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 "exprs/function/match.h" |
19 | | |
20 | | #include <hs/hs.h> |
21 | | |
22 | | #include "core/field.h" |
23 | | #include "runtime/query_context.h" |
24 | | #include "runtime/runtime_state.h" |
25 | | #include "storage/index/index_reader_helper.h" |
26 | | #include "storage/index/inverted/analyzer/analyzer.h" |
27 | | #include "util/debug_points.h" |
28 | | |
29 | | namespace doris { |
30 | | |
31 | | namespace { |
32 | | |
33 | 492 | const InvertedIndexAnalyzerCtx* get_match_analyzer_ctx(FunctionContext* context) { |
34 | 492 | if (context == nullptr) { |
35 | 0 | return nullptr; |
36 | 0 | } |
37 | 492 | auto* analyzer_ctx = reinterpret_cast<const InvertedIndexAnalyzerCtx*>( |
38 | 492 | context->get_function_state(FunctionContext::THREAD_LOCAL)); |
39 | 492 | if (analyzer_ctx == nullptr) { |
40 | 0 | analyzer_ctx = reinterpret_cast<const InvertedIndexAnalyzerCtx*>( |
41 | 0 | context->get_function_state(FunctionContext::FRAGMENT_LOCAL)); |
42 | 0 | } |
43 | 492 | return analyzer_ctx; |
44 | 492 | } |
45 | | |
46 | | } // namespace |
47 | | |
48 | | Status FunctionMatchBase::evaluate_inverted_index( |
49 | | const ColumnsWithTypeAndName& arguments, |
50 | | const std::vector<IndexFieldNameAndTypePair>& data_type_with_names, |
51 | | std::vector<segment_v2::IndexIterator*> iterators, uint32_t num_rows, |
52 | | const InvertedIndexAnalyzerCtx* analyzer_ctx, |
53 | 4.76k | segment_v2::InvertedIndexResultBitmap& bitmap_result) const { |
54 | 4.76k | DCHECK(arguments.size() == 1); |
55 | 4.76k | DCHECK(data_type_with_names.size() == 1); |
56 | 4.76k | DCHECK(iterators.size() == 1); |
57 | 4.76k | auto* iter = iterators[0]; |
58 | 4.76k | auto data_type_with_name = data_type_with_names[0]; |
59 | 4.76k | if (iter == nullptr) { |
60 | 0 | return Status::OK(); |
61 | 0 | } |
62 | 4.76k | const std::string& function_name = get_name(); |
63 | | |
64 | 4.76k | if (function_name == MATCH_PHRASE_FUNCTION || function_name == MATCH_PHRASE_PREFIX_FUNCTION || |
65 | 4.76k | function_name == MATCH_PHRASE_EDGE_FUNCTION) { |
66 | 546 | auto reader = iter->get_reader(InvertedIndexReaderType::FULLTEXT); |
67 | 546 | if (reader && !segment_v2::IndexReaderHelper::is_support_phrase(reader)) { |
68 | 2 | return Status::Error<ErrorCode::INDEX_INVALID_PARAMETERS>( |
69 | 2 | "phrase queries require setting support_phrase = true"); |
70 | 2 | } |
71 | 546 | } |
72 | 4.76k | Field param_value; |
73 | 4.76k | arguments[0].column->get(0, param_value); |
74 | 4.76k | if (param_value.is_null()) { |
75 | | // if query value is null, skip evaluate inverted index |
76 | 0 | return Status::OK(); |
77 | 0 | } |
78 | 4.76k | auto param_type = arguments[0].type->get_primitive_type(); |
79 | 4.76k | if (!is_string_type(param_type)) { |
80 | 0 | return Status::Error<ErrorCode::INDEX_INVALID_PARAMETERS>( |
81 | 0 | "arguments for match must be string"); |
82 | 0 | } |
83 | 4.76k | InvertedIndexParam param; |
84 | 4.76k | param.column_name = data_type_with_name.first; |
85 | 4.76k | param.column_type = data_type_with_name.second; |
86 | 4.76k | param.query_value = param_value; |
87 | 4.76k | param.query_type = get_query_type_from_fn_name(); |
88 | 4.76k | param.num_rows = num_rows; |
89 | 4.76k | param.roaring = std::make_shared<roaring::Roaring>(); |
90 | 4.76k | param.analyzer_ctx = analyzer_ctx; |
91 | 4.76k | if (is_string_type(param_type)) { |
92 | 4.76k | RETURN_IF_ERROR(iter->read_from_index(¶m)); |
93 | 18.4E | } else { |
94 | 18.4E | return Status::Error<ErrorCode::INDEX_INVALID_PARAMETERS>( |
95 | 18.4E | "invalid params type for FunctionMatchBase::evaluate_inverted_index {}", |
96 | 18.4E | param_type); |
97 | 18.4E | } |
98 | 4.76k | std::shared_ptr<roaring::Roaring> null_bitmap = std::make_shared<roaring::Roaring>(); |
99 | 4.77k | if (iter->has_null()) { |
100 | 4.77k | segment_v2::InvertedIndexQueryCacheHandle null_bitmap_cache_handle; |
101 | 4.77k | RETURN_IF_ERROR(iter->read_null_bitmap(&null_bitmap_cache_handle)); |
102 | 4.77k | null_bitmap = null_bitmap_cache_handle.get_bitmap(); |
103 | 4.77k | } |
104 | 4.76k | segment_v2::InvertedIndexResultBitmap result(param.roaring, null_bitmap); |
105 | 4.76k | bitmap_result = result; |
106 | 4.76k | bitmap_result.mask_out_null(); |
107 | | |
108 | 4.76k | return Status::OK(); |
109 | 4.76k | } |
110 | | Status FunctionMatchBase::execute_impl(FunctionContext* context, Block& block, |
111 | | const ColumnNumbers& arguments, uint32_t result, |
112 | 492 | size_t input_rows_count) const { |
113 | 492 | ColumnPtr& column_ptr = block.get_by_position(arguments[1]).column; |
114 | 492 | DataTypePtr& type_ptr = block.get_by_position(arguments[1]).type; |
115 | | |
116 | 492 | auto format_options = DataTypeSerDe::get_default_format_options(); |
117 | 492 | auto time_zone = cctz::utc_time_zone(); |
118 | 492 | format_options.timezone = |
119 | 492 | (context && context->state()) ? &context->state()->timezone_obj() : &time_zone; |
120 | | |
121 | 492 | auto match_query_str = type_ptr->to_string(*column_ptr, 0, format_options); |
122 | 492 | std::string column_name = block.get_by_position(arguments[0]).name; |
123 | 492 | VLOG_DEBUG << "begin to execute match directly, column_name=" << column_name |
124 | 0 | << ", match_query_str=" << match_query_str; |
125 | 492 | auto* analyzer_ctx = get_match_analyzer_ctx(context); |
126 | 492 | const ColumnPtr source_col = |
127 | 492 | block.get_by_position(arguments[0]).column->convert_to_full_column_if_const(); |
128 | 492 | const ColumnString* values = nullptr; |
129 | 492 | const ColumnArray* array_col = nullptr; |
130 | 492 | if (const auto array_col_guard = check_and_get_column<ColumnArray>(source_col.get())) { |
131 | 11 | array_col = array_col_guard.get(); |
132 | 11 | if (array_col && !array_col->get_data().is_column_string()) { |
133 | 0 | return Status::NotSupported(fmt::format( |
134 | 0 | "unsupported nested array of type {} for function {}", |
135 | 0 | is_column_nullable(array_col->get_data()) ? array_col->get_data().get_name() |
136 | 0 | : array_col->get_data().get_name(), |
137 | 0 | get_name())); |
138 | 0 | } |
139 | | |
140 | 11 | if (is_column_nullable(array_col->get_data())) { |
141 | 11 | const auto& array_nested_null_column = |
142 | 11 | reinterpret_cast<const ColumnNullable&>(array_col->get_data()); |
143 | 11 | const auto values_guard = check_and_get_column<ColumnString>( |
144 | 11 | *(array_nested_null_column.get_nested_column_ptr())); |
145 | 11 | if (values_guard) { |
146 | 11 | values = values_guard.get(); |
147 | 11 | } |
148 | 11 | } else { |
149 | | // array column element is always set Nullable for now. |
150 | 0 | const auto values_guard = |
151 | 0 | check_and_get_column<ColumnString>(*(array_col->get_data_ptr())); |
152 | 0 | if (values_guard) { |
153 | 0 | values = values_guard.get(); |
154 | 0 | } |
155 | 0 | } |
156 | 481 | } else if (const auto nullable = check_and_get_column<ColumnNullable>(source_col.get())) { |
157 | 0 | const auto values_guard = |
158 | 0 | check_and_get_column<ColumnString>(*nullable->get_nested_column_ptr()); |
159 | 0 | if (values_guard) { |
160 | 0 | values = values_guard.get(); |
161 | 0 | } |
162 | 481 | } else if (const auto values_guard = check_and_get_column<ColumnString>(source_col.get())) { |
163 | 481 | values = values_guard.get(); |
164 | 481 | } |
165 | | |
166 | 492 | if (!values) { |
167 | 0 | LOG(WARNING) << "Illegal column " << source_col->get_name(); |
168 | 0 | return Status::InternalError("Not supported input column types"); |
169 | 0 | } |
170 | | // result column |
171 | 492 | auto res = ColumnUInt8::create(); |
172 | 492 | ColumnUInt8::Container& vec_res = res->get_data(); |
173 | | // set default value to 0, and match functions only need to set 1/true |
174 | 492 | vec_res.resize_fill(input_rows_count); |
175 | 492 | RETURN_IF_ERROR(execute_match(context, column_name, match_query_str, input_rows_count, values, |
176 | 492 | analyzer_ctx, (array_col ? &(array_col->get_offsets()) : nullptr), |
177 | 492 | vec_res)); |
178 | 489 | block.replace_by_position(result, std::move(res)); |
179 | | |
180 | 489 | return Status::OK(); |
181 | 492 | } |
182 | | |
183 | | inline doris::segment_v2::InvertedIndexQueryType FunctionMatchBase::get_query_type_from_fn_name() |
184 | 4.76k | const { |
185 | 4.76k | std::string fn_name = get_name(); |
186 | 4.76k | if (fn_name == MATCH_ANY_FUNCTION) { |
187 | 3.48k | return doris::segment_v2::InvertedIndexQueryType::MATCH_ANY_QUERY; |
188 | 3.48k | } else if (fn_name == MATCH_ALL_FUNCTION) { |
189 | 718 | return doris::segment_v2::InvertedIndexQueryType::MATCH_ALL_QUERY; |
190 | 718 | } else if (fn_name == MATCH_PHRASE_FUNCTION) { |
191 | 512 | return doris::segment_v2::InvertedIndexQueryType::MATCH_PHRASE_QUERY; |
192 | 512 | } else if (fn_name == MATCH_PHRASE_PREFIX_FUNCTION) { |
193 | 38 | return doris::segment_v2::InvertedIndexQueryType::MATCH_PHRASE_PREFIX_QUERY; |
194 | 38 | } else if (fn_name == MATCH_PHRASE_REGEXP_FUNCTION) { |
195 | 22 | return doris::segment_v2::InvertedIndexQueryType::MATCH_REGEXP_QUERY; |
196 | 18.4E | } else if (fn_name == MATCH_PHRASE_EDGE_FUNCTION) { |
197 | 3 | return doris::segment_v2::InvertedIndexQueryType::MATCH_PHRASE_EDGE_QUERY; |
198 | 3 | } |
199 | 18.4E | return doris::segment_v2::InvertedIndexQueryType::UNKNOWN_QUERY; |
200 | 4.76k | } |
201 | | |
202 | | std::vector<TermInfo> FunctionMatchBase::analyse_query_str_token( |
203 | | const InvertedIndexAnalyzerCtx* analyzer_ctx, const std::string& match_query_str, |
204 | 492 | const std::string& column_name) const { |
205 | 492 | std::vector<TermInfo> query_tokens; |
206 | 492 | if (analyzer_ctx == nullptr) { |
207 | 3 | return query_tokens; |
208 | 3 | } |
209 | | |
210 | 489 | VLOG_DEBUG << "begin to run " << get_name() << ", parser_type: " |
211 | 0 | << inverted_index_parser_type_to_string(analyzer_ctx->parser_type); |
212 | | |
213 | | // Decision is based on parser_type (from index properties): |
214 | | // - PARSER_NONE: no tokenization (keyword/exact match) |
215 | | // - Other parsers: tokenize using the analyzer |
216 | 489 | if (!analyzer_ctx->should_tokenize()) { |
217 | | // Keyword index: all strings (including empty) are valid tokens for exact match. |
218 | | // Empty string is a valid value in keyword index and should be matchable. |
219 | 4 | query_tokens.emplace_back(match_query_str); |
220 | 4 | return query_tokens; |
221 | 4 | } |
222 | | |
223 | | // Safety check: if analyzer is nullptr but tokenization is expected, fall back to no tokenization |
224 | 485 | if (analyzer_ctx->analyzer == nullptr) { |
225 | 0 | VLOG_DEBUG << "Analyzer is nullptr, falling back to no tokenization"; |
226 | | // For fallback case, also allow empty strings to be matched |
227 | 0 | query_tokens.emplace_back(match_query_str); |
228 | 0 | return query_tokens; |
229 | 0 | } |
230 | | |
231 | | // Tokenize using the analyzer |
232 | 485 | auto reader = doris::segment_v2::inverted_index::InvertedIndexAnalyzer::create_reader( |
233 | 485 | analyzer_ctx->char_filter_map); |
234 | 485 | reader->init(match_query_str.data(), (int)match_query_str.size(), true); |
235 | 485 | query_tokens = doris::segment_v2::inverted_index::InvertedIndexAnalyzer::get_analyse_result( |
236 | 485 | reader, analyzer_ctx->analyzer.get()); |
237 | 485 | return query_tokens; |
238 | 485 | } |
239 | | |
240 | | inline std::vector<TermInfo> FunctionMatchBase::analyse_data_token( |
241 | | const std::string& column_name, const InvertedIndexAnalyzerCtx* analyzer_ctx, |
242 | | const ColumnString* string_col, int32_t current_block_row_idx, |
243 | 25.5k | const ColumnArray::Offsets64* array_offsets, int32_t& current_src_array_offset) const { |
244 | 25.5k | std::vector<TermInfo> data_tokens; |
245 | 25.5k | if (analyzer_ctx == nullptr) { |
246 | 0 | return data_tokens; |
247 | 0 | } |
248 | | |
249 | | // Determine tokenization strategy based on parser_type |
250 | 25.5k | const bool should_tokenize = |
251 | 25.5k | analyzer_ctx->should_tokenize() && analyzer_ctx->analyzer != nullptr; |
252 | | |
253 | 25.5k | if (array_offsets) { |
254 | 13 | for (auto next_src_array_offset = (*array_offsets)[current_block_row_idx]; |
255 | 28 | current_src_array_offset < next_src_array_offset; ++current_src_array_offset) { |
256 | 15 | const auto& str_ref = string_col->get_data_at(current_src_array_offset); |
257 | 15 | if (!should_tokenize) { |
258 | 0 | data_tokens.emplace_back(str_ref.to_string()); |
259 | 0 | continue; |
260 | 0 | } |
261 | 15 | auto reader = doris::segment_v2::inverted_index::InvertedIndexAnalyzer::create_reader( |
262 | 15 | analyzer_ctx->char_filter_map); |
263 | 15 | reader->init(str_ref.data, (int)str_ref.size, true); |
264 | 15 | data_tokens = |
265 | 15 | doris::segment_v2::inverted_index::InvertedIndexAnalyzer::get_analyse_result( |
266 | 15 | reader, analyzer_ctx->analyzer.get()); |
267 | 15 | } |
268 | 25.4k | } else { |
269 | 25.4k | const auto& str_ref = string_col->get_data_at(current_block_row_idx); |
270 | 25.4k | if (!should_tokenize) { |
271 | 15 | data_tokens.emplace_back(str_ref.to_string()); |
272 | 25.4k | } else { |
273 | 25.4k | auto reader = doris::segment_v2::inverted_index::InvertedIndexAnalyzer::create_reader( |
274 | 25.4k | analyzer_ctx->char_filter_map); |
275 | 25.4k | reader->init(str_ref.data, (int)str_ref.size, true); |
276 | 25.4k | data_tokens = |
277 | 25.4k | doris::segment_v2::inverted_index::InvertedIndexAnalyzer::get_analyse_result( |
278 | 25.4k | reader, analyzer_ctx->analyzer.get()); |
279 | 25.4k | } |
280 | 25.4k | } |
281 | 25.5k | return data_tokens; |
282 | 25.5k | } |
283 | | |
284 | 492 | Status FunctionMatchBase::check(FunctionContext* context, const std::string& function_name) const { |
285 | 492 | if (!context->state()->query_options().enable_match_without_inverted_index) { |
286 | 2 | return Status::Error<ErrorCode::INVERTED_INDEX_NOT_SUPPORTED>( |
287 | 2 | "{} not support execute_match", function_name); |
288 | 2 | } |
289 | | |
290 | 490 | DBUG_EXECUTE_IF("match.invert_index_not_support_execute_match", { |
291 | 490 | return Status::Error<ErrorCode::INVERTED_INDEX_NOT_SUPPORTED>( |
292 | 490 | "debug point: {} not support execute_match", function_name); |
293 | 490 | }); |
294 | | |
295 | 490 | return Status::OK(); |
296 | 490 | } |
297 | | |
298 | | Status FunctionMatchAny::execute_match(FunctionContext* context, const std::string& column_name, |
299 | | const std::string& match_query_str, size_t input_rows_count, |
300 | | const ColumnString* string_col, |
301 | | const InvertedIndexAnalyzerCtx* analyzer_ctx, |
302 | | const ColumnArray::Offsets64* array_offsets, |
303 | 306 | ColumnUInt8::Container& result) const { |
304 | 306 | RETURN_IF_ERROR(check(context, name)); |
305 | | |
306 | 306 | auto query_tokens = analyse_query_str_token(analyzer_ctx, match_query_str, column_name); |
307 | 306 | if (query_tokens.empty()) { |
308 | 6 | VLOG_DEBUG << fmt::format( |
309 | 0 | "token parser result is empty for query, " |
310 | 0 | "please check your query: '{}' and index parser: '{}'", |
311 | 0 | match_query_str, |
312 | 0 | analyzer_ctx ? inverted_index_parser_type_to_string(analyzer_ctx->parser_type) |
313 | 0 | : "unknown"); |
314 | 6 | return Status::OK(); |
315 | 6 | } |
316 | | |
317 | 300 | auto current_src_array_offset = 0; |
318 | 20.5k | for (int i = 0; i < input_rows_count; i++) { |
319 | 20.2k | auto data_tokens = analyse_data_token(column_name, analyzer_ctx, string_col, i, |
320 | 20.2k | array_offsets, current_src_array_offset); |
321 | | |
322 | | // TODO: more efficient impl |
323 | 21.0k | for (auto& term_info : query_tokens) { |
324 | 21.0k | auto it = |
325 | 74.0k | std::find_if(data_tokens.begin(), data_tokens.end(), [&](const TermInfo& info) { |
326 | 74.0k | return info.get_single_term() == term_info.get_single_term(); |
327 | 74.0k | }); |
328 | 21.0k | if (it != data_tokens.end()) { |
329 | 11.6k | result[i] = true; |
330 | 11.6k | break; |
331 | 11.6k | } |
332 | 21.0k | } |
333 | 20.2k | } |
334 | | |
335 | 300 | return Status::OK(); |
336 | 306 | } |
337 | | |
338 | | Status FunctionMatchAll::execute_match(FunctionContext* context, const std::string& column_name, |
339 | | const std::string& match_query_str, size_t input_rows_count, |
340 | | const ColumnString* string_col, |
341 | | const InvertedIndexAnalyzerCtx* analyzer_ctx, |
342 | | const ColumnArray::Offsets64* array_offsets, |
343 | 91 | ColumnUInt8::Container& result) const { |
344 | 91 | RETURN_IF_ERROR(check(context, name)); |
345 | | |
346 | 91 | auto query_tokens = analyse_query_str_token(analyzer_ctx, match_query_str, column_name); |
347 | 91 | if (query_tokens.empty()) { |
348 | 5 | VLOG_DEBUG << fmt::format( |
349 | 0 | "token parser result is empty for query, " |
350 | 0 | "please check your query: '{}' and index parser: '{}'", |
351 | 0 | match_query_str, |
352 | 0 | analyzer_ctx ? inverted_index_parser_type_to_string(analyzer_ctx->parser_type) |
353 | 0 | : "unknown"); |
354 | 5 | return Status::OK(); |
355 | 5 | } |
356 | | |
357 | 86 | auto current_src_array_offset = 0; |
358 | 1.18k | for (int i = 0; i < input_rows_count; i++) { |
359 | 1.09k | auto data_tokens = analyse_data_token(column_name, analyzer_ctx, string_col, i, |
360 | 1.09k | array_offsets, current_src_array_offset); |
361 | | |
362 | | // TODO: more efficient impl |
363 | 1.09k | auto find_count = 0; |
364 | 1.35k | for (auto& term_info : query_tokens) { |
365 | 1.35k | auto it = |
366 | 9.62k | std::find_if(data_tokens.begin(), data_tokens.end(), [&](const TermInfo& info) { |
367 | 9.62k | return info.get_single_term() == term_info.get_single_term(); |
368 | 9.62k | }); |
369 | 1.35k | if (it != data_tokens.end()) { |
370 | 394 | ++find_count; |
371 | 958 | } else { |
372 | 958 | break; |
373 | 958 | } |
374 | 1.35k | } |
375 | | |
376 | 1.09k | if (find_count == query_tokens.size()) { |
377 | 138 | result[i] = true; |
378 | 138 | } |
379 | 1.09k | } |
380 | | |
381 | 86 | return Status::OK(); |
382 | 91 | } |
383 | | |
384 | | Status FunctionMatchPhrase::execute_match(FunctionContext* context, const std::string& column_name, |
385 | | const std::string& match_query_str, |
386 | | size_t input_rows_count, const ColumnString* string_col, |
387 | | const InvertedIndexAnalyzerCtx* analyzer_ctx, |
388 | | const ColumnArray::Offsets64* array_offsets, |
389 | 84 | ColumnUInt8::Container& result) const { |
390 | 84 | RETURN_IF_ERROR(check(context, name)); |
391 | | |
392 | 83 | auto query_tokens = analyse_query_str_token(analyzer_ctx, match_query_str, column_name); |
393 | 83 | if (query_tokens.empty()) { |
394 | 7 | VLOG_DEBUG << fmt::format( |
395 | 0 | "token parser result is empty for query, " |
396 | 0 | "please check your query: '{}' and index parser: '{}'", |
397 | 0 | match_query_str, |
398 | 0 | analyzer_ctx ? inverted_index_parser_type_to_string(analyzer_ctx->parser_type) |
399 | 0 | : "unknown"); |
400 | 7 | return Status::OK(); |
401 | 7 | } |
402 | | |
403 | 76 | auto current_src_array_offset = 0; |
404 | 1.16k | for (int i = 0; i < input_rows_count; i++) { |
405 | 1.08k | auto data_tokens = analyse_data_token(column_name, analyzer_ctx, string_col, i, |
406 | 1.08k | array_offsets, current_src_array_offset); |
407 | | |
408 | | // TODO: more efficient impl |
409 | 1.08k | bool matched = false; |
410 | 1.08k | auto data_it = data_tokens.begin(); |
411 | 2.16k | while (data_it != data_tokens.end()) { |
412 | | // find position of first token |
413 | 8.63k | data_it = std::find_if(data_it, data_tokens.end(), [&](const TermInfo& info) { |
414 | 8.63k | return info.get_single_term() == query_tokens[0].get_single_term(); |
415 | 8.63k | }); |
416 | 1.19k | if (data_it != data_tokens.end()) { |
417 | 234 | matched = true; |
418 | 234 | auto data_it_next = ++data_it; |
419 | 234 | auto query_it = query_tokens.begin() + 1; |
420 | | // compare query_tokens after the first to data_tokens one by one |
421 | 356 | while (query_it != query_tokens.end()) { |
422 | 238 | if (data_it_next == data_tokens.end() || |
423 | 238 | data_it_next->get_single_term() != query_it->get_single_term()) { |
424 | 116 | matched = false; |
425 | 116 | break; |
426 | 116 | } |
427 | 122 | query_it++; |
428 | 122 | data_it_next++; |
429 | 122 | } |
430 | | |
431 | 234 | if (matched) { |
432 | 118 | break; |
433 | 118 | } |
434 | 234 | } |
435 | 1.19k | } |
436 | | |
437 | | // check matched |
438 | 1.08k | if (matched) { |
439 | 118 | result[i] = true; |
440 | 118 | } |
441 | 1.08k | } |
442 | | |
443 | 76 | return Status::OK(); |
444 | 83 | } |
445 | | |
446 | | Status FunctionMatchPhrasePrefix::execute_match( |
447 | | FunctionContext* context, const std::string& column_name, |
448 | | const std::string& match_query_str, size_t input_rows_count, const ColumnString* string_col, |
449 | | const InvertedIndexAnalyzerCtx* analyzer_ctx, const ColumnArray::Offsets64* array_offsets, |
450 | 5 | ColumnUInt8::Container& result) const { |
451 | 5 | RETURN_IF_ERROR(check(context, name)); |
452 | | |
453 | 4 | auto query_tokens = analyse_query_str_token(analyzer_ctx, match_query_str, column_name); |
454 | 4 | if (query_tokens.empty()) { |
455 | 2 | VLOG_DEBUG << fmt::format( |
456 | 0 | "token parser result is empty for query, " |
457 | 0 | "please check your query: '{}' and index parser: '{}'", |
458 | 0 | match_query_str, |
459 | 0 | analyzer_ctx ? inverted_index_parser_type_to_string(analyzer_ctx->parser_type) |
460 | 0 | : "unknown"); |
461 | 2 | return Status::OK(); |
462 | 2 | } |
463 | | |
464 | 2 | int32_t current_src_array_offset = 0; |
465 | 1.00k | for (int i = 0; i < input_rows_count; i++) { |
466 | 1.00k | auto data_tokens = analyse_data_token(column_name, analyzer_ctx, string_col, i, |
467 | 1.00k | array_offsets, current_src_array_offset); |
468 | | |
469 | 1.00k | int64_t dis_count = data_tokens.size() - query_tokens.size(); |
470 | 1.00k | if (dis_count < 0) { |
471 | 1 | continue; |
472 | 1 | } |
473 | | |
474 | 8.38k | for (size_t j = 0; j < dis_count + 1; j++) { |
475 | 7.50k | if (data_tokens[j].get_single_term() == query_tokens[0].get_single_term() || |
476 | 7.50k | query_tokens.size() == 1) { |
477 | 210 | bool match = true; |
478 | 540 | for (size_t k = 0; k < query_tokens.size(); k++) { |
479 | 420 | const std::string& data_token = data_tokens[j + k].get_single_term(); |
480 | 420 | const std::string& query_token = query_tokens[k].get_single_term(); |
481 | 420 | if (k == query_tokens.size() - 1) { |
482 | 210 | if (data_token.compare(0, query_token.size(), query_token) != 0) { |
483 | 90 | match = false; |
484 | 90 | break; |
485 | 90 | } |
486 | 210 | } else { |
487 | 210 | if (data_token != query_token) { |
488 | 0 | match = false; |
489 | 0 | break; |
490 | 0 | } |
491 | 210 | } |
492 | 420 | } |
493 | 210 | if (match) { |
494 | 120 | result[i] = true; |
495 | 120 | break; |
496 | 120 | } |
497 | 210 | } |
498 | 7.50k | } |
499 | 1.00k | } |
500 | | |
501 | 2 | return Status::OK(); |
502 | 4 | } |
503 | | |
504 | | Status FunctionMatchRegexp::execute_match(FunctionContext* context, const std::string& column_name, |
505 | | const std::string& match_query_str, |
506 | | size_t input_rows_count, const ColumnString* string_col, |
507 | | const InvertedIndexAnalyzerCtx* analyzer_ctx, |
508 | | const ColumnArray::Offsets64* array_offsets, |
509 | 6 | ColumnUInt8::Container& result) const { |
510 | 6 | RETURN_IF_ERROR(check(context, name)); |
511 | | |
512 | 6 | VLOG_DEBUG << "begin to run FunctionMatchRegexp::execute_match, parser_type: " |
513 | 0 | << (analyzer_ctx ? inverted_index_parser_type_to_string(analyzer_ctx->parser_type) |
514 | 0 | : "unknown"); |
515 | | |
516 | 6 | const std::string& pattern = match_query_str; |
517 | | |
518 | 6 | hs_database_t* database = nullptr; |
519 | 6 | hs_compile_error_t* compile_err = nullptr; |
520 | 6 | hs_scratch_t* scratch = nullptr; |
521 | | |
522 | 6 | if (hs_compile(pattern.data(), HS_FLAG_DOTALL | HS_FLAG_ALLOWEMPTY | HS_FLAG_UTF8, |
523 | 6 | HS_MODE_BLOCK, nullptr, &database, &compile_err) != HS_SUCCESS) { |
524 | 1 | std::string err_message = "hyperscan compilation failed: "; |
525 | 1 | err_message.append(compile_err->message); |
526 | 1 | LOG(ERROR) << err_message; |
527 | 1 | hs_free_compile_error(compile_err); |
528 | 1 | return Status::Error<ErrorCode::INDEX_INVALID_PARAMETERS>(err_message); |
529 | 1 | } |
530 | | |
531 | 5 | if (hs_alloc_scratch(database, &scratch) != HS_SUCCESS) { |
532 | 0 | LOG(ERROR) << "hyperscan could not allocate scratch space."; |
533 | 0 | hs_free_database(database); |
534 | 0 | return Status::Error<ErrorCode::INDEX_INVALID_PARAMETERS>( |
535 | 0 | "hyperscan could not allocate scratch space."); |
536 | 0 | } |
537 | | |
538 | 5 | auto on_match = [](unsigned int id, unsigned long long from, unsigned long long to, |
539 | 4.06k | unsigned int flags, void* context) -> int { |
540 | 4.06k | *((bool*)context) = true; |
541 | 4.06k | return 0; |
542 | 4.06k | }; |
543 | | |
544 | 5 | try { |
545 | 5 | auto current_src_array_offset = 0; |
546 | 2.00k | for (int i = 0; i < input_rows_count; i++) { |
547 | 2.00k | auto data_tokens = analyse_data_token(column_name, analyzer_ctx, string_col, i, |
548 | 2.00k | array_offsets, current_src_array_offset); |
549 | | |
550 | 9.62k | for (auto& input : data_tokens) { |
551 | 9.62k | bool is_match = false; |
552 | 9.62k | const auto& input_str = input.get_single_term(); |
553 | 9.62k | if (hs_scan(database, input_str.data(), (uint32_t)input_str.size(), 0, scratch, |
554 | 9.62k | on_match, (void*)&is_match) != HS_SUCCESS) { |
555 | 0 | LOG(ERROR) << "hyperscan match failed: " << input_str; |
556 | 0 | break; |
557 | 0 | } |
558 | | |
559 | 9.62k | if (is_match) { |
560 | 1.06k | result[i] = true; |
561 | 1.06k | break; |
562 | 1.06k | } |
563 | 9.62k | } |
564 | 2.00k | } |
565 | 5 | } |
566 | 5 | _CLFINALLY({ |
567 | 5 | hs_free_scratch(scratch); |
568 | 5 | hs_free_database(database); |
569 | 5 | }) |
570 | | |
571 | 5 | return Status::OK(); |
572 | 5 | } |
573 | | |
574 | | Status FunctionMatchPhraseEdge::execute_match( |
575 | | FunctionContext* context, const std::string& column_name, |
576 | | const std::string& match_query_str, size_t input_rows_count, const ColumnString* string_col, |
577 | | const InvertedIndexAnalyzerCtx* analyzer_ctx, const ColumnArray::Offsets64* array_offsets, |
578 | 0 | ColumnUInt8::Container& result) const { |
579 | 0 | RETURN_IF_ERROR(check(context, name)); |
580 | | |
581 | 0 | auto query_tokens = analyse_query_str_token(analyzer_ctx, match_query_str, column_name); |
582 | 0 | if (query_tokens.empty()) { |
583 | 0 | VLOG_DEBUG << fmt::format( |
584 | 0 | "token parser result is empty for query, " |
585 | 0 | "please check your query: '{}' and index parser: '{}'", |
586 | 0 | match_query_str, |
587 | 0 | analyzer_ctx ? inverted_index_parser_type_to_string(analyzer_ctx->parser_type) |
588 | 0 | : "unknown"); |
589 | 0 | return Status::OK(); |
590 | 0 | } |
591 | | |
592 | 0 | int32_t current_src_array_offset = 0; |
593 | 0 | for (int i = 0; i < input_rows_count; i++) { |
594 | 0 | auto data_tokens = analyse_data_token(column_name, analyzer_ctx, string_col, i, |
595 | 0 | array_offsets, current_src_array_offset); |
596 | |
|
597 | 0 | int64_t dis_count = data_tokens.size() - query_tokens.size(); |
598 | 0 | if (dis_count < 0) { |
599 | 0 | continue; |
600 | 0 | } |
601 | | |
602 | 0 | for (size_t j = 0; j < dis_count + 1; j++) { |
603 | 0 | bool match = true; |
604 | 0 | if (query_tokens.size() == 1) { |
605 | 0 | if (data_tokens[j].get_single_term().find(query_tokens[0].get_single_term()) == |
606 | 0 | std::string::npos) { |
607 | 0 | match = false; |
608 | 0 | } |
609 | 0 | } else { |
610 | 0 | for (size_t k = 0; k < query_tokens.size(); k++) { |
611 | 0 | const std::string& data_token = data_tokens[j + k].get_single_term(); |
612 | 0 | const std::string& query_token = query_tokens[k].get_single_term(); |
613 | 0 | if (k == 0) { |
614 | 0 | if (!data_token.ends_with(query_token)) { |
615 | 0 | match = false; |
616 | 0 | break; |
617 | 0 | } |
618 | 0 | } else if (k == query_tokens.size() - 1) { |
619 | 0 | if (!data_token.starts_with(query_token)) { |
620 | 0 | match = false; |
621 | 0 | break; |
622 | 0 | } |
623 | 0 | } else { |
624 | 0 | if (data_token != query_token) { |
625 | 0 | match = false; |
626 | 0 | break; |
627 | 0 | } |
628 | 0 | } |
629 | 0 | } |
630 | 0 | } |
631 | 0 | if (match) { |
632 | 0 | result[i] = true; |
633 | 0 | break; |
634 | 0 | } |
635 | 0 | } |
636 | 0 | } |
637 | |
|
638 | 0 | return Status::OK(); |
639 | 0 | } |
640 | | |
641 | 8 | void register_function_match(SimpleFunctionFactory& factory) { |
642 | 8 | factory.register_function<FunctionMatchAny>(); |
643 | 8 | factory.register_function<FunctionMatchAll>(); |
644 | 8 | factory.register_function<FunctionMatchPhrase>(); |
645 | 8 | factory.register_function<FunctionMatchPhrasePrefix>(); |
646 | 8 | factory.register_function<FunctionMatchRegexp>(); |
647 | 8 | factory.register_function<FunctionMatchPhraseEdge>(); |
648 | 8 | } |
649 | | } // namespace doris |