Yao Xiao | 7a1995b | 2022-03-09 08:18:55 | [diff] [blame] | 1 | // Copyright 2022 The Chromium Authors. All rights reserved. |
| 2 | // Use of this source code is governed by a BSD-style license that can be |
| 3 | // found in the LICENSE file. |
| 4 | |
| 5 | #include "components/browsing_topics/browsing_topics_service_impl.h" |
| 6 | |
Yao Xiao | cc37939 | 2022-03-25 21:39:06 | [diff] [blame] | 7 | #include <random> |
| 8 | |
| 9 | #include "base/rand_util.h" |
Yao Xiao | 57892a2 | 2022-06-28 19:21:41 | [diff] [blame^] | 10 | #include "base/ranges/algorithm.h" |
Gabriel Charette | d87f10f | 2022-03-31 00:44:22 | [diff] [blame] | 11 | #include "base/time/time.h" |
Yao Xiao | cc37939 | 2022-03-25 21:39:06 | [diff] [blame] | 12 | #include "components/browsing_topics/browsing_topics_calculator.h" |
| 13 | #include "components/browsing_topics/browsing_topics_page_load_data_tracker.h" |
Yao Xiao | 21f1faa | 2022-04-29 06:20:32 | [diff] [blame] | 14 | #include "components/browsing_topics/mojom/browsing_topics_internals.mojom.h" |
Yao Xiao | cc37939 | 2022-03-25 21:39:06 | [diff] [blame] | 15 | #include "components/browsing_topics/util.h" |
| 16 | #include "components/optimization_guide/content/browser/page_content_annotations_service.h" |
| 17 | #include "content/public/browser/browsing_topics_site_data_manager.h" |
| 18 | #include "net/base/registry_controlled_domains/registry_controlled_domain.h" |
Yao Xiao | 716e481 | 2022-04-20 22:57:19 | [diff] [blame] | 19 | #include "services/metrics/public/cpp/ukm_builders.h" |
| 20 | #include "services/metrics/public/cpp/ukm_recorder.h" |
Yao Xiao | cc37939 | 2022-03-25 21:39:06 | [diff] [blame] | 21 | #include "third_party/blink/public/common/features.h" |
| 22 | #include "third_party/blink/public/mojom/browsing_topics/browsing_topics.mojom.h" |
| 23 | |
Yao Xiao | 7a1995b | 2022-03-09 08:18:55 | [diff] [blame] | 24 | namespace browsing_topics { |
| 25 | |
Yao Xiao | cc37939 | 2022-03-25 21:39:06 | [diff] [blame] | 26 | namespace { |
| 27 | |
Yao Xiao | bf39e34d | 2022-03-28 21:48:28 | [diff] [blame] | 28 | // Returns whether the topics should all be cleared given |
| 29 | // `browsing_topics_data_accessible_since` and `is_topic_allowed_by_settings`. |
| 30 | // Returns true if `browsing_topics_data_accessible_since` is greater than the |
| 31 | // last calculation time, or if any top topic is disallowed from the settings. |
| 32 | // The latter could happen if the topic became disallowed when |
| 33 | // `browsing_topics_state` was still loading (and we didn't get a chance to |
| 34 | // clear it). This is an unlikely edge case, so it's fine to over-delete. |
| 35 | bool ShouldClearTopicsOnStartup( |
Yao Xiao | cc37939 | 2022-03-25 21:39:06 | [diff] [blame] | 36 | const BrowsingTopicsState& browsing_topics_state, |
Yao Xiao | bf39e34d | 2022-03-28 21:48:28 | [diff] [blame] | 37 | base::Time browsing_topics_data_accessible_since, |
| 38 | base::RepeatingCallback<bool(const privacy_sandbox::CanonicalTopic&)> |
| 39 | is_topic_allowed_by_settings) { |
| 40 | DCHECK(!is_topic_allowed_by_settings.is_null()); |
| 41 | |
| 42 | if (browsing_topics_state.epochs().empty()) |
| 43 | return false; |
| 44 | |
Yao Xiao | cc37939 | 2022-03-25 21:39:06 | [diff] [blame] | 45 | // Here we rely on the fact that `browsing_topics_data_accessible_since` can |
| 46 | // only be updated to base::Time::Now() due to data deletion. So we'll either |
| 47 | // need to clear all topics data, or no-op. If this assumption no longer |
| 48 | // holds, we'd need to iterate over all epochs, check their calculation time, |
| 49 | // and selectively delete the epochs. |
Yao Xiao | bf39e34d | 2022-03-28 21:48:28 | [diff] [blame] | 50 | if (browsing_topics_data_accessible_since > |
| 51 | browsing_topics_state.epochs().back().calculation_time()) { |
| 52 | return true; |
| 53 | } |
| 54 | |
| 55 | for (const EpochTopics& epoch : browsing_topics_state.epochs()) { |
| 56 | for (const TopicAndDomains& topic_and_domains : |
| 57 | epoch.top_topics_and_observing_domains()) { |
| 58 | if (!topic_and_domains.IsValid()) |
| 59 | continue; |
| 60 | |
| 61 | if (!is_topic_allowed_by_settings.Run(privacy_sandbox::CanonicalTopic( |
| 62 | topic_and_domains.topic(), epoch.taxonomy_version()))) { |
| 63 | return true; |
| 64 | } |
| 65 | } |
| 66 | } |
| 67 | |
| 68 | return false; |
Yao Xiao | cc37939 | 2022-03-25 21:39:06 | [diff] [blame] | 69 | } |
| 70 | |
| 71 | struct StartupCalculateDecision { |
| 72 | bool clear_topics_data = true; |
| 73 | base::TimeDelta next_calculation_delay; |
| 74 | }; |
| 75 | |
| 76 | StartupCalculateDecision GetStartupCalculationDecision( |
| 77 | const BrowsingTopicsState& browsing_topics_state, |
Yao Xiao | bf39e34d | 2022-03-28 21:48:28 | [diff] [blame] | 78 | base::Time browsing_topics_data_accessible_since, |
| 79 | base::RepeatingCallback<bool(const privacy_sandbox::CanonicalTopic&)> |
| 80 | is_topic_allowed_by_settings) { |
Yao Xiao | cc37939 | 2022-03-25 21:39:06 | [diff] [blame] | 81 | // The topics have never been calculated. This could happen with a fresh |
| 82 | // profile or the if the config has updated. In case of a config update, the |
| 83 | // topics should have already been cleared when initializing the |
| 84 | // `BrowsingTopicsState`. |
| 85 | if (browsing_topics_state.next_scheduled_calculation_time().is_null()) { |
| 86 | return StartupCalculateDecision{ |
| 87 | .clear_topics_data = false, |
| 88 | .next_calculation_delay = base::TimeDelta()}; |
| 89 | } |
| 90 | |
| 91 | // This could happen when clear-on-exit is turned on and has caused the |
Yao Xiao | bf39e34d | 2022-03-28 21:48:28 | [diff] [blame] | 92 | // cookies to be deleted on startup, of if a topic became disallowed when |
| 93 | // `browsing_topics_state` was still loading. |
| 94 | bool should_clear_topics_data = ShouldClearTopicsOnStartup( |
| 95 | browsing_topics_state, browsing_topics_data_accessible_since, |
| 96 | is_topic_allowed_by_settings); |
Yao Xiao | cc37939 | 2022-03-25 21:39:06 | [diff] [blame] | 97 | |
| 98 | base::TimeDelta presumed_next_calculation_delay = |
| 99 | browsing_topics_state.next_scheduled_calculation_time() - |
| 100 | base::Time::Now(); |
| 101 | |
| 102 | // The scheduled calculation time was reached before the startup. |
| 103 | if (presumed_next_calculation_delay <= base::TimeDelta()) { |
| 104 | return StartupCalculateDecision{ |
| 105 | .clear_topics_data = should_clear_topics_data, |
| 106 | .next_calculation_delay = base::TimeDelta()}; |
| 107 | } |
| 108 | |
| 109 | // This could happen if the machine time has changed since the last |
| 110 | // calculation. Recalculate immediately to align with the expected schedule |
| 111 | // rather than potentially stop computing for a very long time. |
| 112 | if (presumed_next_calculation_delay >= |
| 113 | 2 * blink::features::kBrowsingTopicsTimePeriodPerEpoch.Get()) { |
| 114 | return StartupCalculateDecision{ |
| 115 | .clear_topics_data = should_clear_topics_data, |
| 116 | .next_calculation_delay = base::TimeDelta()}; |
| 117 | } |
| 118 | |
| 119 | return StartupCalculateDecision{ |
| 120 | .clear_topics_data = should_clear_topics_data, |
| 121 | .next_calculation_delay = presumed_next_calculation_delay}; |
| 122 | } |
| 123 | |
Yao Xiao | 716e481 | 2022-04-20 22:57:19 | [diff] [blame] | 124 | // Represents the different reasons why the topics API returns an empty result. |
| 125 | // These values are persisted to logs. Entries should not be renumbered and |
| 126 | // numeric values should never be reused. |
| 127 | enum class EmptyApiResultReason { |
| 128 | // The topics state hasn't finished loading. |
| 129 | kStateNotReady = 0, |
| 130 | |
| 131 | // Access is disallowed by user settings. |
| 132 | kAccessDisallowedBySettings = 1, |
| 133 | |
| 134 | // There are no candidate topics, e.g. no candidate epochs; epoch calculation |
| 135 | // failed; individual topics were cleared or blocked. |
| 136 | kNoCandicateTopics = 2, |
| 137 | |
| 138 | // The candidate topics were filtered for the requesting context. |
| 139 | kCandicateTopicsFiltered = 3, |
| 140 | |
| 141 | kMaxValue = kCandicateTopicsFiltered, |
| 142 | }; |
| 143 | |
| 144 | void RecordBrowsingTopicsApiResultUkmMetrics( |
| 145 | EmptyApiResultReason empty_reason, |
| 146 | content::RenderFrameHost* main_frame) { |
| 147 | ukm::UkmRecorder* ukm_recorder = ukm::UkmRecorder::Get(); |
| 148 | ukm::builders::BrowsingTopics_DocumentBrowsingTopicsApiResult builder( |
| 149 | main_frame->GetPageUkmSourceId()); |
| 150 | builder.SetEmptyReason(static_cast<int64_t>(empty_reason)); |
| 151 | builder.Record(ukm_recorder->Get()); |
| 152 | } |
| 153 | |
| 154 | void RecordBrowsingTopicsApiResultUkmMetrics( |
| 155 | const std::vector<std::pair<blink::mojom::EpochTopicPtr, bool>>& |
| 156 | topics_with_status, |
| 157 | content::RenderFrameHost* main_frame) { |
| 158 | DCHECK(!topics_with_status.empty()); |
| 159 | |
| 160 | ukm::UkmRecorder* ukm_recorder = ukm::UkmRecorder::Get(); |
| 161 | ukm::builders::BrowsingTopics_DocumentBrowsingTopicsApiResult builder( |
| 162 | main_frame->GetPageUkmSourceId()); |
| 163 | |
| 164 | for (size_t i = 0; i < 3u && topics_with_status.size() > i; ++i) { |
| 165 | const blink::mojom::EpochTopicPtr& topic = topics_with_status[i].first; |
| 166 | bool is_true_topic = topics_with_status[i].second; |
| 167 | |
| 168 | int taxonomy_version = 0; |
| 169 | base::StringToInt(topic->taxonomy_version, &taxonomy_version); |
| 170 | DCHECK(taxonomy_version); |
| 171 | |
| 172 | int64_t model_version = 0; |
| 173 | base::StringToInt64(topic->model_version, &model_version); |
| 174 | DCHECK(model_version); |
| 175 | |
| 176 | if (i == 0) { |
| 177 | builder.SetReturnedTopic0(topic->topic) |
| 178 | .SetReturnedTopic0IsTrueTopTopic(is_true_topic) |
| 179 | .SetReturnedTopic0TaxonomyVersion(taxonomy_version) |
| 180 | .SetReturnedTopic0ModelVersion(model_version); |
| 181 | } else if (i == 1) { |
| 182 | builder.SetReturnedTopic1(topic->topic) |
| 183 | .SetReturnedTopic1IsTrueTopTopic(is_true_topic) |
| 184 | .SetReturnedTopic1TaxonomyVersion(taxonomy_version) |
| 185 | .SetReturnedTopic1ModelVersion(model_version); |
| 186 | } else { |
| 187 | DCHECK_EQ(i, 2u); |
| 188 | builder.SetReturnedTopic2(topic->topic) |
| 189 | .SetReturnedTopic2IsTrueTopTopic(is_true_topic) |
| 190 | .SetReturnedTopic2TaxonomyVersion(taxonomy_version) |
| 191 | .SetReturnedTopic2ModelVersion(model_version); |
| 192 | } |
| 193 | } |
| 194 | |
| 195 | builder.Record(ukm_recorder->Get()); |
| 196 | } |
| 197 | |
Yao Xiao | cc37939 | 2022-03-25 21:39:06 | [diff] [blame] | 198 | } // namespace |
| 199 | |
Yao Xiao | 7a1995b | 2022-03-09 08:18:55 | [diff] [blame] | 200 | BrowsingTopicsServiceImpl::~BrowsingTopicsServiceImpl() = default; |
| 201 | |
Yao Xiao | cc37939 | 2022-03-25 21:39:06 | [diff] [blame] | 202 | BrowsingTopicsServiceImpl::BrowsingTopicsServiceImpl( |
| 203 | const base::FilePath& profile_path, |
| 204 | privacy_sandbox::PrivacySandboxSettings* privacy_sandbox_settings, |
| 205 | history::HistoryService* history_service, |
| 206 | content::BrowsingTopicsSiteDataManager* site_data_manager, |
| 207 | optimization_guide::PageContentAnnotationsService* annotations_service) |
| 208 | : privacy_sandbox_settings_(privacy_sandbox_settings), |
| 209 | history_service_(history_service), |
| 210 | site_data_manager_(site_data_manager), |
| 211 | annotations_service_(annotations_service), |
| 212 | browsing_topics_state_( |
| 213 | profile_path, |
| 214 | base::BindOnce( |
| 215 | &BrowsingTopicsServiceImpl::OnBrowsingTopicsStateLoaded, |
| 216 | base::Unretained(this))) { |
| 217 | privacy_sandbox_settings_observation_.Observe(privacy_sandbox_settings); |
| 218 | history_service_observation_.Observe(history_service); |
| 219 | |
Yao Xiao | bc1241a | 2022-03-29 05:23:37 | [diff] [blame] | 220 | // Greedily request the model to be available to reduce the latency in later |
| 221 | // topics calculation. |
Yao Xiao | cc37939 | 2022-03-25 21:39:06 | [diff] [blame] | 222 | annotations_service_->RequestAndNotifyWhenModelAvailable( |
| 223 | optimization_guide::AnnotationType::kPageTopics, base::DoNothing()); |
| 224 | } |
| 225 | |
| 226 | std::vector<blink::mojom::EpochTopicPtr> |
| 227 | BrowsingTopicsServiceImpl::GetBrowsingTopicsForJsApi( |
| 228 | const url::Origin& context_origin, |
| 229 | content::RenderFrameHost* main_frame) { |
Yao Xiao | 716e481 | 2022-04-20 22:57:19 | [diff] [blame] | 230 | if (!browsing_topics_state_loaded_) { |
| 231 | RecordBrowsingTopicsApiResultUkmMetrics( |
| 232 | EmptyApiResultReason::kStateNotReady, main_frame); |
Yao Xiao | cc37939 | 2022-03-25 21:39:06 | [diff] [blame] | 233 | return {}; |
Yao Xiao | 716e481 | 2022-04-20 22:57:19 | [diff] [blame] | 234 | } |
Yao Xiao | cc37939 | 2022-03-25 21:39:06 | [diff] [blame] | 235 | |
Yao Xiao | 716e481 | 2022-04-20 22:57:19 | [diff] [blame] | 236 | if (!privacy_sandbox_settings_->IsTopicsAllowed()) { |
| 237 | RecordBrowsingTopicsApiResultUkmMetrics( |
| 238 | EmptyApiResultReason::kAccessDisallowedBySettings, main_frame); |
Yao Xiao | cc37939 | 2022-03-25 21:39:06 | [diff] [blame] | 239 | return {}; |
Yao Xiao | 716e481 | 2022-04-20 22:57:19 | [diff] [blame] | 240 | } |
Yao Xiao | cc37939 | 2022-03-25 21:39:06 | [diff] [blame] | 241 | |
| 242 | if (!privacy_sandbox_settings_->IsTopicsAllowedForContext( |
| 243 | context_origin.GetURL(), main_frame->GetLastCommittedOrigin())) { |
Yao Xiao | 716e481 | 2022-04-20 22:57:19 | [diff] [blame] | 244 | RecordBrowsingTopicsApiResultUkmMetrics( |
| 245 | EmptyApiResultReason::kAccessDisallowedBySettings, main_frame); |
Yao Xiao | cc37939 | 2022-03-25 21:39:06 | [diff] [blame] | 246 | return {}; |
| 247 | } |
| 248 | |
| 249 | std::string context_domain = |
| 250 | net::registry_controlled_domains::GetDomainAndRegistry( |
| 251 | context_origin.GetURL(), |
| 252 | net::registry_controlled_domains::INCLUDE_PRIVATE_REGISTRIES); |
| 253 | |
| 254 | HashedDomain hashed_context_domain = HashContextDomainForStorage( |
| 255 | browsing_topics_state_.hmac_key(), context_domain); |
| 256 | |
| 257 | // Track the API usage context after the permissions check. |
| 258 | BrowsingTopicsPageLoadDataTracker::GetOrCreateForPage(main_frame->GetPage()) |
| 259 | ->OnBrowsingTopicsApiUsed(hashed_context_domain, history_service_); |
| 260 | |
| 261 | std::string top_domain = |
| 262 | net::registry_controlled_domains::GetDomainAndRegistry( |
| 263 | main_frame->GetLastCommittedOrigin().GetURL(), |
| 264 | net::registry_controlled_domains::INCLUDE_PRIVATE_REGISTRIES); |
| 265 | |
Yao Xiao | 716e481 | 2022-04-20 22:57:19 | [diff] [blame] | 266 | bool has_filtered_topics = false; |
| 267 | |
| 268 | // The result topics along with flags denoting whether they are true topics. |
| 269 | std::vector<std::pair<blink::mojom::EpochTopicPtr, bool>> topics_with_status; |
| 270 | |
Yao Xiao | cc37939 | 2022-03-25 21:39:06 | [diff] [blame] | 271 | for (const EpochTopics* epoch : |
| 272 | browsing_topics_state_.EpochsForSite(top_domain)) { |
Yao Xiao | 716e481 | 2022-04-20 22:57:19 | [diff] [blame] | 273 | bool output_is_true_topic = false; |
| 274 | bool candidate_topic_filtered = false; |
Yao Xiao | cc37939 | 2022-03-25 21:39:06 | [diff] [blame] | 275 | absl::optional<Topic> topic = epoch->TopicForSite( |
Yao Xiao | 716e481 | 2022-04-20 22:57:19 | [diff] [blame] | 276 | top_domain, hashed_context_domain, browsing_topics_state_.hmac_key(), |
| 277 | output_is_true_topic, candidate_topic_filtered); |
| 278 | |
| 279 | if (candidate_topic_filtered) |
| 280 | has_filtered_topics = true; |
Yao Xiao | cc37939 | 2022-03-25 21:39:06 | [diff] [blame] | 281 | |
| 282 | // Only add a non-empty topic to the result. |
| 283 | if (!topic) |
| 284 | continue; |
| 285 | |
Yao Xiao | bf39e34d | 2022-03-28 21:48:28 | [diff] [blame] | 286 | // Although a top topic can never be in the disallowed state, the returned |
| 287 | // `topic` may be the random one. Thus we still need this check. |
Yao Xiao | cc37939 | 2022-03-25 21:39:06 | [diff] [blame] | 288 | if (!privacy_sandbox_settings_->IsTopicAllowed( |
| 289 | privacy_sandbox::CanonicalTopic(*topic, |
| 290 | epoch->taxonomy_version()))) { |
Yao Xiao | bf39e34d | 2022-03-28 21:48:28 | [diff] [blame] | 291 | continue; |
Yao Xiao | cc37939 | 2022-03-25 21:39:06 | [diff] [blame] | 292 | } |
| 293 | |
Yao Xiao | 21f1faa | 2022-04-29 06:20:32 | [diff] [blame] | 294 | auto result_topic = blink::mojom::EpochTopic::New(); |
Yao Xiao | cc37939 | 2022-03-25 21:39:06 | [diff] [blame] | 295 | result_topic->topic = topic.value().value(); |
| 296 | result_topic->config_version = base::StrCat( |
| 297 | {"chrome.", base::NumberToString( |
| 298 | blink::features::kBrowsingTopicsConfigVersion.Get())}); |
| 299 | result_topic->model_version = base::NumberToString(epoch->model_version()); |
| 300 | result_topic->taxonomy_version = |
| 301 | base::NumberToString(epoch->taxonomy_version()); |
| 302 | result_topic->version = base::StrCat({result_topic->config_version, ":", |
| 303 | result_topic->taxonomy_version, ":", |
| 304 | result_topic->model_version}); |
Yao Xiao | 716e481 | 2022-04-20 22:57:19 | [diff] [blame] | 305 | topics_with_status.emplace_back(std::move(result_topic), |
| 306 | output_is_true_topic); |
Yao Xiao | cc37939 | 2022-03-25 21:39:06 | [diff] [blame] | 307 | } |
| 308 | |
Yao Xiao | 716e481 | 2022-04-20 22:57:19 | [diff] [blame] | 309 | // Sort `topics_with_status` based on `EpochTopicPtr` first, and if the |
| 310 | // `EpochTopicPtr` parts are equal, then a true topic will be ordered before a |
| 311 | // random topic. This ensures that when we later deduplicate based on the |
| 312 | // `EpochTopicPtr` field only, the associated is-true-topic status will be |
| 313 | // true as long as there is one true topic for that topic in |
| 314 | // `topics_with_status`. |
| 315 | std::sort(topics_with_status.begin(), topics_with_status.end(), |
| 316 | [](const auto& left, const auto& right) { |
| 317 | if (left.first < right.first) |
| 318 | return true; |
| 319 | if (left.first > right.first) |
| 320 | return false; |
| 321 | return right.second < left.second; |
| 322 | }); |
| 323 | |
| 324 | // Remove duplicate `EpochTopicPtr` entries. |
| 325 | topics_with_status.erase( |
| 326 | std::unique(topics_with_status.begin(), topics_with_status.end(), |
| 327 | [](const auto& left, const auto& right) { |
| 328 | return left.first == right.first; |
| 329 | }), |
| 330 | topics_with_status.end()); |
Yao Xiao | cc37939 | 2022-03-25 21:39:06 | [diff] [blame] | 331 | |
| 332 | // Shuffle the entries. |
Yao Xiao | 716e481 | 2022-04-20 22:57:19 | [diff] [blame] | 333 | base::RandomShuffle(topics_with_status.begin(), topics_with_status.end()); |
| 334 | |
| 335 | if (topics_with_status.empty()) { |
| 336 | if (has_filtered_topics) { |
| 337 | RecordBrowsingTopicsApiResultUkmMetrics( |
| 338 | EmptyApiResultReason::kCandicateTopicsFiltered, main_frame); |
| 339 | } else { |
| 340 | RecordBrowsingTopicsApiResultUkmMetrics( |
| 341 | EmptyApiResultReason::kNoCandicateTopics, main_frame); |
| 342 | } |
| 343 | return {}; |
| 344 | } |
| 345 | |
| 346 | RecordBrowsingTopicsApiResultUkmMetrics(topics_with_status, main_frame); |
| 347 | |
| 348 | std::vector<blink::mojom::EpochTopicPtr> result_topics; |
| 349 | result_topics.reserve(topics_with_status.size()); |
| 350 | std::transform(topics_with_status.begin(), topics_with_status.end(), |
| 351 | std::back_inserter(result_topics), |
| 352 | [](auto& topic_with_status) { |
| 353 | return std::move(topic_with_status.first); |
| 354 | }); |
Yao Xiao | cc37939 | 2022-03-25 21:39:06 | [diff] [blame] | 355 | |
| 356 | return result_topics; |
| 357 | } |
Yao Xiao | 7a1995b | 2022-03-09 08:18:55 | [diff] [blame] | 358 | |
Yao Xiao | 57892a2 | 2022-06-28 19:21:41 | [diff] [blame^] | 359 | void BrowsingTopicsServiceImpl::GetBrowsingTopicsStateForWebUi( |
| 360 | bool calculate_now, |
| 361 | mojom::PageHandler::GetBrowsingTopicsStateCallback callback) { |
Yao Xiao | 21f1faa | 2022-04-29 06:20:32 | [diff] [blame] | 362 | if (!browsing_topics_state_loaded_) { |
Yao Xiao | 57892a2 | 2022-06-28 19:21:41 | [diff] [blame^] | 363 | std::move(callback).Run( |
| 364 | mojom::WebUIGetBrowsingTopicsStateResult::NewOverrideStatusMessage( |
| 365 | "State loading hasn't finished. Please retry shortly.")); |
| 366 | return; |
Yao Xiao | 21f1faa | 2022-04-29 06:20:32 | [diff] [blame] | 367 | } |
| 368 | |
Yao Xiao | 57892a2 | 2022-06-28 19:21:41 | [diff] [blame^] | 369 | // If a calculation is already in progress, get the webui topics state after |
| 370 | // the calculation is done. Do this regardless of whether `calculate_now` is |
| 371 | // true, i.e. if `calculate_now` is true, this request is effectively merged |
| 372 | // with the in progress calculation. |
| 373 | if (topics_calculator_) { |
| 374 | get_state_for_webui_callbacks_.push_back(std::move(callback)); |
| 375 | return; |
Yao Xiao | 21f1faa | 2022-04-29 06:20:32 | [diff] [blame] | 376 | } |
| 377 | |
Yao Xiao | 57892a2 | 2022-06-28 19:21:41 | [diff] [blame^] | 378 | DCHECK(schedule_calculate_timer_.IsRunning()); |
Yao Xiao | 21f1faa | 2022-04-29 06:20:32 | [diff] [blame] | 379 | |
Yao Xiao | 57892a2 | 2022-06-28 19:21:41 | [diff] [blame^] | 380 | if (calculate_now) { |
| 381 | get_state_for_webui_callbacks_.push_back(std::move(callback)); |
Yao Xiao | 21f1faa | 2022-04-29 06:20:32 | [diff] [blame] | 382 | |
Yao Xiao | 57892a2 | 2022-06-28 19:21:41 | [diff] [blame^] | 383 | schedule_calculate_timer_.AbandonAndStop(); |
| 384 | CalculateBrowsingTopics(); |
| 385 | return; |
Yao Xiao | 21f1faa | 2022-04-29 06:20:32 | [diff] [blame] | 386 | } |
| 387 | |
Yao Xiao | 57892a2 | 2022-06-28 19:21:41 | [diff] [blame^] | 388 | std::move(callback).Run(GetBrowsingTopicsStateForWebUiHelper()); |
Yao Xiao | 21f1faa | 2022-04-29 06:20:32 | [diff] [blame] | 389 | } |
| 390 | |
Yao Xiao | 7a1995b | 2022-03-09 08:18:55 | [diff] [blame] | 391 | std::vector<privacy_sandbox::CanonicalTopic> |
| 392 | BrowsingTopicsServiceImpl::GetTopicsForSiteForDisplay( |
| 393 | const url::Origin& top_origin) const { |
Yao Xiao | cc37939 | 2022-03-25 21:39:06 | [diff] [blame] | 394 | if (!browsing_topics_state_loaded_) |
| 395 | return {}; |
| 396 | |
| 397 | std::string top_domain = |
| 398 | net::registry_controlled_domains::GetDomainAndRegistry( |
| 399 | top_origin.GetURL(), |
| 400 | net::registry_controlled_domains::INCLUDE_PRIVATE_REGISTRIES); |
| 401 | |
| 402 | std::vector<privacy_sandbox::CanonicalTopic> result; |
| 403 | |
| 404 | for (const EpochTopics* epoch : |
| 405 | browsing_topics_state_.EpochsForSite(top_domain)) { |
Yao Xiao | bf39e34d | 2022-03-28 21:48:28 | [diff] [blame] | 406 | absl::optional<Topic> topic = epoch->TopicForSiteForDisplay( |
Yao Xiao | cc37939 | 2022-03-25 21:39:06 | [diff] [blame] | 407 | top_domain, browsing_topics_state_.hmac_key()); |
| 408 | |
| 409 | if (!topic) |
| 410 | continue; |
| 411 | |
Yao Xiao | bf39e34d | 2022-03-28 21:48:28 | [diff] [blame] | 412 | // `epoch->TopicForSiteForDisplay()` shall only return a top topic, and a |
| 413 | // top topic can never be in the disallowed state (i.e. it will be cleared |
| 414 | // when it becomes diallowed). |
| 415 | DCHECK(privacy_sandbox_settings_->IsTopicAllowed( |
| 416 | privacy_sandbox::CanonicalTopic(*topic, epoch->taxonomy_version()))); |
| 417 | |
Yao Xiao | cc37939 | 2022-03-25 21:39:06 | [diff] [blame] | 418 | result.emplace_back(*topic, epoch->taxonomy_version()); |
| 419 | } |
| 420 | |
| 421 | return result; |
Yao Xiao | 7a1995b | 2022-03-09 08:18:55 | [diff] [blame] | 422 | } |
| 423 | |
| 424 | std::vector<privacy_sandbox::CanonicalTopic> |
| 425 | BrowsingTopicsServiceImpl::GetTopTopicsForDisplay() const { |
Yao Xiao | cc37939 | 2022-03-25 21:39:06 | [diff] [blame] | 426 | if (!browsing_topics_state_loaded_) |
| 427 | return {}; |
| 428 | |
| 429 | std::vector<privacy_sandbox::CanonicalTopic> result; |
| 430 | |
| 431 | for (const EpochTopics& epoch : browsing_topics_state_.epochs()) { |
Yao Xiao | bf39e34d | 2022-03-28 21:48:28 | [diff] [blame] | 432 | DCHECK_LE(epoch.padded_top_topics_start_index(), |
| 433 | epoch.top_topics_and_observing_domains().size()); |
| 434 | |
| 435 | for (size_t i = 0; i < epoch.padded_top_topics_start_index(); ++i) { |
| 436 | const TopicAndDomains& topic_and_domains = |
| 437 | epoch.top_topics_and_observing_domains()[i]; |
| 438 | |
| 439 | if (!topic_and_domains.IsValid()) |
Yao Xiao | cc37939 | 2022-03-25 21:39:06 | [diff] [blame] | 440 | continue; |
| 441 | |
Yao Xiao | bf39e34d | 2022-03-28 21:48:28 | [diff] [blame] | 442 | // A top topic can never be in the disallowed state (i.e. it will be |
| 443 | // cleared when it becomes diallowed). |
| 444 | DCHECK(privacy_sandbox_settings_->IsTopicAllowed( |
| 445 | privacy_sandbox::CanonicalTopic(topic_and_domains.topic(), |
| 446 | epoch.taxonomy_version()))); |
| 447 | |
| 448 | result.emplace_back(topic_and_domains.topic(), epoch.taxonomy_version()); |
Yao Xiao | cc37939 | 2022-03-25 21:39:06 | [diff] [blame] | 449 | } |
| 450 | } |
| 451 | |
| 452 | return result; |
| 453 | } |
| 454 | |
Yao Xiao | bf39e34d | 2022-03-28 21:48:28 | [diff] [blame] | 455 | void BrowsingTopicsServiceImpl::ClearTopic( |
| 456 | const privacy_sandbox::CanonicalTopic& canonical_topic) { |
| 457 | if (!browsing_topics_state_loaded_) |
| 458 | return; |
| 459 | |
| 460 | browsing_topics_state_.ClearTopic(canonical_topic.topic_id(), |
| 461 | canonical_topic.taxonomy_version()); |
| 462 | } |
| 463 | |
| 464 | void BrowsingTopicsServiceImpl::ClearTopicsDataForOrigin( |
| 465 | const url::Origin& origin) { |
| 466 | if (!browsing_topics_state_loaded_) |
| 467 | return; |
| 468 | |
| 469 | std::string context_domain = |
| 470 | net::registry_controlled_domains::GetDomainAndRegistry( |
| 471 | origin.GetURL(), |
| 472 | net::registry_controlled_domains::INCLUDE_PRIVATE_REGISTRIES); |
| 473 | |
| 474 | HashedDomain hashed_context_domain = HashContextDomainForStorage( |
| 475 | browsing_topics_state_.hmac_key(), context_domain); |
| 476 | |
| 477 | browsing_topics_state_.ClearContextDomain(hashed_context_domain); |
| 478 | site_data_manager_->ClearContextDomain(hashed_context_domain); |
| 479 | } |
| 480 | |
| 481 | void BrowsingTopicsServiceImpl::ClearAllTopicsData() { |
| 482 | if (!browsing_topics_state_loaded_) |
| 483 | return; |
| 484 | |
| 485 | browsing_topics_state_.ClearAllTopics(); |
| 486 | site_data_manager_->ExpireDataBefore(base::Time::Now()); |
| 487 | } |
| 488 | |
Yao Xiao | cc37939 | 2022-03-25 21:39:06 | [diff] [blame] | 489 | std::unique_ptr<BrowsingTopicsCalculator> |
| 490 | BrowsingTopicsServiceImpl::CreateCalculator( |
| 491 | privacy_sandbox::PrivacySandboxSettings* privacy_sandbox_settings, |
| 492 | history::HistoryService* history_service, |
| 493 | content::BrowsingTopicsSiteDataManager* site_data_manager, |
| 494 | optimization_guide::PageContentAnnotationsService* annotations_service, |
Yao Xiao | 57892a2 | 2022-06-28 19:21:41 | [diff] [blame^] | 495 | const base::circular_deque<EpochTopics>& epochs, |
Yao Xiao | cc37939 | 2022-03-25 21:39:06 | [diff] [blame] | 496 | BrowsingTopicsCalculator::CalculateCompletedCallback callback) { |
| 497 | return std::make_unique<BrowsingTopicsCalculator>( |
| 498 | privacy_sandbox_settings, history_service, site_data_manager, |
Yao Xiao | 57892a2 | 2022-06-28 19:21:41 | [diff] [blame^] | 499 | annotations_service, epochs, std::move(callback)); |
Yao Xiao | cc37939 | 2022-03-25 21:39:06 | [diff] [blame] | 500 | } |
| 501 | |
| 502 | const BrowsingTopicsState& BrowsingTopicsServiceImpl::browsing_topics_state() { |
| 503 | return browsing_topics_state_; |
| 504 | } |
| 505 | |
| 506 | void BrowsingTopicsServiceImpl::ScheduleBrowsingTopicsCalculation( |
| 507 | base::TimeDelta delay) { |
| 508 | DCHECK(browsing_topics_state_loaded_); |
| 509 | |
| 510 | // `this` owns the timer, which is automatically cancelled on destruction, so |
| 511 | // base::Unretained(this) is safe. |
| 512 | schedule_calculate_timer_.Start( |
| 513 | FROM_HERE, delay, |
| 514 | base::BindOnce(&BrowsingTopicsServiceImpl::CalculateBrowsingTopics, |
| 515 | base::Unretained(this))); |
| 516 | } |
| 517 | |
| 518 | void BrowsingTopicsServiceImpl::CalculateBrowsingTopics() { |
| 519 | DCHECK(browsing_topics_state_loaded_); |
| 520 | |
| 521 | DCHECK(!topics_calculator_); |
| 522 | |
| 523 | // `this` owns `topics_calculator_` so `topics_calculator_` should not invoke |
| 524 | // the callback once it's destroyed. |
| 525 | topics_calculator_ = CreateCalculator( |
| 526 | privacy_sandbox_settings_, history_service_, site_data_manager_, |
Yao Xiao | 57892a2 | 2022-06-28 19:21:41 | [diff] [blame^] | 527 | annotations_service_, browsing_topics_state_.epochs(), |
Yao Xiao | cc37939 | 2022-03-25 21:39:06 | [diff] [blame] | 528 | base::BindOnce( |
| 529 | &BrowsingTopicsServiceImpl::OnCalculateBrowsingTopicsCompleted, |
| 530 | base::Unretained(this))); |
| 531 | } |
| 532 | |
| 533 | void BrowsingTopicsServiceImpl::OnCalculateBrowsingTopicsCompleted( |
| 534 | EpochTopics epoch_topics) { |
| 535 | DCHECK(browsing_topics_state_loaded_); |
| 536 | |
| 537 | DCHECK(topics_calculator_); |
| 538 | topics_calculator_.reset(); |
| 539 | |
| 540 | browsing_topics_state_.AddEpoch(std::move(epoch_topics)); |
| 541 | browsing_topics_state_.UpdateNextScheduledCalculationTime(); |
| 542 | |
| 543 | ScheduleBrowsingTopicsCalculation( |
| 544 | blink::features::kBrowsingTopicsTimePeriodPerEpoch.Get()); |
Yao Xiao | 57892a2 | 2022-06-28 19:21:41 | [diff] [blame^] | 545 | |
| 546 | if (!get_state_for_webui_callbacks_.empty()) { |
| 547 | mojom::WebUIGetBrowsingTopicsStateResultPtr webui_state = |
| 548 | GetBrowsingTopicsStateForWebUiHelper(); |
| 549 | |
| 550 | for (auto& callback : get_state_for_webui_callbacks_) { |
| 551 | std::move(callback).Run(webui_state->Clone()); |
| 552 | } |
| 553 | |
| 554 | get_state_for_webui_callbacks_.clear(); |
| 555 | } |
Yao Xiao | cc37939 | 2022-03-25 21:39:06 | [diff] [blame] | 556 | } |
| 557 | |
| 558 | void BrowsingTopicsServiceImpl::OnBrowsingTopicsStateLoaded() { |
| 559 | DCHECK(!browsing_topics_state_loaded_); |
| 560 | browsing_topics_state_loaded_ = true; |
| 561 | |
| 562 | base::Time browsing_topics_data_sccessible_since = |
| 563 | privacy_sandbox_settings_->TopicsDataAccessibleSince(); |
| 564 | |
| 565 | StartupCalculateDecision decision = GetStartupCalculationDecision( |
Yao Xiao | bf39e34d | 2022-03-28 21:48:28 | [diff] [blame] | 566 | browsing_topics_state_, browsing_topics_data_sccessible_since, |
| 567 | base::BindRepeating( |
| 568 | &privacy_sandbox::PrivacySandboxSettings::IsTopicAllowed, |
| 569 | base::Unretained(privacy_sandbox_settings_))); |
Yao Xiao | cc37939 | 2022-03-25 21:39:06 | [diff] [blame] | 570 | |
| 571 | if (decision.clear_topics_data) |
| 572 | browsing_topics_state_.ClearAllTopics(); |
| 573 | |
| 574 | site_data_manager_->ExpireDataBefore(browsing_topics_data_sccessible_since); |
| 575 | |
| 576 | ScheduleBrowsingTopicsCalculation(decision.next_calculation_delay); |
| 577 | } |
| 578 | |
| 579 | void BrowsingTopicsServiceImpl::Shutdown() { |
| 580 | privacy_sandbox_settings_observation_.Reset(); |
| 581 | history_service_observation_.Reset(); |
| 582 | } |
| 583 | |
| 584 | void BrowsingTopicsServiceImpl::OnTopicsDataAccessibleSinceUpdated() { |
| 585 | if (!browsing_topics_state_loaded_) |
| 586 | return; |
| 587 | |
Yao Xiao | bf39e34d | 2022-03-28 21:48:28 | [diff] [blame] | 588 | // Here we rely on the fact that `browsing_topics_data_accessible_since` can |
| 589 | // only be updated to base::Time::Now() due to data deletion. In this case, we |
| 590 | // should just clear all topics. |
| 591 | browsing_topics_state_.ClearAllTopics(); |
| 592 | site_data_manager_->ExpireDataBefore( |
| 593 | privacy_sandbox_settings_->TopicsDataAccessibleSince()); |
Yao Xiao | cc37939 | 2022-03-25 21:39:06 | [diff] [blame] | 594 | |
| 595 | // Abort the outstanding topics calculation and restart immediately. |
| 596 | if (topics_calculator_) { |
| 597 | DCHECK(!schedule_calculate_timer_.IsRunning()); |
| 598 | |
| 599 | topics_calculator_.reset(); |
| 600 | CalculateBrowsingTopics(); |
| 601 | } |
| 602 | } |
| 603 | |
| 604 | void BrowsingTopicsServiceImpl::OnURLsDeleted( |
| 605 | history::HistoryService* history_service, |
| 606 | const history::DeletionInfo& deletion_info) { |
| 607 | if (!browsing_topics_state_loaded_) |
| 608 | return; |
| 609 | |
| 610 | // Ignore invalid time_range. |
| 611 | if (!deletion_info.IsAllHistory() && !deletion_info.time_range().IsValid()) |
| 612 | return; |
| 613 | |
| 614 | for (size_t i = 0; i < browsing_topics_state_.epochs().size(); ++i) { |
| 615 | const EpochTopics& epoch_topics = browsing_topics_state_.epochs()[i]; |
| 616 | |
| 617 | if (epoch_topics.empty()) |
| 618 | continue; |
| 619 | |
Yao Xiao | 57892a2 | 2022-06-28 19:21:41 | [diff] [blame^] | 620 | // The typical case is assumed here. We cannot always derive the original |
| 621 | // history start time, as the necessary data (e.g. its previous epoch's |
| 622 | // calculation time) may have been gone. |
| 623 | base::Time history_data_start_time = |
| 624 | epoch_topics.calculation_time() - |
| 625 | blink::features::kBrowsingTopicsTimePeriodPerEpoch.Get(); |
| 626 | |
Yao Xiao | cc37939 | 2022-03-25 21:39:06 | [diff] [blame] | 627 | bool time_range_overlap = |
| 628 | epoch_topics.calculation_time() >= deletion_info.time_range().begin() && |
Yao Xiao | 57892a2 | 2022-06-28 19:21:41 | [diff] [blame^] | 629 | history_data_start_time <= deletion_info.time_range().end(); |
Yao Xiao | cc37939 | 2022-03-25 21:39:06 | [diff] [blame] | 630 | |
| 631 | if (time_range_overlap) |
| 632 | browsing_topics_state_.ClearOneEpoch(i); |
| 633 | } |
| 634 | |
| 635 | // If there's an outstanding topics calculation, abort and restart it. |
| 636 | if (topics_calculator_) { |
| 637 | DCHECK(!schedule_calculate_timer_.IsRunning()); |
| 638 | |
| 639 | topics_calculator_.reset(); |
| 640 | CalculateBrowsingTopics(); |
| 641 | } |
Yao Xiao | 7a1995b | 2022-03-09 08:18:55 | [diff] [blame] | 642 | } |
| 643 | |
Yao Xiao | 57892a2 | 2022-06-28 19:21:41 | [diff] [blame^] | 644 | mojom::WebUIGetBrowsingTopicsStateResultPtr |
| 645 | BrowsingTopicsServiceImpl::GetBrowsingTopicsStateForWebUiHelper() { |
| 646 | DCHECK(browsing_topics_state_loaded_); |
| 647 | DCHECK(!topics_calculator_); |
| 648 | |
| 649 | auto webui_state = mojom::WebUIBrowsingTopicsState::New(); |
| 650 | |
| 651 | webui_state->next_scheduled_calculation_time = |
| 652 | browsing_topics_state_.next_scheduled_calculation_time(); |
| 653 | |
| 654 | for (const EpochTopics& epoch : browsing_topics_state_.epochs()) { |
| 655 | DCHECK_LE(epoch.padded_top_topics_start_index(), |
| 656 | epoch.top_topics_and_observing_domains().size()); |
| 657 | |
| 658 | // Note: for a failed epoch calculation, the default zero-initialized values |
| 659 | // will be displayed in the Web UI. |
| 660 | auto webui_epoch = mojom::WebUIEpoch::New(); |
| 661 | webui_epoch->calculation_time = epoch.calculation_time(); |
| 662 | webui_epoch->model_version = base::NumberToString(epoch.model_version()); |
| 663 | webui_epoch->taxonomy_version = |
| 664 | base::NumberToString(epoch.taxonomy_version()); |
| 665 | |
| 666 | for (size_t i = 0; i < epoch.top_topics_and_observing_domains().size(); |
| 667 | ++i) { |
| 668 | const TopicAndDomains& topic_and_domains = |
| 669 | epoch.top_topics_and_observing_domains()[i]; |
| 670 | |
| 671 | privacy_sandbox::CanonicalTopic canonical_topic = |
| 672 | privacy_sandbox::CanonicalTopic(topic_and_domains.topic(), |
| 673 | epoch.taxonomy_version()); |
| 674 | |
| 675 | std::vector<std::string> webui_observed_by_domains; |
| 676 | webui_observed_by_domains.reserve( |
| 677 | topic_and_domains.hashed_domains().size()); |
| 678 | for (const auto& domain : topic_and_domains.hashed_domains()) { |
| 679 | webui_observed_by_domains.push_back( |
| 680 | base::NumberToString(domain.value())); |
| 681 | } |
| 682 | |
| 683 | // Note: if the topic is invalid (i.e. cleared), the output `topic_id` |
| 684 | // will be 0; if the topic is invalid, or if the taxonomy version isn't |
| 685 | // recognized by this Chrome binary, the output `topic_name` will be |
| 686 | // "Unknown". |
| 687 | auto webui_topic = mojom::WebUITopic::New(); |
| 688 | webui_topic->topic_id = topic_and_domains.topic().value(); |
| 689 | webui_topic->topic_name = canonical_topic.GetLocalizedRepresentation(); |
| 690 | webui_topic->is_real_topic = (i < epoch.padded_top_topics_start_index()); |
| 691 | webui_topic->observed_by_domains = std::move(webui_observed_by_domains); |
| 692 | |
| 693 | webui_epoch->topics.push_back(std::move(webui_topic)); |
| 694 | } |
| 695 | |
| 696 | webui_state->epochs.push_back(std::move(webui_epoch)); |
| 697 | } |
| 698 | |
| 699 | // Reorder the epochs from latest to oldest. |
| 700 | base::ranges::reverse(webui_state->epochs); |
| 701 | |
| 702 | return mojom::WebUIGetBrowsingTopicsStateResult::NewBrowsingTopicsState( |
| 703 | std::move(webui_state)); |
| 704 | } |
| 705 | |
Yao Xiao | 7a1995b | 2022-03-09 08:18:55 | [diff] [blame] | 706 | } // namespace browsing_topics |