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