Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Remove arrow from npy reader #2086

Merged
merged 1 commit into from
Sep 26, 2023
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
12 changes: 1 addition & 11 deletions src/binder/bind/bind_copy.cpp
Original file line number Diff line number Diff line change
Expand Up @@ -197,6 +197,7 @@ expression_vector Binder::bindExpectedNodeFileColumns(
columns.push_back(createVariable(columnName, stringType));
}
} break;
case FileType::NPY:
case FileType::PARQUET:
case FileType::CSV: {
for (auto& property : tableSchema->properties) {
Expand All @@ -208,17 +209,6 @@ expression_vector Binder::bindExpectedNodeFileColumns(
columns.push_back(createVariable(property->getName(), *property->getDataType()));
}
} break;
case FileType::NPY: {
for (auto& property : tableSchema->properties) {
if (skipPropertyInFile(*property)) {
continue;
}
readerConfig.columnNames.push_back(property->getName());
readerConfig.columnTypes.push_back(property->getDataType()->copy());
columns.push_back(
createVariable(property->getName(), LogicalType{LogicalTypeID::ARROW_COLUMN}));
}
} break;
default: {
throw NotImplementedException{"Binder::bindCopyNodeColumns"};
}
Expand Down
9 changes: 3 additions & 6 deletions src/include/processor/operator/persistent/reader/npy_reader.h
Original file line number Diff line number Diff line change
Expand Up @@ -4,11 +4,9 @@
#include <unordered_map>
#include <vector>

#include "common/data_chunk/data_chunk.h"
#include "common/types/internal_id_t.h"
#include "common/types/types.h"
#include <arrow/array.h>
#include <arrow/buffer.h>
#include <arrow/record_batch.h>

namespace kuzu {
namespace processor {
Expand All @@ -25,8 +23,7 @@ class NpyReader {

inline size_t getNumRows() const { return shape[0]; }

std::shared_ptr<arrow::DataType> getArrowType() const;
std::shared_ptr<arrow::RecordBatch> readBlock(common::block_idx_t blockIdx) const;
void readBlock(common::block_idx_t blockIdx, common::ValueVector* vectorToRead) const;

// Used in tests only.
inline common::LogicalTypeID getType() const { return type; }
Expand Down Expand Up @@ -54,7 +51,7 @@ class NpyMultiFileReader {
public:
explicit NpyMultiFileReader(const std::vector<std::string>& filePaths);

std::shared_ptr<arrow::RecordBatch> readBlock(common::block_idx_t blockIdx) const;
void readBlock(common::block_idx_t blockIdx, common::DataChunk* dataChunkToRead) const;

private:
std::vector<std::unique_ptr<NpyReader>> fileReaders;
Expand Down
8 changes: 7 additions & 1 deletion src/include/storage/store/column_chunk.h
Original file line number Diff line number Diff line change
Expand Up @@ -78,6 +78,7 @@ class ColumnChunk {
assert(idx < childrenChunks.size());
return childrenChunks[idx].get();
}
virtual inline uint64_t getBufferSize() const { return numBytesPerValue * capacity; }

virtual void resetToEmpty();

Expand Down Expand Up @@ -157,7 +158,7 @@ class ColumnChunk {

common::offset_t getOffsetInBuffer(common::offset_t pos) const;

void copyVectorToBuffer(common::ValueVector* vector, common::offset_t startPosInChunk);
virtual void copyVectorToBuffer(common::ValueVector* vector, common::offset_t startPosInChunk);

protected:
common::LogicalType dataType;
Expand Down Expand Up @@ -261,6 +262,11 @@ class FixedListColumnChunk : public ColumnChunk {
common::offset_t startPosInChunk, uint32_t numValuesToAppend) final;

void write(const common::Value& fixedListVal, uint64_t posToWrite) final;

void copyVectorToBuffer(common::ValueVector* vector, common::offset_t startPosInChunk) final;

private:
uint64_t getBufferSize() const final;
};

class SerialColumnChunk : public ColumnChunk {
Expand Down
57 changes: 7 additions & 50 deletions src/processor/operator/persistent/reader/npy/npy_reader.cpp
Original file line number Diff line number Diff line change
Expand Up @@ -207,47 +207,13 @@ void NpyReader::validate(const LogicalType& type_, offset_t numRows) {
}
}

std::shared_ptr<arrow::DataType> NpyReader::getArrowType() const {
auto thisType = getType();
if (thisType == LogicalTypeID::DOUBLE) {
return arrow::float64();
} else if (thisType == LogicalTypeID::FLOAT) {
return arrow::float32();
} else if (thisType == LogicalTypeID::INT64) {
return arrow::int64();
} else if (thisType == LogicalTypeID::INT32) {
return arrow::int32();
} else if (thisType == LogicalTypeID::INT16) {
return arrow::int16();
} else {
throw CopyException("File type does not match any Arrow data type");
}
}

std::shared_ptr<arrow::RecordBatch> NpyReader::readBlock(block_idx_t blockIdx) const {
void NpyReader::readBlock(block_idx_t blockIdx, common::ValueVector* vectorToRead) const {
uint64_t rowNumber = DEFAULT_VECTOR_CAPACITY * blockIdx;
auto rowPointer = getPointerToRow(rowNumber);
auto arrowType = getArrowType();
auto numRowsToRead = std::min(DEFAULT_VECTOR_CAPACITY, getNumRows() - rowNumber);
auto buffer = std::make_shared<arrow::Buffer>(
rowPointer, numRowsToRead * arrowType->byte_width() * getNumElementsPerRow());
std::shared_ptr<arrow::Field> field;
std::shared_ptr<arrow::Array> arr;
if (getNumDimensions() > 1) {
auto elementField = std::make_shared<arrow::Field>(defaultFieldName, arrowType);
auto fixedListArrowType = arrow::fixed_size_list(elementField, (int32_t)numRowsToRead);
field = std::make_shared<arrow::Field>(defaultFieldName, fixedListArrowType);
auto valuesArr = std::make_shared<arrow::PrimitiveArray>(
arrowType, numRowsToRead * getNumElementsPerRow(), buffer);
arr = arrow::FixedSizeListArray::FromArrays(valuesArr, (int32_t)getNumElementsPerRow())
.ValueOrDie();
} else {
field = std::make_shared<arrow::Field>(defaultFieldName, arrowType);
arr = std::make_shared<arrow::PrimitiveArray>(arrowType, numRowsToRead, buffer);
}
auto schema =
std::make_shared<arrow::Schema>(std::vector<std::shared_ptr<arrow::Field>>{field});
return arrow::RecordBatch::Make(schema, (int64_t)numRowsToRead, {arr});
memcpy(
vectorToRead->getData(), rowPointer, numRowsToRead * vectorToRead->getNumBytesPerValue());
vectorToRead->state->selVector->selectedSize = numRowsToRead;
}

NpyMultiFileReader::NpyMultiFileReader(const std::vector<std::string>& filePaths) {
Expand All @@ -256,20 +222,11 @@ NpyMultiFileReader::NpyMultiFileReader(const std::vector<std::string>& filePaths
}
}

std::shared_ptr<arrow::RecordBatch> NpyMultiFileReader::readBlock(block_idx_t blockIdx) const {
void NpyMultiFileReader::readBlock(block_idx_t blockIdx, common::DataChunk* dataChunkToRead) const {
assert(fileReaders.size() > 1);
auto resultArrowBatch = fileReaders[0]->readBlock(blockIdx);
for (int fileIdx = 1; fileIdx < fileReaders.size(); fileIdx++) {
auto nextArrowBatch = fileReaders[fileIdx]->readBlock(blockIdx);
auto result = resultArrowBatch->AddColumn(
fileIdx, std::to_string(fileIdx), nextArrowBatch->column(0));
if (result.ok()) {
resultArrowBatch = result.ValueOrDie();
} else {
throw CopyException("Failed to read NPY file.");
}
for (auto i = 0u; i < fileReaders.size(); i++) {
fileReaders[i]->readBlock(blockIdx, dataChunkToRead->getValueVector(i).get());
}
return resultArrowBatch;
}

} // namespace processor
Expand Down
7 changes: 1 addition & 6 deletions src/processor/operator/persistent/reader_functions.cpp
Original file line number Diff line number Diff line change
Expand Up @@ -375,12 +375,7 @@ void ReaderFunctions::readRowsFromNodeParquetFile(const ReaderFunctionData& func
void ReaderFunctions::readRowsFromNPYFile(const ReaderFunctionData& functionData,
common::block_idx_t blockIdx, common::DataChunk* dataChunkToRead) {
auto& readerData = reinterpret_cast<const NPYReaderFunctionData&>(functionData);
auto recordBatch = readerData.reader->readBlock(blockIdx);
for (auto i = 0u; i < dataChunkToRead->getNumValueVectors(); i++) {
ArrowColumnVector::setArrowColumn(dataChunkToRead->getValueVector(i).get(),
std::make_shared<arrow::ChunkedArray>(recordBatch->column((int)i)));
}
dataChunkToRead->state->selVector->selectedSize = recordBatch->num_rows();
readerData.reader->readBlock(blockIdx, dataChunkToRead);
}

void ReaderFunctions::readRowsFromRDFFile(const ReaderFunctionData& functionData,
Expand Down
2 changes: 2 additions & 0 deletions src/processor/operator/persistent/reader_state.cpp
Original file line number Diff line number Diff line change
Expand Up @@ -4,6 +4,8 @@ using namespace kuzu::catalog;
using namespace kuzu::common;
using namespace kuzu::storage;

#include "arrow/array.h"

namespace kuzu {
namespace processor {

Expand Down
39 changes: 33 additions & 6 deletions src/storage/store/column_chunk.cpp
Original file line number Diff line number Diff line change
Expand Up @@ -143,8 +143,8 @@ ColumnChunk::ColumnChunk(
}

void ColumnChunk::initialize(offset_t capacity) {
bufferSize = numBytesPerValue * capacity;
this->capacity = capacity;
bufferSize = getBufferSize();
buffer = std::make_unique<uint8_t[]>(bufferSize);
if (nullChunk) {
static_cast<ColumnChunk*>(nullChunk.get())->initialize(capacity);
Expand Down Expand Up @@ -598,6 +598,25 @@ void FixedListColumnChunk::write(const Value& fixedListVal, uint64_t posToWrite)
}
}

void FixedListColumnChunk::copyVectorToBuffer(
common::ValueVector* vector, common::offset_t startPosInChunk) {
auto vectorDataToWriteFrom = vector->getData();
for (auto i = 0u; i < vector->state->selVector->selectedSize; i++) {
auto pos = vector->state->selVector->selectedPositions[i];
nullChunk->setNull(startPosInChunk + i, vector->isNull(pos));
auto offset = getOffsetInBuffer(startPosInChunk + i);
memcpy(buffer.get() + getOffsetInBuffer(startPosInChunk + i),
vectorDataToWriteFrom + pos * numBytesPerValue, numBytesPerValue);
}
}

uint64_t FixedListColumnChunk::getBufferSize() const {
auto numElementsInAPage =
PageUtils::getNumElementsInAPage(numBytesPerValue, false /* hasNull */);
auto numPages = capacity / numElementsInAPage + (capacity % numElementsInAPage ? 1 : 0);
return common::BufferPoolConstants::PAGE_4KB_SIZE * numPages;
}

std::unique_ptr<ColumnChunk> ColumnChunkFactory::createColumnChunk(
const LogicalType& dataType, CSVReaderConfig* csvReaderConfig) {
auto csvReaderConfigCopy = csvReaderConfig ? csvReaderConfig->copy() : nullptr;
Expand Down Expand Up @@ -698,11 +717,19 @@ void ColumnChunk::copyVectorToBuffer(
common::ValueVector* vector, common::offset_t startPosInChunk) {
auto bufferToWrite = buffer.get() + startPosInChunk * numBytesPerValue;
auto vectorDataToWriteFrom = vector->getData();
for (auto i = 0u; i < vector->state->selVector->selectedSize; i++) {
auto pos = vector->state->selVector->selectedPositions[i];
nullChunk->setNull(startPosInChunk + i, vector->isNull(pos));
memcpy(bufferToWrite, vectorDataToWriteFrom + pos * numBytesPerValue, numBytesPerValue);
bufferToWrite += numBytesPerValue;
if (vector->state->selVector->isUnfiltered()) {
memcpy(bufferToWrite, vectorDataToWriteFrom,
vector->state->selVector->selectedSize * numBytesPerValue);
for (auto i = 0u; i < vector->state->selVector->selectedSize; i++) {
nullChunk->setNull(startPosInChunk + i, vector->isNull(i));
}
} else {
for (auto i = 0u; i < vector->state->selVector->selectedSize; i++) {
auto pos = vector->state->selVector->selectedPositions[i];
nullChunk->setNull(startPosInChunk + i, vector->isNull(pos));
memcpy(bufferToWrite, vectorDataToWriteFrom + pos * numBytesPerValue, numBytesPerValue);
bufferToWrite += numBytesPerValue;
}
}
}

Expand Down