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