Add dictionary header writing methods.

Bug: 6669677
Change-Id: I392ac4776b55779903cbaa17e683005d80017a41
main
Keisuke Kuroyanagi 2013-09-18 18:08:33 +09:00
parent 59418672a1
commit 484fa7b59c
12 changed files with 181 additions and 67 deletions

View File

@ -250,7 +250,7 @@ void DynamicPatriciaTriePolicy::flush(const char *const filePath) {
}
DynamicPatriciaTrieWritingHelper writingHelper(&mBufferWithExtendableBuffer,
&mBigramListPolicy, &mShortcutListPolicy);
writingHelper.writeToDictFile(filePath, mBuffer->getBuffer(), mHeaderPolicy.getSize());
writingHelper.writeToDictFile(filePath, &mHeaderPolicy);
}
void DynamicPatriciaTriePolicy::flushWithGC(const char *const filePath) {

View File

@ -33,7 +33,7 @@ class DicNodeVector;
class DynamicPatriciaTriePolicy : public DictionaryStructureWithBufferPolicy {
public:
DynamicPatriciaTriePolicy(const MmappedBuffer *const buffer)
: mBuffer(buffer), mHeaderPolicy(mBuffer->getBuffer()),
: mBuffer(buffer), mHeaderPolicy(mBuffer->getBuffer(), buffer->getBufferSize()),
mBufferWithExtendableBuffer(mBuffer->getBuffer() + mHeaderPolicy.getSize(),
mBuffer->getBufferSize() - mHeaderPolicy.getSize()),
mShortcutListPolicy(&mBufferWithExtendableBuffer),

View File

@ -24,6 +24,7 @@
#include "suggest/policyimpl/dictionary/dynamic_patricia_trie_reading_helper.h"
#include "suggest/policyimpl/dictionary/dynamic_patricia_trie_reading_utils.h"
#include "suggest/policyimpl/dictionary/dynamic_patricia_trie_writing_utils.h"
#include "suggest/policyimpl/dictionary/header/header_policy.h"
#include "suggest/policyimpl/dictionary/patricia_trie_reading_utils.h"
#include "suggest/policyimpl/dictionary/shortcut/dynamic_shortcut_list_policy.h"
@ -137,7 +138,11 @@ bool DynamicPatriciaTrieWritingHelper::removeBigramWords(const int word0Pos, con
}
void DynamicPatriciaTrieWritingHelper::writeToDictFile(const char *const fileName,
const uint8_t *const headerBuf, const int headerSize) {
const HeaderPolicy *const headerPolicy) {
BufferWithExtendableBuffer headerBuffer(0 /* originalBuffer */, 0 /* originalBufferSize */);
if (!headerPolicy->writeHeaderToBuffer(&headerBuffer, false /* updatesLastUpdatedTime */)) {
return;
}
const int tmpFileNameBufSize = strlen(fileName)
+ strlen(TEMP_FILE_SUFFIX_FOR_WRITING_DICT_FILE) + 1;
char tmpFileName[tmpFileNameBufSize];
@ -148,7 +153,8 @@ void DynamicPatriciaTrieWritingHelper::writeToDictFile(const char *const fileNam
return;
}
// Write header.
if (fwrite(headerBuf, headerSize, 1, file) < 1) {
if (fwrite(headerBuffer.getBuffer(true /* usesAdditionalBuffer */),
headerBuffer.getTailPosition(), 1, file) < 1) {
fclose(file);
remove(tmpFileName);
return;

View File

@ -28,6 +28,7 @@ class DynamicBigramListPolicy;
class DynamicPatriciaTrieNodeReader;
class DynamicPatriciaTrieReadingHelper;
class DynamicShortcutListPolicy;
class HeaderPolicy;
class DynamicPatriciaTrieWritingHelper {
public:
@ -48,8 +49,9 @@ class DynamicPatriciaTrieWritingHelper {
// Remove a bigram relation from word0Pos to word1Pos.
bool removeBigramWords(const int word0Pos, const int word1Pos);
void writeToDictFile(const char *const fileName, const uint8_t *const headerBuf,
const int headerSize);
void writeToDictFile(const char *const fileName, const HeaderPolicy *const headerPolicy);
void writeToDictFileWithGC(const char *const fileName, const HeaderPolicy *const headerPolicy);
private:
DISALLOW_IMPLICIT_CONSTRUCTORS(DynamicPatriciaTrieWritingHelper);

View File

@ -17,6 +17,8 @@
#include "suggest/policyimpl/dictionary/header/header_policy.h"
#include <cstddef>
#include <cstdio>
#include <ctime>
namespace latinime {
@ -36,7 +38,7 @@ void HeaderPolicy::readHeaderValueOrQuestionMark(const char *const key, int *out
}
std::vector<int> keyCodePointVector;
insertCharactersIntoVector(key, &keyCodePointVector);
HeaderReadingUtils::AttributeMap::const_iterator it = mAttributeMap.find(keyCodePointVector);
HeaderReadWriteUtils::AttributeMap::const_iterator it = mAttributeMap.find(keyCodePointVector);
if (it == mAttributeMap.end()) {
// The key was not found.
outValue[0] = '?';
@ -85,7 +87,7 @@ int HeaderPolicy::readLastUpdatedTime() const {
bool HeaderPolicy::getAttributeValueAsInt(const char *const key, int *const outValue) const {
std::vector<int> keyVector;
insertCharactersIntoVector(key, &keyVector);
HeaderReadingUtils::AttributeMap::const_iterator it = mAttributeMap.find(keyVector);
HeaderReadWriteUtils::AttributeMap::const_iterator it = mAttributeMap.find(keyVector);
if (it == mAttributeMap.end()) {
// The key was not found.
return false;
@ -94,10 +96,56 @@ bool HeaderPolicy::getAttributeValueAsInt(const char *const key, int *const outV
return true;
}
/* static */ HeaderReadingUtils::AttributeMap HeaderPolicy::createAttributeMapAndReadAllAttributes(
const uint8_t *const dictBuf) {
HeaderReadingUtils::AttributeMap attributeMap;
HeaderReadingUtils::fetchAllHeaderAttributes(dictBuf, &attributeMap);
bool HeaderPolicy::writeHeaderToBuffer(BufferWithExtendableBuffer *const bufferToWrite,
const bool updatesLastUpdatedTime) const {
int writingPos = 0;
if (!HeaderReadWriteUtils::writeDictionaryVersion(bufferToWrite, mDictFormatVersion,
&writingPos)) {
return false;
}
if (!HeaderReadWriteUtils::writeDictionaryFlags(bufferToWrite, mDictionaryFlags,
&writingPos)) {
return false;
}
// Temporarily writes a dummy header size.
int headerSizeFieldPos = writingPos;
if (!HeaderReadWriteUtils::writeDictionaryHeaderSize(bufferToWrite, 0 /* size */,
&writingPos)) {
return false;
}
if (updatesLastUpdatedTime) {
// Set current time as a last updated time.
HeaderReadWriteUtils::AttributeMap attributeMapTowrite(mAttributeMap);
std::vector<int> updatedTimekey;
insertCharactersIntoVector(LAST_UPDATED_TIME_KEY, &updatedTimekey);
const time_t currentTime = time(NULL);
std::vector<int> updatedTimeValue;
char charBuf[LARGEST_INT_DIGIT_COUNT + 1];
snprintf(charBuf, LARGEST_INT_DIGIT_COUNT + 1, "%ld", currentTime);
insertCharactersIntoVector(charBuf, &updatedTimeValue);
attributeMapTowrite[updatedTimekey] = updatedTimeValue;
if (!HeaderReadWriteUtils::writeHeaderAttributes(bufferToWrite, &attributeMapTowrite,
&writingPos)) {
return false;
}
} else {
if (!HeaderReadWriteUtils::writeHeaderAttributes(bufferToWrite, &mAttributeMap,
&writingPos)) {
return false;
}
}
// Writes an actual header size.
if (!HeaderReadWriteUtils::writeDictionaryHeaderSize(bufferToWrite, writingPos,
&headerSizeFieldPos)) {
return false;
}
return true;
}
/* static */ HeaderReadWriteUtils::AttributeMap
HeaderPolicy::createAttributeMapAndReadAllAttributes(const uint8_t *const dictBuf) {
HeaderReadWriteUtils::AttributeMap attributeMap;
HeaderReadWriteUtils::fetchAllHeaderAttributes(dictBuf, &attributeMap);
return attributeMap;
}

View File

@ -23,14 +23,17 @@
#include "defines.h"
#include "suggest/core/policy/dictionary_header_structure_policy.h"
#include "suggest/policyimpl/dictionary/header/header_reading_utils.h"
#include "suggest/policyimpl/dictionary/utils/format_utils.h"
namespace latinime {
class HeaderPolicy : public DictionaryHeaderStructurePolicy {
public:
explicit HeaderPolicy(const uint8_t *const dictBuf)
: mDictBuf(dictBuf), mDictionaryFlags(HeaderReadingUtils::getFlags(dictBuf)),
mSize(HeaderReadingUtils::getHeaderSize(dictBuf)),
explicit HeaderPolicy(const uint8_t *const dictBuf, const int dictSize)
: mDictBuf(dictBuf),
mDictFormatVersion(FormatUtils::detectFormatVersion(dictBuf, dictSize)),
mDictionaryFlags(HeaderReadWriteUtils::getFlags(dictBuf)),
mSize(HeaderReadWriteUtils::getHeaderSize(dictBuf)),
mAttributeMap(createAttributeMapAndReadAllAttributes(mDictBuf)),
mMultiWordCostMultiplier(readMultipleWordCostMultiplier()),
mUsesForgettingCurve(readUsesForgettingCurveFlag()),
@ -43,16 +46,15 @@ class HeaderPolicy : public DictionaryHeaderStructurePolicy {
}
AK_FORCE_INLINE bool supportsDynamicUpdate() const {
return HeaderReadingUtils::supportsDynamicUpdate(mDictionaryFlags);
return HeaderReadWriteUtils::supportsDynamicUpdate(mDictionaryFlags);
}
AK_FORCE_INLINE bool requiresGermanUmlautProcessing() const {
return HeaderReadingUtils::requiresGermanUmlautProcessing(mDictionaryFlags);
return HeaderReadWriteUtils::requiresGermanUmlautProcessing(mDictionaryFlags);
}
AK_FORCE_INLINE bool requiresFrenchLigatureProcessing() const {
return HeaderReadingUtils::requiresFrenchLigatureProcessing(
mDictionaryFlags);
return HeaderReadWriteUtils::requiresFrenchLigatureProcessing(mDictionaryFlags);
}
AK_FORCE_INLINE float getMultiWordCostMultiplier() const {
@ -70,6 +72,9 @@ class HeaderPolicy : public DictionaryHeaderStructurePolicy {
void readHeaderValueOrQuestionMark(const char *const key,
int *outValue, int outValueSize) const;
bool writeHeaderToBuffer(BufferWithExtendableBuffer *const bufferToWrite,
const bool updatesLastUpdatedTime) const;
private:
DISALLOW_IMPLICIT_CONSTRUCTORS(HeaderPolicy);
@ -80,9 +85,10 @@ class HeaderPolicy : public DictionaryHeaderStructurePolicy {
static const float MULTIPLE_WORD_COST_MULTIPLIER_SCALE;
const uint8_t *const mDictBuf;
const HeaderReadingUtils::DictionaryFlags mDictionaryFlags;
const FormatUtils::FORMAT_VERSION mDictFormatVersion;
const HeaderReadWriteUtils::DictionaryFlags mDictionaryFlags;
const int mSize;
HeaderReadingUtils::AttributeMap mAttributeMap;
HeaderReadWriteUtils::AttributeMap mAttributeMap;
const float mMultiWordCostMultiplier;
const bool mUsesForgettingCurve;
const int mLastUpdatedTime;
@ -95,7 +101,7 @@ class HeaderPolicy : public DictionaryHeaderStructurePolicy {
bool getAttributeValueAsInt(const char *const key, int *const outValue) const;
static HeaderReadingUtils::AttributeMap createAttributeMapAndReadAllAttributes(
static HeaderReadWriteUtils::AttributeMap createAttributeMapAndReadAllAttributes(
const uint8_t *const dictBuf);
static int parseIntAttributeValue(const std::vector<int> *const attributeValue);

View File

@ -19,43 +19,44 @@
#include <vector>
#include "defines.h"
#include "suggest/policyimpl/dictionary/utils/buffer_with_extendable_buffer.h"
#include "suggest/policyimpl/dictionary/utils/byte_array_utils.h"
namespace latinime {
const int HeaderReadingUtils::MAX_ATTRIBUTE_KEY_LENGTH = 256;
const int HeaderReadingUtils::MAX_ATTRIBUTE_VALUE_LENGTH = 256;
const int HeaderReadWriteUtils::MAX_ATTRIBUTE_KEY_LENGTH = 256;
const int HeaderReadWriteUtils::MAX_ATTRIBUTE_VALUE_LENGTH = 256;
const int HeaderReadingUtils::HEADER_MAGIC_NUMBER_SIZE = 4;
const int HeaderReadingUtils::HEADER_DICTIONARY_VERSION_SIZE = 2;
const int HeaderReadingUtils::HEADER_FLAG_SIZE = 2;
const int HeaderReadingUtils::HEADER_SIZE_FIELD_SIZE = 4;
const int HeaderReadWriteUtils::HEADER_MAGIC_NUMBER_SIZE = 4;
const int HeaderReadWriteUtils::HEADER_DICTIONARY_VERSION_SIZE = 2;
const int HeaderReadWriteUtils::HEADER_FLAG_SIZE = 2;
const int HeaderReadWriteUtils::HEADER_SIZE_FIELD_SIZE = 4;
const HeaderReadingUtils::DictionaryFlags HeaderReadingUtils::NO_FLAGS = 0;
const HeaderReadWriteUtils::DictionaryFlags HeaderReadWriteUtils::NO_FLAGS = 0;
// Flags for special processing
// Those *must* match the flags in makedict (FormatSpec#*_PROCESSING_FLAG) or
// something very bad (like, the apocalypse) will happen. Please update both at the same time.
const HeaderReadingUtils::DictionaryFlags
HeaderReadingUtils::GERMAN_UMLAUT_PROCESSING_FLAG = 0x1;
const HeaderReadingUtils::DictionaryFlags
HeaderReadingUtils::SUPPORTS_DYNAMIC_UPDATE_FLAG = 0x2;
const HeaderReadingUtils::DictionaryFlags
HeaderReadingUtils::FRENCH_LIGATURE_PROCESSING_FLAG = 0x4;
const HeaderReadWriteUtils::DictionaryFlags
HeaderReadWriteUtils::GERMAN_UMLAUT_PROCESSING_FLAG = 0x1;
const HeaderReadWriteUtils::DictionaryFlags
HeaderReadWriteUtils::SUPPORTS_DYNAMIC_UPDATE_FLAG = 0x2;
const HeaderReadWriteUtils::DictionaryFlags
HeaderReadWriteUtils::FRENCH_LIGATURE_PROCESSING_FLAG = 0x4;
/* static */ int HeaderReadingUtils::getHeaderSize(const uint8_t *const dictBuf) {
/* static */ int HeaderReadWriteUtils::getHeaderSize(const uint8_t *const dictBuf) {
// See the format of the header in the comment in
// BinaryDictionaryFormatUtils::detectFormatVersion()
return ByteArrayUtils::readUint32(dictBuf, HEADER_MAGIC_NUMBER_SIZE
+ HEADER_DICTIONARY_VERSION_SIZE + HEADER_FLAG_SIZE);
}
/* static */ HeaderReadingUtils::DictionaryFlags
HeaderReadingUtils::getFlags(const uint8_t *const dictBuf) {
/* static */ HeaderReadWriteUtils::DictionaryFlags
HeaderReadWriteUtils::getFlags(const uint8_t *const dictBuf) {
return ByteArrayUtils::readUint16(dictBuf,
HEADER_MAGIC_NUMBER_SIZE + HEADER_DICTIONARY_VERSION_SIZE);
}
/* static */ void HeaderReadingUtils::fetchAllHeaderAttributes(const uint8_t *const dictBuf,
/* static */ void HeaderReadWriteUtils::fetchAllHeaderAttributes(const uint8_t *const dictBuf,
AttributeMap *const headerAttributes) {
const int headerSize = getHeaderSize(dictBuf);
int pos = getHeaderOptionsPosition();
@ -78,4 +79,53 @@ const HeaderReadingUtils::DictionaryFlags
}
}
/* static */ bool HeaderReadWriteUtils::writeDictionaryVersion(
BufferWithExtendableBuffer *const buffer, const FormatUtils::FORMAT_VERSION version,
int *const writingPos) {
if (!buffer->writeUintAndAdvancePosition(FormatUtils::MAGIC_NUMBER, HEADER_MAGIC_NUMBER_SIZE,
writingPos)) {
return false;
}
switch (version) {
case FormatUtils::VERSION_2:
// Version 2 dictionary writing is not supported.
return false;
case FormatUtils::VERSION_3:
return buffer->writeUintAndAdvancePosition(3 /* data */,
HEADER_DICTIONARY_VERSION_SIZE, writingPos);
default:
return false;
}
}
/* static */ bool HeaderReadWriteUtils::writeDictionaryFlags(
BufferWithExtendableBuffer *const buffer, const DictionaryFlags flags,
int *const writingPos) {
return buffer->writeUintAndAdvancePosition(flags, HEADER_FLAG_SIZE, writingPos);
}
/* static */ bool HeaderReadWriteUtils::writeDictionaryHeaderSize(
BufferWithExtendableBuffer *const buffer, const int size, int *const writingPos) {
return buffer->writeUintAndAdvancePosition(size, HEADER_SIZE_FIELD_SIZE, writingPos);
}
/* static */ bool HeaderReadWriteUtils::writeHeaderAttributes(
BufferWithExtendableBuffer *const buffer, const AttributeMap *const headerAttributes,
int *const writingPos) {
for (AttributeMap::const_iterator it = headerAttributes->begin();
it != headerAttributes->end(); ++it) {
// Write a key.
if (!buffer->writeCodePointsAndAdvancePosition(&(it->first.at(0)), it->first.size(),
true /* writesTerminator */, writingPos)) {
return false;
}
// Write a value.
if (!buffer->writeCodePointsAndAdvancePosition(&(it->second.at(0)), it->second.size(),
true /* writesTerminator */, writingPos)) {
return false;
}
}
return true;
}
} // namespace latinime

View File

@ -22,10 +22,14 @@
#include <vector>
#include "defines.h"
#include "suggest/policyimpl/dictionary/utils/format_utils.h"
namespace latinime {
class HeaderReadingUtils {
class BufferWithExtendableBuffer;
// TODO: Change this file name to header_read_write_utils.h.
class HeaderReadWriteUtils {
public:
typedef uint16_t DictionaryFlags;
typedef std::map<std::vector<int>, std::vector<int> > AttributeMap;
@ -54,8 +58,20 @@ class HeaderReadingUtils {
static void fetchAllHeaderAttributes(const uint8_t *const dictBuf,
AttributeMap *const headerAttributes);
static bool writeDictionaryVersion(BufferWithExtendableBuffer *const buffer,
const FormatUtils::FORMAT_VERSION version, int *const writingPos);
static bool writeDictionaryFlags(BufferWithExtendableBuffer *const buffer,
const DictionaryFlags flags, int *const writingPos);
static bool writeDictionaryHeaderSize(BufferWithExtendableBuffer *const buffer,
const int size, int *const writingPos);
static bool writeHeaderAttributes(BufferWithExtendableBuffer *const buffer,
const AttributeMap *const headerAttributes, int *const writingPos);
private:
DISALLOW_IMPLICIT_CONSTRUCTORS(HeaderReadingUtils);
DISALLOW_IMPLICIT_CONSTRUCTORS(HeaderReadWriteUtils);
static const int MAX_ATTRIBUTE_KEY_LENGTH;
static const int MAX_ATTRIBUTE_VALUE_LENGTH;

View File

@ -34,7 +34,7 @@ class DicNodeVector;
class PatriciaTriePolicy : public DictionaryStructureWithBufferPolicy {
public:
PatriciaTriePolicy(const MmappedBuffer *const buffer)
: mBuffer(buffer), mHeaderPolicy(mBuffer->getBuffer()),
: mBuffer(buffer), mHeaderPolicy(mBuffer->getBuffer(), buffer->getBufferSize()),
mDictRoot(mBuffer->getBuffer() + mHeaderPolicy.getSize()),
mBigramListPolicy(mDictRoot), mShortcutListPolicy(mDictRoot) {}

View File

@ -20,20 +20,10 @@
namespace latinime {
/**
* Dictionary size
*/
// Any file smaller than this is not a dictionary.
const int FormatUtils::DICTIONARY_MINIMUM_SIZE = 4;
const uint32_t FormatUtils::MAGIC_NUMBER = 0x9BC13AFE;
/**
* Format versions
*/
// 32 bit magic number is stored at the beginning of the dictionary header to reject unsupported
// or obsolete dictionary formats.
const uint32_t FormatUtils::HEADER_VERSION_2_MAGIC_NUMBER = 0x9BC13AFE;
// Magic number (4 bytes), version (2 bytes), options (2 bytes), header size (4 bytes) = 12
const int FormatUtils::HEADER_VERSION_2_MINIMUM_SIZE = 12;
// Magic number (4 bytes), version (2 bytes), flags (2 bytes), header size (4 bytes) = 12
const int FormatUtils::DICTIONARY_MINIMUM_SIZE = 12;
/* static */ FormatUtils::FORMAT_VERSION FormatUtils::detectFormatVersion(
const uint8_t *const dict, const int dictSize) {
@ -45,16 +35,10 @@ const int FormatUtils::HEADER_VERSION_2_MINIMUM_SIZE = 12;
}
const uint32_t magicNumber = ByteArrayUtils::readUint32(dict, 0);
switch (magicNumber) {
case HEADER_VERSION_2_MAGIC_NUMBER:
// Version 2 header are at least 12 bytes long.
// If this header has the version 2 magic number but is less than 12 bytes long,
// then it's an unknown format and we need to avoid confidently reading the next bytes.
if (dictSize < HEADER_VERSION_2_MINIMUM_SIZE) {
return UNKNOWN_VERSION;
}
case MAGIC_NUMBER:
// Version 2 header is as follows:
// Magic number (4 bytes) 0x9B 0xC1 0x3A 0xFE
// Version number (2 bytes)
// Dictionary format version number (2 bytes)
// Options (2 bytes)
// Header size (4 bytes) : integer, big endian
if (ByteArrayUtils::readUint16(dict, 4) == 2) {

View File

@ -34,14 +34,16 @@ class FormatUtils {
UNKNOWN_VERSION
};
// 32 bit magic number is stored at the beginning of the dictionary header to reject
// unsupported or obsolete dictionary formats.
static const uint32_t MAGIC_NUMBER;
static FORMAT_VERSION detectFormatVersion(const uint8_t *const dict, const int dictSize);
private:
DISALLOW_IMPLICIT_CONSTRUCTORS(FormatUtils);
static const int DICTIONARY_MINIMUM_SIZE;
static const uint32_t HEADER_VERSION_2_MAGIC_NUMBER;
static const int HEADER_VERSION_2_MINIMUM_SIZE;
};
} // namespace latinime
#endif /* LATINIME_FORMAT_UTILS_H */

View File

@ -358,8 +358,8 @@ public class BinaryDictionaryTests extends AndroidTestCase {
0 /* offset */, dictFile.length(), true /* useFullEditDistance */,
Locale.getDefault(), TEST_LOCALE, true /* isUpdatable */);
assertEquals(-1, binaryDictionary.getFrequency("aaa"));
assertEquals(-1, binaryDictionary.getFrequency("abcd"));
assertEquals(Dictionary.NOT_A_PROBABILITY, binaryDictionary.getFrequency("aaa"));
assertEquals(Dictionary.NOT_A_PROBABILITY, binaryDictionary.getFrequency("abcd"));
binaryDictionary.addUnigramWord("aaa", probability);
binaryDictionary.addUnigramWord("abcd", probability);