2022-05-23 00:16:32 +04:00

461 lines
16 KiB
C++

// 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 <cstring>
#include <memory>
#include <vector>
#include "arrow/util/spaced.h"
#include "parquet/exception.h"
#include "parquet/platform.h"
#include "parquet/types.h"
namespace arrow {
class Array;
class ArrayBuilder;
class BinaryArray;
class BinaryBuilder;
class BooleanBuilder;
class Int32Type;
class Int64Type;
class FloatType;
class DoubleType;
class FixedSizeBinaryType;
template <typename T>
class NumericBuilder;
class FixedSizeBinaryBuilder;
template <typename T>
class Dictionary32Builder;
} // namespace arrow
namespace parquet {
template <typename DType>
class TypedEncoder;
using BooleanEncoder = TypedEncoder<BooleanType>;
using Int32Encoder = TypedEncoder<Int32Type>;
using Int64Encoder = TypedEncoder<Int64Type>;
using Int96Encoder = TypedEncoder<Int96Type>;
using FloatEncoder = TypedEncoder<FloatType>;
using DoubleEncoder = TypedEncoder<DoubleType>;
using ByteArrayEncoder = TypedEncoder<ByteArrayType>;
using FLBAEncoder = TypedEncoder<FLBAType>;
template <typename DType>
class TypedDecoder;
class BooleanDecoder;
using Int32Decoder = TypedDecoder<Int32Type>;
using Int64Decoder = TypedDecoder<Int64Type>;
using Int96Decoder = TypedDecoder<Int96Type>;
using FloatDecoder = TypedDecoder<FloatType>;
using DoubleDecoder = TypedDecoder<DoubleType>;
using ByteArrayDecoder = TypedDecoder<ByteArrayType>;
class FLBADecoder;
template <typename T>
struct EncodingTraits;
template <>
struct EncodingTraits<BooleanType> {
using Encoder = BooleanEncoder;
using Decoder = BooleanDecoder;
using ArrowType = ::arrow::BooleanType;
using Accumulator = ::arrow::BooleanBuilder;
struct DictAccumulator {};
};
template <>
struct EncodingTraits<Int32Type> {
using Encoder = Int32Encoder;
using Decoder = Int32Decoder;
using ArrowType = ::arrow::Int32Type;
using Accumulator = ::arrow::NumericBuilder<::arrow::Int32Type>;
using DictAccumulator = ::arrow::Dictionary32Builder<::arrow::Int32Type>;
};
template <>
struct EncodingTraits<Int64Type> {
using Encoder = Int64Encoder;
using Decoder = Int64Decoder;
using ArrowType = ::arrow::Int64Type;
using Accumulator = ::arrow::NumericBuilder<::arrow::Int64Type>;
using DictAccumulator = ::arrow::Dictionary32Builder<::arrow::Int64Type>;
};
template <>
struct EncodingTraits<Int96Type> {
using Encoder = Int96Encoder;
using Decoder = Int96Decoder;
struct Accumulator {};
struct DictAccumulator {};
};
template <>
struct EncodingTraits<FloatType> {
using Encoder = FloatEncoder;
using Decoder = FloatDecoder;
using ArrowType = ::arrow::FloatType;
using Accumulator = ::arrow::NumericBuilder<::arrow::FloatType>;
using DictAccumulator = ::arrow::Dictionary32Builder<::arrow::FloatType>;
};
template <>
struct EncodingTraits<DoubleType> {
using Encoder = DoubleEncoder;
using Decoder = DoubleDecoder;
using ArrowType = ::arrow::DoubleType;
using Accumulator = ::arrow::NumericBuilder<::arrow::DoubleType>;
using DictAccumulator = ::arrow::Dictionary32Builder<::arrow::DoubleType>;
};
template <>
struct EncodingTraits<ByteArrayType> {
using Encoder = ByteArrayEncoder;
using Decoder = ByteArrayDecoder;
/// \brief Internal helper class for decoding BYTE_ARRAY data where we can
/// overflow the capacity of a single arrow::BinaryArray
struct Accumulator {
std::unique_ptr<::arrow::BinaryBuilder> builder;
std::vector<std::shared_ptr<::arrow::Array>> chunks;
};
using ArrowType = ::arrow::BinaryType;
using DictAccumulator = ::arrow::Dictionary32Builder<::arrow::BinaryType>;
};
template <>
struct EncodingTraits<FLBAType> {
using Encoder = FLBAEncoder;
using Decoder = FLBADecoder;
using ArrowType = ::arrow::FixedSizeBinaryType;
using Accumulator = ::arrow::FixedSizeBinaryBuilder;
using DictAccumulator = ::arrow::Dictionary32Builder<::arrow::FixedSizeBinaryType>;
};
class ColumnDescriptor;
// Untyped base for all encoders
class Encoder {
public:
virtual ~Encoder() = default;
virtual int64_t EstimatedDataEncodedSize() = 0;
virtual std::shared_ptr<Buffer> FlushValues() = 0;
virtual Encoding::type encoding() const = 0;
virtual void Put(const ::arrow::Array& values) = 0;
virtual MemoryPool* memory_pool() const = 0;
};
// Base class for value encoders. Since encoders may or not have state (e.g.,
// dictionary encoding) we use a class instance to maintain any state.
//
// Encode interfaces are internal, subject to change without deprecation.
template <typename DType>
class TypedEncoder : virtual public Encoder {
public:
typedef typename DType::c_type T;
using Encoder::Put;
virtual void Put(const T* src, int num_values) = 0;
virtual void Put(const std::vector<T>& src, int num_values = -1);
virtual void PutSpaced(const T* src, int num_values, const uint8_t* valid_bits,
int64_t valid_bits_offset) = 0;
};
template <typename DType>
void TypedEncoder<DType>::Put(const std::vector<T>& src, int num_values) {
if (num_values == -1) {
num_values = static_cast<int>(src.size());
}
Put(src.data(), num_values);
}
template <>
inline void TypedEncoder<BooleanType>::Put(const std::vector<bool>& src, int num_values) {
// NOTE(wesm): This stub is here only to satisfy the compiler; it is
// overridden later with the actual implementation
}
// Base class for dictionary encoders
template <typename DType>
class DictEncoder : virtual public TypedEncoder<DType> {
public:
/// Writes out any buffered indices to buffer preceded by the bit width of this data.
/// Returns the number of bytes written.
/// If the supplied buffer is not big enough, returns -1.
/// buffer must be preallocated with buffer_len bytes. Use EstimatedDataEncodedSize()
/// to size buffer.
virtual int WriteIndices(uint8_t* buffer, int buffer_len) = 0;
virtual int dict_encoded_size() = 0;
// virtual int dict_encoded_size() { return dict_encoded_size_; }
virtual int bit_width() const = 0;
/// Writes out the encoded dictionary to buffer. buffer must be preallocated to
/// dict_encoded_size() bytes.
virtual void WriteDict(uint8_t* buffer) = 0;
virtual int num_entries() const = 0;
/// \brief EXPERIMENTAL: Append dictionary indices into the encoder. It is
/// assumed (without any boundschecking) that the indices reference
/// pre-existing dictionary values
/// \param[in] indices the dictionary index values. Only Int32Array currently
/// supported
virtual void PutIndices(const ::arrow::Array& indices) = 0;
/// \brief EXPERIMENTAL: Append dictionary into encoder, inserting indices
/// separately. Currently throws exception if the current dictionary memo is
/// non-empty
/// \param[in] values the dictionary values. Only valid for certain
/// Parquet/Arrow type combinations, like BYTE_ARRAY/BinaryArray
virtual void PutDictionary(const ::arrow::Array& values) = 0;
};
// ----------------------------------------------------------------------
// Value decoding
class Decoder {
public:
virtual ~Decoder() = default;
// Sets the data for a new page. This will be called multiple times on the same
// decoder and should reset all internal state.
virtual void SetData(int num_values, const uint8_t* data, int len) = 0;
// Returns the number of values left (for the last call to SetData()). This is
// the number of values left in this page.
virtual int values_left() const = 0;
virtual Encoding::type encoding() const = 0;
};
template <typename DType>
class TypedDecoder : virtual public Decoder {
public:
using T = typename DType::c_type;
/// \brief Decode values into a buffer
///
/// Subclasses may override the more specialized Decode methods below.
///
/// \param[in] buffer destination for decoded values
/// \param[in] max_values maximum number of values to decode
/// \return The number of values decoded. Should be identical to max_values except
/// at the end of the current data page.
virtual int Decode(T* buffer, int max_values) = 0;
/// \brief Decode the values in this data page but leave spaces for null entries.
///
/// \param[in] buffer destination for decoded values
/// \param[in] num_values size of the def_levels and buffer arrays including the number
/// of null slots
/// \param[in] null_count number of null slots
/// \param[in] valid_bits bitmap data indicating position of valid slots
/// \param[in] valid_bits_offset offset into valid_bits
/// \return The number of values decoded, including nulls.
virtual int DecodeSpaced(T* buffer, int num_values, int null_count,
const uint8_t* valid_bits, int64_t valid_bits_offset) {
if (null_count > 0) {
int values_to_read = num_values - null_count;
int values_read = Decode(buffer, values_to_read);
if (values_read != values_to_read) {
throw ParquetException("Number of values / definition_levels read did not match");
}
return ::arrow::util::internal::SpacedExpand<T>(buffer, num_values, null_count,
valid_bits, valid_bits_offset);
} else {
return Decode(buffer, num_values);
}
}
/// \brief Decode into an ArrayBuilder or other accumulator
///
/// This function assumes the definition levels were already decoded
/// as a validity bitmap in the given `valid_bits`. `null_count`
/// is the number of 0s in `valid_bits`.
/// As a space optimization, it is allowed for `valid_bits` to be null
/// if `null_count` is zero.
///
/// \return number of values decoded
virtual int DecodeArrow(int num_values, int null_count, const uint8_t* valid_bits,
int64_t valid_bits_offset,
typename EncodingTraits<DType>::Accumulator* out) = 0;
/// \brief Decode into an ArrayBuilder or other accumulator ignoring nulls
///
/// \return number of values decoded
int DecodeArrowNonNull(int num_values,
typename EncodingTraits<DType>::Accumulator* out) {
return DecodeArrow(num_values, 0, /*valid_bits=*/NULLPTR, 0, out);
}
/// \brief Decode into a DictionaryBuilder
///
/// This function assumes the definition levels were already decoded
/// as a validity bitmap in the given `valid_bits`. `null_count`
/// is the number of 0s in `valid_bits`.
/// As a space optimization, it is allowed for `valid_bits` to be null
/// if `null_count` is zero.
///
/// \return number of values decoded
virtual int DecodeArrow(int num_values, int null_count, const uint8_t* valid_bits,
int64_t valid_bits_offset,
typename EncodingTraits<DType>::DictAccumulator* builder) = 0;
/// \brief Decode into a DictionaryBuilder ignoring nulls
///
/// \return number of values decoded
int DecodeArrowNonNull(int num_values,
typename EncodingTraits<DType>::DictAccumulator* builder) {
return DecodeArrow(num_values, 0, /*valid_bits=*/NULLPTR, 0, builder);
}
};
template <typename DType>
class DictDecoder : virtual public TypedDecoder<DType> {
public:
using T = typename DType::c_type;
virtual void SetDict(TypedDecoder<DType>* dictionary) = 0;
/// \brief Insert dictionary values into the Arrow dictionary builder's memo,
/// but do not append any indices
virtual void InsertDictionary(::arrow::ArrayBuilder* builder) = 0;
/// \brief Decode only dictionary indices and append to dictionary
/// builder. The builder must have had the dictionary from this decoder
/// inserted already.
///
/// \warning Remember to reset the builder each time the dict decoder is initialized
/// with a new dictionary page
virtual int DecodeIndicesSpaced(int num_values, int null_count,
const uint8_t* valid_bits, int64_t valid_bits_offset,
::arrow::ArrayBuilder* builder) = 0;
/// \brief Decode only dictionary indices (no nulls)
///
/// \warning Remember to reset the builder each time the dict decoder is initialized
/// with a new dictionary page
virtual int DecodeIndices(int num_values, ::arrow::ArrayBuilder* builder) = 0;
/// \brief Decode only dictionary indices (no nulls). Same as above
/// DecodeIndices but target is an array instead of a builder.
///
/// \note API EXPERIMENTAL
virtual int DecodeIndices(int num_values, int32_t* indices) = 0;
/// \brief Get dictionary. The reader will call this API when it encounters a
/// new dictionary.
///
/// @param[out] dictionary The pointer to dictionary values. Dictionary is owned by
/// the decoder and is destroyed when the decoder is destroyed.
/// @param[out] dictionary_length The dictionary length.
///
/// \note API EXPERIMENTAL
virtual void GetDictionary(const T** dictionary, int32_t* dictionary_length) = 0;
};
// ----------------------------------------------------------------------
// TypedEncoder specializations, traits, and factory functions
class BooleanDecoder : virtual public TypedDecoder<BooleanType> {
public:
using TypedDecoder<BooleanType>::Decode;
virtual int Decode(uint8_t* buffer, int max_values) = 0;
};
class FLBADecoder : virtual public TypedDecoder<FLBAType> {
public:
using TypedDecoder<FLBAType>::DecodeSpaced;
// TODO(wesm): As possible follow-up to PARQUET-1508, we should examine if
// there is value in adding specialized read methods for
// FIXED_LEN_BYTE_ARRAY. If only Decimal data can occur with this data type
// then perhaps not
};
PARQUET_EXPORT
std::unique_ptr<Encoder> MakeEncoder(
Type::type type_num, Encoding::type encoding, bool use_dictionary = false,
const ColumnDescriptor* descr = NULLPTR,
::arrow::MemoryPool* pool = ::arrow::default_memory_pool());
template <typename DType>
std::unique_ptr<typename EncodingTraits<DType>::Encoder> MakeTypedEncoder(
Encoding::type encoding, bool use_dictionary = false,
const ColumnDescriptor* descr = NULLPTR,
::arrow::MemoryPool* pool = ::arrow::default_memory_pool()) {
using OutType = typename EncodingTraits<DType>::Encoder;
std::unique_ptr<Encoder> base =
MakeEncoder(DType::type_num, encoding, use_dictionary, descr, pool);
return std::unique_ptr<OutType>(dynamic_cast<OutType*>(base.release()));
}
PARQUET_EXPORT
std::unique_ptr<Decoder> MakeDecoder(Type::type type_num, Encoding::type encoding,
const ColumnDescriptor* descr = NULLPTR);
namespace detail {
PARQUET_EXPORT
std::unique_ptr<Decoder> MakeDictDecoder(Type::type type_num,
const ColumnDescriptor* descr,
::arrow::MemoryPool* pool);
} // namespace detail
template <typename DType>
std::unique_ptr<DictDecoder<DType>> MakeDictDecoder(
const ColumnDescriptor* descr = NULLPTR,
::arrow::MemoryPool* pool = ::arrow::default_memory_pool()) {
using OutType = DictDecoder<DType>;
auto decoder = detail::MakeDictDecoder(DType::type_num, descr, pool);
return std::unique_ptr<OutType>(dynamic_cast<OutType*>(decoder.release()));
}
template <typename DType>
std::unique_ptr<typename EncodingTraits<DType>::Decoder> MakeTypedDecoder(
Encoding::type encoding, const ColumnDescriptor* descr = NULLPTR) {
using OutType = typename EncodingTraits<DType>::Decoder;
std::unique_ptr<Decoder> base = MakeDecoder(DType::type_num, encoding, descr);
return std::unique_ptr<OutType>(dynamic_cast<OutType*>(base.release()));
}
} // namespace parquet