jamtur01's picture
Upload folder using huggingface_hub
9c6594c verified
// Licensed to the Apache Software Foundation (ASF) under one
// or more contributor license agreements. See the NOTICE file
// distributed with this work for additional information
// regarding copyright ownership. The ASF licenses this file
// to you 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.
#pragma once
#include <cstdint>
#include <memory>
#include <utility>
#include <vector>
#include "parquet/exception.h"
#include "parquet/level_conversion.h"
#include "parquet/metadata.h"
#include "parquet/platform.h"
#include "parquet/properties.h"
#include "parquet/schema.h"
#include "parquet/types.h"
namespace arrow {
class Array;
class ChunkedArray;
namespace bit_util {
class BitReader;
} // namespace bit_util
namespace util {
class RleDecoder;
} // namespace util
} // namespace arrow
namespace parquet {
class Decryptor;
class Page;
// 16 MB is the default maximum page header size
static constexpr uint32_t kDefaultMaxPageHeaderSize = 16 * 1024 * 1024;
// 16 KB is the default expected page header size
static constexpr uint32_t kDefaultPageHeaderSize = 16 * 1024;
// \brief DataPageStats stores encoded statistics and number of values/rows for
// a page.
struct PARQUET_EXPORT DataPageStats {
DataPageStats(const EncodedStatistics* encoded_statistics, int32_t num_values,
std::optional<int32_t> num_rows)
: encoded_statistics(encoded_statistics),
num_values(num_values),
num_rows(num_rows) {}
// Encoded statistics extracted from the page header.
// Nullptr if there are no statistics in the page header.
const EncodedStatistics* encoded_statistics;
// Number of values stored in the page. Filled for both V1 and V2 data pages.
// For repeated fields, this can be greater than number of rows. For
// non-repeated fields, this will be the same as the number of rows.
int32_t num_values;
// Number of rows stored in the page. std::nullopt if not available.
std::optional<int32_t> num_rows;
};
class PARQUET_EXPORT LevelDecoder {
public:
LevelDecoder();
~LevelDecoder();
// Initialize the LevelDecoder state with new data
// and return the number of bytes consumed
int SetData(Encoding::type encoding, int16_t max_level, int num_buffered_values,
const uint8_t* data, int32_t data_size);
void SetDataV2(int32_t num_bytes, int16_t max_level, int num_buffered_values,
const uint8_t* data);
// Decodes a batch of levels into an array and returns the number of levels decoded
int Decode(int batch_size, int16_t* levels);
private:
int bit_width_;
int num_values_remaining_;
Encoding::type encoding_;
std::unique_ptr<::arrow::util::RleDecoder> rle_decoder_;
std::unique_ptr<::arrow::bit_util::BitReader> bit_packed_decoder_;
int16_t max_level_;
};
struct CryptoContext {
bool start_decrypt_with_dictionary_page = false;
int16_t row_group_ordinal = -1;
int16_t column_ordinal = -1;
std::function<std::unique_ptr<Decryptor>()> meta_decryptor_factory;
std::function<std::unique_ptr<Decryptor>()> data_decryptor_factory;
};
// Abstract page iterator interface. This way, we can feed column pages to the
// ColumnReader through whatever mechanism we choose
class PARQUET_EXPORT PageReader {
using DataPageFilter = std::function<bool(const DataPageStats&)>;
public:
virtual ~PageReader() = default;
static std::unique_ptr<PageReader> Open(
std::shared_ptr<ArrowInputStream> stream, int64_t total_num_values,
Compression::type codec, bool always_compressed = false,
::arrow::MemoryPool* pool = ::arrow::default_memory_pool(),
const CryptoContext* ctx = NULLPTR);
static std::unique_ptr<PageReader> Open(std::shared_ptr<ArrowInputStream> stream,
int64_t total_num_values,
Compression::type codec,
const ReaderProperties& properties,
bool always_compressed = false,
const CryptoContext* ctx = NULLPTR);
// If data_page_filter is present (not null), NextPage() will call the
// callback function exactly once per page in the order the pages appear in
// the column. If the callback function returns true the page will be
// skipped. The callback will be called only if the page type is DATA_PAGE or
// DATA_PAGE_V2. Dictionary pages will not be skipped.
// Caller is responsible for checking that statistics are correct using
// ApplicationVersion::HasCorrectStatistics().
// \note API EXPERIMENTAL
void set_data_page_filter(DataPageFilter data_page_filter) {
data_page_filter_ = std::move(data_page_filter);
}
// @returns: shared_ptr<Page>(nullptr) on EOS, std::shared_ptr<Page>
// containing new Page otherwise
//
// The returned Page may contain references that aren't guaranteed to live
// beyond the next call to NextPage().
virtual std::shared_ptr<Page> NextPage() = 0;
virtual void set_max_page_header_size(uint32_t size) = 0;
protected:
// Callback that decides if we should skip a page or not.
DataPageFilter data_page_filter_;
};
class PARQUET_EXPORT ColumnReader {
public:
virtual ~ColumnReader() = default;
static std::shared_ptr<ColumnReader> Make(
const ColumnDescriptor* descr, std::unique_ptr<PageReader> pager,
::arrow::MemoryPool* pool = ::arrow::default_memory_pool());
// Returns true if there are still values in this column.
virtual bool HasNext() = 0;
virtual Type::type type() const = 0;
virtual const ColumnDescriptor* descr() const = 0;
// Get the encoding that can be exposed by this reader. If it returns
// dictionary encoding, then ReadBatchWithDictionary can be used to read data.
//
// \note API EXPERIMENTAL
virtual ExposedEncoding GetExposedEncoding() = 0;
protected:
friend class RowGroupReader;
// Set the encoding that can be exposed by this reader.
//
// \note API EXPERIMENTAL
virtual void SetExposedEncoding(ExposedEncoding encoding) = 0;
};
// API to read values from a single column. This is a main client facing API.
template <typename DType>
class TypedColumnReader : public ColumnReader {
public:
using T = typename DType::c_type;
// Read a batch of repetition levels, definition levels, and values from the
// column.
//
// Since null values are not stored in the values, the number of values read
// may be less than the number of repetition and definition levels. With
// nested data this is almost certainly true.
//
// Set def_levels or rep_levels to nullptr if you want to skip reading them.
// This is only safe if you know through some other source that there are no
// undefined values.
//
// To fully exhaust a row group, you must read batches until the number of
// values read reaches the number of stored values according to the metadata.
//
// This API is the same for both V1 and V2 of the DataPage
//
// @returns: actual number of levels read (see values_read for number of values read)
virtual int64_t ReadBatch(int64_t batch_size, int16_t* def_levels, int16_t* rep_levels,
T* values, int64_t* values_read) = 0;
// Skip reading values. This method will work for both repeated and
// non-repeated fields. Note that this method is skipping values and not
// records. This distinction is important for repeated fields, meaning that
// we are not skipping over the values to the next record. For example,
// consider the following two consecutive records containing one repeated field:
// {[1, 2, 3]}, {[4, 5]}. If we Skip(2), our next read value will be 3, which
// is inside the first record.
// Returns the number of values skipped.
virtual int64_t Skip(int64_t num_values_to_skip) = 0;
// Read a batch of repetition levels, definition levels, and indices from the
// column. And read the dictionary if a dictionary page is encountered during
// reading pages. This API is similar to ReadBatch(), with ability to read
// dictionary and indices. It is only valid to call this method when the reader can
// expose dictionary encoding. (i.e., the reader's GetExposedEncoding() returns
// DICTIONARY).
//
// The dictionary is read along with the data page. When there's no data page,
// the dictionary won't be returned.
//
// @param batch_size The batch size to read
// @param[out] def_levels The Parquet definition levels.
// @param[out] rep_levels The Parquet repetition levels.
// @param[out] indices The dictionary indices.
// @param[out] indices_read The number of indices read.
// @param[out] dict The pointer to dictionary values. It will return nullptr if
// there's no data page. Each column chunk only has one dictionary page. The dictionary
// is owned by the reader, so the caller is responsible for copying the dictionary
// values before the reader gets destroyed.
// @param[out] dict_len The dictionary length. It will return 0 if there's no data
// page.
// @returns: actual number of levels read (see indices_read for number of
// indices read
//
// \note API EXPERIMENTAL
virtual int64_t ReadBatchWithDictionary(int64_t batch_size, int16_t* def_levels,
int16_t* rep_levels, int32_t* indices,
int64_t* indices_read, const T** dict,
int32_t* dict_len) = 0;
};
namespace internal {
/// \brief Stateful column reader that delimits semantic records for both flat
/// and nested columns
///
/// \note API EXPERIMENTAL
/// \since 1.3.0
class PARQUET_EXPORT RecordReader {
public:
/// \brief Creates a record reader.
/// @param descr Column descriptor
/// @param leaf_info Level info, used to determine if a column is nullable or not
/// @param pool Memory pool to use for buffering values and rep/def levels
/// @param read_dictionary True if reading directly as Arrow dictionary-encoded
/// @param read_dense_for_nullable True if reading dense and not leaving space for null
/// values
static std::shared_ptr<RecordReader> Make(
const ColumnDescriptor* descr, LevelInfo leaf_info,
::arrow::MemoryPool* pool = ::arrow::default_memory_pool(),
bool read_dictionary = false, bool read_dense_for_nullable = false);
virtual ~RecordReader() = default;
/// \brief Attempt to read indicated number of records from column chunk
/// Note that for repeated fields, a record may have more than one value
/// and all of them are read. If read_dense_for_nullable() it will
/// not leave any space for null values. Otherwise, it will read spaced.
/// \return number of records read
virtual int64_t ReadRecords(int64_t num_records) = 0;
/// \brief Attempt to skip indicated number of records from column chunk.
/// Note that for repeated fields, a record may have more than one value
/// and all of them are skipped.
/// \return number of records skipped
virtual int64_t SkipRecords(int64_t num_records) = 0;
/// \brief Pre-allocate space for data. Results in better flat read performance
virtual void Reserve(int64_t num_values) = 0;
/// \brief Clear consumed values and repetition/definition levels as the
/// result of calling ReadRecords
/// For FLBA and ByteArray types, call GetBuilderChunks() to reset them.
virtual void Reset() = 0;
/// \brief Transfer filled values buffer to caller. A new one will be
/// allocated in subsequent ReadRecords calls
virtual std::shared_ptr<ResizableBuffer> ReleaseValues() = 0;
/// \brief Transfer filled validity bitmap buffer to caller. A new one will
/// be allocated in subsequent ReadRecords calls
virtual std::shared_ptr<ResizableBuffer> ReleaseIsValid() = 0;
/// \brief Return true if the record reader has more internal data yet to
/// process
virtual bool HasMoreData() const = 0;
/// \brief Advance record reader to the next row group. Must be set before
/// any records could be read/skipped.
/// \param[in] reader obtained from RowGroupReader::GetColumnPageReader
virtual void SetPageReader(std::unique_ptr<PageReader> reader) = 0;
/// \brief Returns the underlying column reader's descriptor.
virtual const ColumnDescriptor* descr() const = 0;
virtual void DebugPrintState() = 0;
/// \brief Returns the dictionary owned by the current decoder. Throws an
/// exception if the current decoder is not for dictionary encoding. The caller is
/// responsible for casting the returned pointer to proper type depending on the
/// column's physical type. An example:
/// const ByteArray* dict = reinterpret_cast<const ByteArray*>(ReadDictionary(&len));
/// or:
/// const float* dict = reinterpret_cast<const float*>(ReadDictionary(&len));
/// \param[out] dictionary_length The number of dictionary entries.
virtual const void* ReadDictionary(int32_t* dictionary_length) = 0;
/// \brief Decoded definition levels
int16_t* def_levels() const {
return reinterpret_cast<int16_t*>(def_levels_->mutable_data());
}
/// \brief Decoded repetition levels
int16_t* rep_levels() const {
return reinterpret_cast<int16_t*>(rep_levels_->mutable_data());
}
/// \brief Decoded values, including nulls, if any
/// FLBA and ByteArray types do not use this array and read into their own
/// builders.
uint8_t* values() const { return values_->mutable_data(); }
/// \brief Number of values written, including space left for nulls if any.
/// If this Reader was constructed with read_dense_for_nullable(), there is no space for
/// nulls and null_count() will be 0. There is no read-ahead/buffering for values. For
/// FLBA and ByteArray types this value reflects the values written with the last
/// ReadRecords call since those readers will reset the values after each call.
int64_t values_written() const { return values_written_; }
/// \brief Number of definition / repetition levels (from those that have
/// been decoded) that have been consumed inside the reader.
int64_t levels_position() const { return levels_position_; }
/// \brief Number of definition / repetition levels that have been written
/// internally in the reader. This may be larger than values_written() because
/// for repeated fields we need to look at the levels in advance to figure out
/// the record boundaries.
int64_t levels_written() const { return levels_written_; }
/// \brief Number of nulls in the leaf that we have read so far into the
/// values vector. This is only valid when !read_dense_for_nullable(). When
/// read_dense_for_nullable() it will always be 0.
int64_t null_count() const { return null_count_; }
/// \brief True if the leaf values are nullable
bool nullable_values() const { return nullable_values_; }
/// \brief True if reading directly as Arrow dictionary-encoded
bool read_dictionary() const { return read_dictionary_; }
/// \brief True if reading dense for nullable columns.
bool read_dense_for_nullable() const { return read_dense_for_nullable_; }
protected:
/// \brief Indicates if we can have nullable values. Note that repeated fields
/// may or may not be nullable.
bool nullable_values_;
bool at_record_start_;
int64_t records_read_;
/// \brief Stores values. These values are populated based on each ReadRecords
/// call. No extra values are buffered for the next call. SkipRecords will not
/// add any value to this buffer.
std::shared_ptr<::arrow::ResizableBuffer> values_;
/// \brief False for BYTE_ARRAY, in which case we don't allocate the values
/// buffer and we directly read into builder classes.
bool uses_values_;
/// \brief Values that we have read into 'values_' + 'null_count_'.
int64_t values_written_;
int64_t values_capacity_;
int64_t null_count_;
/// \brief Each bit corresponds to one element in 'values_' and specifies if it
/// is null or not null.
///
/// Not set if leaf type is not nullable or read_dense_for_nullable_ is true.
std::shared_ptr<::arrow::ResizableBuffer> valid_bits_;
/// \brief Buffer for definition levels. May contain more levels than
/// is actually read. This is because we read levels ahead to
/// figure out record boundaries for repeated fields.
/// For flat required fields, 'def_levels_' and 'rep_levels_' are not
/// populated. For non-repeated fields 'rep_levels_' is not populated.
/// 'def_levels_' and 'rep_levels_' must be of the same size if present.
std::shared_ptr<::arrow::ResizableBuffer> def_levels_;
/// \brief Buffer for repetition levels. Only populated for repeated
/// fields.
std::shared_ptr<::arrow::ResizableBuffer> rep_levels_;
/// \brief Number of definition / repetition levels that have been written
/// internally in the reader. This may be larger than values_written() since
/// for repeated fields we need to look at the levels in advance to figure out
/// the record boundaries.
int64_t levels_written_;
/// \brief Position of the next level that should be consumed.
int64_t levels_position_;
int64_t levels_capacity_;
bool read_dictionary_ = false;
// If true, we will not leave any space for the null values in the values_
// vector or fill nulls values in BinaryRecordReader/DictionaryRecordReader.
//
// If read_dense_for_nullable_ is true, the BinaryRecordReader/DictionaryRecordReader
// might still populate the validity bitmap buffer.
bool read_dense_for_nullable_ = false;
};
class BinaryRecordReader : virtual public RecordReader {
public:
virtual std::vector<std::shared_ptr<::arrow::Array>> GetBuilderChunks() = 0;
};
/// \brief Read records directly to dictionary-encoded Arrow form (int32
/// indices). Only valid for BYTE_ARRAY columns
class DictionaryRecordReader : virtual public RecordReader {
public:
virtual std::shared_ptr<::arrow::ChunkedArray> GetResult() = 0;
};
} // namespace internal
using BoolReader = TypedColumnReader<BooleanType>;
using Int32Reader = TypedColumnReader<Int32Type>;
using Int64Reader = TypedColumnReader<Int64Type>;
using Int96Reader = TypedColumnReader<Int96Type>;
using FloatReader = TypedColumnReader<FloatType>;
using DoubleReader = TypedColumnReader<DoubleType>;
using ByteArrayReader = TypedColumnReader<ByteArrayType>;
using FixedLenByteArrayReader = TypedColumnReader<FLBAType>;
} // namespace parquet