Adrià Vilanova MartÃnez | f19ea43 | 2024-01-23 20:20:52 +0100 | [diff] [blame^] | 1 | # Copyright 2016 The Chromium Authors |
| 2 | # Use of this source code is governed by a BSD-style license that can be |
| 3 | # found in the LICENSE file. |
Copybara | 854996b | 2021-09-07 19:36:02 +0000 | [diff] [blame] | 4 | |
| 5 | """A set of functions that provide fulltext search for issues.""" |
| 6 | from __future__ import print_function |
| 7 | from __future__ import division |
| 8 | from __future__ import absolute_import |
| 9 | |
| 10 | import collections |
| 11 | import logging |
| 12 | import time |
| 13 | |
| 14 | from six import string_types |
| 15 | |
| 16 | from google.appengine.api import search |
| 17 | |
| 18 | import settings |
| 19 | from framework import framework_constants |
| 20 | from framework import framework_helpers |
| 21 | from framework import framework_views |
| 22 | from services import fulltext_helpers |
| 23 | from tracker import tracker_bizobj |
| 24 | |
| 25 | |
| 26 | # When updating and re-indexing all issues in a project, work in batches |
| 27 | # of this size to manage memory usage and avoid rpc timeouts. |
| 28 | _INDEX_BATCH_SIZE = 40 |
| 29 | |
| 30 | |
| 31 | # The user can search for text that occurs specifically in these |
| 32 | # parts of an issue. |
| 33 | ISSUE_FULLTEXT_FIELDS = ['summary', 'description', 'comment'] |
| 34 | # Note: issue documents also contain a "metadata" field, but we do not |
| 35 | # expose that to users. Issue metadata can be searched in a structured way |
| 36 | # by giving a specific field name such as "owner:" or "status:". The metadata |
| 37 | # search field exists only for fulltext queries that do not specify any field. |
| 38 | |
| 39 | |
| 40 | def IndexIssues(cnxn, issues, user_service, issue_service, config_service): |
| 41 | """(Re)index all the given issues. |
| 42 | |
| 43 | Args: |
| 44 | cnxn: connection to SQL database. |
| 45 | issues: list of Issue PBs to index. |
| 46 | user_service: interface to user data storage. |
| 47 | issue_service: interface to issue data storage. |
| 48 | config_service: interface to configuration data storage. |
| 49 | """ |
| 50 | issues = list(issues) |
| 51 | config_dict = config_service.GetProjectConfigs( |
| 52 | cnxn, {issue.project_id for issue in issues}) |
| 53 | for start in range(0, len(issues), _INDEX_BATCH_SIZE): |
| 54 | logging.info('indexing issues: %d remaining', len(issues) - start) |
| 55 | _IndexIssueBatch( |
| 56 | cnxn, issues[start:start + _INDEX_BATCH_SIZE], user_service, |
| 57 | issue_service, config_dict) |
| 58 | |
| 59 | |
| 60 | def _IndexIssueBatch(cnxn, issues, user_service, issue_service, config_dict): |
| 61 | """Internal method to (re)index the given batch of issues. |
| 62 | |
| 63 | Args: |
| 64 | cnxn: connection to SQL database. |
| 65 | issues: list of Issue PBs to index. |
| 66 | user_service: interface to user data storage. |
| 67 | issue_service: interface to issue data storage. |
| 68 | config_dict: dict {project_id: config} for all the projects that |
| 69 | the given issues are in. |
| 70 | """ |
| 71 | user_ids = tracker_bizobj.UsersInvolvedInIssues(issues) |
| 72 | comments_dict = issue_service.GetCommentsForIssues( |
| 73 | cnxn, [issue.issue_id for issue in issues]) |
| 74 | for comments in comments_dict.values(): |
| 75 | user_ids.update([ic.user_id for ic in comments]) |
| 76 | |
| 77 | users_by_id = framework_views.MakeAllUserViews( |
| 78 | cnxn, user_service, user_ids) |
| 79 | _CreateIssueSearchDocuments(issues, comments_dict, users_by_id, config_dict) |
| 80 | |
| 81 | |
| 82 | def _CreateIssueSearchDocuments( |
| 83 | issues, comments_dict, users_by_id, config_dict): |
| 84 | """Make the GAE search index documents for the given issue batch. |
| 85 | |
| 86 | Args: |
| 87 | issues: list of issues to index. |
| 88 | comments_dict: prefetched dictionary of comments on those issues. |
| 89 | users_by_id: dictionary {user_id: UserView} so that the email |
| 90 | addresses of users who left comments can be found via search. |
| 91 | config_dict: dict {project_id: config} for all the projects that |
| 92 | the given issues are in. |
| 93 | """ |
| 94 | documents_by_shard = collections.defaultdict(list) |
| 95 | for issue in issues: |
| 96 | summary = issue.summary |
| 97 | # TODO(jrobbins): allow search specifically on explicit vs derived |
| 98 | # fields. |
| 99 | owner_id = tracker_bizobj.GetOwnerId(issue) |
| 100 | owner_email = users_by_id[owner_id].email |
| 101 | config = config_dict[issue.project_id] |
| 102 | component_paths = [] |
| 103 | for component_id in issue.component_ids: |
| 104 | cd = tracker_bizobj.FindComponentDefByID(component_id, config) |
| 105 | if cd: |
| 106 | component_paths.append(cd.path) |
| 107 | |
| 108 | field_values = [tracker_bizobj.GetFieldValue(fv, users_by_id) |
| 109 | for fv in issue.field_values] |
| 110 | # Convert to string only the values that are not strings already. |
| 111 | # This is done because the default encoding in appengine seems to be 'ascii' |
| 112 | # and string values might contain unicode characters, so str will fail to |
| 113 | # encode them. |
| 114 | field_values = [value if isinstance(value, string_types) else str(value) |
| 115 | for value in field_values] |
| 116 | |
| 117 | metadata = '%s %s %s %s %s %s' % ( |
| 118 | tracker_bizobj.GetStatus(issue), |
| 119 | owner_email, |
| 120 | [users_by_id[cc_id].email for cc_id in |
| 121 | tracker_bizobj.GetCcIds(issue)], |
| 122 | ' '.join(component_paths), |
| 123 | ' '.join(field_values), |
| 124 | ' '.join(tracker_bizobj.GetLabels(issue))) |
| 125 | custom_fields = _BuildCustomFTSFields(issue) |
| 126 | |
| 127 | comments = comments_dict.get(issue.issue_id, []) |
| 128 | room_for_comments = (framework_constants.MAX_FTS_FIELD_SIZE - |
| 129 | len(summary) - |
| 130 | len(metadata) - |
| 131 | sum(len(cf.value) for cf in custom_fields)) |
| 132 | comments = _IndexableComments( |
| 133 | comments, users_by_id, remaining_chars=room_for_comments) |
| 134 | logging.info('len(comments) is %r', len(comments)) |
| 135 | if comments: |
| 136 | description = _ExtractCommentText(comments[0], users_by_id) |
| 137 | description = description[:framework_constants.MAX_FTS_FIELD_SIZE] |
| 138 | all_comments = ' '. join( |
| 139 | _ExtractCommentText(c, users_by_id) for c in comments[1:]) |
| 140 | all_comments = all_comments[:framework_constants.MAX_FTS_FIELD_SIZE] |
| 141 | else: |
| 142 | description = '' |
| 143 | all_comments = '' |
| 144 | logging.info( |
| 145 | 'Issue %s:%r has zero indexable comments', |
| 146 | issue.project_name, issue.local_id) |
| 147 | |
| 148 | logging.info('Building document for %s:%d', |
| 149 | issue.project_name, issue.local_id) |
| 150 | logging.info('len(summary) = %d', len(summary)) |
| 151 | logging.info('len(metadata) = %d', len(metadata)) |
| 152 | logging.info('len(description) = %d', len(description)) |
| 153 | logging.info('len(comment) = %d', len(all_comments)) |
| 154 | for cf in custom_fields: |
| 155 | logging.info('len(%s) = %d', cf.name, len(cf.value)) |
| 156 | |
| 157 | doc = search.Document( |
| 158 | doc_id=str(issue.issue_id), |
| 159 | fields=[ |
| 160 | search.NumberField(name='project_id', value=issue.project_id), |
| 161 | search.TextField(name='summary', value=summary), |
| 162 | search.TextField(name='metadata', value=metadata), |
| 163 | search.TextField(name='description', value=description), |
| 164 | search.TextField(name='comment', value=all_comments), |
| 165 | ] + custom_fields) |
| 166 | |
| 167 | shard_id = issue.issue_id % settings.num_logical_shards |
| 168 | documents_by_shard[shard_id].append(doc) |
| 169 | |
| 170 | start_time = time.time() |
| 171 | promises = [] |
| 172 | for shard_id, documents in documents_by_shard.items(): |
| 173 | if documents: |
| 174 | promises.append(framework_helpers.Promise( |
| 175 | _IndexDocsInShard, shard_id, documents)) |
| 176 | |
| 177 | for promise in promises: |
| 178 | promise.WaitAndGetValue() |
| 179 | |
| 180 | logging.info('Finished %d indexing in shards in %d ms', |
| 181 | len(documents_by_shard), int((time.time() - start_time) * 1000)) |
| 182 | |
| 183 | |
| 184 | def _IndexableComments(comments, users_by_id, remaining_chars=None): |
| 185 | """We only index the comments that are not deleted or banned. |
| 186 | |
| 187 | Args: |
| 188 | comments: list of Comment PBs for one issue. |
| 189 | users_by_id: Dict of (user_id -> UserView) for all users. |
| 190 | remaining_chars: number of characters available for comment text |
| 191 | without hitting the GAE search index max document size. |
| 192 | |
| 193 | Returns: |
| 194 | A list of comments filtered to not have any deleted comments or |
| 195 | comments from banned users. If the issue has a huge number of |
| 196 | comments, only a certain number of the first and last comments |
| 197 | are actually indexed. |
| 198 | """ |
| 199 | if remaining_chars is None: |
| 200 | remaining_chars = framework_constants.MAX_FTS_FIELD_SIZE |
| 201 | allowed_comments = [] |
| 202 | for comment in comments: |
| 203 | user_view = users_by_id.get(comment.user_id) |
| 204 | if not (comment.deleted_by or (user_view and user_view.banned)): |
| 205 | if comment.is_description and allowed_comments: |
| 206 | # index the latest description, but not older descriptions |
| 207 | allowed_comments[0] = comment |
| 208 | else: |
| 209 | allowed_comments.append(comment) |
| 210 | |
| 211 | reasonable_size = (framework_constants.INITIAL_COMMENTS_TO_INDEX + |
| 212 | framework_constants.FINAL_COMMENTS_TO_INDEX) |
| 213 | if len(allowed_comments) <= reasonable_size: |
| 214 | candidates = allowed_comments |
| 215 | else: |
| 216 | candidates = ( # Prioritize the description and recent comments. |
| 217 | allowed_comments[0:1] + |
| 218 | allowed_comments[-framework_constants.FINAL_COMMENTS_TO_INDEX:] + |
| 219 | allowed_comments[1:framework_constants.INITIAL_COMMENTS_TO_INDEX]) |
| 220 | |
| 221 | total_length = 0 |
| 222 | result = [] |
| 223 | for comment in candidates: |
| 224 | total_length += len(comment.content) |
| 225 | if total_length > remaining_chars: |
| 226 | break |
| 227 | result.append(comment) |
| 228 | |
| 229 | return result |
| 230 | |
| 231 | |
| 232 | def _IndexDocsInShard(shard_id, documents): |
| 233 | search_index = search.Index( |
| 234 | name=settings.search_index_name_format % shard_id) |
| 235 | search_index.put(documents) |
| 236 | logging.info('FTS indexed %d docs in shard %d', len(documents), shard_id) |
| 237 | # TODO(jrobbins): catch OverQuotaError and add the issues to the |
| 238 | # ReindexQueue table instead. |
| 239 | |
| 240 | |
| 241 | def _ExtractCommentText(comment, users_by_id): |
| 242 | """Return a string with all the searchable text of the given Comment PB.""" |
| 243 | commenter_email = users_by_id[comment.user_id].email |
| 244 | return '%s %s %s' % ( |
| 245 | commenter_email, |
| 246 | comment.content, |
| 247 | ' '.join(attach.filename |
| 248 | for attach in comment.attachments |
| 249 | if not attach.deleted)) |
| 250 | |
| 251 | |
| 252 | def _BuildCustomFTSFields(issue): |
| 253 | """Return a list of FTS Fields to index string-valued custom fields.""" |
| 254 | fts_fields = [] |
| 255 | for fv in issue.field_values: |
| 256 | if fv.str_value: |
| 257 | # TODO(jrobbins): also indicate which were derived vs. explicit. |
| 258 | # TODO(jrobbins): also toss in the email addresses of any users in |
| 259 | # user-valued custom fields, ints for int-valued fields, etc. |
| 260 | fts_field = search.TextField( |
| 261 | name='custom_%d' % fv.field_id, value=fv.str_value) |
| 262 | fts_fields.append(fts_field) |
| 263 | |
| 264 | return fts_fields |
| 265 | |
| 266 | |
| 267 | def UnindexIssues(issue_ids): |
| 268 | """Remove many issues from the sharded search indexes.""" |
| 269 | iids_by_shard = {} |
| 270 | for issue_id in issue_ids: |
| 271 | shard_id = issue_id % settings.num_logical_shards |
| 272 | iids_by_shard.setdefault(shard_id, []) |
| 273 | iids_by_shard[shard_id].append(issue_id) |
| 274 | |
| 275 | for shard_id, iids_in_shard in iids_by_shard.items(): |
| 276 | try: |
| 277 | logging.info( |
| 278 | 'unindexing %r issue_ids in %r', len(iids_in_shard), shard_id) |
| 279 | search_index = search.Index( |
| 280 | name=settings.search_index_name_format % shard_id) |
| 281 | search_index.delete([str(iid) for iid in iids_in_shard]) |
| 282 | except search.Error: |
| 283 | logging.exception('FTS deletion failed') |
| 284 | |
| 285 | |
| 286 | def SearchIssueFullText(project_ids, query_ast_conj, shard_id): |
| 287 | """Do full-text search in GAE FTS. |
| 288 | |
| 289 | Args: |
| 290 | project_ids: list of project ID numbers to consider. |
| 291 | query_ast_conj: One conjuctive clause from the AST parsed |
| 292 | from the user's query. |
| 293 | shard_id: int shard ID for the shard to consider. |
| 294 | |
| 295 | Returns: |
| 296 | (issue_ids, capped) where issue_ids is a list of issue issue_ids that match |
| 297 | the full-text query. And, capped is True if the results were capped due to |
| 298 | an implementation limitation. Or, return (None, False) if the given AST |
| 299 | conjunction contains no full-text conditions. |
| 300 | """ |
| 301 | fulltext_query = fulltext_helpers.BuildFTSQuery( |
| 302 | query_ast_conj, ISSUE_FULLTEXT_FIELDS) |
| 303 | if fulltext_query is None: |
| 304 | return None, False |
| 305 | |
| 306 | if project_ids: |
| 307 | project_clause = ' OR '.join( |
| 308 | 'project_id:%d' % pid for pid in project_ids) |
| 309 | fulltext_query = '(%s) %s' % (project_clause, fulltext_query) |
| 310 | |
| 311 | # TODO(jrobbins): it would be good to also include some other |
| 312 | # structured search terms to narrow down the set of index |
| 313 | # documents considered. E.g., most queries are only over the |
| 314 | # open issues. |
| 315 | logging.info('FTS query is %r', fulltext_query) |
| 316 | issue_ids = fulltext_helpers.ComprehensiveSearch( |
| 317 | fulltext_query, settings.search_index_name_format % shard_id) |
| 318 | capped = len(issue_ids) >= settings.fulltext_limit_per_shard |
| 319 | return issue_ids, capped |