/root/doris/be/src/vec/exprs/vsearch.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 "vec/exprs/vsearch.h" |
19 | | |
20 | | #include <memory> |
21 | | #include <roaring/roaring.hh> |
22 | | |
23 | | #include "common/logging.h" |
24 | | #include "common/status.h" |
25 | | #include "glog/logging.h" |
26 | | #include "olap/rowset/segment_v2/inverted_index_reader.h" |
27 | | #include "vec/columns/column_const.h" |
28 | | #include "vec/exprs/vexpr_context.h" |
29 | | #include "vec/exprs/vliteral.h" |
30 | | #include "vec/exprs/vslot_ref.h" |
31 | | #include "vec/functions/function_search.h" |
32 | | |
33 | | namespace doris::vectorized { |
34 | | using namespace segment_v2; |
35 | | |
36 | | namespace { |
37 | | |
38 | | struct SearchInputBundle { |
39 | | std::unordered_map<std::string, IndexIterator*> iterators; |
40 | | std::unordered_map<std::string, vectorized::IndexFieldNameAndTypePair> field_types; |
41 | | std::vector<int> column_ids; |
42 | | vectorized::ColumnsWithTypeAndName literal_args; |
43 | | }; |
44 | | |
45 | | Status collect_search_inputs(const VSearchExpr& expr, VExprContext* context, |
46 | 4 | SearchInputBundle* bundle) { |
47 | 4 | DCHECK(bundle != nullptr); |
48 | | |
49 | 4 | auto index_context = context->get_inverted_index_context(); |
50 | 4 | if (index_context == nullptr) { |
51 | 0 | LOG(WARNING) << "collect_search_inputs: No inverted index context available"; |
52 | 0 | return Status::InternalError("No inverted index context available"); |
53 | 0 | } |
54 | | |
55 | | // Get field bindings for variant subcolumn support |
56 | 4 | const auto& search_param = expr.get_search_param(); |
57 | 4 | const auto& field_bindings = search_param.field_bindings; |
58 | | |
59 | 4 | int child_index = 0; // Index for iterating through children |
60 | 4 | for (const auto& child : expr.children()) { |
61 | 4 | if (child->is_slot_ref()) { |
62 | 3 | auto* column_slot_ref = assert_cast<VSlotRef*>(child.get()); |
63 | 3 | int column_id = column_slot_ref->column_id(); |
64 | 3 | auto* iterator = index_context->get_inverted_index_iterator_by_column_id(column_id); |
65 | | |
66 | | // Determine the field_name from field_bindings (for variant subcolumns) |
67 | | // field_bindings and children should have the same order |
68 | 3 | std::string field_name; |
69 | 3 | if (child_index < field_bindings.size()) { |
70 | | // Use field_name from binding (may include "parent.subcolumn" for variant) |
71 | 3 | field_name = field_bindings[child_index].field_name; |
72 | 3 | } else { |
73 | | // Fallback to column_name if binding not found |
74 | 0 | field_name = column_slot_ref->column_name(); |
75 | 0 | } |
76 | | |
77 | | // Only collect fields that have iterators (materialized columns with indexes) |
78 | 3 | if (iterator != nullptr) { |
79 | 2 | const auto* storage_name_type = |
80 | 2 | index_context->get_storage_name_and_type_by_column_id(column_id); |
81 | 2 | if (storage_name_type == nullptr) { |
82 | 1 | return Status::InternalError("storage_name_type not found for column {} in {}", |
83 | 1 | column_id, expr.expr_name()); |
84 | 1 | } |
85 | | |
86 | 1 | bundle->iterators.emplace(field_name, iterator); |
87 | 1 | bundle->field_types.emplace(field_name, *storage_name_type); |
88 | 1 | bundle->column_ids.emplace_back(column_id); |
89 | 1 | } |
90 | | |
91 | 2 | child_index++; |
92 | 2 | } else if (child->is_literal()) { |
93 | 0 | auto* literal = assert_cast<VLiteral*>(child.get()); |
94 | 0 | bundle->literal_args.emplace_back(literal->get_column_ptr(), literal->get_data_type(), |
95 | 0 | literal->expr_name()); |
96 | 1 | } else { |
97 | | // Check if this is ElementAt expression (for variant subcolumn access) |
98 | 1 | if (child->expr_name() == "element_at" && child_index < field_bindings.size() && |
99 | 1 | field_bindings[child_index].__isset.is_variant_subcolumn && |
100 | 1 | field_bindings[child_index].is_variant_subcolumn) { |
101 | | // Variant subcolumn not materialized - skip, will create empty BitmapQuery in function_search |
102 | 0 | child_index++; |
103 | 0 | continue; |
104 | 0 | } |
105 | | |
106 | | // Not a supported child type |
107 | 1 | return Status::InvalidArgument("Unsupported child node type: {}", child->expr_name()); |
108 | 1 | } |
109 | 4 | } |
110 | | |
111 | 2 | return Status::OK(); |
112 | 4 | } |
113 | | |
114 | | } // namespace |
115 | | |
116 | 77 | VSearchExpr::VSearchExpr(const TExprNode& node) : VExpr(node) { |
117 | 77 | if (node.__isset.search_param) { |
118 | 73 | _search_param = node.search_param; |
119 | 73 | _original_dsl = _search_param.original_dsl; |
120 | 73 | } |
121 | 77 | } |
122 | | |
123 | 4 | const std::string& VSearchExpr::expr_name() const { |
124 | 4 | static const std::string name = "VSearchExpr"; |
125 | 4 | return name; |
126 | 4 | } |
127 | | |
128 | 3 | Status VSearchExpr::execute(VExprContext* context, Block* block, int* result_column_id) { |
129 | 3 | if (fast_execute(context, block, result_column_id)) { |
130 | 1 | return Status::OK(); |
131 | 1 | } |
132 | | |
133 | 2 | return Status::InternalError("SearchExpr should not be executed without inverted index"); |
134 | 3 | } |
135 | | |
136 | 22 | Status VSearchExpr::evaluate_inverted_index(VExprContext* context, uint32_t segment_num_rows) { |
137 | 22 | LOG(INFO) << "VSearchExpr::evaluate_inverted_index called, DSL: " << _search_param.original_dsl; |
138 | | |
139 | 22 | if (_search_param.original_dsl.empty()) { |
140 | 3 | return Status::InvalidArgument("search DSL is empty"); |
141 | 3 | } |
142 | | |
143 | 19 | auto index_context = context->get_inverted_index_context(); |
144 | 19 | if (!index_context) { |
145 | 15 | LOG(WARNING) << "VSearchExpr: No inverted index context available"; |
146 | 15 | return Status::OK(); |
147 | 15 | } |
148 | | |
149 | 4 | SearchInputBundle bundle; |
150 | 4 | RETURN_IF_ERROR(collect_search_inputs(*this, context, &bundle)); |
151 | | |
152 | 2 | VLOG_DEBUG << "VSearchExpr: bundle.iterators.size()=" << bundle.iterators.size(); |
153 | | |
154 | 2 | if (bundle.iterators.empty()) { |
155 | 1 | LOG(WARNING) << "VSearchExpr: No indexed columns available for evaluation, DSL: " |
156 | 1 | << _original_dsl; |
157 | 1 | auto empty_bitmap = InvertedIndexResultBitmap(std::make_shared<roaring::Roaring>(), |
158 | 1 | std::make_shared<roaring::Roaring>()); |
159 | 1 | index_context->set_inverted_index_result_for_expr(this, std::move(empty_bitmap)); |
160 | 1 | return Status::OK(); |
161 | 1 | } |
162 | | |
163 | 1 | auto function = std::make_shared<FunctionSearch>(); |
164 | 1 | auto result_bitmap = InvertedIndexResultBitmap(); |
165 | 1 | auto status = function->evaluate_inverted_index_with_search_param( |
166 | 1 | _search_param, bundle.field_types, bundle.iterators, segment_num_rows, result_bitmap); |
167 | | |
168 | 1 | if (!status.ok()) { |
169 | 1 | LOG(WARNING) << "VSearchExpr: Function evaluation failed: " << status.to_string(); |
170 | 1 | return status; |
171 | 1 | } |
172 | | |
173 | 0 | index_context->set_inverted_index_result_for_expr(this, result_bitmap); |
174 | 0 | for (int column_id : bundle.column_ids) { |
175 | 0 | index_context->set_true_for_inverted_index_status(this, column_id); |
176 | 0 | } |
177 | |
|
178 | 0 | return Status::OK(); |
179 | 1 | } |
180 | | |
181 | | } // namespace doris::vectorized |