *codes, int codesSize, unsigned short *outWords,
+ int *frequencies, int *nextLetters, int nextLettersSize)
+{
+
+ initSuggestions(codes, codesSize, outWords, frequencies);
+
+ int suggestedWordsCount = getSuggestionCandidates(codesSize, -1, nextLetters,
+ nextLettersSize);
+
+ // If there aren't sufficient suggestions, search for words by allowing wild cards at
+ // the different character positions. This feature is not ready for prime-time as we need
+ // to figure out the best ranking for such words compared to proximity corrections and
+ // completions.
+ if (SUGGEST_MISSING_CHARACTERS && suggestedWordsCount < SUGGEST_MISSING_CHARACTERS_THRESHOLD) {
+ for (int i = 0; i < codesSize; ++i) {
+ int tempCount = getSuggestionCandidates(codesSize, i, NULL, 0);
+ if (tempCount > suggestedWordsCount) {
+ suggestedWordsCount = tempCount;
+ break;
+ }
+ }
+ }
+
+ if (DEBUG_DICT) {
+ LOGI("Returning %d words", suggestedWordsCount);
+ LOGI("Next letters: ");
+ for (int k = 0; k < nextLettersSize; k++) {
+ if (nextLetters[k] > 0) {
+ LOGI("%c = %d,", k, nextLetters[k]);
+ }
+ }
+ LOGI("\n");
+ }
+ return suggestedWordsCount;
+}
+
+void UnigramDictionary::initSuggestions(int *codes, int codesSize, unsigned short *outWords,
+ int *frequencies) {
+ mFrequencies = frequencies;
+ mOutputChars = outWords;
+ mInputCodes = codes;
+ mInputLength = codesSize;
+ mMaxEditDistance = mInputLength < 5 ? 2 : mInputLength / 2;
+}
+
+int UnigramDictionary::getSuggestionCandidates(int inputLength, int skipPos,
+ int *nextLetters, int nextLettersSize) {
+ int initialPos = 0;
+ if (IS_LATEST_DICT_VERSION) {
+ initialPos = DICTIONARY_HEADER_SIZE;
+ }
+ getWords(initialPos, inputLength, skipPos, nextLetters, nextLettersSize);
+
+ // Get the word count
+ int suggestedWordsCount = 0;
+ while (suggestedWordsCount < MAX_WORDS && mFrequencies[suggestedWordsCount] > 0) {
+ suggestedWordsCount++;
+ }
+ return suggestedWordsCount;
+}
+
+void UnigramDictionary::registerNextLetter(
+ unsigned short c, int *nextLetters, int nextLettersSize) {
+ if (c < nextLettersSize) {
+ nextLetters[c]++;
+ }
+}
+
+bool UnigramDictionary::addWord(unsigned short *word, int length, int frequency) {
+ word[length] = 0;
+ if (DEBUG_DICT) {
+ char s[length + 1];
+ for (int i = 0; i <= length; i++) s[i] = word[i];
+ LOGI("Found word = %s, freq = %d : \n", s, frequency);
+ }
+
+ // Find the right insertion point
+ int insertAt = 0;
+ while (insertAt < MAX_WORDS) {
+ if (frequency > mFrequencies[insertAt] || (mFrequencies[insertAt] == frequency
+ && length < Dictionary::wideStrLen(mOutputChars + insertAt * MAX_WORD_LENGTH))) {
+ break;
+ }
+ insertAt++;
+ }
+ if (insertAt < MAX_WORDS) {
+ 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),
+ (char*) mOutputChars + insertAt * MAX_WORD_LENGTH * sizeof(short),
+ (MAX_WORDS - insertAt - 1) * sizeof(short) * MAX_WORD_LENGTH);
+ unsigned short *dest = mOutputChars + insertAt * MAX_WORD_LENGTH;
+ while (length--) {
+ *dest++ = *word++;
+ }
+ *dest = 0; // NULL terminate
+ if (DEBUG_DICT) LOGI("Added word at %d\n", insertAt);
+ return true;
+ }
+ return false;
+}
+
+unsigned short UnigramDictionary::toLowerCase(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;
+}
+
+bool UnigramDictionary::sameAsTyped(unsigned short *word, int length) {
+ if (length != mInputLength) {
+ return false;
+ }
+ int *inputCodes = mInputCodes;
+ while (length--) {
+ if ((unsigned int) *inputCodes != (unsigned int) *word) {
+ return false;
+ }
+ inputCodes += MAX_ALTERNATIVES;
+ word++;
+ }
+ return true;
+}
+
+static const char QUOTE = '\'';
+
+void UnigramDictionary::getWords(const int initialPos, const int inputLength, const int skipPos,
+ int *nextLetters, const int nextLettersSize) {
+ int initialPosition = initialPos;
+ const int count = Dictionary::getCount(DICT, &initialPosition);
+ getWordsRec(count, initialPosition, 0, inputLength * MAX_DEPTH_MULTIPLIER,
+ mInputLength <= 0, 1, 0, 0, skipPos, nextLetters, nextLettersSize);
+}
+
+// snr : frequency?
+void UnigramDictionary::getWordsRec(const int childrenCount, const int pos, const int depth,
+ const int maxDepth, const bool traverseAllNodes, const int snr, const int inputIndex,
+ const int diffs, const int skipPos, int *nextLetters, const int nextLettersSize) {
+ int siblingPos = pos;
+ for (int i = 0; i < childrenCount; ++i) {
+ int newCount;
+ int newChildPosition;
+ int newDepth;
+ bool newTraverseAllNodes;
+ int newSnr;
+ int newInputIndex;
+ int newDiffs;
+ int newSiblingPos;
+ const bool needsToTraverseChildrenNodes = processCurrentNode(siblingPos, depth, maxDepth,
+ traverseAllNodes, snr, inputIndex, diffs, skipPos, nextLetters, nextLettersSize,
+ &newCount, &newChildPosition, &newDepth, &newTraverseAllNodes, &newSnr,
+ &newInputIndex, &newDiffs, &newSiblingPos);
+ siblingPos = newSiblingPos;
+
+ if (needsToTraverseChildrenNodes) {
+ getWordsRec(newCount, newChildPosition, newDepth, maxDepth, newTraverseAllNodes,
+ newSnr, newInputIndex, newDiffs, skipPos, nextLetters, nextLettersSize);
+ }
+ }
+}
+
+inline void UnigramDictionary::onTerminalWhenUserTypedLengthIsGreaterThanInputLength(
+ unsigned short *word, const int inputLength, const int depth, const int snr,
+ int *nextLetters, const int nextLettersSize, const int skipPos, const int freq) {
+ addWord(word, depth + 1, freq * snr);
+ if (depth >= inputLength && skipPos < 0) {
+ registerNextLetter(mWord[mInputLength], nextLetters, nextLettersSize);
+ }
+}
+
+inline void UnigramDictionary::onTerminalWhenUserTypedLengthIsSameAsInputLength(
+ unsigned short *word, const int depth, const int snr, const int skipPos, const int freq,
+ const int addedWeight) {
+ if (!sameAsTyped(word, depth + 1)) {
+ int finalFreq = freq * snr * addedWeight;
+ // Proximity collection will promote a word of the same length as
+ // what user typed.
+ if (skipPos < 0) finalFreq *= FULL_WORD_MULTIPLIER;
+ addWord(word, depth + 1, finalFreq);
+ }
+}
+
+inline bool UnigramDictionary::needsToSkipCurrentNode(const unsigned short c,
+ const int inputIndex, const int skipPos, const int depth) {
+ const unsigned short userTypedChar = (mInputCodes + (inputIndex * MAX_ALTERNATIVES))[0];
+ // Skip the ' or other letter and continue deeper
+ return (c == QUOTE && userTypedChar != QUOTE) || skipPos == depth;
+}
+
+inline int UnigramDictionary::getMatchedProximityId(const int *currentChars,
+ const unsigned short c, const int skipPos) {
+ const unsigned short lowerC = toLowerCase(c);
+ int j = 0;
+ while (currentChars[j] > 0) {
+ const bool matched = (currentChars[j] == lowerC || currentChars[j] == c);
+ // If skipPos is defined, not to search proximity collections.
+ // First char is what user typed.
+ if (matched) {
+ return j;
+ } else if (skipPos >= 0) {
+ return -1;
+ }
+ ++j;
+ }
+ return -1;
+}
+
+inline bool UnigramDictionary::processCurrentNode(const int pos, const int depth,
+ const int maxDepth, const bool traverseAllNodes, const int snr, const int inputIndex,
+ const int diffs, const int skipPos, int *nextLetters, const int nextLettersSize,
+ int *newCount, int *newChildPosition, int *newDepth, bool *newTraverseAllNodes,
+ int *newSnr, int*newInputIndex, int *newDiffs, int *nextSiblingPosition) {
+ unsigned short c;
+ int childPosition;
+ bool terminal;
+ int freq;
+ *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) {
+ onTerminalWhenUserTypedLengthIsGreaterThanInputLength(mWord, mInputLength, depth,
+ snr, nextLetters, nextLettersSize, skipPos, freq);
+ }
+ if (!needsToTraverseChildrenNodes) return false;
+ *newTraverseAllNodes = traverseAllNodes;
+ *newSnr = snr;
+ *newDiffs = diffs;
+ *newInputIndex = inputIndex;
+ *newDepth = depth + 1;
+ } else {
+ int *currentChars = mInputCodes + (inputIndex * MAX_ALTERNATIVES);
+ int matchedProximityCharId = getMatchedProximityId(currentChars, c, skipPos);
+ if (matchedProximityCharId < 0) 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.
+ const int addedWeight = matchedProximityCharId == 0 ? TYPED_LETTER_MULTIPLIER : 1;
+ const bool isSameAsUserTypedLength = mInputLength == inputIndex + 1;
+ if (isSameAsUserTypedLength && terminal) {
+ onTerminalWhenUserTypedLengthIsSameAsInputLength(mWord, depth, snr,
+ skipPos, freq, addedWeight);
+ }
+ if (!needsToTraverseChildrenNodes) return false;
+ // Start traversing all nodes after the index exceeds the user typed length
+ *newTraverseAllNodes = isSameAsUserTypedLength;
+ *newSnr = snr * addedWeight;
+ *newDiffs = diffs + (matchedProximityCharId > 0);
+ *newInputIndex = inputIndex + 1;
+ *newDepth = depth + 1;
+ }
+ // Optimization: Prune out words that are too long compared to how much was typed.
+ if (*newDepth > maxDepth || *newDiffs > mMaxEditDistance) {
+ return false;
+ }
+
+ // If inputIndex is greater than mInputLength, that means there are no proximity chars.
+ if (mInputLength <= *newInputIndex) {
+ *newTraverseAllNodes = true;
+ }
+ // get the count of nodes and increment childAddress.
+ *newCount = Dictionary::getCount(DICT, &childPosition);
+ *newChildPosition = childPosition;
+ if (DEBUG_DICT) assert(needsToTraverseChildrenNodes);
+ return needsToTraverseChildrenNodes;
+}
+
+} // namespace latinime
diff --git a/native/src/unigram_dictionary.h b/native/src/unigram_dictionary.h
new file mode 100644
index 000000000..c02d366de
--- /dev/null
+++ b/native/src/unigram_dictionary.h
@@ -0,0 +1,84 @@
+/*
+ * Copyright (C) 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_UNIGRAM_DICTIONARY_H
+#define LATINIME_UNIGRAM_DICTIONARY_H
+
+#include "defines.h"
+
+namespace latinime {
+
+class UnigramDictionary {
+public:
+ UnigramDictionary(const unsigned char *dict, int typedLetterMultipler, int fullWordMultiplier,
+ int maxWordLength, int maxWords, int maxAlternatives, const bool isLatestDictVersion);
+ int getSuggestions(int *codes, int codesSize, unsigned short *outWords, int *frequencies,
+ int *nextLetters, int nextLettersSize);
+ ~UnigramDictionary();
+
+private:
+ void initSuggestions(int *codes, int codesSize, unsigned short *outWords, int *frequencies);
+ int getSuggestionCandidates(int inputLength, int skipPos, int *nextLetters,
+ int nextLettersSize);
+ void getVersionNumber();
+ bool checkIfDictVersionIsLatest();
+ int getAddress(int *pos);
+ int getFreq(int *pos);
+ int wideStrLen(unsigned short *str);
+ bool sameAsTyped(unsigned short *word, int length);
+ bool addWord(unsigned short *word, int length, int frequency);
+ unsigned short toLowerCase(unsigned short c);
+ void getWordsRec(const int childrenCount, const int pos, const int depth, const int maxDepth,
+ const bool traverseAllNodes, const int snr, const int inputIndex, const int diffs,
+ const int skipPos, int *nextLetters, const int nextLettersSize);
+ void getWords(const int initialPos, const int inputLength, const int skipPos, int *nextLetters,
+ const int nextLettersSize);
+ void registerNextLetter(unsigned short c, int *nextLetters, int nextLettersSize);
+ void onTerminalWhenUserTypedLengthIsGreaterThanInputLength(unsigned short *word,
+ const int mInputLength, const int depth, const int snr, int *nextLetters,
+ const int nextLettersSize, const int skipPos, const int freq);
+ void onTerminalWhenUserTypedLengthIsSameAsInputLength(unsigned short *word, const int depth,
+ const int snr, const int skipPos, const int freq, const int addedWeight);
+ bool needsToSkipCurrentNode(const unsigned short c,
+ const int inputIndex, const int skipPos, const int depth);
+ int getMatchedProximityId(const int *currentChars, const unsigned short c, const int skipPos);
+ bool processCurrentNode(const int pos, const int depth,
+ const int maxDepth, const bool traverseAllNodes, const int snr, const int inputIndex,
+ const int diffs, const int skipPos, int *nextLetters, const int nextLettersSize,
+ int *newCount, int *newChildPosition, int *newDepth, bool *newTraverseAllNodes,
+ int *newSnr, int*newInputIndex, int *newDiffs, int *nextSiblingPosition);
+ const unsigned char *DICT;
+ const int MAX_WORDS;
+ const int MAX_WORD_LENGTH;
+ const int MAX_ALTERNATIVES;
+ const bool IS_LATEST_DICT_VERSION;
+ const int TYPED_LETTER_MULTIPLIER;
+ const int FULL_WORD_MULTIPLIER;
+
+ int *mFrequencies;
+ unsigned short *mOutputChars;
+ int *mInputCodes;
+ int mInputLength;
+ // MAX_WORD_LENGTH_INTERNAL must be bigger than MAX_WORD_LENGTH
+ unsigned short mWord[MAX_WORD_LENGTH_INTERNAL];
+ int mMaxEditDistance;
+};
+
+// ----------------------------------------------------------------------------
+
+}; // namespace latinime
+
+#endif // LATINIME_UNIGRAM_DICTIONARY_H
generated by cgit v1.2.3-83-g751a (git 2.47.1) at 2025-08-15 19:55:03 +0000