/root/doris/be/src/vec/exprs/vexpr.cpp
Line | Count | Source (jump to first uncovered line) |
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/vexpr.h" |
19 | | |
20 | | #include <fmt/format.h> |
21 | | #include <gen_cpp/Exprs_types.h> |
22 | | #include <gen_cpp/FrontendService_types.h> |
23 | | #include <thrift/protocol/TDebugProtocol.h> |
24 | | |
25 | | #include <algorithm> |
26 | | #include <boost/algorithm/string/split.hpp> |
27 | | #include <boost/iterator/iterator_facade.hpp> |
28 | | #include <memory> |
29 | | #include <stack> |
30 | | #include <utility> |
31 | | |
32 | | #include "common/config.h" |
33 | | #include "common/exception.h" |
34 | | #include "common/status.h" |
35 | | #include "pipeline/pipeline_task.h" |
36 | | #include "runtime/define_primitive_type.h" |
37 | | #include "vec/columns/column_vector.h" |
38 | | #include "vec/columns/columns_number.h" |
39 | | #include "vec/data_types/data_type_array.h" |
40 | | #include "vec/data_types/data_type_factory.hpp" |
41 | | #include "vec/data_types/data_type_nullable.h" |
42 | | #include "vec/data_types/data_type_number.h" |
43 | | #include "vec/exprs/varray_literal.h" |
44 | | #include "vec/exprs/vcase_expr.h" |
45 | | #include "vec/exprs/vcast_expr.h" |
46 | | #include "vec/exprs/vcolumn_ref.h" |
47 | | #include "vec/exprs/vcompound_pred.h" |
48 | | #include "vec/exprs/vectorized_fn_call.h" |
49 | | #include "vec/exprs/vexpr_context.h" |
50 | | #include "vec/exprs/vin_predicate.h" |
51 | | #include "vec/exprs/vinfo_func.h" |
52 | | #include "vec/exprs/vlambda_function_call_expr.h" |
53 | | #include "vec/exprs/vlambda_function_expr.h" |
54 | | #include "vec/exprs/vliteral.h" |
55 | | #include "vec/exprs/vmap_literal.h" |
56 | | #include "vec/exprs/vmatch_predicate.h" |
57 | | #include "vec/exprs/vslot_ref.h" |
58 | | #include "vec/exprs/vstruct_literal.h" |
59 | | #include "vec/exprs/vtuple_is_null_predicate.h" |
60 | | #include "vec/utils/util.hpp" |
61 | | |
62 | | namespace doris { |
63 | | class RowDescriptor; |
64 | | class RuntimeState; |
65 | | |
66 | | // NOLINTBEGIN(readability-function-cognitive-complexity) |
67 | | // NOLINTBEGIN(readability-function-size) |
68 | | TExprNode create_texpr_node_from(const void* data, const PrimitiveType& type, int precision, |
69 | 0 | int scale) { |
70 | 0 | TExprNode node; |
71 | |
|
72 | 0 | switch (type) { |
73 | 0 | case TYPE_BOOLEAN: { Branch (73:5): [True: 0, False: 0]
|
74 | 0 | THROW_IF_ERROR(create_texpr_literal_node<TYPE_BOOLEAN>(data, &node)); Line | Count | Source | 652 | 0 | do { \ | 653 | 0 | Status _status_ = (stmt); \ | 654 | 0 | if (UNLIKELY(!_status_.ok())) { \ Line | Count | Source | 36 | 0 | #define UNLIKELY(expr) __builtin_expect(!!(expr), 0) Branch (36:24): [True: 0, False: 0]
|
| 655 | 0 | throw Exception(_status_); \ | 656 | 0 | } \ | 657 | 0 | } while (false) Branch (657:14): [Folded - Ignored]
|
|
75 | 0 | break; |
76 | 0 | } |
77 | 0 | case TYPE_TINYINT: { Branch (77:5): [True: 0, False: 0]
|
78 | 0 | THROW_IF_ERROR(create_texpr_literal_node<TYPE_TINYINT>(data, &node)); Line | Count | Source | 652 | 0 | do { \ | 653 | 0 | Status _status_ = (stmt); \ | 654 | 0 | if (UNLIKELY(!_status_.ok())) { \ Line | Count | Source | 36 | 0 | #define UNLIKELY(expr) __builtin_expect(!!(expr), 0) Branch (36:24): [True: 0, False: 0]
|
| 655 | 0 | throw Exception(_status_); \ | 656 | 0 | } \ | 657 | 0 | } while (false) Branch (657:14): [Folded - Ignored]
|
|
79 | 0 | break; |
80 | 0 | } |
81 | 0 | case TYPE_SMALLINT: { Branch (81:5): [True: 0, False: 0]
|
82 | 0 | THROW_IF_ERROR(create_texpr_literal_node<TYPE_SMALLINT>(data, &node)); Line | Count | Source | 652 | 0 | do { \ | 653 | 0 | Status _status_ = (stmt); \ | 654 | 0 | if (UNLIKELY(!_status_.ok())) { \ Line | Count | Source | 36 | 0 | #define UNLIKELY(expr) __builtin_expect(!!(expr), 0) Branch (36:24): [True: 0, False: 0]
|
| 655 | 0 | throw Exception(_status_); \ | 656 | 0 | } \ | 657 | 0 | } while (false) Branch (657:14): [Folded - Ignored]
|
|
83 | 0 | break; |
84 | 0 | } |
85 | 0 | case TYPE_INT: { Branch (85:5): [True: 0, False: 0]
|
86 | 0 | THROW_IF_ERROR(create_texpr_literal_node<TYPE_INT>(data, &node)); Line | Count | Source | 652 | 0 | do { \ | 653 | 0 | Status _status_ = (stmt); \ | 654 | 0 | if (UNLIKELY(!_status_.ok())) { \ Line | Count | Source | 36 | 0 | #define UNLIKELY(expr) __builtin_expect(!!(expr), 0) Branch (36:24): [True: 0, False: 0]
|
| 655 | 0 | throw Exception(_status_); \ | 656 | 0 | } \ | 657 | 0 | } while (false) Branch (657:14): [Folded - Ignored]
|
|
87 | 0 | break; |
88 | 0 | } |
89 | 0 | case TYPE_BIGINT: { Branch (89:5): [True: 0, False: 0]
|
90 | 0 | THROW_IF_ERROR(create_texpr_literal_node<TYPE_BIGINT>(data, &node)); Line | Count | Source | 652 | 0 | do { \ | 653 | 0 | Status _status_ = (stmt); \ | 654 | 0 | if (UNLIKELY(!_status_.ok())) { \ Line | Count | Source | 36 | 0 | #define UNLIKELY(expr) __builtin_expect(!!(expr), 0) Branch (36:24): [True: 0, False: 0]
|
| 655 | 0 | throw Exception(_status_); \ | 656 | 0 | } \ | 657 | 0 | } while (false) Branch (657:14): [Folded - Ignored]
|
|
91 | 0 | break; |
92 | 0 | } |
93 | 0 | case TYPE_LARGEINT: { Branch (93:5): [True: 0, False: 0]
|
94 | 0 | THROW_IF_ERROR(create_texpr_literal_node<TYPE_LARGEINT>(data, &node)); Line | Count | Source | 652 | 0 | do { \ | 653 | 0 | Status _status_ = (stmt); \ | 654 | 0 | if (UNLIKELY(!_status_.ok())) { \ Line | Count | Source | 36 | 0 | #define UNLIKELY(expr) __builtin_expect(!!(expr), 0) Branch (36:24): [True: 0, False: 0]
|
| 655 | 0 | throw Exception(_status_); \ | 656 | 0 | } \ | 657 | 0 | } while (false) Branch (657:14): [Folded - Ignored]
|
|
95 | 0 | break; |
96 | 0 | } |
97 | 0 | case TYPE_FLOAT: { Branch (97:5): [True: 0, False: 0]
|
98 | 0 | THROW_IF_ERROR(create_texpr_literal_node<TYPE_FLOAT>(data, &node)); Line | Count | Source | 652 | 0 | do { \ | 653 | 0 | Status _status_ = (stmt); \ | 654 | 0 | if (UNLIKELY(!_status_.ok())) { \ Line | Count | Source | 36 | 0 | #define UNLIKELY(expr) __builtin_expect(!!(expr), 0) Branch (36:24): [True: 0, False: 0]
|
| 655 | 0 | throw Exception(_status_); \ | 656 | 0 | } \ | 657 | 0 | } while (false) Branch (657:14): [Folded - Ignored]
|
|
99 | 0 | break; |
100 | 0 | } |
101 | 0 | case TYPE_DOUBLE: { Branch (101:5): [True: 0, False: 0]
|
102 | 0 | THROW_IF_ERROR(create_texpr_literal_node<TYPE_DOUBLE>(data, &node)); Line | Count | Source | 652 | 0 | do { \ | 653 | 0 | Status _status_ = (stmt); \ | 654 | 0 | if (UNLIKELY(!_status_.ok())) { \ Line | Count | Source | 36 | 0 | #define UNLIKELY(expr) __builtin_expect(!!(expr), 0) Branch (36:24): [True: 0, False: 0]
|
| 655 | 0 | throw Exception(_status_); \ | 656 | 0 | } \ | 657 | 0 | } while (false) Branch (657:14): [Folded - Ignored]
|
|
103 | 0 | break; |
104 | 0 | } |
105 | 0 | case TYPE_DATEV2: { Branch (105:5): [True: 0, False: 0]
|
106 | 0 | THROW_IF_ERROR(create_texpr_literal_node<TYPE_DATEV2>(data, &node)); Line | Count | Source | 652 | 0 | do { \ | 653 | 0 | Status _status_ = (stmt); \ | 654 | 0 | if (UNLIKELY(!_status_.ok())) { \ Line | Count | Source | 36 | 0 | #define UNLIKELY(expr) __builtin_expect(!!(expr), 0) Branch (36:24): [True: 0, False: 0]
|
| 655 | 0 | throw Exception(_status_); \ | 656 | 0 | } \ | 657 | 0 | } while (false) Branch (657:14): [Folded - Ignored]
|
|
107 | 0 | break; |
108 | 0 | } |
109 | 0 | case TYPE_DATETIMEV2: { Branch (109:5): [True: 0, False: 0]
|
110 | 0 | THROW_IF_ERROR(create_texpr_literal_node<TYPE_DATETIMEV2>(data, &node, precision, scale)); Line | Count | Source | 652 | 0 | do { \ | 653 | 0 | Status _status_ = (stmt); \ | 654 | 0 | if (UNLIKELY(!_status_.ok())) { \ Line | Count | Source | 36 | 0 | #define UNLIKELY(expr) __builtin_expect(!!(expr), 0) Branch (36:24): [True: 0, False: 0]
|
| 655 | 0 | throw Exception(_status_); \ | 656 | 0 | } \ | 657 | 0 | } while (false) Branch (657:14): [Folded - Ignored]
|
|
111 | 0 | break; |
112 | 0 | } |
113 | 0 | case TYPE_DATE: { Branch (113:5): [True: 0, False: 0]
|
114 | 0 | THROW_IF_ERROR(create_texpr_literal_node<TYPE_DATE>(data, &node)); Line | Count | Source | 652 | 0 | do { \ | 653 | 0 | Status _status_ = (stmt); \ | 654 | 0 | if (UNLIKELY(!_status_.ok())) { \ Line | Count | Source | 36 | 0 | #define UNLIKELY(expr) __builtin_expect(!!(expr), 0) Branch (36:24): [True: 0, False: 0]
|
| 655 | 0 | throw Exception(_status_); \ | 656 | 0 | } \ | 657 | 0 | } while (false) Branch (657:14): [Folded - Ignored]
|
|
115 | 0 | break; |
116 | 0 | } |
117 | 0 | case TYPE_DATETIME: { Branch (117:5): [True: 0, False: 0]
|
118 | 0 | THROW_IF_ERROR(create_texpr_literal_node<TYPE_DATETIME>(data, &node)); Line | Count | Source | 652 | 0 | do { \ | 653 | 0 | Status _status_ = (stmt); \ | 654 | 0 | if (UNLIKELY(!_status_.ok())) { \ Line | Count | Source | 36 | 0 | #define UNLIKELY(expr) __builtin_expect(!!(expr), 0) Branch (36:24): [True: 0, False: 0]
|
| 655 | 0 | throw Exception(_status_); \ | 656 | 0 | } \ | 657 | 0 | } while (false) Branch (657:14): [Folded - Ignored]
|
|
119 | 0 | break; |
120 | 0 | } |
121 | 0 | case TYPE_DECIMALV2: { Branch (121:5): [True: 0, False: 0]
|
122 | 0 | THROW_IF_ERROR(create_texpr_literal_node<TYPE_DECIMALV2>(data, &node, precision, scale)); Line | Count | Source | 652 | 0 | do { \ | 653 | 0 | Status _status_ = (stmt); \ | 654 | 0 | if (UNLIKELY(!_status_.ok())) { \ Line | Count | Source | 36 | 0 | #define UNLIKELY(expr) __builtin_expect(!!(expr), 0) Branch (36:24): [True: 0, False: 0]
|
| 655 | 0 | throw Exception(_status_); \ | 656 | 0 | } \ | 657 | 0 | } while (false) Branch (657:14): [Folded - Ignored]
|
|
123 | 0 | break; |
124 | 0 | } |
125 | 0 | case TYPE_DECIMAL32: { Branch (125:5): [True: 0, False: 0]
|
126 | 0 | THROW_IF_ERROR(create_texpr_literal_node<TYPE_DECIMAL32>(data, &node, precision, scale)); Line | Count | Source | 652 | 0 | do { \ | 653 | 0 | Status _status_ = (stmt); \ | 654 | 0 | if (UNLIKELY(!_status_.ok())) { \ Line | Count | Source | 36 | 0 | #define UNLIKELY(expr) __builtin_expect(!!(expr), 0) Branch (36:24): [True: 0, False: 0]
|
| 655 | 0 | throw Exception(_status_); \ | 656 | 0 | } \ | 657 | 0 | } while (false) Branch (657:14): [Folded - Ignored]
|
|
127 | 0 | break; |
128 | 0 | } |
129 | 0 | case TYPE_DECIMAL64: { Branch (129:5): [True: 0, False: 0]
|
130 | 0 | THROW_IF_ERROR(create_texpr_literal_node<TYPE_DECIMAL64>(data, &node, precision, scale)); Line | Count | Source | 652 | 0 | do { \ | 653 | 0 | Status _status_ = (stmt); \ | 654 | 0 | if (UNLIKELY(!_status_.ok())) { \ Line | Count | Source | 36 | 0 | #define UNLIKELY(expr) __builtin_expect(!!(expr), 0) Branch (36:24): [True: 0, False: 0]
|
| 655 | 0 | throw Exception(_status_); \ | 656 | 0 | } \ | 657 | 0 | } while (false) Branch (657:14): [Folded - Ignored]
|
|
131 | 0 | break; |
132 | 0 | } |
133 | 0 | case TYPE_DECIMAL128I: { Branch (133:5): [True: 0, False: 0]
|
134 | 0 | THROW_IF_ERROR(create_texpr_literal_node<TYPE_DECIMAL128I>(data, &node, precision, scale)); Line | Count | Source | 652 | 0 | do { \ | 653 | 0 | Status _status_ = (stmt); \ | 654 | 0 | if (UNLIKELY(!_status_.ok())) { \ Line | Count | Source | 36 | 0 | #define UNLIKELY(expr) __builtin_expect(!!(expr), 0) Branch (36:24): [True: 0, False: 0]
|
| 655 | 0 | throw Exception(_status_); \ | 656 | 0 | } \ | 657 | 0 | } while (false) Branch (657:14): [Folded - Ignored]
|
|
135 | 0 | break; |
136 | 0 | } |
137 | 0 | case TYPE_DECIMAL256: { Branch (137:5): [True: 0, False: 0]
|
138 | 0 | THROW_IF_ERROR(create_texpr_literal_node<TYPE_DECIMAL256>(data, &node, precision, scale)); Line | Count | Source | 652 | 0 | do { \ | 653 | 0 | Status _status_ = (stmt); \ | 654 | 0 | if (UNLIKELY(!_status_.ok())) { \ Line | Count | Source | 36 | 0 | #define UNLIKELY(expr) __builtin_expect(!!(expr), 0) Branch (36:24): [True: 0, False: 0]
|
| 655 | 0 | throw Exception(_status_); \ | 656 | 0 | } \ | 657 | 0 | } while (false) Branch (657:14): [Folded - Ignored]
|
|
139 | 0 | break; |
140 | 0 | } |
141 | 0 | case TYPE_CHAR: { Branch (141:5): [True: 0, False: 0]
|
142 | 0 | THROW_IF_ERROR(create_texpr_literal_node<TYPE_CHAR>(data, &node)); Line | Count | Source | 652 | 0 | do { \ | 653 | 0 | Status _status_ = (stmt); \ | 654 | 0 | if (UNLIKELY(!_status_.ok())) { \ Line | Count | Source | 36 | 0 | #define UNLIKELY(expr) __builtin_expect(!!(expr), 0) Branch (36:24): [True: 0, False: 0]
|
| 655 | 0 | throw Exception(_status_); \ | 656 | 0 | } \ | 657 | 0 | } while (false) Branch (657:14): [Folded - Ignored]
|
|
143 | 0 | break; |
144 | 0 | } |
145 | 0 | case TYPE_VARCHAR: { Branch (145:5): [True: 0, False: 0]
|
146 | 0 | THROW_IF_ERROR(create_texpr_literal_node<TYPE_VARCHAR>(data, &node)); Line | Count | Source | 652 | 0 | do { \ | 653 | 0 | Status _status_ = (stmt); \ | 654 | 0 | if (UNLIKELY(!_status_.ok())) { \ Line | Count | Source | 36 | 0 | #define UNLIKELY(expr) __builtin_expect(!!(expr), 0) Branch (36:24): [True: 0, False: 0]
|
| 655 | 0 | throw Exception(_status_); \ | 656 | 0 | } \ | 657 | 0 | } while (false) Branch (657:14): [Folded - Ignored]
|
|
147 | 0 | break; |
148 | 0 | } |
149 | 0 | case TYPE_STRING: { Branch (149:5): [True: 0, False: 0]
|
150 | 0 | THROW_IF_ERROR(create_texpr_literal_node<TYPE_STRING>(data, &node)); Line | Count | Source | 652 | 0 | do { \ | 653 | 0 | Status _status_ = (stmt); \ | 654 | 0 | if (UNLIKELY(!_status_.ok())) { \ Line | Count | Source | 36 | 0 | #define UNLIKELY(expr) __builtin_expect(!!(expr), 0) Branch (36:24): [True: 0, False: 0]
|
| 655 | 0 | throw Exception(_status_); \ | 656 | 0 | } \ | 657 | 0 | } while (false) Branch (657:14): [Folded - Ignored]
|
|
151 | 0 | break; |
152 | 0 | } |
153 | 0 | case TYPE_IPV4: { Branch (153:5): [True: 0, False: 0]
|
154 | 0 | THROW_IF_ERROR(create_texpr_literal_node<TYPE_IPV4>(data, &node)); Line | Count | Source | 652 | 0 | do { \ | 653 | 0 | Status _status_ = (stmt); \ | 654 | 0 | if (UNLIKELY(!_status_.ok())) { \ Line | Count | Source | 36 | 0 | #define UNLIKELY(expr) __builtin_expect(!!(expr), 0) Branch (36:24): [True: 0, False: 0]
|
| 655 | 0 | throw Exception(_status_); \ | 656 | 0 | } \ | 657 | 0 | } while (false) Branch (657:14): [Folded - Ignored]
|
|
155 | 0 | break; |
156 | 0 | } |
157 | 0 | case TYPE_IPV6: { Branch (157:5): [True: 0, False: 0]
|
158 | 0 | THROW_IF_ERROR(create_texpr_literal_node<TYPE_IPV6>(data, &node)); Line | Count | Source | 652 | 0 | do { \ | 653 | 0 | Status _status_ = (stmt); \ | 654 | 0 | if (UNLIKELY(!_status_.ok())) { \ Line | Count | Source | 36 | 0 | #define UNLIKELY(expr) __builtin_expect(!!(expr), 0) Branch (36:24): [True: 0, False: 0]
|
| 655 | 0 | throw Exception(_status_); \ | 656 | 0 | } \ | 657 | 0 | } while (false) Branch (657:14): [Folded - Ignored]
|
|
159 | 0 | break; |
160 | 0 | } |
161 | 0 | default: Branch (161:5): [True: 0, False: 0]
|
162 | 0 | throw Exception(ErrorCode::INTERNAL_ERROR, "runtime filter meet invalid type {}", |
163 | 0 | int(type)); |
164 | 0 | } |
165 | 0 | return node; |
166 | 0 | } |
167 | | // NOLINTEND(readability-function-size) |
168 | | // NOLINTEND(readability-function-cognitive-complexity) |
169 | | } // namespace doris |
170 | | |
171 | | namespace doris::vectorized { |
172 | | |
173 | 0 | bool VExpr::is_acting_on_a_slot(const VExpr& expr) { |
174 | 0 | const auto& children = expr.children(); |
175 | |
|
176 | 0 | auto is_a_slot = std::any_of(children.begin(), children.end(), |
177 | 0 | [](const auto& child) { return is_acting_on_a_slot(*child); }); |
178 | |
|
179 | 0 | return is_a_slot ? true : (expr.node_type() == TExprNodeType::SLOT_REF); Branch (179:12): [True: 0, False: 0]
|
180 | 0 | } |
181 | | |
182 | | VExpr::VExpr(const TExprNode& node) |
183 | | : _node_type(node.node_type), |
184 | | _opcode(node.__isset.opcode ? node.opcode : TExprOpcode::INVALID_OPCODE), |
185 | 146 | _type(TypeDescriptor::from_thrift(node.type)) { |
186 | 146 | if (node.__isset.fn) { Branch (186:9): [True: 33, False: 113]
|
187 | 33 | _fn = node.fn; |
188 | 33 | } |
189 | | |
190 | 146 | bool is_nullable = true; |
191 | 146 | if (node.__isset.is_nullable) { Branch (191:9): [True: 120, False: 26]
|
192 | 120 | is_nullable = node.is_nullable; |
193 | 120 | } |
194 | | // If we define null literal ,should make nullable data type to get correct field instead of undefined ptr |
195 | 146 | if (node.node_type == TExprNodeType::NULL_LITERAL) { Branch (195:9): [True: 1, False: 145]
|
196 | 1 | CHECK(is_nullable); |
197 | 1 | } |
198 | 146 | _data_type = DataTypeFactory::instance().create_data_type(_type, is_nullable); |
199 | 146 | } |
200 | | |
201 | 0 | VExpr::VExpr(const VExpr& vexpr) = default; |
202 | | |
203 | | VExpr::VExpr(TypeDescriptor type, bool is_slotref, bool is_nullable) |
204 | 0 | : _opcode(TExprOpcode::INVALID_OPCODE), _type(std::move(type)) { |
205 | 0 | if (is_slotref) { Branch (205:9): [True: 0, False: 0]
|
206 | 0 | _node_type = TExprNodeType::SLOT_REF; |
207 | 0 | } |
208 | |
|
209 | 0 | _data_type = DataTypeFactory::instance().create_data_type(_type, is_nullable); |
210 | 0 | } |
211 | | |
212 | 132 | Status VExpr::prepare(RuntimeState* state, const RowDescriptor& row_desc, VExprContext* context) { |
213 | 132 | ++context->_depth_num; |
214 | 132 | if (context->_depth_num > config::max_depth_of_expr_tree) { Branch (214:9): [True: 0, False: 132]
|
215 | 0 | return Status::Error<ErrorCode::EXCEEDED_LIMIT>( |
216 | 0 | "The depth of the expression tree is too big, make it less than {}", |
217 | 0 | config::max_depth_of_expr_tree); |
218 | 0 | } |
219 | | |
220 | 132 | for (auto& i : _children) { Branch (220:18): [True: 108, False: 132]
|
221 | 108 | RETURN_IF_ERROR(i->prepare(state, row_desc, context)); Line | Count | Source | 637 | 108 | do { \ | 638 | 108 | Status _status_ = (stmt); \ | 639 | 108 | if (UNLIKELY(!_status_.ok())) { \ Line | Count | Source | 36 | 108 | #define UNLIKELY(expr) __builtin_expect(!!(expr), 0) Branch (36:24): [True: 0, False: 108]
|
| 640 | 0 | return _status_; \ | 641 | 0 | } \ | 642 | 108 | } while (false) Branch (642:14): [Folded - Ignored]
|
|
222 | 108 | } |
223 | 132 | --context->_depth_num; |
224 | 132 | return Status::OK(); |
225 | 132 | } |
226 | | |
227 | | Status VExpr::open(RuntimeState* state, VExprContext* context, |
228 | 2 | FunctionContext::FunctionStateScope scope) { |
229 | 2 | for (auto& i : _children) { Branch (229:18): [True: 0, False: 2]
|
230 | 0 | RETURN_IF_ERROR(i->open(state, context, scope)); Line | Count | Source | 637 | 0 | do { \ | 638 | 0 | Status _status_ = (stmt); \ | 639 | 0 | if (UNLIKELY(!_status_.ok())) { \ Line | Count | Source | 36 | 0 | #define UNLIKELY(expr) __builtin_expect(!!(expr), 0) Branch (36:24): [True: 0, False: 0]
|
| 640 | 0 | return _status_; \ | 641 | 0 | } \ | 642 | 0 | } while (false) Branch (642:14): [Folded - Ignored]
|
|
231 | 0 | } |
232 | 2 | if (scope == FunctionContext::FRAGMENT_LOCAL) { Branch (232:9): [True: 2, False: 0]
|
233 | 2 | RETURN_IF_ERROR(VExpr::get_const_col(context, nullptr)); Line | Count | Source | 637 | 2 | do { \ | 638 | 2 | Status _status_ = (stmt); \ | 639 | 2 | if (UNLIKELY(!_status_.ok())) { \ Line | Count | Source | 36 | 2 | #define UNLIKELY(expr) __builtin_expect(!!(expr), 0) Branch (36:24): [True: 0, False: 2]
|
| 640 | 0 | return _status_; \ | 641 | 0 | } \ | 642 | 2 | } while (false) Branch (642:14): [Folded - Ignored]
|
|
234 | 2 | } |
235 | 2 | return Status::OK(); |
236 | 2 | } |
237 | | |
238 | 4 | void VExpr::close(VExprContext* context, FunctionContext::FunctionStateScope scope) { |
239 | 4 | for (auto& i : _children) { Branch (239:18): [True: 2, False: 4]
|
240 | 2 | i->close(context, scope); |
241 | 2 | } |
242 | 4 | } |
243 | | |
244 | | // NOLINTBEGIN(readability-function-size) |
245 | 132 | Status VExpr::create_expr(const TExprNode& expr_node, VExprSPtr& expr) { |
246 | 132 | try { |
247 | 132 | switch (expr_node.node_type) { |
248 | 0 | case TExprNodeType::BOOL_LITERAL: Branch (248:9): [True: 0, False: 132]
|
249 | 25 | case TExprNodeType::INT_LITERAL: Branch (249:9): [True: 25, False: 107]
|
250 | 25 | case TExprNodeType::LARGE_INT_LITERAL: Branch (250:9): [True: 0, False: 132]
|
251 | 26 | case TExprNodeType::IPV4_LITERAL: Branch (251:9): [True: 1, False: 131]
|
252 | 26 | case TExprNodeType::IPV6_LITERAL: Branch (252:9): [True: 0, False: 132]
|
253 | 27 | case TExprNodeType::FLOAT_LITERAL: Branch (253:9): [True: 1, False: 131]
|
254 | 28 | case TExprNodeType::DECIMAL_LITERAL: Branch (254:9): [True: 1, False: 131]
|
255 | 31 | case TExprNodeType::DATE_LITERAL: Branch (255:9): [True: 3, False: 129]
|
256 | 37 | case TExprNodeType::STRING_LITERAL: Branch (256:9): [True: 6, False: 126]
|
257 | 37 | case TExprNodeType::JSON_LITERAL: Branch (257:9): [True: 0, False: 132]
|
258 | 38 | case TExprNodeType::NULL_LITERAL: { Branch (258:9): [True: 1, False: 131]
|
259 | 38 | expr = VLiteral::create_shared(expr_node); |
260 | 38 | break; |
261 | 37 | } |
262 | 0 | case TExprNodeType::ARRAY_LITERAL: { Branch (262:9): [True: 0, False: 132]
|
263 | 0 | expr = VArrayLiteral::create_shared(expr_node); |
264 | 0 | break; |
265 | 37 | } |
266 | 0 | case TExprNodeType::MAP_LITERAL: { Branch (266:9): [True: 0, False: 132]
|
267 | 0 | expr = VMapLiteral::create_shared(expr_node); |
268 | 0 | break; |
269 | 37 | } |
270 | 0 | case TExprNodeType::STRUCT_LITERAL: { Branch (270:9): [True: 0, False: 132]
|
271 | 0 | expr = VStructLiteral::create_shared(expr_node); |
272 | 0 | break; |
273 | 37 | } |
274 | 39 | case TExprNodeType::SLOT_REF: { Branch (274:9): [True: 39, False: 93]
|
275 | 39 | expr = VSlotRef::create_shared(expr_node); |
276 | 39 | break; |
277 | 37 | } |
278 | 0 | case TExprNodeType::COLUMN_REF: { Branch (278:9): [True: 0, False: 132]
|
279 | 0 | expr = VColumnRef::create_shared(expr_node); |
280 | 0 | break; |
281 | 37 | } |
282 | 17 | case TExprNodeType::COMPOUND_PRED: { Branch (282:9): [True: 17, False: 115]
|
283 | 17 | expr = VCompoundPred::create_shared(expr_node); |
284 | 17 | break; |
285 | 37 | } |
286 | 0 | case TExprNodeType::LAMBDA_FUNCTION_EXPR: { Branch (286:9): [True: 0, False: 132]
|
287 | 0 | expr = VLambdaFunctionExpr::create_shared(expr_node); |
288 | 0 | break; |
289 | 37 | } |
290 | 0 | case TExprNodeType::LAMBDA_FUNCTION_CALL_EXPR: { Branch (290:9): [True: 0, False: 132]
|
291 | 0 | expr = VLambdaFunctionCallExpr::create_shared(expr_node); |
292 | 0 | break; |
293 | 37 | } |
294 | 1 | case TExprNodeType::ARITHMETIC_EXPR: Branch (294:9): [True: 1, False: 131]
|
295 | 23 | case TExprNodeType::BINARY_PRED: Branch (295:9): [True: 22, False: 110]
|
296 | 23 | case TExprNodeType::NULL_AWARE_BINARY_PRED: Branch (296:9): [True: 0, False: 132]
|
297 | 31 | case TExprNodeType::FUNCTION_CALL: Branch (297:9): [True: 8, False: 124]
|
298 | 31 | case TExprNodeType::COMPUTE_FUNCTION_CALL: { Branch (298:9): [True: 0, False: 132]
|
299 | 31 | expr = VectorizedFnCall::create_shared(expr_node); |
300 | 31 | break; |
301 | 31 | } |
302 | 0 | case TExprNodeType::MATCH_PRED: { Branch (302:9): [True: 0, False: 132]
|
303 | 0 | expr = VMatchPredicate::create_shared(expr_node); |
304 | 0 | break; |
305 | 31 | } |
306 | 2 | case TExprNodeType::CAST_EXPR: { Branch (306:9): [True: 2, False: 130]
|
307 | 2 | expr = VCastExpr::create_shared(expr_node); |
308 | 2 | break; |
309 | 31 | } |
310 | 5 | case TExprNodeType::IN_PRED: { Branch (310:9): [True: 5, False: 127]
|
311 | 5 | expr = VInPredicate::create_shared(expr_node); |
312 | 5 | break; |
313 | 31 | } |
314 | 0 | case TExprNodeType::CASE_EXPR: { Branch (314:9): [True: 0, False: 132]
|
315 | 0 | if (!expr_node.__isset.case_expr) { Branch (315:17): [True: 0, False: 0]
|
316 | 0 | return Status::InternalError("Case expression not set in thrift node"); |
317 | 0 | } |
318 | 0 | expr = VCaseExpr::create_shared(expr_node); |
319 | 0 | break; |
320 | 0 | } |
321 | 0 | case TExprNodeType::INFO_FUNC: { Branch (321:9): [True: 0, False: 132]
|
322 | 0 | expr = VInfoFunc::create_shared(expr_node); |
323 | 0 | break; |
324 | 0 | } |
325 | 0 | case TExprNodeType::TUPLE_IS_NULL_PRED: { Branch (325:9): [True: 0, False: 132]
|
326 | 0 | expr = VTupleIsNullPredicate::create_shared(expr_node); |
327 | 0 | break; |
328 | 0 | } |
329 | 0 | default: Branch (329:9): [True: 0, False: 132]
|
330 | 0 | return Status::InternalError("Unknown expr node type: {}", expr_node.node_type); |
331 | 132 | } |
332 | 132 | } catch (const Exception& e) { |
333 | 0 | if (e.code() == ErrorCode::INTERNAL_ERROR) { Branch (333:13): [True: 0, False: 0]
|
334 | 0 | return Status::InternalError("Create Expr failed because {}\nTExprNode={}", e.what(), |
335 | 0 | apache::thrift::ThriftDebugString(expr_node)); |
336 | 0 | } |
337 | 0 | return Status::Error<false>(e.code(), "Create Expr failed because {}", e.what()); |
338 | 0 | LOG(WARNING) << "create expr failed, TExprNode={}, reason={}" |
339 | 0 | << apache::thrift::ThriftDebugString(expr_node) << e.what(); |
340 | 0 | } |
341 | 132 | if (!expr->data_type()) { Branch (341:9): [True: 0, False: 132]
|
342 | 0 | return Status::InvalidArgument("Unknown expr type: {}", expr_node.node_type); |
343 | 0 | } |
344 | 132 | return Status::OK(); |
345 | 132 | } |
346 | | // NOLINTEND(readability-function-size) |
347 | | |
348 | | Status VExpr::create_tree_from_thrift(const std::vector<TExprNode>& nodes, int* node_idx, |
349 | 23 | VExprSPtr& root_expr, VExprContextSPtr& ctx) { |
350 | | // propagate error case |
351 | 23 | if (*node_idx >= nodes.size()) { Branch (351:9): [True: 0, False: 23]
|
352 | 0 | return Status::InternalError("Failed to reconstruct expression tree from thrift."); |
353 | 0 | } |
354 | | |
355 | | // create root expr |
356 | 23 | int root_children = nodes[*node_idx].num_children; |
357 | 23 | VExprSPtr root; |
358 | 23 | RETURN_IF_ERROR(create_expr(nodes[*node_idx], root)); Line | Count | Source | 637 | 23 | do { \ | 638 | 23 | Status _status_ = (stmt); \ | 639 | 23 | if (UNLIKELY(!_status_.ok())) { \ Line | Count | Source | 36 | 23 | #define UNLIKELY(expr) __builtin_expect(!!(expr), 0) Branch (36:24): [True: 0, False: 23]
|
| 640 | 0 | return _status_; \ | 641 | 0 | } \ | 642 | 23 | } while (false) Branch (642:14): [Folded - Ignored]
|
|
359 | 23 | DCHECK(root != nullptr); |
360 | 23 | root_expr = root; |
361 | 23 | ctx = std::make_shared<VExprContext>(root); |
362 | | // short path for leaf node |
363 | 23 | if (root_children <= 0) { Branch (363:9): [True: 8, False: 15]
|
364 | 8 | return Status::OK(); |
365 | 8 | } |
366 | | |
367 | | // non-recursive traversal |
368 | 15 | using VExprSPtrCountPair = std::pair<VExprSPtr, int>; |
369 | 15 | std::stack<std::shared_ptr<VExprSPtrCountPair>> s; |
370 | 15 | s.emplace(std::make_shared<VExprSPtrCountPair>(root, root_children)); |
371 | 120 | while (!s.empty()) { Branch (371:12): [True: 105, False: 15]
|
372 | | // copy the shared ptr resource to avoid dangling reference |
373 | 105 | auto parent = s.top(); |
374 | | // Decrement or pop |
375 | 105 | if (parent->second > 1) { Branch (375:13): [True: 53, False: 52]
|
376 | 53 | parent->second -= 1; |
377 | 53 | } else { |
378 | 52 | s.pop(); |
379 | 52 | } |
380 | | |
381 | 105 | DCHECK(parent->first != nullptr); |
382 | 105 | if (++*node_idx >= nodes.size()) { Branch (382:13): [True: 0, False: 105]
|
383 | 0 | return Status::InternalError("Failed to reconstruct expression tree from thrift."); |
384 | 0 | } |
385 | | |
386 | 105 | VExprSPtr expr; |
387 | 105 | RETURN_IF_ERROR(create_expr(nodes[*node_idx], expr)); Line | Count | Source | 637 | 105 | do { \ | 638 | 105 | Status _status_ = (stmt); \ | 639 | 105 | if (UNLIKELY(!_status_.ok())) { \ Line | Count | Source | 36 | 105 | #define UNLIKELY(expr) __builtin_expect(!!(expr), 0) Branch (36:24): [True: 0, False: 105]
|
| 640 | 0 | return _status_; \ | 641 | 0 | } \ | 642 | 105 | } while (false) Branch (642:14): [Folded - Ignored]
|
|
388 | 105 | DCHECK(expr != nullptr); |
389 | 105 | parent->first->add_child(expr); |
390 | | // push to stack if has children |
391 | 105 | int num_children = nodes[*node_idx].num_children; |
392 | 105 | if (num_children > 0) { Branch (392:13): [True: 37, False: 68]
|
393 | 37 | s.emplace(std::make_shared<VExprSPtrCountPair>(expr, num_children)); |
394 | 37 | } |
395 | 105 | } |
396 | 15 | return Status::OK(); |
397 | 15 | } |
398 | | |
399 | 23 | Status VExpr::create_expr_tree(const TExpr& texpr, VExprContextSPtr& ctx) { |
400 | 23 | if (texpr.nodes.empty()) { Branch (400:9): [True: 0, False: 23]
|
401 | 0 | ctx = nullptr; |
402 | 0 | return Status::OK(); |
403 | 0 | } |
404 | 23 | int node_idx = 0; |
405 | 23 | VExprSPtr e; |
406 | 23 | Status status = create_tree_from_thrift(texpr.nodes, &node_idx, e, ctx); |
407 | 23 | if (status.ok() && node_idx + 1 != texpr.nodes.size()) { Branch (407:9): [True: 23, False: 0]
Branch (407:24): [True: 0, False: 23]
|
408 | 0 | status = Status::InternalError( |
409 | 0 | "Expression tree only partially reconstructed. Not all thrift nodes were " |
410 | 0 | "used."); |
411 | 0 | } |
412 | 23 | if (!status.ok()) { Branch (412:9): [True: 0, False: 23]
|
413 | 0 | LOG(ERROR) << "Could not construct expr tree.\n" |
414 | 0 | << status << "\n" |
415 | 0 | << apache::thrift::ThriftDebugString(texpr); |
416 | 0 | } |
417 | 23 | return status; |
418 | 23 | } |
419 | | |
420 | 0 | Status VExpr::create_expr_trees(const std::vector<TExpr>& texprs, VExprContextSPtrs& ctxs) { |
421 | 0 | ctxs.clear(); |
422 | 0 | for (const auto& texpr : texprs) { Branch (422:28): [True: 0, False: 0]
|
423 | 0 | VExprContextSPtr ctx; |
424 | 0 | RETURN_IF_ERROR(create_expr_tree(texpr, ctx)); Line | Count | Source | 637 | 0 | do { \ | 638 | 0 | Status _status_ = (stmt); \ | 639 | 0 | if (UNLIKELY(!_status_.ok())) { \ Line | Count | Source | 36 | 0 | #define UNLIKELY(expr) __builtin_expect(!!(expr), 0) Branch (36:24): [True: 0, False: 0]
|
| 640 | 0 | return _status_; \ | 641 | 0 | } \ | 642 | 0 | } while (false) Branch (642:14): [Folded - Ignored]
|
|
425 | 0 | ctxs.push_back(ctx); |
426 | 0 | } |
427 | 0 | return Status::OK(); |
428 | 0 | } |
429 | | |
430 | | Status VExpr::check_expr_output_type(const VExprContextSPtrs& ctxs, |
431 | 0 | const RowDescriptor& output_row_desc) { |
432 | 0 | if (ctxs.empty()) { Branch (432:9): [True: 0, False: 0]
|
433 | 0 | return Status::OK(); |
434 | 0 | } |
435 | 0 | auto name_and_types = VectorizedUtils::create_name_and_data_types(output_row_desc); |
436 | 0 | if (ctxs.size() != name_and_types.size()) { Branch (436:9): [True: 0, False: 0]
|
437 | 0 | return Status::InternalError( |
438 | 0 | "output type size not match expr size {} , expected output size {} ", ctxs.size(), |
439 | 0 | name_and_types.size()); |
440 | 0 | } |
441 | 0 | auto check_type_can_be_converted = [](DataTypePtr& from, DataTypePtr& to) -> bool { |
442 | 0 | if (to->equals(*from)) { Branch (442:13): [True: 0, False: 0]
|
443 | 0 | return true; |
444 | 0 | } |
445 | 0 | if (to->is_nullable() && !from->is_nullable()) { Branch (445:13): [True: 0, False: 0]
Branch (445:34): [True: 0, False: 0]
|
446 | 0 | return remove_nullable(to)->equals(*from); |
447 | 0 | } |
448 | 0 | return false; |
449 | 0 | }; |
450 | 0 | for (int i = 0; i < ctxs.size(); i++) { Branch (450:21): [True: 0, False: 0]
|
451 | 0 | auto real_expr_type = ctxs[i]->root()->data_type(); |
452 | 0 | auto&& [name, expected_type] = name_and_types[i]; |
453 | 0 | if (!check_type_can_be_converted(real_expr_type, expected_type)) { Branch (453:13): [True: 0, False: 0]
|
454 | 0 | return Status::InternalError( |
455 | 0 | "output type not match expr type , col name {} , expected type {} , real type " |
456 | 0 | "{}", |
457 | 0 | name, expected_type->get_name(), real_expr_type->get_name()); |
458 | 0 | } |
459 | 0 | } |
460 | 0 | return Status::OK(); |
461 | 0 | } |
462 | | |
463 | | Status VExpr::prepare(const VExprContextSPtrs& ctxs, RuntimeState* state, |
464 | 4 | const RowDescriptor& row_desc) { |
465 | 4 | for (auto ctx : ctxs) { Branch (465:19): [True: 0, False: 4]
|
466 | 0 | RETURN_IF_ERROR(ctx->prepare(state, row_desc)); Line | Count | Source | 637 | 0 | do { \ | 638 | 0 | Status _status_ = (stmt); \ | 639 | 0 | if (UNLIKELY(!_status_.ok())) { \ Line | Count | Source | 36 | 0 | #define UNLIKELY(expr) __builtin_expect(!!(expr), 0) Branch (36:24): [True: 0, False: 0]
|
| 640 | 0 | return _status_; \ | 641 | 0 | } \ | 642 | 0 | } while (false) Branch (642:14): [Folded - Ignored]
|
|
467 | 0 | } |
468 | 4 | return Status::OK(); |
469 | 4 | } |
470 | | |
471 | 4 | Status VExpr::open(const VExprContextSPtrs& ctxs, RuntimeState* state) { |
472 | 4 | for (const auto& ctx : ctxs) { Branch (472:26): [True: 0, False: 4]
|
473 | 0 | RETURN_IF_ERROR(ctx->open(state)); Line | Count | Source | 637 | 0 | do { \ | 638 | 0 | Status _status_ = (stmt); \ | 639 | 0 | if (UNLIKELY(!_status_.ok())) { \ Line | Count | Source | 36 | 0 | #define UNLIKELY(expr) __builtin_expect(!!(expr), 0) Branch (36:24): [True: 0, False: 0]
|
| 640 | 0 | return _status_; \ | 641 | 0 | } \ | 642 | 0 | } while (false) Branch (642:14): [Folded - Ignored]
|
|
474 | 0 | } |
475 | 4 | return Status::OK(); |
476 | 4 | } |
477 | | |
478 | | Status VExpr::clone_if_not_exists(const VExprContextSPtrs& ctxs, RuntimeState* state, |
479 | 0 | VExprContextSPtrs& new_ctxs) { |
480 | 0 | if (!new_ctxs.empty()) { Branch (480:9): [True: 0, False: 0]
|
481 | | // 'ctxs' was already cloned into '*new_ctxs', nothing to do. |
482 | 0 | DCHECK_EQ(new_ctxs.size(), ctxs.size()); |
483 | 0 | for (auto& new_ctx : new_ctxs) { Branch (483:28): [True: 0, False: 0]
|
484 | 0 | DCHECK(new_ctx->_is_clone); |
485 | 0 | } |
486 | 0 | return Status::OK(); |
487 | 0 | } |
488 | 0 | new_ctxs.resize(ctxs.size()); |
489 | 0 | for (int i = 0; i < ctxs.size(); ++i) { Branch (489:21): [True: 0, False: 0]
|
490 | 0 | RETURN_IF_ERROR(ctxs[i]->clone(state, new_ctxs[i])); Line | Count | Source | 637 | 0 | do { \ | 638 | 0 | Status _status_ = (stmt); \ | 639 | 0 | if (UNLIKELY(!_status_.ok())) { \ Line | Count | Source | 36 | 0 | #define UNLIKELY(expr) __builtin_expect(!!(expr), 0) Branch (36:24): [True: 0, False: 0]
|
| 640 | 0 | return _status_; \ | 641 | 0 | } \ | 642 | 0 | } while (false) Branch (642:14): [Folded - Ignored]
|
|
491 | 0 | } |
492 | 0 | return Status::OK(); |
493 | 0 | } |
494 | | |
495 | 0 | std::string VExpr::debug_string() const { |
496 | | // TODO: implement partial debug string for member vars |
497 | 0 | std::stringstream out; |
498 | 0 | out << " type=" << _type.debug_string(); |
499 | |
|
500 | 0 | if (!_children.empty()) { Branch (500:9): [True: 0, False: 0]
|
501 | 0 | out << " children=" << debug_string(_children); |
502 | 0 | } |
503 | |
|
504 | 0 | return out.str(); |
505 | 0 | } |
506 | | |
507 | 0 | std::string VExpr::debug_string(const VExprSPtrs& exprs) { |
508 | 0 | std::stringstream out; |
509 | 0 | out << "["; |
510 | |
|
511 | 0 | for (int i = 0; i < exprs.size(); ++i) { Branch (511:21): [True: 0, False: 0]
|
512 | 0 | out << (i == 0 ? "" : " ") << exprs[i]->debug_string(); Branch (512:17): [True: 0, False: 0]
|
513 | 0 | } |
514 | |
|
515 | 0 | out << "]"; |
516 | 0 | return out.str(); |
517 | 0 | } |
518 | | |
519 | 0 | std::string VExpr::debug_string(const VExprContextSPtrs& ctxs) { |
520 | 0 | VExprSPtrs exprs; |
521 | 0 | for (const auto& ctx : ctxs) { Branch (521:26): [True: 0, False: 0]
|
522 | 0 | exprs.push_back(ctx->root()); |
523 | 0 | } |
524 | 0 | return debug_string(exprs); |
525 | 0 | } |
526 | | |
527 | 2 | bool VExpr::is_constant() const { |
528 | 2 | return std::all_of(_children.begin(), _children.end(), |
529 | 2 | [](const VExprSPtr& expr) { return expr->is_constant(); }); |
530 | 2 | } |
531 | | |
532 | | Status VExpr::get_const_col(VExprContext* context, |
533 | 6 | std::shared_ptr<ColumnPtrWrapper>* column_wrapper) { |
534 | 6 | if (!is_constant()) { Branch (534:9): [True: 6, False: 0]
|
535 | 6 | return Status::OK(); |
536 | 6 | } |
537 | | |
538 | 0 | if (_constant_col != nullptr) { Branch (538:9): [True: 0, False: 0]
|
539 | 0 | DCHECK(column_wrapper != nullptr); |
540 | 0 | *column_wrapper = _constant_col; |
541 | 0 | return Status::OK(); |
542 | 0 | } |
543 | | |
544 | 0 | int result = -1; |
545 | 0 | Block block; |
546 | | // If block is empty, some functions will produce no result. So we insert a column with |
547 | | // single value here. |
548 | 0 | block.insert({ColumnUInt8::create(1), std::make_shared<DataTypeUInt8>(), ""}); |
549 | |
|
550 | 0 | _getting_const_col = true; |
551 | 0 | RETURN_IF_ERROR(execute(context, &block, &result)); Line | Count | Source | 637 | 0 | do { \ | 638 | 0 | Status _status_ = (stmt); \ | 639 | 0 | if (UNLIKELY(!_status_.ok())) { \ Line | Count | Source | 36 | 0 | #define UNLIKELY(expr) __builtin_expect(!!(expr), 0) Branch (36:24): [True: 0, False: 0]
|
| 640 | 0 | return _status_; \ | 641 | 0 | } \ | 642 | 0 | } while (false) Branch (642:14): [Folded - Ignored]
|
|
552 | 0 | _getting_const_col = false; |
553 | |
|
554 | 0 | DCHECK(result != -1); |
555 | 0 | const auto& column = block.get_by_position(result).column; |
556 | 0 | _constant_col = std::make_shared<ColumnPtrWrapper>(column); |
557 | 0 | if (column_wrapper != nullptr) { Branch (557:9): [True: 0, False: 0]
|
558 | 0 | *column_wrapper = _constant_col; |
559 | 0 | } |
560 | |
|
561 | 0 | return Status::OK(); |
562 | 0 | } |
563 | | |
564 | 55 | void VExpr::register_function_context(RuntimeState* state, VExprContext* context) { |
565 | 55 | std::vector<TypeDescriptor> arg_types; |
566 | 108 | for (auto& i : _children) { Branch (566:18): [True: 108, False: 55]
|
567 | 108 | arg_types.push_back(i->type()); |
568 | 108 | } |
569 | | |
570 | 55 | _fn_context_index = context->register_function_context(state, _type, arg_types); |
571 | 55 | } |
572 | | |
573 | | Status VExpr::init_function_context(RuntimeState* state, VExprContext* context, |
574 | | FunctionContext::FunctionStateScope scope, |
575 | 2 | const FunctionBasePtr& function) const { |
576 | 2 | FunctionContext* fn_ctx = context->fn_context(_fn_context_index); |
577 | 2 | if (scope == FunctionContext::FRAGMENT_LOCAL) { Branch (577:9): [True: 2, False: 0]
|
578 | 2 | std::vector<std::shared_ptr<ColumnPtrWrapper>> constant_cols; |
579 | 2 | for (auto c : _children) { Branch (579:21): [True: 2, False: 2]
|
580 | 2 | std::shared_ptr<ColumnPtrWrapper> const_col; |
581 | 2 | RETURN_IF_ERROR(c->get_const_col(context, &const_col)); Line | Count | Source | 637 | 2 | do { \ | 638 | 2 | Status _status_ = (stmt); \ | 639 | 2 | if (UNLIKELY(!_status_.ok())) { \ Line | Count | Source | 36 | 2 | #define UNLIKELY(expr) __builtin_expect(!!(expr), 0) Branch (36:24): [True: 0, False: 2]
|
| 640 | 0 | return _status_; \ | 641 | 0 | } \ | 642 | 2 | } while (false) Branch (642:14): [Folded - Ignored]
|
|
582 | 2 | constant_cols.push_back(const_col); |
583 | 2 | } |
584 | 2 | fn_ctx->set_constant_cols(constant_cols); |
585 | 2 | } else { |
586 | 0 | if (function->is_udf_function()) { Branch (586:13): [True: 0, False: 0]
|
587 | 0 | auto* timer = ADD_TIMER(state->get_task()->get_task_profile(), Line | Count | Source | 60 | 0 | #define ADD_TIMER(profile, name) (profile)->add_counter(name, TUnit::TIME_NS) |
|
588 | 0 | "UDF[" + function->get_name() + "]"); |
589 | 0 | fn_ctx->set_udf_execute_timer(timer); |
590 | 0 | } |
591 | 0 | } |
592 | | |
593 | 2 | if (scope == FunctionContext::FRAGMENT_LOCAL) { Branch (593:9): [True: 2, False: 0]
|
594 | 2 | RETURN_IF_ERROR(function->open(fn_ctx, FunctionContext::FRAGMENT_LOCAL)); Line | Count | Source | 637 | 2 | do { \ | 638 | 2 | Status _status_ = (stmt); \ | 639 | 2 | if (UNLIKELY(!_status_.ok())) { \ Line | Count | Source | 36 | 2 | #define UNLIKELY(expr) __builtin_expect(!!(expr), 0) Branch (36:24): [True: 0, False: 2]
|
| 640 | 0 | return _status_; \ | 641 | 0 | } \ | 642 | 2 | } while (false) Branch (642:14): [Folded - Ignored]
|
|
595 | 2 | } |
596 | 2 | RETURN_IF_ERROR(function->open(fn_ctx, FunctionContext::THREAD_LOCAL)); Line | Count | Source | 637 | 2 | do { \ | 638 | 2 | Status _status_ = (stmt); \ | 639 | 2 | if (UNLIKELY(!_status_.ok())) { \ Line | Count | Source | 36 | 2 | #define UNLIKELY(expr) __builtin_expect(!!(expr), 0) Branch (36:24): [True: 0, False: 2]
|
| 640 | 0 | return _status_; \ | 641 | 0 | } \ | 642 | 2 | } while (false) Branch (642:14): [Folded - Ignored]
|
|
597 | 2 | return Status::OK(); |
598 | 2 | } |
599 | | |
600 | | void VExpr::close_function_context(VExprContext* context, FunctionContext::FunctionStateScope scope, |
601 | 2 | const FunctionBasePtr& function) const { |
602 | 2 | if (_fn_context_index != -1) { Branch (602:9): [True: 2, False: 0]
|
603 | 2 | FunctionContext* fn_ctx = context->fn_context(_fn_context_index); |
604 | | // `close_function_context` is called in VExprContext's destructor so do not throw exceptions here. |
605 | 2 | static_cast<void>(function->close(fn_ctx, FunctionContext::THREAD_LOCAL)); |
606 | 2 | if (scope == FunctionContext::FRAGMENT_LOCAL) { Branch (606:13): [True: 2, False: 0]
|
607 | 2 | static_cast<void>(function->close(fn_ctx, FunctionContext::FRAGMENT_LOCAL)); |
608 | 2 | } |
609 | 2 | } |
610 | 2 | } |
611 | | |
612 | 0 | Status VExpr::check_constant(const Block& block, ColumnNumbers arguments) const { |
613 | 0 | if (is_constant() && !VectorizedUtils::all_arguments_are_constant(block, arguments)) { Branch (613:9): [True: 0, False: 0]
Branch (613:26): [True: 0, False: 0]
|
614 | 0 | return Status::InternalError("const check failed, expr={}", debug_string()); |
615 | 0 | } |
616 | 0 | return Status::OK(); |
617 | 0 | } |
618 | | |
619 | | Status VExpr::get_result_from_const(vectorized::Block* block, const std::string& expr_name, |
620 | 0 | int* result_column_id) { |
621 | 0 | *result_column_id = block->columns(); |
622 | 0 | auto column = ColumnConst::create(_constant_col->column_ptr, block->rows()); |
623 | 0 | block->insert({std::move(column), _data_type, expr_name}); |
624 | 0 | return Status::OK(); |
625 | 0 | } |
626 | | |
627 | | Status VExpr::_evaluate_inverted_index(VExprContext* context, const FunctionBasePtr& function, |
628 | 1 | uint32_t segment_num_rows) { |
629 | | // Pre-allocate vectors based on an estimated or known size |
630 | 1 | std::vector<segment_v2::InvertedIndexIterator*> iterators; |
631 | 1 | std::vector<vectorized::IndexFieldNameAndTypePair> data_type_with_names; |
632 | 1 | std::vector<int> column_ids; |
633 | 1 | vectorized::ColumnsWithTypeAndName arguments; |
634 | 1 | VExprSPtrs children_exprs; |
635 | | |
636 | | // Reserve space to avoid multiple reallocations |
637 | 1 | const size_t estimated_size = children().size(); |
638 | 1 | iterators.reserve(estimated_size); |
639 | 1 | data_type_with_names.reserve(estimated_size); |
640 | 1 | column_ids.reserve(estimated_size); |
641 | 1 | children_exprs.reserve(estimated_size); |
642 | | |
643 | 1 | auto index_context = context->get_inverted_index_context(); |
644 | | |
645 | | // if child is cast expr, we need to ensure target data type is the same with storage data type. |
646 | | // or they are all string type |
647 | | // and if data type is array, we need to get the nested data type to ensure that. |
648 | 2 | for (const auto& child : children()) { Branch (648:28): [True: 2, False: 0]
|
649 | 2 | if (child->node_type() == TExprNodeType::CAST_EXPR) { Branch (649:13): [True: 1, False: 1]
|
650 | 1 | auto* cast_expr = assert_cast<VCastExpr*>(child.get()); |
651 | 1 | DCHECK_EQ(cast_expr->children().size(), 1); |
652 | 1 | if (cast_expr->get_child(0)->is_slot_ref()) { Branch (652:17): [True: 0, False: 1]
|
653 | 0 | auto* column_slot_ref = assert_cast<VSlotRef*>(cast_expr->get_child(0).get()); |
654 | 0 | auto column_id = column_slot_ref->column_id(); |
655 | 0 | const auto* storage_name_type = |
656 | 0 | context->get_inverted_index_context() |
657 | 0 | ->get_storage_name_and_type_by_column_id(column_id); |
658 | 0 | auto storage_type = remove_nullable(storage_name_type->second); |
659 | 0 | auto target_type = remove_nullable(cast_expr->get_target_type()); |
660 | 0 | auto origin_primitive_type = storage_type->get_type_as_type_descriptor().type; |
661 | 0 | auto target_primitive_type = target_type->get_type_as_type_descriptor().type; |
662 | 0 | if (is_complex_type(storage_type)) { Branch (662:21): [True: 0, False: 0]
|
663 | 0 | if (is_array(storage_type) && is_array(target_type)) { Branch (663:25): [True: 0, False: 0]
Branch (663:51): [True: 0, False: 0]
|
664 | 0 | auto nested_storage_type = |
665 | 0 | (assert_cast<const DataTypeArray*>(storage_type.get())) |
666 | 0 | ->get_nested_type(); |
667 | 0 | origin_primitive_type = |
668 | 0 | nested_storage_type->get_type_as_type_descriptor().type; |
669 | 0 | auto nested_target_type = |
670 | 0 | (assert_cast<const DataTypeArray*>(target_type.get())) |
671 | 0 | ->get_nested_type(); |
672 | 0 | target_primitive_type = |
673 | 0 | nested_target_type->get_type_as_type_descriptor().type; |
674 | 0 | } else { |
675 | 0 | continue; |
676 | 0 | } |
677 | 0 | } |
678 | 0 | if (origin_primitive_type != TYPE_VARIANT && Branch (678:21): [True: 0, False: 0]
|
679 | 0 | (storage_type->equals(*target_type) || Branch (679:22): [True: 0, False: 0]
|
680 | 0 | (is_string_type(target_primitive_type) && Branch (680:23): [True: 0, False: 0]
|
681 | 0 | is_string_type(origin_primitive_type)))) { Branch (681:23): [True: 0, False: 0]
|
682 | 0 | children_exprs.emplace_back(expr_without_cast(child)); |
683 | 0 | } |
684 | 1 | } else { |
685 | 1 | return Status::OK(); // for example: cast("abc") as ipv4 case |
686 | 1 | } |
687 | 1 | } else { |
688 | 1 | children_exprs.emplace_back(child); |
689 | 1 | } |
690 | 2 | } |
691 | | |
692 | 0 | if (children_exprs.empty()) { Branch (692:9): [True: 0, False: 0]
|
693 | 0 | return Status::OK(); // Early exit if no children to process |
694 | 0 | } |
695 | | |
696 | 0 | for (const auto& child : children_exprs) { Branch (696:28): [True: 0, False: 0]
|
697 | 0 | if (child->is_slot_ref()) { Branch (697:13): [True: 0, False: 0]
|
698 | 0 | auto* column_slot_ref = assert_cast<VSlotRef*>(child.get()); |
699 | 0 | auto column_id = column_slot_ref->column_id(); |
700 | 0 | auto* iter = |
701 | 0 | context->get_inverted_index_context()->get_inverted_index_iterator_by_column_id( |
702 | 0 | column_id); |
703 | | //column does not have inverted index |
704 | 0 | if (iter == nullptr) { Branch (704:17): [True: 0, False: 0]
|
705 | 0 | continue; |
706 | 0 | } |
707 | 0 | const auto* storage_name_type = |
708 | 0 | context->get_inverted_index_context()->get_storage_name_and_type_by_column_id( |
709 | 0 | column_id); |
710 | 0 | if (storage_name_type == nullptr) { Branch (710:17): [True: 0, False: 0]
|
711 | 0 | auto err_msg = fmt::format( |
712 | 0 | "storage_name_type cannot be found for column {} while in {} " |
713 | 0 | "evaluate_inverted_index", |
714 | 0 | column_id, expr_name()); |
715 | 0 | LOG(ERROR) << err_msg; |
716 | 0 | return Status::InternalError(err_msg); |
717 | 0 | } |
718 | 0 | iterators.emplace_back(iter); |
719 | 0 | data_type_with_names.emplace_back(*storage_name_type); |
720 | 0 | column_ids.emplace_back(column_id); |
721 | 0 | } else if (child->is_literal()) { Branch (721:20): [True: 0, False: 0]
|
722 | 0 | auto* column_literal = assert_cast<VLiteral*>(child.get()); |
723 | 0 | arguments.emplace_back(column_literal->get_column_ptr(), |
724 | 0 | column_literal->get_data_type(), column_literal->expr_name()); |
725 | 0 | } else { |
726 | 0 | return Status::OK(); // others cases |
727 | 0 | } |
728 | 0 | } |
729 | | |
730 | | // is null or is not null has no arguments |
731 | 0 | if (iterators.empty() || (arguments.empty() && !(function->get_name() == "is_not_null_pred" || Branch (731:9): [True: 0, False: 0]
Branch (731:9): [True: 0, False: 0]
Branch (731:31): [True: 0, False: 0]
Branch (731:54): [True: 0, False: 0]
|
732 | 0 | function->get_name() == "is_null_pred"))) { Branch (732:54): [True: 0, False: 0]
|
733 | 0 | return Status::OK(); // Nothing to evaluate or no literals to compare against |
734 | 0 | } |
735 | | |
736 | 0 | auto result_bitmap = segment_v2::InvertedIndexResultBitmap(); |
737 | 0 | auto res = function->evaluate_inverted_index(arguments, data_type_with_names, iterators, |
738 | 0 | segment_num_rows, result_bitmap); |
739 | 0 | if (!res.ok()) { Branch (739:9): [True: 0, False: 0]
|
740 | 0 | return res; |
741 | 0 | } |
742 | 0 | if (!result_bitmap.is_empty()) { Branch (742:9): [True: 0, False: 0]
|
743 | 0 | index_context->set_inverted_index_result_for_expr(this, result_bitmap); |
744 | 0 | for (int column_id : column_ids) { Branch (744:28): [True: 0, False: 0]
|
745 | 0 | index_context->set_true_for_inverted_index_status(this, column_id); |
746 | 0 | } |
747 | 0 | } |
748 | 0 | return Status::OK(); |
749 | 0 | } |
750 | | |
751 | | bool VExpr::fast_execute(doris::vectorized::VExprContext* context, doris::vectorized::Block* block, |
752 | 0 | int* result_column_id) { |
753 | 0 | if (context->get_inverted_index_context() && Branch (753:9): [True: 0, False: 0]
Branch (753:9): [True: 0, False: 0]
|
754 | 0 | context->get_inverted_index_context()->get_inverted_index_result_column().contains(this)) { Branch (754:9): [True: 0, False: 0]
|
755 | 0 | size_t num_columns_without_result = block->columns(); |
756 | | // prepare a column to save result |
757 | 0 | auto result_column = |
758 | 0 | context->get_inverted_index_context()->get_inverted_index_result_column()[this]; |
759 | 0 | if (_data_type->is_nullable()) { Branch (759:13): [True: 0, False: 0]
|
760 | 0 | block->insert( |
761 | 0 | {ColumnNullable::create(result_column, ColumnUInt8::create(block->rows(), 0)), |
762 | 0 | _data_type, expr_name()}); |
763 | 0 | } else { |
764 | 0 | block->insert({result_column, _data_type, expr_name()}); |
765 | 0 | } |
766 | 0 | *result_column_id = num_columns_without_result; |
767 | 0 | return true; |
768 | 0 | } |
769 | 0 | return false; |
770 | 0 | } |
771 | | |
772 | 0 | bool VExpr::equals(const VExpr& other) { |
773 | 0 | return false; |
774 | 0 | } |
775 | | |
776 | | } // namespace doris::vectorized |