components/dom_distiller/standalone/content_extractor_browsertest.cc

   1 // Copyright 2014 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 <sstream>
   6
   7 #include "base/command_line.h"
   8 #include "base/files/scoped_temp_dir.h"
   9 #include "base/id_map.h"
  10 #include "base/location.h"
  11 #include "base/path_service.h"
  12 #include "base/run_loop.h"
  13 #include "base/single_thread_task_runner.h"
  14 #include "base/strings/string_number_conversions.h"
  15 #include "base/strings/string_split.h"
  16 #include "base/thread_task_runner_handle.h"
  17 #include "components/dom_distiller/content/distiller_javascript_utils.h"
  18 #include "components/dom_distiller/content/distiller_page_web_contents.h"
  19 #include "components/dom_distiller/core/article_entry.h"
  20 #include "components/dom_distiller/core/distilled_page_prefs.h"
  21 #include "components/dom_distiller/core/distiller.h"
  22 #include "components/dom_distiller/core/dom_distiller_service.h"
  23 #include "components/dom_distiller/core/dom_distiller_store.h"
  24 #include "components/dom_distiller/core/proto/distilled_article.pb.h"
  25 #include "components/dom_distiller/core/proto/distilled_page.pb.h"
  26 #include "components/dom_distiller/core/task_tracker.h"
  27 #include "components/leveldb_proto/proto_database.h"
  28 #include "components/leveldb_proto/proto_database_impl.h"
  29 #include "components/pref_registry/testing_pref_service_syncable.h"
  30 #include "content/public/browser/browser_context.h"
  31 #include "content/public/browser/browser_thread.h"
  32 #include "content/public/common/isolated_world_ids.h"
  33 #include "content/public/test/content_browser_test.h"
  34 #include "content/shell/browser/shell.h"
  35 #include "google/protobuf/io/coded_stream.h"
  36 #include "google/protobuf/io/zero_copy_stream_impl_lite.h"
  37 #include "net/dns/mock_host_resolver.h"
  38 #include "third_party/dom_distiller_js/dom_distiller.pb.h"
  39 #include "ui/base/resource/resource_bundle.h"
  40
  41 using content::ContentBrowserTest;
  42
  43 namespace dom_distiller {
  44
  45 namespace {
  46
  47 typedef base::hash_map<std::string, std::string> FileToUrlMap;
  48
  49 }
  50
  51 // Factory for creating a Distiller that creates different DomDistillerOptions
  52 // for different URLs, i.e. a specific kOriginalUrl option for each URL.
  53 class TestDistillerFactoryImpl : public DistillerFactory {
  54  public:
  55   TestDistillerFactoryImpl(
  56       scoped_ptr<DistillerURLFetcherFactory> distiller_url_fetcher_factory,
  57       const dom_distiller::proto::DomDistillerOptions& dom_distiller_options,
  58       const FileToUrlMap& file_to_url_map)
  59       : distiller_url_fetcher_factory_(distiller_url_fetcher_factory.Pass()),
  60         dom_distiller_options_(dom_distiller_options),
  61         file_to_url_map_(file_to_url_map) {
  62   }
  63
  64   ~TestDistillerFactoryImpl() override {}
  65
  66   scoped_ptr<Distiller> CreateDistillerForUrl(const GURL& url) override {
  67     dom_distiller::proto::DomDistillerOptions options;
  68     options = dom_distiller_options_;
  69     FileToUrlMap::const_iterator it = file_to_url_map_.find(url.spec());
  70     if (it != file_to_url_map_.end()) {
  71       options.set_original_url(it->second);
  72     }
  73     scoped_ptr<DistillerImpl> distiller(new DistillerImpl(
  74         *distiller_url_fetcher_factory_, options));
  75     return distiller.Pass();
  76   }
  77
  78  private:
  79   scoped_ptr<DistillerURLFetcherFactory> distiller_url_fetcher_factory_;
  80   dom_distiller::proto::DomDistillerOptions dom_distiller_options_;
  81   FileToUrlMap file_to_url_map_;
  82 };
  83
  84 namespace {
  85
  86 // The url to distill.
  87 const char* kUrlSwitch = "url";
  88
  89 // A space-separated list of urls to distill.
  90 const char* kUrlsSwitch = "urls";
  91
  92 // Indicates that DNS resolution should be disabled for this test.
  93 const char* kDisableDnsSwitch = "disable-dns";
  94
  95 // Will write the distilled output to the given file instead of to stdout.
  96 const char* kOutputFile = "output-file";
  97
  98 // Indicates to output a serialized protocol buffer instead of human-readable
  99 // output.
 100 const char* kShouldOutputBinary = "output-binary";
 101
 102 // Indicates to output only the text of the article and not the enclosing html.
 103 const char* kExtractTextOnly = "extract-text-only";
 104
 105 // Indicates to include debug output.
 106 const char* kDebugLevel = "debug-level";
 107
 108 // The original URL of the page if |kUrlSwitch| is a file.
 109 const char* kOriginalUrl = "original-url";
 110
 111 // A semi-colon-separated (i.e. ';') list of original URLs corresponding to
 112 // "kUrlsSwitch".
 113 const char* kOriginalUrls = "original-urls";
 114
 115 // Maximum number of concurrent started extractor requests.
 116 const int kMaxExtractorTasks = 8;
 117
 118 scoped_ptr<DomDistillerService> CreateDomDistillerService(
 119     content::BrowserContext* context,
 120     const base::FilePath& db_path,
 121     const FileToUrlMap& file_to_url_map) {
 122   scoped_refptr<base::SequencedTaskRunner> background_task_runner =
 123       content::BrowserThread::GetBlockingPool()->GetSequencedTaskRunner(
 124           content::BrowserThread::GetBlockingPool()->GetSequenceToken());
 125
 126   // TODO(cjhopman): use an in-memory database instead of an on-disk one with
 127   // temporary directory.
 128   scoped_ptr<leveldb_proto::ProtoDatabaseImpl<ArticleEntry> > db(
 129       new leveldb_proto::ProtoDatabaseImpl<ArticleEntry>(
 130           background_task_runner));
 131   scoped_ptr<DomDistillerStore> dom_distiller_store(
 132       new DomDistillerStore(db.Pass(), db_path));
 133
 134   scoped_ptr<DistillerPageFactory> distiller_page_factory(
 135       new DistillerPageWebContentsFactory(context));
 136   scoped_ptr<DistillerURLFetcherFactory> distiller_url_fetcher_factory(
 137       new DistillerURLFetcherFactory(context->GetRequestContext()));
 138
 139   dom_distiller::proto::DomDistillerOptions options;
 140   if (base::CommandLine::ForCurrentProcess()->HasSwitch(kExtractTextOnly)) {
 141     options.set_extract_text_only(true);
 142   }
 143   int debug_level = 0;
 144   if (base::CommandLine::ForCurrentProcess()->HasSwitch(kDebugLevel) &&
 145       base::StringToInt(
 146           base::CommandLine::ForCurrentProcess()->GetSwitchValueASCII(
 147               kDebugLevel),
 148           &debug_level)) {
 149     options.set_debug_level(debug_level);
 150   }
 151   scoped_ptr<DistillerFactory> distiller_factory(
 152       new TestDistillerFactoryImpl(distiller_url_fetcher_factory.Pass(),
 153                                    options,
 154                                    file_to_url_map));
 155
 156   // Setting up PrefService for DistilledPagePrefs.
 157   user_prefs::TestingPrefServiceSyncable* pref_service =
 158       new user_prefs::TestingPrefServiceSyncable();
 159   DistilledPagePrefs::RegisterProfilePrefs(pref_service->registry());
 160
 161   return scoped_ptr<DomDistillerService>(new DomDistillerService(
 162       dom_distiller_store.Pass(),
 163       distiller_factory.Pass(),
 164       distiller_page_factory.Pass(),
 165       scoped_ptr<DistilledPagePrefs>(new DistilledPagePrefs(pref_service))));
 166 }
 167
 168 void AddComponentsTestResources() {
 169   base::FilePath pak_file;
 170   base::FilePath pak_dir;
 171   PathService::Get(base::DIR_MODULE, &pak_dir);
 172   pak_file =
 173       pak_dir.Append(FILE_PATH_LITERAL("components_tests_resources.pak"));
 174   ui::ResourceBundle::GetSharedInstance().AddDataPackFromPath(
 175       pak_file, ui::SCALE_FACTOR_NONE);
 176 }
 177
 178 bool WriteProtobufWithSize(
 179     const google::protobuf::MessageLite& message,
 180     google::protobuf::io::ZeroCopyOutputStream* output_stream) {
 181   google::protobuf::io::CodedOutputStream coded_output(output_stream);
 182
 183   // Write the size.
 184   const int size = message.ByteSize();
 185   coded_output.WriteLittleEndian32(size);
 186   message.SerializeWithCachedSizes(&coded_output);
 187   return !coded_output.HadError();
 188 }
 189
 190 std::string GetReadableArticleString(
 191     const DistilledArticleProto& article_proto) {
 192   std::stringstream output;
 193   output << "Article Title: " << article_proto.title() << std::endl;
 194   output << "# of pages: " << article_proto.pages_size() << std::endl;
 195   for (int i = 0; i < article_proto.pages_size(); ++i) {
 196     if (i > 0) output << std::endl;
 197     const DistilledPageProto& page = article_proto.pages(i);
 198     output << "Page " << i << std::endl;
 199     output << "URL: " << page.url() << std::endl;
 200     output << "Content: " << page.html() << std::endl;
 201     if (page.has_debug_info() && page.debug_info().has_log())
 202       output << "Log: " << page.debug_info().log() << std::endl;
 203     if (page.has_pagination_info()) {
 204       if (page.pagination_info().has_next_page()) {
 205         output << "Next Page: " << page.pagination_info().next_page()
 206                << std::endl;
 207       }
 208       if (page.pagination_info().has_prev_page()) {
 209         output << "Prev Page: " << page.pagination_info().prev_page()
 210                << std::endl;
 211       }
 212     }
 213   }
 214   return output.str();
 215 }
 216
 217 }  // namespace
 218
 219 class ContentExtractionRequest : public ViewRequestDelegate {
 220  public:
 221   void Start(DomDistillerService* service, const gfx::Size& render_view_size,
 222              base::Closure finished_callback) {
 223     finished_callback_ = finished_callback;
 224     viewer_handle_ =
 225         service->ViewUrl(this,
 226                          service->CreateDefaultDistillerPage(render_view_size),
 227                          url_);
 228   }
 229
 230   DistilledArticleProto GetArticleCopy() {
 231     return *article_proto_;
 232   }
 233
 234   static ScopedVector<ContentExtractionRequest> CreateForCommandLine(
 235       const base::CommandLine& command_line,
 236       FileToUrlMap* file_to_url_map) {
 237     ScopedVector<ContentExtractionRequest> requests;
 238     if (command_line.HasSwitch(kUrlSwitch)) {
 239       GURL url;
 240       std::string url_string = command_line.GetSwitchValueASCII(kUrlSwitch);
 241       url = GURL(url_string);
 242       if (url.is_valid()) {
 243         requests.push_back(new ContentExtractionRequest(url));
 244         if (command_line.HasSwitch(kOriginalUrl)) {
 245           (*file_to_url_map)[url.spec()] =
 246               command_line.GetSwitchValueASCII(kOriginalUrl);
 247         }
 248       }
 249     } else if (command_line.HasSwitch(kUrlsSwitch)) {
 250       std::string urls_string = command_line.GetSwitchValueASCII(kUrlsSwitch);
 251       std::vector<std::string> urls;
 252       base::SplitString(urls_string, ' ', &urls);
 253       // Check for original-urls switch, which must exactly pair up with
 254       // |kUrlsSwitch| i.e. number of original urls must be same as that of
 255       // urls.
 256       std::vector<std::string> original_urls;
 257       if (command_line.HasSwitch(kOriginalUrls)) {
 258         std::string original_urls_string =
 259             command_line.GetSwitchValueASCII(kOriginalUrls);
 260         base::SplitString(original_urls_string, ' ', &original_urls);
 261         if (original_urls.size() != urls.size()) original_urls.clear();
 262       }
 263       for (size_t i = 0; i < urls.size(); ++i) {
 264         GURL url(urls[i]);
 265         if (url.is_valid()) {
 266           requests.push_back(new ContentExtractionRequest(url));
 267           // Only regard non-empty original urls.
 268           if (!original_urls.empty() && !original_urls[i].empty()) {
 269               (*file_to_url_map)[url.spec()] = original_urls[i];
 270           }
 271         } else {
 272           ADD_FAILURE() << "Bad url";
 273         }
 274       }
 275     }
 276     if (requests.empty()) {
 277       ADD_FAILURE() << "No valid url provided";
 278     }
 279
 280     return requests.Pass();
 281   }
 282
 283  private:
 284   ContentExtractionRequest(const GURL& url) : url_(url) {}
 285
 286   void OnArticleUpdated(ArticleDistillationUpdate article_update) override {}
 287
 288   void OnArticleReady(const DistilledArticleProto* article_proto) override {
 289     article_proto_ = article_proto;
 290     CHECK(article_proto->pages_size()) << "Failed extracting " << url_;
 291     base::ThreadTaskRunnerHandle::Get()->PostTask(FROM_HERE,
 292                                                   finished_callback_);
 293   }
 294
 295   const DistilledArticleProto* article_proto_;
 296   scoped_ptr<ViewerHandle> viewer_handle_;
 297   GURL url_;
 298   base::Closure finished_callback_;
 299 };
 300
 301 class ContentExtractor : public ContentBrowserTest {
 302  public:
 303   ContentExtractor()
 304       : pending_tasks_(0),
 305         max_tasks_(kMaxExtractorTasks),
 306         next_request_(0),
 307         output_data_(),
 308         protobuf_output_stream_(
 309             new google::protobuf::io::StringOutputStream(&output_data_)) {}
 310
 311   // Change behavior of the default host resolver to avoid DNS lookup errors, so
 312   // we can make network calls.
 313   void SetUpOnMainThread() override {
 314     if (!base::CommandLine::ForCurrentProcess()->HasSwitch(kDisableDnsSwitch)) {
 315       EnableDNSLookupForThisTest();
 316     }
 317     CHECK(db_dir_.CreateUniqueTempDir());
 318     AddComponentsTestResources();
 319   }
 320
 321   void TearDownOnMainThread() override { DisableDNSLookupForThisTest(); }
 322
 323  protected:
 324   // Creates the DomDistillerService and creates and starts the extraction
 325   // request.
 326   void Start() {
 327     const base::CommandLine& command_line =
 328         *base::CommandLine::ForCurrentProcess();
 329     FileToUrlMap file_to_url_map;
 330     requests_ = ContentExtractionRequest::CreateForCommandLine(
 331         command_line, &file_to_url_map);
 332     content::BrowserContext* context =
 333         shell()->web_contents()->GetBrowserContext();
 334     service_ = CreateDomDistillerService(context,
 335                                          db_dir_.path(),
 336                                          file_to_url_map);
 337     PumpQueue();
 338   }
 339
 340   void PumpQueue() {
 341     while (pending_tasks_ < max_tasks_ && next_request_ < requests_.size()) {
 342       requests_[next_request_]->Start(
 343           service_.get(),
 344           shell()->web_contents()->GetContainerBounds().size(),
 345           base::Bind(&ContentExtractor::FinishRequest, base::Unretained(this)));
 346       ++next_request_;
 347       ++pending_tasks_;
 348     }
 349   }
 350
 351  private:
 352   // Change behavior of the default host resolver to allow DNS lookup
 353   // to proceed instead of being blocked by the test infrastructure.
 354   void EnableDNSLookupForThisTest() {
 355     // mock_host_resolver_override_ takes ownership of the resolver.
 356     scoped_refptr<net::RuleBasedHostResolverProc> resolver =
 357         new net::RuleBasedHostResolverProc(host_resolver());
 358     resolver->AllowDirectLookup("*");
 359     mock_host_resolver_override_.reset(
 360         new net::ScopedDefaultHostResolverProc(resolver.get()));
 361   }
 362
 363   // We need to reset the DNS lookup when we finish, or the test will fail.
 364   void DisableDNSLookupForThisTest() {
 365     mock_host_resolver_override_.reset();
 366   }
 367
 368   void FinishRequest() {
 369     --pending_tasks_;
 370     if (next_request_ == requests_.size() && pending_tasks_ == 0) {
 371       Finish();
 372     } else {
 373       PumpQueue();
 374     }
 375   }
 376
 377   void DoArticleOutput() {
 378     const base::CommandLine& command_line =
 379         *base::CommandLine::ForCurrentProcess();
 380     for (size_t i = 0; i < requests_.size(); ++i) {
 381       const DistilledArticleProto& article = requests_[i]->GetArticleCopy();
 382       if (command_line.HasSwitch(kShouldOutputBinary)) {
 383         WriteProtobufWithSize(article, protobuf_output_stream_.get());
 384       } else {
 385         output_data_ += GetReadableArticleString(article) + "\n";
 386       }
 387     }
 388
 389     if (command_line.HasSwitch(kOutputFile)) {
 390       base::FilePath filename = command_line.GetSwitchValuePath(kOutputFile);
 391       ASSERT_EQ(
 392           (int)output_data_.size(),
 393           base::WriteFile(filename, output_data_.c_str(), output_data_.size()));
 394     } else {
 395       VLOG(0) << output_data_;
 396     }
 397   }
 398
 399   void Finish() {
 400     DoArticleOutput();
 401     requests_.clear();
 402     service_.reset();
 403     base::ThreadTaskRunnerHandle::Get()->PostTask(
 404         FROM_HERE, base::MessageLoop::QuitWhenIdleClosure());
 405   }
 406
 407   size_t pending_tasks_;
 408   size_t max_tasks_;
 409   size_t next_request_;
 410
 411   base::ScopedTempDir db_dir_;
 412   scoped_ptr<net::ScopedDefaultHostResolverProc> mock_host_resolver_override_;
 413   scoped_ptr<DomDistillerService> service_;
 414   ScopedVector<ContentExtractionRequest> requests_;
 415
 416   std::string output_data_;
 417   scoped_ptr<google::protobuf::io::StringOutputStream> protobuf_output_stream_;
 418 };
 419
 420 IN_PROC_BROWSER_TEST_F(ContentExtractor, MANUAL_ExtractUrl) {
 421   SetDistillerJavaScriptWorldId(content::ISOLATED_WORLD_ID_CONTENT_END);
 422   Start();
 423   base::RunLoop().Run();
 424 }
 425
 426 }  // namespace dom_distiller