be/src/storage/predicate_collector.cpp
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 | | #include "storage/predicate_collector.h" |
19 | | |
20 | | #include <glog/logging.h> |
21 | | |
22 | | #include "exprs/vexpr.h" |
23 | | #include "exprs/vexpr_context.h" |
24 | | #include "exprs/vliteral.h" |
25 | | #include "exprs/vsearch.h" |
26 | | #include "exprs/vslot_ref.h" |
27 | | #include "gen_cpp/Exprs_types.h" |
28 | | #include "storage/index/index_reader_helper.h" |
29 | | #include "storage/index/inverted/analyzer/analyzer.h" |
30 | | #include "storage/index/inverted/util/string_helper.h" |
31 | | #include "storage/tablet/tablet_schema.h" |
32 | | |
33 | | namespace doris { |
34 | | |
35 | | using namespace segment_v2; |
36 | | |
37 | 23 | VSlotRef* PredicateCollector::find_slot_ref(const VExprSPtr& expr) const { |
38 | 23 | if (!expr) { |
39 | 0 | return nullptr; |
40 | 0 | } |
41 | | |
42 | 23 | auto cur = VExpr::expr_without_cast(expr); |
43 | 23 | if (cur->node_type() == TExprNodeType::SLOT_REF) { |
44 | 23 | return static_cast<VSlotRef*>(cur.get()); |
45 | 23 | } |
46 | | |
47 | 0 | for (const auto& ch : cur->children()) { |
48 | 0 | if (auto* s = find_slot_ref(ch)) { |
49 | 0 | return s; |
50 | 0 | } |
51 | 0 | } |
52 | | |
53 | 0 | return nullptr; |
54 | 0 | } |
55 | | |
56 | | std::string PredicateCollector::build_field_name(int32_t col_unique_id, |
57 | 14 | const std::string& suffix_path) const { |
58 | 14 | std::string field_name = std::to_string(col_unique_id); |
59 | 14 | if (!suffix_path.empty()) { |
60 | 0 | field_name += "." + suffix_path; |
61 | 0 | } |
62 | 14 | return field_name; |
63 | 14 | } |
64 | | |
65 | | Status MatchPredicateCollector::collect(RuntimeState* state, const TabletSchemaSPtr& tablet_schema, |
66 | 14 | const VExprSPtr& expr, CollectInfoMap* collect_infos) { |
67 | 14 | DCHECK(collect_infos != nullptr); |
68 | | |
69 | 14 | auto* left_slot_ref = find_slot_ref(expr->children()[0]); |
70 | 14 | if (left_slot_ref == nullptr) { |
71 | 0 | return Status::Error<ErrorCode::INVERTED_INDEX_NOT_SUPPORTED>( |
72 | 0 | "Index statistics collection failed: Cannot find slot reference in match predicate " |
73 | 0 | "left expression"); |
74 | 0 | } |
75 | | |
76 | 14 | auto* right_literal = static_cast<VLiteral*>(expr->children()[1].get()); |
77 | 14 | DCHECK(right_literal != nullptr); |
78 | | |
79 | 14 | const auto* sd = state->desc_tbl().get_slot_descriptor(left_slot_ref->slot_id()); |
80 | 14 | if (sd == nullptr) { |
81 | 0 | return Status::Error<ErrorCode::INVERTED_INDEX_NOT_SUPPORTED>( |
82 | 0 | "Index statistics collection failed: Cannot find slot descriptor for slot_id={}", |
83 | 0 | left_slot_ref->slot_id()); |
84 | 0 | } |
85 | | |
86 | 14 | int32_t col_idx = tablet_schema->field_index(left_slot_ref->column_name()); |
87 | 14 | if (col_idx == -1) { |
88 | 0 | return Status::Error<ErrorCode::INVERTED_INDEX_NOT_SUPPORTED>( |
89 | 0 | "Index statistics collection failed: Cannot find column index for column={}", |
90 | 0 | left_slot_ref->column_name()); |
91 | 0 | } |
92 | | |
93 | 14 | const auto& column = tablet_schema->column(col_idx); |
94 | 14 | auto index_metas = tablet_schema->inverted_indexs(sd->col_unique_id(), column.suffix_path()); |
95 | | |
96 | 14 | #ifndef BE_TEST |
97 | 14 | if (index_metas.empty()) { |
98 | 0 | return Status::Error<ErrorCode::INVERTED_INDEX_NOT_SUPPORTED>( |
99 | 0 | "Index statistics collection failed: Score query is not supported without inverted " |
100 | 0 | "index for column={}", |
101 | 0 | left_slot_ref->column_name()); |
102 | 0 | } |
103 | 14 | #endif |
104 | | |
105 | 14 | for (const auto* index_meta : index_metas) { |
106 | 14 | if (!InvertedIndexAnalyzer::should_analyzer(index_meta->properties())) { |
107 | 0 | continue; |
108 | 0 | } |
109 | | |
110 | 14 | if (!IndexReaderHelper::is_need_similarity_score(expr->op(), index_meta)) { |
111 | 0 | continue; |
112 | 0 | } |
113 | | |
114 | 14 | auto options = DataTypeSerDe::get_default_format_options(); |
115 | 14 | options.timezone = &state->timezone_obj(); |
116 | 14 | auto term_infos = InvertedIndexAnalyzer::get_analyse_result(right_literal->value(options), |
117 | 14 | index_meta->properties()); |
118 | | |
119 | 14 | std::string field_name = |
120 | 14 | build_field_name(index_meta->col_unique_ids()[0], column.suffix_path()); |
121 | 14 | std::wstring ws_field_name = StringHelper::to_wstring(field_name); |
122 | | |
123 | 14 | auto iter = collect_infos->find(ws_field_name); |
124 | 14 | if (iter == collect_infos->end()) { |
125 | 14 | CollectInfo collect_info; |
126 | 14 | collect_info.term_infos.insert(term_infos.begin(), term_infos.end()); |
127 | 14 | collect_info.index_meta = index_meta; |
128 | 14 | (*collect_infos)[ws_field_name] = std::move(collect_info); |
129 | 14 | } else { |
130 | 0 | iter->second.term_infos.insert(term_infos.begin(), term_infos.end()); |
131 | 0 | } |
132 | 14 | } |
133 | | |
134 | 14 | return Status::OK(); |
135 | 14 | } |
136 | | |
137 | | Status SearchPredicateCollector::collect(RuntimeState* state, const TabletSchemaSPtr& tablet_schema, |
138 | 0 | const VExprSPtr& expr, CollectInfoMap* collect_infos) { |
139 | 0 | DCHECK(collect_infos != nullptr); |
140 | |
|
141 | 0 | auto* search_expr = dynamic_cast<VSearchExpr*>(expr.get()); |
142 | 0 | if (search_expr == nullptr) { |
143 | 0 | return Status::InternalError("SearchPredicateCollector: expr is not VSearchExpr type"); |
144 | 0 | } |
145 | | |
146 | 0 | const TSearchParam& search_param = search_expr->get_search_param(); |
147 | |
|
148 | 0 | RETURN_IF_ERROR(collect_from_clause(search_param.root, state, tablet_schema, collect_infos)); |
149 | | |
150 | 0 | return Status::OK(); |
151 | 0 | } |
152 | | |
153 | | Status SearchPredicateCollector::collect_from_clause(const TSearchClause& clause, |
154 | | RuntimeState* state, |
155 | | const TabletSchemaSPtr& tablet_schema, |
156 | 0 | CollectInfoMap* collect_infos) { |
157 | 0 | const std::string& clause_type = clause.clause_type; |
158 | 0 | ClauseTypeCategory category = get_clause_type_category(clause_type); |
159 | |
|
160 | 0 | if (category == ClauseTypeCategory::COMPOUND) { |
161 | 0 | if (clause.__isset.children) { |
162 | 0 | for (const auto& child_clause : clause.children) { |
163 | 0 | RETURN_IF_ERROR( |
164 | 0 | collect_from_clause(child_clause, state, tablet_schema, collect_infos)); |
165 | 0 | } |
166 | 0 | } |
167 | 0 | return Status::OK(); |
168 | 0 | } |
169 | | |
170 | 0 | return collect_from_leaf(clause, state, tablet_schema, collect_infos); |
171 | 0 | } |
172 | | |
173 | | Status SearchPredicateCollector::collect_from_leaf(const TSearchClause& clause, RuntimeState* state, |
174 | | const TabletSchemaSPtr& tablet_schema, |
175 | 0 | CollectInfoMap* collect_infos) { |
176 | 0 | if (!clause.__isset.field_name || !clause.__isset.value) { |
177 | 0 | return Status::InvalidArgument("Search clause missing field_name or value"); |
178 | 0 | } |
179 | | |
180 | 0 | const std::string& field_name = clause.field_name; |
181 | 0 | const std::string& value = clause.value; |
182 | 0 | const std::string& clause_type = clause.clause_type; |
183 | |
|
184 | 0 | if (!is_score_query_type(clause_type)) { |
185 | 0 | return Status::OK(); |
186 | 0 | } |
187 | | |
188 | 0 | int32_t col_idx = tablet_schema->field_index(field_name); |
189 | 0 | if (col_idx == -1) { |
190 | 0 | return Status::OK(); |
191 | 0 | } |
192 | | |
193 | 0 | const auto& column = tablet_schema->column(col_idx); |
194 | |
|
195 | 0 | auto index_metas = tablet_schema->inverted_indexs(column.unique_id(), column.suffix_path()); |
196 | 0 | if (index_metas.empty()) { |
197 | 0 | return Status::OK(); |
198 | 0 | } |
199 | | |
200 | 0 | ClauseTypeCategory category = get_clause_type_category(clause_type); |
201 | 0 | for (const auto* index_meta : index_metas) { |
202 | 0 | std::set<TermInfo, TermInfoComparer> term_infos; |
203 | |
|
204 | 0 | if (category == ClauseTypeCategory::TOKENIZED) { |
205 | 0 | if (InvertedIndexAnalyzer::should_analyzer(index_meta->properties())) { |
206 | 0 | auto analyzed_terms = |
207 | 0 | InvertedIndexAnalyzer::get_analyse_result(value, index_meta->properties()); |
208 | 0 | term_infos.insert(analyzed_terms.begin(), analyzed_terms.end()); |
209 | 0 | } else { |
210 | 0 | term_infos.insert(TermInfo(value)); |
211 | 0 | } |
212 | 0 | } else if (category == ClauseTypeCategory::NON_TOKENIZED) { |
213 | 0 | if (clause_type == "TERM" && |
214 | 0 | InvertedIndexAnalyzer::should_analyzer(index_meta->properties())) { |
215 | 0 | auto analyzed_terms = |
216 | 0 | InvertedIndexAnalyzer::get_analyse_result(value, index_meta->properties()); |
217 | 0 | term_infos.insert(analyzed_terms.begin(), analyzed_terms.end()); |
218 | 0 | } else { |
219 | 0 | term_infos.insert(TermInfo(value)); |
220 | 0 | } |
221 | 0 | } |
222 | |
|
223 | 0 | std::string lucene_field_name = |
224 | 0 | build_field_name(index_meta->col_unique_ids()[0], column.suffix_path()); |
225 | 0 | std::wstring ws_field_name = StringHelper::to_wstring(lucene_field_name); |
226 | |
|
227 | 0 | auto iter = collect_infos->find(ws_field_name); |
228 | 0 | if (iter == collect_infos->end()) { |
229 | 0 | CollectInfo collect_info; |
230 | 0 | collect_info.term_infos = std::move(term_infos); |
231 | 0 | collect_info.index_meta = index_meta; |
232 | 0 | (*collect_infos)[ws_field_name] = std::move(collect_info); |
233 | 0 | } else { |
234 | 0 | iter->second.term_infos.insert(term_infos.begin(), term_infos.end()); |
235 | 0 | } |
236 | 0 | } |
237 | |
|
238 | 0 | return Status::OK(); |
239 | 0 | } |
240 | | |
241 | 0 | bool SearchPredicateCollector::is_score_query_type(const std::string& clause_type) const { |
242 | 0 | return clause_type == "TERM" || clause_type == "EXACT" || clause_type == "PHRASE" || |
243 | 0 | clause_type == "MATCH" || clause_type == "ANY" || clause_type == "ALL"; |
244 | 0 | } |
245 | | |
246 | | SearchPredicateCollector::ClauseTypeCategory SearchPredicateCollector::get_clause_type_category( |
247 | 0 | const std::string& clause_type) const { |
248 | 0 | if (clause_type == "AND" || clause_type == "OR" || clause_type == "NOT" || |
249 | 0 | clause_type == "OCCUR_BOOLEAN") { |
250 | 0 | return ClauseTypeCategory::COMPOUND; |
251 | 0 | } else if (clause_type == "TERM" || clause_type == "EXACT") { |
252 | 0 | return ClauseTypeCategory::NON_TOKENIZED; |
253 | 0 | } else if (clause_type == "PHRASE" || clause_type == "MATCH" || clause_type == "ANY" || |
254 | 0 | clause_type == "ALL") { |
255 | 0 | return ClauseTypeCategory::TOKENIZED; |
256 | 0 | } else { |
257 | | LOG(WARNING) << "Unknown clause type '" << clause_type |
258 | 0 | << "', defaulting to NON_TOKENIZED category"; |
259 | 0 | return ClauseTypeCategory::NON_TOKENIZED; |
260 | 0 | } |
261 | 0 | } |
262 | | |
263 | | } // namespace doris |