* commit '155fa8c6f7030001906da4b6a97e4d6d26f96c6b': Make MmappedBuffer use byte array view.
This commit is contained in:
commit
cac2ff7495
13 changed files with 76 additions and 53 deletions
|
@ -30,6 +30,7 @@
|
|||
#include "suggest/policyimpl/dictionary/utils/buffer_with_extendable_buffer.h"
|
||||
#include "suggest/policyimpl/dictionary/utils/dict_file_writing_utils.h"
|
||||
#include "suggest/policyimpl/dictionary/utils/mmapped_buffer.h"
|
||||
#include "utils/byte_array_view.h"
|
||||
|
||||
namespace latinime {
|
||||
namespace backward {
|
||||
|
@ -40,8 +41,9 @@ class SingleDictContent : public DictContent {
|
|||
SingleDictContent(const char *const dictPath, const char *const contentFileName,
|
||||
const bool isUpdatable)
|
||||
: mMmappedBuffer(MmappedBuffer::openBuffer(dictPath, contentFileName, isUpdatable)),
|
||||
mExpandableContentBuffer(mMmappedBuffer ? mMmappedBuffer->getBuffer() : nullptr,
|
||||
mMmappedBuffer ? mMmappedBuffer->getBufferSize() : 0,
|
||||
mExpandableContentBuffer(
|
||||
mMmappedBuffer ? mMmappedBuffer->getReadWriteByteArrayView() :
|
||||
ReadWriteByteArrayView(),
|
||||
BufferWithExtendableBuffer::DEFAULT_MAX_ADDITIONAL_BUFFER_SIZE),
|
||||
mIsValid(mMmappedBuffer) {}
|
||||
|
||||
|
|
|
@ -31,6 +31,7 @@
|
|||
#include "suggest/policyimpl/dictionary/utils/dict_file_writing_utils.h"
|
||||
#include "suggest/policyimpl/dictionary/utils/mmapped_buffer.h"
|
||||
#include "suggest/policyimpl/dictionary/utils/sparse_table.h"
|
||||
#include "utils/byte_array_view.h"
|
||||
|
||||
namespace latinime {
|
||||
namespace backward {
|
||||
|
@ -50,15 +51,16 @@ class SparseTableDictContent : public DictContent {
|
|||
mContentBuffer(
|
||||
MmappedBuffer::openBuffer(dictPath, contentFileName, isUpdatable)),
|
||||
mExpandableLookupTableBuffer(
|
||||
mLookupTableBuffer ? mLookupTableBuffer->getBuffer() : nullptr,
|
||||
mLookupTableBuffer ? mLookupTableBuffer->getBufferSize() : 0,
|
||||
mLookupTableBuffer ? mLookupTableBuffer->getReadWriteByteArrayView() :
|
||||
ReadWriteByteArrayView(),
|
||||
BufferWithExtendableBuffer::DEFAULT_MAX_ADDITIONAL_BUFFER_SIZE),
|
||||
mExpandableAddressTableBuffer(
|
||||
mAddressTableBuffer ? mAddressTableBuffer->getBuffer() : nullptr,
|
||||
mAddressTableBuffer ? mAddressTableBuffer->getBufferSize() : 0,
|
||||
mAddressTableBuffer ? mAddressTableBuffer->getReadWriteByteArrayView() :
|
||||
ReadWriteByteArrayView(),
|
||||
BufferWithExtendableBuffer::DEFAULT_MAX_ADDITIONAL_BUFFER_SIZE),
|
||||
mExpandableContentBuffer(mContentBuffer ? mContentBuffer->getBuffer() : nullptr,
|
||||
mContentBuffer ? mContentBuffer->getBufferSize() : 0,
|
||||
mExpandableContentBuffer(
|
||||
mContentBuffer ? mContentBuffer->getReadWriteByteArrayView() :
|
||||
ReadWriteByteArrayView(),
|
||||
BufferWithExtendableBuffer::DEFAULT_MAX_ADDITIONAL_BUFFER_SIZE),
|
||||
mAddressLookupTable(&mExpandableLookupTableBuffer, &mExpandableAddressTableBuffer,
|
||||
sparseTableBlockSize, sparseTableDataSize),
|
||||
|
|
|
@ -30,6 +30,7 @@
|
|||
|
||||
#include "suggest/policyimpl/dictionary/utils/dict_file_writing_utils.h"
|
||||
#include "suggest/policyimpl/dictionary/utils/file_utils.h"
|
||||
#include "utils/byte_array_view.h"
|
||||
|
||||
namespace latinime {
|
||||
namespace backward {
|
||||
|
@ -130,12 +131,12 @@ Ver4DictBuffers::Ver4DictBuffers(const char *const dictPath,
|
|||
: mHeaderBuffer(std::move(headerBuffer)),
|
||||
mDictBuffer(MmappedBuffer::openBuffer(dictPath,
|
||||
Ver4DictConstants::TRIE_FILE_EXTENSION, isUpdatable)),
|
||||
mHeaderPolicy(mHeaderBuffer->getBuffer(), formatVersion),
|
||||
mExpandableHeaderBuffer(mHeaderBuffer ? mHeaderBuffer->getBuffer() : nullptr,
|
||||
mHeaderPolicy.getSize(),
|
||||
mHeaderPolicy(mHeaderBuffer->getReadOnlyByteArrayView().data(), formatVersion),
|
||||
mExpandableHeaderBuffer(mHeaderBuffer->getReadWriteByteArrayView(),
|
||||
BufferWithExtendableBuffer::DEFAULT_MAX_ADDITIONAL_BUFFER_SIZE),
|
||||
mExpandableTrieBuffer(mDictBuffer ? mDictBuffer->getBuffer() : nullptr,
|
||||
mDictBuffer ? mDictBuffer->getBufferSize() : 0,
|
||||
mExpandableTrieBuffer(
|
||||
mDictBuffer ? mDictBuffer->getReadWriteByteArrayView() :
|
||||
ReadWriteByteArrayView(),
|
||||
BufferWithExtendableBuffer::DEFAULT_MAX_ADDITIONAL_BUFFER_SIZE),
|
||||
mTerminalPositionLookupTable(dictPath, isUpdatable),
|
||||
mProbabilityDictContent(dictPath, mHeaderPolicy.hasHistoricalInfoOfWords(), isUpdatable),
|
||||
|
|
|
@ -31,6 +31,7 @@
|
|||
#include "suggest/policyimpl/dictionary/utils/file_utils.h"
|
||||
#include "suggest/policyimpl/dictionary/utils/format_utils.h"
|
||||
#include "suggest/policyimpl/dictionary/utils/mmapped_buffer.h"
|
||||
#include "utils/byte_array_view.h"
|
||||
|
||||
namespace latinime {
|
||||
|
||||
|
@ -110,7 +111,8 @@ template<class DictConstants, class DictBuffers, class DictBuffersPtr, class Str
|
|||
return nullptr;
|
||||
}
|
||||
const FormatUtils::FORMAT_VERSION formatVersion = FormatUtils::detectFormatVersion(
|
||||
mmappedBuffer->getBuffer(), mmappedBuffer->getBufferSize());
|
||||
mmappedBuffer->getReadOnlyByteArrayView().data(),
|
||||
mmappedBuffer->getReadOnlyByteArrayView().size());
|
||||
switch (formatVersion) {
|
||||
case FormatUtils::VERSION_2:
|
||||
AKLOGE("Given path is a directory but the format is version 2. path: %s", path);
|
||||
|
@ -172,8 +174,8 @@ template<class DictConstants, class DictBuffers, class DictBuffersPtr, class Str
|
|||
if (!mmappedBuffer) {
|
||||
return nullptr;
|
||||
}
|
||||
switch (FormatUtils::detectFormatVersion(mmappedBuffer->getBuffer(),
|
||||
mmappedBuffer->getBufferSize())) {
|
||||
switch (FormatUtils::detectFormatVersion(mmappedBuffer->getReadOnlyByteArrayView().data(),
|
||||
mmappedBuffer->getReadOnlyByteArrayView().size())) {
|
||||
case FormatUtils::VERSION_2:
|
||||
return DictionaryStructureWithBufferPolicy::StructurePolicyPtr(
|
||||
new PatriciaTriePolicy(std::move(mmappedBuffer)));
|
||||
|
|
|
@ -29,6 +29,7 @@
|
|||
#include "suggest/policyimpl/dictionary/structure/v2/ver2_pt_node_array_reader.h"
|
||||
#include "suggest/policyimpl/dictionary/utils/format_utils.h"
|
||||
#include "suggest/policyimpl/dictionary/utils/mmapped_buffer.h"
|
||||
#include "utils/byte_array_view.h"
|
||||
|
||||
namespace latinime {
|
||||
|
||||
|
@ -39,9 +40,12 @@ class PatriciaTriePolicy : public DictionaryStructureWithBufferPolicy {
|
|||
public:
|
||||
PatriciaTriePolicy(MmappedBuffer::MmappedBufferPtr mmappedBuffer)
|
||||
: mMmappedBuffer(std::move(mmappedBuffer)),
|
||||
mHeaderPolicy(mMmappedBuffer->getBuffer(), FormatUtils::VERSION_2),
|
||||
mDictRoot(mMmappedBuffer->getBuffer() + mHeaderPolicy.getSize()),
|
||||
mDictBufferSize(mMmappedBuffer->getBufferSize() - mHeaderPolicy.getSize()),
|
||||
mHeaderPolicy(mMmappedBuffer->getReadOnlyByteArrayView().data(),
|
||||
FormatUtils::VERSION_2),
|
||||
mDictRoot(mMmappedBuffer->getReadOnlyByteArrayView().data()
|
||||
+ mHeaderPolicy.getSize()),
|
||||
mDictBufferSize(mMmappedBuffer->getReadOnlyByteArrayView().size()
|
||||
- mHeaderPolicy.getSize()),
|
||||
mBigramListPolicy(mDictRoot, mDictBufferSize), mShortcutListPolicy(mDictRoot),
|
||||
mPtNodeReader(mDictRoot, mDictBufferSize, &mBigramListPolicy, &mShortcutListPolicy),
|
||||
mPtNodeArrayReader(mDictRoot, mDictBufferSize),
|
||||
|
|
|
@ -24,13 +24,14 @@
|
|||
#include "suggest/policyimpl/dictionary/structure/v4/ver4_dict_constants.h"
|
||||
#include "suggest/policyimpl/dictionary/utils/buffer_with_extendable_buffer.h"
|
||||
#include "suggest/policyimpl/dictionary/utils/dict_file_writing_utils.h"
|
||||
#include "utils/byte_array_view.h"
|
||||
|
||||
namespace latinime {
|
||||
|
||||
class SingleDictContent {
|
||||
public:
|
||||
SingleDictContent(uint8_t *const buffer, const int bufferSize)
|
||||
: mExpandableContentBuffer(buffer, bufferSize,
|
||||
: mExpandableContentBuffer(ReadWriteByteArrayView(buffer, bufferSize),
|
||||
BufferWithExtendableBuffer::DEFAULT_MAX_ADDITIONAL_BUFFER_SIZE) {}
|
||||
|
||||
SingleDictContent()
|
||||
|
|
|
@ -24,6 +24,7 @@
|
|||
#include "suggest/policyimpl/dictionary/structure/v4/ver4_dict_constants.h"
|
||||
#include "suggest/policyimpl/dictionary/utils/buffer_with_extendable_buffer.h"
|
||||
#include "suggest/policyimpl/dictionary/utils/sparse_table.h"
|
||||
#include "utils/byte_array_view.h"
|
||||
|
||||
namespace latinime {
|
||||
|
||||
|
@ -32,14 +33,17 @@ class SparseTableDictContent {
|
|||
public:
|
||||
AK_FORCE_INLINE SparseTableDictContent(uint8_t *const *buffers, const int *bufferSizes,
|
||||
const int sparseTableBlockSize, const int sparseTableDataSize)
|
||||
: mExpandableLookupTableBuffer(buffers[LOOKUP_TABLE_BUFFER_INDEX],
|
||||
bufferSizes[LOOKUP_TABLE_BUFFER_INDEX],
|
||||
: mExpandableLookupTableBuffer(
|
||||
ReadWriteByteArrayView(buffers[LOOKUP_TABLE_BUFFER_INDEX],
|
||||
bufferSizes[LOOKUP_TABLE_BUFFER_INDEX]),
|
||||
BufferWithExtendableBuffer::DEFAULT_MAX_ADDITIONAL_BUFFER_SIZE),
|
||||
mExpandableAddressTableBuffer(buffers[ADDRESS_TABLE_BUFFER_INDEX],
|
||||
bufferSizes[ADDRESS_TABLE_BUFFER_INDEX],
|
||||
mExpandableAddressTableBuffer(
|
||||
ReadWriteByteArrayView(buffers[ADDRESS_TABLE_BUFFER_INDEX],
|
||||
bufferSizes[ADDRESS_TABLE_BUFFER_INDEX]),
|
||||
BufferWithExtendableBuffer::DEFAULT_MAX_ADDITIONAL_BUFFER_SIZE),
|
||||
mExpandableContentBuffer(buffers[CONTENT_BUFFER_INDEX],
|
||||
bufferSizes[CONTENT_BUFFER_INDEX],
|
||||
mExpandableContentBuffer(
|
||||
ReadWriteByteArrayView(buffers[CONTENT_BUFFER_INDEX],
|
||||
bufferSizes[CONTENT_BUFFER_INDEX]),
|
||||
BufferWithExtendableBuffer::DEFAULT_MAX_ADDITIONAL_BUFFER_SIZE),
|
||||
mAddressLookupTable(&mExpandableLookupTableBuffer, &mExpandableAddressTableBuffer,
|
||||
sparseTableBlockSize, sparseTableDataSize) {}
|
||||
|
|
|
@ -26,6 +26,7 @@
|
|||
#include "suggest/policyimpl/dictionary/utils/byte_array_utils.h"
|
||||
#include "suggest/policyimpl/dictionary/utils/dict_file_writing_utils.h"
|
||||
#include "suggest/policyimpl/dictionary/utils/file_utils.h"
|
||||
#include "utils/byte_array_view.h"
|
||||
|
||||
namespace latinime {
|
||||
|
||||
|
@ -46,14 +47,16 @@ namespace latinime {
|
|||
}
|
||||
std::vector<uint8_t *> buffers;
|
||||
std::vector<int> bufferSizes;
|
||||
uint8_t *const buffer = bodyBuffer->getBuffer();
|
||||
const ReadWriteByteArrayView buffer = bodyBuffer->getReadWriteByteArrayView();
|
||||
int position = 0;
|
||||
while (position < bodyBuffer->getBufferSize()) {
|
||||
const int bufferSize = ByteArrayUtils::readUint32AndAdvancePosition(buffer, &position);
|
||||
buffers.push_back(buffer + position);
|
||||
bufferSizes.push_back(bufferSize);
|
||||
while (position < static_cast<int>(buffer.size())) {
|
||||
const int bufferSize = ByteArrayUtils::readUint32AndAdvancePosition(
|
||||
buffer.data(), &position);
|
||||
const ReadWriteByteArrayView subBuffer = buffer.subView(position, bufferSize);
|
||||
buffers.push_back(subBuffer.data());
|
||||
bufferSizes.push_back(subBuffer.size());
|
||||
position += bufferSize;
|
||||
if (bufferSize < 0 || position < 0 || position > bodyBuffer->getBufferSize()) {
|
||||
if (bufferSize < 0 || position < 0 || position > static_cast<int>(buffer.size())) {
|
||||
AKLOGE("The dict body file is corrupted.");
|
||||
return Ver4DictBuffersPtr(nullptr);
|
||||
}
|
||||
|
@ -177,12 +180,12 @@ Ver4DictBuffers::Ver4DictBuffers(MmappedBuffer::MmappedBufferPtr &&headerBuffer,
|
|||
const FormatUtils::FORMAT_VERSION formatVersion,
|
||||
const std::vector<uint8_t *> &contentBuffers, const std::vector<int> &contentBufferSizes)
|
||||
: mHeaderBuffer(std::move(headerBuffer)), mDictBuffer(std::move(bodyBuffer)),
|
||||
mHeaderPolicy(mHeaderBuffer->getBuffer(), formatVersion),
|
||||
mExpandableHeaderBuffer(mHeaderBuffer ? mHeaderBuffer->getBuffer() : nullptr,
|
||||
mHeaderPolicy.getSize(),
|
||||
mHeaderPolicy(mHeaderBuffer->getReadOnlyByteArrayView().data(), formatVersion),
|
||||
mExpandableHeaderBuffer(mHeaderBuffer->getReadWriteByteArrayView(),
|
||||
BufferWithExtendableBuffer::DEFAULT_MAX_ADDITIONAL_BUFFER_SIZE),
|
||||
mExpandableTrieBuffer(contentBuffers[Ver4DictConstants::TRIE_BUFFER_INDEX],
|
||||
contentBufferSizes[Ver4DictConstants::TRIE_BUFFER_INDEX],
|
||||
mExpandableTrieBuffer(
|
||||
ReadWriteByteArrayView(contentBuffers[Ver4DictConstants::TRIE_BUFFER_INDEX],
|
||||
contentBufferSizes[Ver4DictConstants::TRIE_BUFFER_INDEX]),
|
||||
BufferWithExtendableBuffer::DEFAULT_MAX_ADDITIONAL_BUFFER_SIZE),
|
||||
mTerminalPositionLookupTable(
|
||||
contentBuffers[Ver4DictConstants::TERMINAL_ADDRESS_LOOKUP_TABLE_BUFFER_INDEX],
|
||||
|
|
|
@ -35,15 +35,14 @@ class BufferWithExtendableBuffer {
|
|||
public:
|
||||
static const size_t DEFAULT_MAX_ADDITIONAL_BUFFER_SIZE;
|
||||
|
||||
BufferWithExtendableBuffer(uint8_t *const originalBuffer, const int originalBufferSize,
|
||||
BufferWithExtendableBuffer(const ReadWriteByteArrayView originalBuffer,
|
||||
const int maxAdditionalBufferSize)
|
||||
: mOriginalBuffer(originalBuffer, originalBufferSize),
|
||||
mAdditionalBuffer(0), mUsedAdditionalBufferSize(0),
|
||||
: mOriginalBuffer(originalBuffer), mAdditionalBuffer(), mUsedAdditionalBufferSize(0),
|
||||
mMaxAdditionalBufferSize(maxAdditionalBufferSize) {}
|
||||
|
||||
// Without original buffer.
|
||||
BufferWithExtendableBuffer(const int maxAdditionalBufferSize)
|
||||
: mOriginalBuffer(), mAdditionalBuffer(0), mUsedAdditionalBufferSize(0),
|
||||
: mOriginalBuffer(), mAdditionalBuffer(), mUsedAdditionalBufferSize(0),
|
||||
mMaxAdditionalBufferSize(maxAdditionalBufferSize) {}
|
||||
|
||||
AK_FORCE_INLINE int getTailPosition() const {
|
||||
|
|
|
@ -21,6 +21,7 @@
|
|||
#include <memory>
|
||||
|
||||
#include "defines.h"
|
||||
#include "utils/byte_array_view.h"
|
||||
|
||||
namespace latinime {
|
||||
|
||||
|
@ -39,12 +40,12 @@ class MmappedBuffer {
|
|||
|
||||
~MmappedBuffer();
|
||||
|
||||
AK_FORCE_INLINE uint8_t *getBuffer() const {
|
||||
return mBuffer;
|
||||
ReadWriteByteArrayView getReadWriteByteArrayView() const {
|
||||
return mByteArrayView;
|
||||
}
|
||||
|
||||
AK_FORCE_INLINE int getBufferSize() const {
|
||||
return mBufferSize;
|
||||
ReadOnlyByteArrayView getReadOnlyByteArrayView() const {
|
||||
return mByteArrayView.getReadOnlyView();
|
||||
}
|
||||
|
||||
AK_FORCE_INLINE bool isUpdatable() const {
|
||||
|
@ -55,18 +56,17 @@ class MmappedBuffer {
|
|||
AK_FORCE_INLINE MmappedBuffer(uint8_t *const buffer, const int bufferSize,
|
||||
void *const mmappedBuffer, const int alignedSize, const int mmapFd,
|
||||
const bool isUpdatable)
|
||||
: mBuffer(buffer), mBufferSize(bufferSize), mMmappedBuffer(mmappedBuffer),
|
||||
: mByteArrayView(buffer, bufferSize), mMmappedBuffer(mmappedBuffer),
|
||||
mAlignedSize(alignedSize), mMmapFd(mmapFd), mIsUpdatable(isUpdatable) {}
|
||||
|
||||
// Empty file. We have to handle an empty file as a valid part of a dictionary.
|
||||
AK_FORCE_INLINE MmappedBuffer(const bool isUpdatable)
|
||||
: mBuffer(nullptr), mBufferSize(0), mMmappedBuffer(nullptr), mAlignedSize(0),
|
||||
: mByteArrayView(), mMmappedBuffer(nullptr), mAlignedSize(0),
|
||||
mMmapFd(0), mIsUpdatable(isUpdatable) {}
|
||||
|
||||
DISALLOW_IMPLICIT_CONSTRUCTORS(MmappedBuffer);
|
||||
|
||||
uint8_t *const mBuffer;
|
||||
const int mBufferSize;
|
||||
const ReadWriteByteArrayView mByteArrayView;
|
||||
void *const mMmappedBuffer;
|
||||
const int mAlignedSize;
|
||||
const int mMmapFd;
|
||||
|
|
|
@ -43,9 +43,8 @@ TrieMap::TrieMap() : mBuffer(MAX_BUFFER_SIZE) {
|
|||
writeEntry(EMPTY_BITMAP_ENTRY, ROOT_BITMAP_ENTRY_INDEX);
|
||||
}
|
||||
|
||||
TrieMap::TrieMap(uint8_t *const buffer, const int bufferSize)
|
||||
: mBuffer(buffer, bufferSize,
|
||||
BufferWithExtendableBuffer::DEFAULT_MAX_ADDITIONAL_BUFFER_SIZE) {}
|
||||
TrieMap::TrieMap(const ReadWriteByteArrayView buffer)
|
||||
: mBuffer(buffer, BufferWithExtendableBuffer::DEFAULT_MAX_ADDITIONAL_BUFFER_SIZE) {}
|
||||
|
||||
void TrieMap::dump(const int from, const int to) const {
|
||||
AKLOGI("BufSize: %d", mBuffer.getTailPosition());
|
||||
|
|
|
@ -24,6 +24,7 @@
|
|||
|
||||
#include "defines.h"
|
||||
#include "suggest/policyimpl/dictionary/utils/buffer_with_extendable_buffer.h"
|
||||
#include "utils/byte_array_view.h"
|
||||
|
||||
namespace latinime {
|
||||
|
||||
|
@ -161,7 +162,7 @@ class TrieMap {
|
|||
|
||||
TrieMap();
|
||||
// Construct TrieMap using existing data in the memory region written by save().
|
||||
TrieMap(uint8_t *const buffer, const int bufferSize);
|
||||
TrieMap(const ReadWriteByteArrayView buffer);
|
||||
void dump(const int from = 0, const int to = 0) const;
|
||||
|
||||
bool isNearSizeLimit() const {
|
||||
|
|
|
@ -71,6 +71,11 @@ class ReadWriteByteArrayView {
|
|||
return ReadOnlyByteArrayView(mPtr, mSize);
|
||||
}
|
||||
|
||||
ReadWriteByteArrayView subView(const size_t start, const size_t n) const {
|
||||
ASSERT(start + n <= mSize);
|
||||
return ReadWriteByteArrayView(mPtr + start, n);
|
||||
}
|
||||
|
||||
private:
|
||||
DISALLOW_ASSIGNMENT_OPERATOR(ReadWriteByteArrayView);
|
||||
|
||||
|
|
Loading…
Reference in a new issue