OmniSciDB  2e3a973ef4
ParquetStringNoneEncoder.h
Go to the documentation of this file.
1 /*
2  * Copyright 2020 OmniSci, Inc.
3  *
4  * Licensed under the Apache License, Version 2.0 (the "License");
5  * you may not use this file except in compliance with the License.
6  * You may obtain a copy of the License at
7  *
8  * http://www.apache.org/licenses/LICENSE-2.0
9  *
10  * Unless required by applicable law or agreed to in writing, software
11  * distributed under the License is distributed on an "AS IS" BASIS,
12  * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
13  * See the License for the specific language governing permissions and
14  * limitations under the License.
15  */
16 
17 #pragma once
18 
19 #include "LazyParquetChunkLoader.h"
20 #include "ParquetEncoder.h"
21 
22 #include <parquet/schema.h>
23 #include <parquet/types.h>
24 
25 namespace foreign_storage {
26 
28  public:
30  Data_Namespace::AbstractBuffer* index_buffer)
31  : ParquetEncoder(buffer)
32  , index_buffer_(index_buffer)
33  , encode_buffer_(LazyParquetChunkLoader::batch_reader_num_elements *
34  sizeof(StringOffsetT)) {}
35 
36  void appendData(const int16_t* def_levels,
37  const int16_t* rep_levels,
38  const int64_t values_read,
39  const int64_t levels_read,
40  const bool is_last_batch,
41  int8_t* values) override {
42  CHECK(levels_read > 0);
44 
45  auto parquet_data_ptr = reinterpret_cast<const parquet::ByteArray*>(values);
46  auto offsets = reinterpret_cast<StringOffsetT*>(encode_buffer_.data());
47  auto last_offset = buffer_->size();
48 
49  size_t total_len = 0;
50  for (int64_t i = 0, j = 0; i < levels_read; ++i) {
51  if (def_levels[i]) {
52  CHECK(j < values_read);
53  auto& byte_array = parquet_data_ptr[j++];
54  total_len += byte_array.len;
55  }
56  offsets[i] = last_offset + total_len;
57  }
58  index_buffer_->append(encode_buffer_.data(), levels_read * sizeof(StringOffsetT));
59 
60  encode_buffer_.resize(std::max<size_t>(total_len, encode_buffer_.size()));
61  buffer_->reserve(total_len);
62  total_len = 0;
63  for (int64_t i = 0, j = 0; i < levels_read; ++i) {
64  if (def_levels[i]) {
65  CHECK(j < values_read);
66  auto& byte_array = parquet_data_ptr[j++];
67  memcpy(encode_buffer_.data() + total_len, byte_array.ptr, byte_array.len);
68  total_len += byte_array.len;
69  }
70  }
71  buffer_->append(encode_buffer_.data(), total_len);
72  }
73 
74  private:
76  if (!index_buffer_->size()) {
77  // write the initial starting offset
78  StringOffsetT zero = 0;
79  index_buffer_->append(reinterpret_cast<int8_t*>(&zero), sizeof(StringOffsetT));
80  }
81  }
82 
84  std::vector<int8_t> encode_buffer_;
85 };
86 
87 } // namespace foreign_storage
Data_Namespace::AbstractBuffer * index_buffer_
int32_t StringOffsetT
Definition: sqltypes.h:868
An AbstractBuffer is a unit of data management for a data manager.
virtual void append(int8_t *src, const size_t num_bytes, const MemoryLevel src_buffer_type=CPU_LEVEL, const int device_id=-1)=0
void appendData(const int16_t *def_levels, const int16_t *rep_levels, const int64_t values_read, const int64_t levels_read, const bool is_last_batch, int8_t *values) override
ParquetStringNoneEncoder(Data_Namespace::AbstractBuffer *buffer, Data_Namespace::AbstractBuffer *index_buffer)
#define CHECK(condition)
Definition: Logger.h:197
Data_Namespace::AbstractBuffer * buffer_
virtual void reserve(size_t num_bytes)=0