be/src/exec/scan/split_source_connector.h
Line | Count | Source |
1 | | // Licensed to the Apache Software Foundation (ASF) under one |
2 | | // or more contributor license agreements. See the NOTICE file |
3 | | // distributed with this work for additional information |
4 | | // regarding copyright ownership. The ASF licenses this file |
5 | | // to you under the Apache License, Version 2.0 (the |
6 | | // "License"); you may not use this file except in compliance |
7 | | // with the License. You may obtain a copy of the License at |
8 | | // |
9 | | // http://www.apache.org/licenses/LICENSE-2.0 |
10 | | // |
11 | | // Unless required by applicable law or agreed to in writing, |
12 | | // software distributed under the License is distributed on an |
13 | | // "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY |
14 | | // KIND, either express or implied. See the License for the |
15 | | // specific language governing permissions and limitations |
16 | | // under the License. |
17 | | |
18 | | #pragma once |
19 | | |
20 | | #include "common/config.h" |
21 | | #include "runtime/runtime_state.h" |
22 | | #include "util/client_cache.h" |
23 | | |
24 | | namespace doris { |
25 | | #include "common/compile_check_begin.h" |
26 | | |
27 | | /* |
28 | | * Multiple scanners within a scan node share a split source. |
29 | | * Each scanner call `get_next` to get the next scan range. A fast scanner will immediately obtain |
30 | | * the next scan range, so there is no situation of data skewing. |
31 | | */ |
32 | | class SplitSourceConnector { |
33 | | public: |
34 | 3 | SplitSourceConnector() = default; |
35 | 3 | virtual ~SplitSourceConnector() = default; |
36 | | |
37 | | /** |
38 | | * Get the next scan range. has_next should be to true to fetch the next scan range. |
39 | | * @param has_next whether exists the next scan range |
40 | | * @param range the obtained next scan range |
41 | | */ |
42 | | virtual Status get_next(bool* has_next, TFileRangeDesc* range) = 0; |
43 | | |
44 | | virtual int num_scan_ranges() = 0; |
45 | | |
46 | | virtual TFileScanRangeParams* get_params() = 0; |
47 | | |
48 | | protected: |
49 | | template <typename T> |
50 | 2 | void _merge_ranges(std::vector<T>& merged_ranges, const std::vector<T>& scan_ranges) { |
51 | 2 | if (scan_ranges.size() <= _max_scanners) { |
52 | 2 | merged_ranges = scan_ranges; |
53 | 2 | return; |
54 | 2 | } |
55 | | |
56 | | // There is no need for the number of scanners to exceed the number of threads in thread pool. |
57 | | // scan_ranges is sorted by path(as well as partition path) in FE, so merge scan ranges in order. |
58 | | // In the insert statement, reading data in partition order can reduce the memory usage of BE |
59 | | // and prevent the generation of smaller tables. |
60 | 0 | merged_ranges.resize(_max_scanners); |
61 | 0 | int num_ranges = static_cast<int>(scan_ranges.size()) / _max_scanners; |
62 | 0 | int num_add_one = static_cast<int>(scan_ranges.size()) - num_ranges * _max_scanners; |
63 | 0 | int scan_index = 0; |
64 | 0 | int range_index = 0; |
65 | 0 | for (int i = 0; i < num_add_one; ++i) { |
66 | 0 | merged_ranges[scan_index] = scan_ranges[range_index++]; |
67 | 0 | auto& ranges = |
68 | 0 | merged_ranges[scan_index++].scan_range.ext_scan_range.file_scan_range.ranges; |
69 | 0 | for (int j = 0; j < num_ranges; j++) { |
70 | 0 | auto& tmp_merged_ranges = |
71 | 0 | scan_ranges[range_index++].scan_range.ext_scan_range.file_scan_range.ranges; |
72 | 0 | ranges.insert(ranges.end(), tmp_merged_ranges.begin(), tmp_merged_ranges.end()); |
73 | 0 | } |
74 | 0 | } |
75 | 0 | for (int i = num_add_one; i < _max_scanners; ++i) { |
76 | 0 | merged_ranges[scan_index] = scan_ranges[range_index++]; |
77 | 0 | auto& ranges = |
78 | 0 | merged_ranges[scan_index++].scan_range.ext_scan_range.file_scan_range.ranges; |
79 | 0 | for (int j = 0; j < num_ranges - 1; j++) { |
80 | 0 | auto& tmp_merged_ranges = |
81 | 0 | scan_ranges[range_index++].scan_range.ext_scan_range.file_scan_range.ranges; |
82 | 0 | ranges.insert(ranges.end(), tmp_merged_ranges.begin(), tmp_merged_ranges.end()); |
83 | 0 | } |
84 | 0 | } |
85 | 0 | LOG(INFO) << "Merge " << scan_ranges.size() << " scan ranges to " << merged_ranges.size(); |
86 | 0 | } _ZN5doris20SplitSourceConnector13_merge_rangesINS_16TScanRangeParamsEEEvRSt6vectorIT_SaIS4_EERKS6_ Line | Count | Source | 50 | 2 | void _merge_ranges(std::vector<T>& merged_ranges, const std::vector<T>& scan_ranges) { | 51 | 2 | if (scan_ranges.size() <= _max_scanners) { | 52 | 2 | merged_ranges = scan_ranges; | 53 | 2 | return; | 54 | 2 | } | 55 | | | 56 | | // There is no need for the number of scanners to exceed the number of threads in thread pool. | 57 | | // scan_ranges is sorted by path(as well as partition path) in FE, so merge scan ranges in order. | 58 | | // In the insert statement, reading data in partition order can reduce the memory usage of BE | 59 | | // and prevent the generation of smaller tables. | 60 | 0 | merged_ranges.resize(_max_scanners); | 61 | 0 | int num_ranges = static_cast<int>(scan_ranges.size()) / _max_scanners; | 62 | 0 | int num_add_one = static_cast<int>(scan_ranges.size()) - num_ranges * _max_scanners; | 63 | 0 | int scan_index = 0; | 64 | 0 | int range_index = 0; | 65 | 0 | for (int i = 0; i < num_add_one; ++i) { | 66 | 0 | merged_ranges[scan_index] = scan_ranges[range_index++]; | 67 | 0 | auto& ranges = | 68 | 0 | merged_ranges[scan_index++].scan_range.ext_scan_range.file_scan_range.ranges; | 69 | 0 | for (int j = 0; j < num_ranges; j++) { | 70 | 0 | auto& tmp_merged_ranges = | 71 | 0 | scan_ranges[range_index++].scan_range.ext_scan_range.file_scan_range.ranges; | 72 | 0 | ranges.insert(ranges.end(), tmp_merged_ranges.begin(), tmp_merged_ranges.end()); | 73 | 0 | } | 74 | 0 | } | 75 | 0 | for (int i = num_add_one; i < _max_scanners; ++i) { | 76 | 0 | merged_ranges[scan_index] = scan_ranges[range_index++]; | 77 | 0 | auto& ranges = | 78 | 0 | merged_ranges[scan_index++].scan_range.ext_scan_range.file_scan_range.ranges; | 79 | 0 | for (int j = 0; j < num_ranges - 1; j++) { | 80 | 0 | auto& tmp_merged_ranges = | 81 | 0 | scan_ranges[range_index++].scan_range.ext_scan_range.file_scan_range.ranges; | 82 | 0 | ranges.insert(ranges.end(), tmp_merged_ranges.begin(), tmp_merged_ranges.end()); | 83 | 0 | } | 84 | 0 | } | 85 | | LOG(INFO) << "Merge " << scan_ranges.size() << " scan ranges to " << merged_ranges.size(); | 86 | 0 | } |
Unexecuted instantiation: _ZN5doris20SplitSourceConnector13_merge_rangesINS_19TScanRangeLocationsEEEvRSt6vectorIT_SaIS4_EERKS6_ |
87 | | |
88 | | protected: |
89 | | int _max_scanners; |
90 | | }; |
91 | | |
92 | | /** |
93 | | * The file splits are already assigned in `TFileScanRange.ranges`. Scan node has need to |
94 | | * fetch the scan ranges from frontend. |
95 | | * |
96 | | * In cases where the number of files is small, the splits are directly transmitted to backend. |
97 | | */ |
98 | | class LocalSplitSourceConnector : public SplitSourceConnector { |
99 | | private: |
100 | | std::mutex _range_lock; |
101 | | std::vector<TScanRangeParams> _scan_ranges; |
102 | | int _scan_index = 0; |
103 | | int _range_index = 0; |
104 | | |
105 | | public: |
106 | 2 | LocalSplitSourceConnector(const std::vector<TScanRangeParams>& scan_ranges, int max_scanners) { |
107 | 2 | _max_scanners = max_scanners; |
108 | 2 | _merge_ranges<TScanRangeParams>(_scan_ranges, scan_ranges); |
109 | 2 | } |
110 | | |
111 | | Status get_next(bool* has_next, TFileRangeDesc* range) override; |
112 | | |
113 | 2 | int num_scan_ranges() override { return static_cast<int>(_scan_ranges.size()); } |
114 | | |
115 | 0 | TFileScanRangeParams* get_params() override { |
116 | 0 | if (_scan_ranges.size() > 0 && |
117 | 0 | _scan_ranges[0].scan_range.ext_scan_range.file_scan_range.__isset.params) { |
118 | | // for compatibility. |
119 | 0 | return &_scan_ranges[0].scan_range.ext_scan_range.file_scan_range.params; |
120 | 0 | } |
121 | 0 | throw Exception( |
122 | 0 | Status::FatalError("Unreachable, params is got by file_scan_range_params_map")); |
123 | 0 | } |
124 | | }; |
125 | | |
126 | | /** |
127 | | * The file splits are lazily generated in frontend, and saved as a split source in frontend. |
128 | | * The scan node needs to fetch file splits from the frontend service. Each split source is identified by |
129 | | * a unique ID, and the ID is stored in `TFileScanRange.split_source.split_source_id` |
130 | | * |
131 | | * In the case of a large number of files, backend can scan data while obtaining splits information. |
132 | | */ |
133 | | class RemoteSplitSourceConnector : public SplitSourceConnector { |
134 | | private: |
135 | | std::mutex _range_lock; |
136 | | RuntimeState* _state; |
137 | | RuntimeProfile::Counter* _get_split_timer; |
138 | | int64_t _split_source_id; |
139 | | int _num_splits; |
140 | | |
141 | | std::vector<TScanRangeLocations> _scan_ranges; |
142 | | bool _last_batch = false; |
143 | | int _scan_index = 0; |
144 | | int _range_index = 0; |
145 | | |
146 | | public: |
147 | | RemoteSplitSourceConnector(RuntimeState* state, RuntimeProfile::Counter* get_split_timer, |
148 | | int64_t split_source_id, int num_splits, int max_scanners) |
149 | 0 | : _state(state), |
150 | 0 | _get_split_timer(get_split_timer), |
151 | 0 | _split_source_id(split_source_id), |
152 | 0 | _num_splits(num_splits) { |
153 | 0 | _max_scanners = max_scanners; |
154 | 0 | } |
155 | | |
156 | | Status get_next(bool* has_next, TFileRangeDesc* range) override; |
157 | | |
158 | | /* |
159 | | * Remote split source is fetched in batch mode, and the splits are generated while scanning, |
160 | | * so the number of scan ranges may not be accurate. |
161 | | */ |
162 | 0 | int num_scan_ranges() override { return _num_splits; } |
163 | | |
164 | 0 | TFileScanRangeParams* get_params() override { |
165 | 0 | throw Exception( |
166 | 0 | Status::FatalError("Unreachable, params is got by file_scan_range_params_map")); |
167 | 0 | } |
168 | | }; |
169 | | |
170 | | #include "common/compile_check_end.h" |
171 | | } // namespace doris |