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