1 | #pragma once |
2 | |
3 | #include <memory> |
4 | #include <time.h> |
5 | #include <IO/ReadBufferFromFileBase.h> |
6 | #include "CompressedReadBufferBase.h" |
7 | #include <IO/UncompressedCache.h> |
8 | #include <port/clock.h> |
9 | |
10 | |
11 | namespace DB |
12 | { |
13 | |
14 | |
15 | /** A buffer for reading from a compressed file using the cache of decompressed blocks. |
16 | * The external cache is passed as an argument to the constructor. |
17 | * Allows you to increase performance in cases where the same blocks are often read. |
18 | * Disadvantages: |
19 | * - in case you need to read a lot of data in a row, but of them only a part is cached, you have to do seek-and. |
20 | */ |
21 | class CachedCompressedReadBuffer : public CompressedReadBufferBase, public ReadBuffer |
22 | { |
23 | private: |
24 | const std::string path; |
25 | UncompressedCache * cache; |
26 | size_t buf_size; |
27 | size_t estimated_size; |
28 | size_t aio_threshold; |
29 | |
30 | std::unique_ptr<ReadBufferFromFileBase> file_in; |
31 | size_t file_pos; |
32 | |
33 | /// A piece of data from the cache, or a piece of read data that we put into the cache. |
34 | UncompressedCache::MappedPtr owned_cell; |
35 | |
36 | void initInput(); |
37 | bool nextImpl() override; |
38 | |
39 | /// Passed into file_in. |
40 | ReadBufferFromFileBase::ProfileCallback profile_callback; |
41 | clockid_t clock_type {}; |
42 | |
43 | public: |
44 | CachedCompressedReadBuffer( |
45 | const std::string & path_, UncompressedCache * cache_, size_t estimated_size_, size_t aio_threshold_, |
46 | size_t buf_size_ = DBMS_DEFAULT_BUFFER_SIZE); |
47 | |
48 | |
49 | void seek(size_t offset_in_compressed_file, size_t offset_in_decompressed_block); |
50 | |
51 | void setProfileCallback(const ReadBufferFromFileBase::ProfileCallback & profile_callback_, clockid_t clock_type_ = CLOCK_MONOTONIC_COARSE) |
52 | { |
53 | profile_callback = profile_callback_; |
54 | clock_type = clock_type_; |
55 | } |
56 | }; |
57 | |
58 | } |
59 | |