aboutsummaryrefslogtreecommitdiffstats
path: root/native/src
diff options
context:
space:
mode:
Diffstat (limited to 'native/src')
-rw-r--r--native/src/defines.h7
-rw-r--r--native/src/proximity_info.cpp42
-rw-r--r--native/src/proximity_info.h15
-rw-r--r--native/src/unigram_dictionary.cpp124
-rw-r--r--native/src/unigram_dictionary.h7
5 files changed, 139 insertions, 56 deletions
diff --git a/native/src/defines.h b/native/src/defines.h
index 0d1f037e1..16927e5bb 100644
--- a/native/src/defines.h
+++ b/native/src/defines.h
@@ -129,10 +129,13 @@ static void prof_out(void) {
#define DICTIONARY_HEADER_SIZE 2
#define NOT_VALID_WORD -99
+#define KEYCODE_SPACE ' '
+
#define SUGGEST_WORDS_WITH_MISSING_CHARACTER true
#define SUGGEST_WORDS_WITH_MISSING_SPACE_CHARACTER true
#define SUGGEST_WORDS_WITH_EXCESSIVE_CHARACTER true
#define SUGGEST_WORDS_WITH_TRANSPOSED_CHARACTERS true
+#define SUGGEST_WORDS_WITH_SPACE_PROXIMITY true
// The following "rate"s are used as a multiplier before dividing by 100, so they are in percent.
#define WORDS_WITH_MISSING_CHARACTER_DEMOTION_RATE 100
@@ -148,6 +151,10 @@ static void prof_out(void) {
#define MAX_DEPTH_MULTIPLIER 3
+// TODO: Reduce this constant if possible; check the maximum number of umlauts in the same German
+// word in the dictionary
+#define DEFAULT_MAX_UMLAUT_SEARCH_DEPTH 5
+
// Minimum suggest depth for one word for all cases except for missing space suggestions.
#define MIN_SUGGEST_DEPTH 1
#define MIN_USER_TYPED_LENGTH_FOR_MISSING_SPACE_SUGGESTION 3
diff --git a/native/src/proximity_info.cpp b/native/src/proximity_info.cpp
index d0cba3eb6..5f2d09f5c 100644
--- a/native/src/proximity_info.cpp
+++ b/native/src/proximity_info.cpp
@@ -17,18 +17,48 @@
#include <stdio.h>
#include <string.h>
+#define LOG_TAG "LatinIME: proximity_info.cpp"
+
#include "proximity_info.h"
namespace latinime {
-ProximityInfo::ProximityInfo(int maxProximityCharsSize, int displayWidth, int displayHeight,
- int gridWidth, int gridHeight, uint32_t const *proximityCharsArray)
- : MAX_PROXIMITY_CHARS_SIZE(maxProximityCharsSize), DISPLAY_WIDTH(displayWidth),
- DISPLAY_HEIGHT(displayHeight), GRID_WIDTH(gridWidth), GRID_HEIGHT(gridHeight) {
- mProximityCharsArray = new uint32_t[GRID_WIDTH * GRID_HEIGHT * MAX_PROXIMITY_CHARS_SIZE];
- memcpy(mProximityCharsArray, proximityCharsArray, sizeof(mProximityCharsArray));
+ProximityInfo::ProximityInfo(const int maxProximityCharsSize, const int keyboardWidth,
+ const int keyboardHeight, const int gridWidth, const int gridHeight,
+ const uint32_t *proximityCharsArray)
+ : MAX_PROXIMITY_CHARS_SIZE(maxProximityCharsSize), KEYBOARD_WIDTH(keyboardWidth),
+ KEYBOARD_HEIGHT(keyboardHeight), GRID_WIDTH(gridWidth), GRID_HEIGHT(gridHeight),
+ CELL_WIDTH((keyboardWidth + gridWidth - 1) / gridWidth),
+ CELL_HEIGHT((keyboardHeight + gridHeight - 1) / gridHeight) {
+ const int len = GRID_WIDTH * GRID_HEIGHT * MAX_PROXIMITY_CHARS_SIZE;
+ mProximityCharsArray = new uint32_t[len];
+ if (DEBUG_PROXIMITY_INFO) {
+ LOGI("Create proximity info array %d", len);
+ }
+ memcpy(mProximityCharsArray, proximityCharsArray, len * sizeof(mProximityCharsArray[0]));
}
ProximityInfo::~ProximityInfo() {
delete[] mProximityCharsArray;
}
+
+inline int ProximityInfo::getStartIndexFromCoordinates(const int x, const int y) const {
+ return (y / CELL_HEIGHT) * GRID_WIDTH + (x / CELL_WIDTH)
+ * MAX_PROXIMITY_CHARS_SIZE;
+}
+
+bool ProximityInfo::hasSpaceProximity(const int x, const int y) const {
+ const int startIndex = getStartIndexFromCoordinates(x, y);
+ if (DEBUG_PROXIMITY_INFO) {
+ LOGI("hasSpaceProximity: index %d", startIndex);
+ }
+ for (int i = 0; i < MAX_PROXIMITY_CHARS_SIZE; ++i) {
+ if (DEBUG_PROXIMITY_INFO) {
+ LOGI("Index: %d", mProximityCharsArray[startIndex + i]);
+ }
+ if (mProximityCharsArray[startIndex + i] == KEYCODE_SPACE) {
+ return true;
+ }
+ }
+ return false;
}
+} // namespace latinime
diff --git a/native/src/proximity_info.h b/native/src/proximity_info.h
index 54e995059..0f1201866 100644
--- a/native/src/proximity_info.h
+++ b/native/src/proximity_info.h
@@ -25,15 +25,20 @@ namespace latinime {
class ProximityInfo {
public:
- ProximityInfo(int maxProximityCharsSize, int displayWidth, int displayHeight, int gridWidth,
- int gridHeight, uint32_t const *proximityCharsArray);
+ ProximityInfo(const int maxProximityCharsSize, const int keyboardWidth,
+ const int keybaordHeight, const int gridWidth, const int gridHeight,
+ const uint32_t *proximityCharsArray);
~ProximityInfo();
+ bool hasSpaceProximity(const int x, const int y) const;
private:
- const int MAX_PROXIMITY_CHARS_SIZE;
- const int DISPLAY_WIDTH;
- const int DISPLAY_HEIGHT;
+ int getStartIndexFromCoordinates(const int x, const int y) const;
+ const int CELL_WIDTH;
+ const int CELL_HEIGHT;
+ const int KEYBOARD_WIDTH;
+ const int KEYBOARD_HEIGHT;
const int GRID_WIDTH;
const int GRID_HEIGHT;
+ const int MAX_PROXIMITY_CHARS_SIZE;
uint32_t *mProximityCharsArray;
};
}; // namespace latinime
diff --git a/native/src/unigram_dictionary.cpp b/native/src/unigram_dictionary.cpp
index 17a87a708..3487d4f11 100644
--- a/native/src/unigram_dictionary.cpp
+++ b/native/src/unigram_dictionary.cpp
@@ -41,7 +41,8 @@ UnigramDictionary::UnigramDictionary(const unsigned char *dict, int typedLetterM
MAX_PROXIMITY_CHARS(maxProximityChars), IS_LATEST_DICT_VERSION(isLatestDictVersion),
TYPED_LETTER_MULTIPLIER(typedLetterMultiplier), FULL_WORD_MULTIPLIER(fullWordMultiplier),
ROOT_POS(isLatestDictVersion ? DICTIONARY_HEADER_SIZE : 0),
- BYTES_IN_ONE_CHAR(MAX_PROXIMITY_CHARS * sizeof(*mInputCodes)) {
+ BYTES_IN_ONE_CHAR(MAX_PROXIMITY_CHARS * sizeof(*mInputCodes)),
+ MAX_UMLAUT_SEARCH_DEPTH(DEFAULT_MAX_UMLAUT_SEARCH_DEPTH) {
if (DEBUG_DICT) LOGI("UnigramDictionary - constructor");
}
@@ -80,30 +81,37 @@ bool UnigramDictionary::isDigraph(const int* codes, const int i, const int codes
void UnigramDictionary::getWordWithDigraphSuggestionsRec(const ProximityInfo *proximityInfo,
const int *xcoordinates, const int* ycoordinates, const int *codesBuffer,
const int codesBufferSize, const int flags, const int* codesSrc, const int codesRemain,
- int* codesDest, unsigned short* outWords, int* frequencies) {
-
- for (int i = 0; i < codesRemain; ++i) {
- if (isDigraph(codesSrc, i, codesRemain)) {
- // Found a digraph. We will try both spellings. eg. the word is "pruefen"
-
- // Copy the word up to the first char of the digraph, then continue processing
- // on the remaining part of the word, skipping the second char of the digraph.
- // In our example, copy "pru" and continue running on "fen"
- memcpy(codesDest, codesSrc, i * BYTES_IN_ONE_CHAR);
- getWordWithDigraphSuggestionsRec(proximityInfo, xcoordinates, ycoordinates, codesBuffer,
- codesBufferSize, flags, codesSrc + (i + 1) * MAX_PROXIMITY_CHARS,
- codesRemain - i - 1, codesDest + i * MAX_PROXIMITY_CHARS,
- outWords, frequencies);
-
- // Copy the second char of the digraph in place, then continue processing on
- // the remaining part of the word.
- // In our example, after "pru" in the buffer copy the "e", and continue running on "fen"
- memcpy(codesDest + i * MAX_PROXIMITY_CHARS, codesSrc + i * MAX_PROXIMITY_CHARS,
- BYTES_IN_ONE_CHAR);
- getWordWithDigraphSuggestionsRec(proximityInfo, xcoordinates, ycoordinates, codesBuffer,
- codesBufferSize, flags, codesSrc + i * MAX_PROXIMITY_CHARS, codesRemain - i,
- codesDest + i * MAX_PROXIMITY_CHARS, outWords, frequencies);
- return;
+ int currentDepth, int* codesDest, unsigned short* outWords, int* frequencies) {
+
+ if (currentDepth < MAX_UMLAUT_SEARCH_DEPTH) {
+ for (int i = 0; i < codesRemain; ++i) {
+ if (isDigraph(codesSrc, i, codesRemain)) {
+ // Found a digraph. We will try both spellings. eg. the word is "pruefen"
+
+ // Copy the word up to the first char of the digraph, then continue processing
+ // on the remaining part of the word, skipping the second char of the digraph.
+ // In our example, copy "pru" and continue running on "fen"
+ // Make i the index of the second char of the digraph for simplicity. Forgetting
+ // to do that results in an infinite recursion so take care!
+ ++i;
+ memcpy(codesDest, codesSrc, i * BYTES_IN_ONE_CHAR);
+ getWordWithDigraphSuggestionsRec(proximityInfo, xcoordinates, ycoordinates,
+ codesBuffer, codesBufferSize, flags,
+ codesSrc + (i + 1) * MAX_PROXIMITY_CHARS, codesRemain - i - 1,
+ currentDepth + 1, codesDest + i * MAX_PROXIMITY_CHARS, outWords,
+ frequencies);
+
+ // Copy the second char of the digraph in place, then continue processing on
+ // the remaining part of the word.
+ // In our example, after "pru" in the buffer copy the "e", and continue on "fen"
+ memcpy(codesDest + i * MAX_PROXIMITY_CHARS, codesSrc + i * MAX_PROXIMITY_CHARS,
+ BYTES_IN_ONE_CHAR);
+ getWordWithDigraphSuggestionsRec(proximityInfo, xcoordinates, ycoordinates,
+ codesBuffer, codesBufferSize, flags, codesSrc + i * MAX_PROXIMITY_CHARS,
+ codesRemain - i, currentDepth + 1, codesDest + i * MAX_PROXIMITY_CHARS,
+ outWords, frequencies);
+ return;
+ }
}
}
@@ -128,13 +136,13 @@ int UnigramDictionary::getSuggestions(const ProximityInfo *proximityInfo, const
{ // Incrementally tune the word and try all possibilities
int codesBuffer[getCodesBufferSize(codes, codesSize, MAX_PROXIMITY_CHARS)];
getWordWithDigraphSuggestionsRec(proximityInfo, xcoordinates, ycoordinates, codesBuffer,
- codesSize, flags, codes, codesSize, codesBuffer, outWords, frequencies);
+ codesSize, flags, codes, codesSize, 0, codesBuffer, outWords, frequencies);
} else { // Normal processing
getWordSuggestions(proximityInfo, xcoordinates, ycoordinates, codes, codesSize,
outWords, frequencies);
}
- PROF_START(6);
+ PROF_START(20);
// Get the word count
int suggestedWordsCount = 0;
while (suggestedWordsCount < MAX_WORDS && mFrequencies[suggestedWordsCount] > 0) {
@@ -150,7 +158,7 @@ int UnigramDictionary::getSuggestions(const ProximityInfo *proximityInfo, const
}
}
}
- PROF_END(6);
+ PROF_END(20);
PROF_CLOSE;
return suggestedWordsCount;
}
@@ -164,12 +172,6 @@ void UnigramDictionary::getWordSuggestions(const ProximityInfo *proximityInfo,
initSuggestions(codes, codesSize, outWords, frequencies);
if (DEBUG_DICT) assert(codesSize == mInputLength);
- if (DEBUG_PROXIMITY_INFO) {
- for (int i = 0; i < codesSize; ++i) {
- LOGI("Input[%d] x = %d, y = %d", i, xcoordinates[i], ycoordinates[i]);
- }
- }
-
const int MAX_DEPTH = min(mInputLength * MAX_DEPTH_MULTIPLIER, MAX_WORD_LENGTH);
PROF_END(0);
@@ -219,6 +221,25 @@ void UnigramDictionary::getWordSuggestions(const ProximityInfo *proximityInfo,
}
}
PROF_END(5);
+
+ PROF_START(6);
+ if (SUGGEST_WORDS_WITH_SPACE_PROXIMITY) {
+ // The first and last "mistyped spaces" are taken care of by excessive character handling
+ for (int i = 1; i < codesSize - 1; ++i) {
+ if (DEBUG_DICT) LOGI("--- Suggest words with proximity space %d", i);
+ const int x = xcoordinates[i];
+ const int y = ycoordinates[i];
+ if (DEBUG_PROXIMITY_INFO)
+ LOGI("Input[%d] x = %d, y = %d, has space proximity = %d",
+ i, x, y, proximityInfo->hasSpaceProximity(x, y));
+
+ if (proximityInfo->hasSpaceProximity(x, y)) {
+ getMistypedSpaceWords(mInputLength, i);
+ }
+
+ }
+ }
+ PROF_END(6);
}
void UnigramDictionary::initSuggestions(const int *codes, const int codesSize,
@@ -379,27 +400,31 @@ inline static void multiplyRate(const int rate, int *freq) {
}
}
-bool UnigramDictionary::getMissingSpaceWords(const int inputLength, const int missingSpacePos) {
- if (missingSpacePos <= 0 || missingSpacePos >= inputLength
- || inputLength >= MAX_WORD_LENGTH) return false;
- const int newWordLength = inputLength + 1;
+bool UnigramDictionary::getSplitTwoWordsSuggestion(const int inputLength,
+ const int firstWordStartPos, const int firstWordLength, const int secondWordStartPos,
+ const int secondWordLength) {
+ if (inputLength >= MAX_WORD_LENGTH) return false;
+ if (0 >= firstWordLength || 0 >= secondWordLength || firstWordStartPos >= secondWordStartPos
+ || firstWordStartPos < 0 || secondWordStartPos >= inputLength)
+ return false;
+ const int newWordLength = firstWordLength + secondWordLength + 1;
// Allocating variable length array on stack
unsigned short word[newWordLength];
- const int firstFreq = getBestWordFreq(0, missingSpacePos, mWord);
+ const int firstFreq = getBestWordFreq(firstWordStartPos, firstWordLength, mWord);
if (DEBUG_DICT) LOGI("First freq: %d", firstFreq);
if (firstFreq <= 0) return false;
- for (int i = 0; i < missingSpacePos; ++i) {
+ for (int i = 0; i < firstWordLength; ++i) {
word[i] = mWord[i];
}
- const int secondFreq = getBestWordFreq(missingSpacePos, inputLength - missingSpacePos, mWord);
+ const int secondFreq = getBestWordFreq(secondWordStartPos, secondWordLength, mWord);
if (DEBUG_DICT) LOGI("Second freq: %d", secondFreq);
if (secondFreq <= 0) return false;
- word[missingSpacePos] = SPACE;
- for (int i = (missingSpacePos + 1); i < newWordLength; ++i) {
- word[i] = mWord[i - missingSpacePos - 1];
+ word[firstWordLength] = SPACE;
+ for (int i = (firstWordLength + 1); i < newWordLength; ++i) {
+ word[i] = mWord[i - firstWordLength - 1];
}
int pairFreq = ((firstFreq + secondFreq) / 2);
@@ -409,6 +434,17 @@ bool UnigramDictionary::getMissingSpaceWords(const int inputLength, const int mi
return true;
}
+bool UnigramDictionary::getMissingSpaceWords(const int inputLength, const int missingSpacePos) {
+ return getSplitTwoWordsSuggestion(
+ inputLength, 0, missingSpacePos, missingSpacePos, inputLength - missingSpacePos);
+}
+
+bool UnigramDictionary::getMistypedSpaceWords(const int inputLength, const int spaceProximityPos) {
+ return getSplitTwoWordsSuggestion(
+ inputLength, 0, spaceProximityPos, spaceProximityPos + 1,
+ inputLength - spaceProximityPos - 1);
+}
+
// Keep this for comparing spec to new getWords
void UnigramDictionary::getWordsOld(const int initialPos, const int inputLength, const int skipPos,
const int excessivePos, const int transposedPos,int *nextLetters,
diff --git a/native/src/unigram_dictionary.h b/native/src/unigram_dictionary.h
index 2912cb0b7..ef820cba5 100644
--- a/native/src/unigram_dictionary.h
+++ b/native/src/unigram_dictionary.h
@@ -46,7 +46,7 @@ private:
void getWordWithDigraphSuggestionsRec(const ProximityInfo *proximityInfo,
const int *xcoordinates, const int* ycoordinates, const int *codesBuffer,
const int codesBufferSize, const int flags, const int* codesSrc, const int codesRemain,
- int* codesDest, unsigned short* outWords, int* frequencies);
+ int currentDepth, int* codesDest, unsigned short* outWords, int* frequencies);
void initSuggestions(const int *codes, const int codesSize, unsigned short *outWords,
int *frequencies);
void getSuggestionCandidates(const int skipPos, const int excessivePos,
@@ -64,7 +64,11 @@ private:
const bool traverseAllNodes, const int snr, const int inputIndex, const int diffs,
const int skipPos, const int excessivePos, const int transposedPos, int *nextLetters,
const int nextLettersSize);
+ bool getSplitTwoWordsSuggestion(const int inputLength,
+ const int firstWordStartPos, const int firstWordLength,
+ const int secondWordStartPos, const int secondWordLength);
bool getMissingSpaceWords(const int inputLength, const int missingSpacePos);
+ bool getMistypedSpaceWords(const int inputLength, const int spaceProximityPos);
// Keep getWordsOld for comparing performance between getWords and getWordsOld
void getWordsOld(const int initialPos, const int inputLength, const int skipPos,
const int excessivePos, const int transposedPos, int *nextLetters,
@@ -109,6 +113,7 @@ private:
const int FULL_WORD_MULTIPLIER;
const int ROOT_POS;
const unsigned int BYTES_IN_ONE_CHAR;
+ const unsigned int MAX_UMLAUT_SEARCH_DEPTH;
// Flags for special processing
// Those *must* match the flags in BinaryDictionary.Flags.ALL_FLAGS in BinaryDictionary.java
generated by cgit v1.2.3-83-g751a (git 2.47.1) at 2025-06-03 18:37:31 +0000