// Setting up bot config
public void setup_abot()
{
CrawlConfiguration crawlConfig = new CrawlConfiguration();
crawlConfig.CrawlTimeoutSeconds = 150;
crawlConfig.MaxConcurrentThreads = 25;
crawlConfig.IsExternalPageCrawlingEnabled = false;
crawlConfig.MaxCrawlDepth = 1;
crawlConfig.MaxPagesToCrawl = 1000;
crawlConfig.UserAgentString = "abot v1.0 http://code.google.com/p/abot";
crawler = new PoliteWebCrawler(crawlConfig, null, null, null, null, null, null, null, null);
crawler.PageCrawlStartingAsync += crawler_ProcessPageCrawlStarting;
crawler.PageCrawlCompletedAsync += crawler_ProcessPageCrawlCompleted;
crawler.ShouldCrawlPage((pageToCrawl, crawlContext) =>
{
Regex rx = new Regex(@"\d{5}");
if (!rx.IsMatch(pageToCrawl.Uri.ToString()) && !pageToCrawl.Uri.ToString().Contains("text="))
return new CrawlDecision { Allow = false, Reason = "Want only comlinks" };
return new CrawlDecision { Allow = true, Reason = "OK Link" }; ;
});
}
private static IWebCrawler GetManuallyConfiguredWebCrawler()
{
//Create a config object manually
CrawlConfiguration config = new CrawlConfiguration();
config.CrawlTimeoutSeconds = 0;
config.DownloadableContentTypes = "text/html, text/plain";
config.IsExternalPageCrawlingEnabled = false;
config.IsExternalPageLinksCrawlingEnabled = false;
config.IsRespectRobotsDotTextEnabled = false;
config.IsUriRecrawlingEnabled = false;
config.MaxConcurrentThreads = 10;
config.MaxPagesToCrawl = 10;
config.MaxPagesToCrawlPerDomain = 0;
config.MinCrawlDelayPerDomainMilliSeconds = 1000;
config.UserAgentString = "abot v@ABOTASSEMBLYVERSION@ http://code.google.com/p/abot";
//Add you own values without modifying Abot's source code.
//These are accessible in CrawlContext.CrawlConfuration.ConfigurationException object throughout the crawl
config.ConfigurationExtensions.Add("Somekey1", "SomeValue1");
config.ConfigurationExtensions.Add("Somekey2", "SomeValue2");
//Initialize the crawler with custom configuration created above.
//This override the app.config file values
return new PoliteWebCrawler(config, null, null, null, null, null, null, null, null);
}
public void Constructor_ValidUri_CreatesInstance()
{
CrawlConfiguration unitUnderTest = new CrawlConfiguration();
Assert.IsNotNull(unitUnderTest.ConfigurationExtensions);
Assert.AreEqual(0, unitUnderTest.ConfigurationExtensions.Count);
Assert.AreEqual(0, unitUnderTest.CrawlTimeoutSeconds);
Assert.AreEqual("text/html", unitUnderTest.DownloadableContentTypes);
Assert.AreEqual(false, unitUnderTest.IsExternalPageCrawlingEnabled);
Assert.AreEqual(false, unitUnderTest.IsExternalPageLinksCrawlingEnabled);
Assert.AreEqual(false, unitUnderTest.IsRespectRobotsDotTextEnabled);
Assert.AreEqual(false, unitUnderTest.IsUriRecrawlingEnabled);
Assert.AreEqual(10, unitUnderTest.MaxConcurrentThreads);
Assert.AreEqual(5, unitUnderTest.MaxRobotsDotTextCrawlDelayInSeconds);
Assert.AreEqual(1000, unitUnderTest.MaxPagesToCrawl);
Assert.AreEqual(0, unitUnderTest.MaxPagesToCrawlPerDomain);
Assert.AreEqual(0, unitUnderTest.MinCrawlDelayPerDomainMilliSeconds);
Assert.AreEqual("Mozilla/5.0 (compatible; MSIE 9.0; Windows NT 6.1; abot v@ABOTASSEMBLYVERSION@ http://code.google.com/p/abot)", unitUnderTest.UserAgentString);
Assert.AreEqual("abot", unitUnderTest.RobotsDotTextUserAgentString);
Assert.AreEqual(0, unitUnderTest.MaxPageSizeInBytes);
Assert.AreEqual(0, unitUnderTest.HttpServicePointConnectionLimit);
Assert.AreEqual(0, unitUnderTest.HttpRequestTimeoutInSeconds);
Assert.AreEqual(7, unitUnderTest.HttpRequestMaxAutoRedirects);
Assert.AreEqual(true, unitUnderTest.IsHttpRequestAutoRedirectsEnabled);
Assert.AreEqual(false, unitUnderTest.IsHttpRequestAutomaticDecompressionEnabled);
Assert.AreEqual(0, unitUnderTest.MaxMemoryUsageCacheTimeInSeconds);
Assert.AreEqual(0, unitUnderTest.MaxMemoryUsageInMb);
Assert.AreEqual(0, unitUnderTest.MinAvailableMemoryRequiredInMb);
Assert.AreEqual(100, unitUnderTest.MaxCrawlDepth);
}
public static void CrawlerInit()
{
_crawledPages = new Dictionary<string, CrawledWebPage>();
_crawlConfig = new CrawlConfiguration
{
CrawlTimeoutSeconds = 100,
MaxConcurrentThreads = 10,
MaxPagesToCrawl = 1000,
UserAgentString = "abot v1.0 http://code.google.com/p/abot",
DownloadableContentTypes = "text/html, text/plain",
IsUriRecrawlingEnabled = false,
IsExternalPageCrawlingEnabled = true,
IsExternalPageLinksCrawlingEnabled = true,
HttpServicePointConnectionLimit = 200,
HttpRequestTimeoutInSeconds = 15,
HttpRequestMaxAutoRedirects = 7,
IsHttpRequestAutoRedirectsEnabled = true,
IsHttpRequestAutomaticDecompressionEnabled = true,
MinAvailableMemoryRequiredInMb = 0,
MaxMemoryUsageInMb = 200,
MaxMemoryUsageCacheTimeInSeconds = 2,
MaxCrawlDepth = 5,//10,
IsRespectRobotsDotTextEnabled = true
};
}
public void Crawl(Uri uri, Action<Page> callback)
{
var crawlConfig = new CrawlConfiguration
{
CrawlTimeoutSeconds = 0,
MaxConcurrentThreads = 5,
UserAgentString = "InspectionCrawler v1.0",
MinCrawlDelayPerDomainMilliSeconds = 1000,
MaxPagesToCrawl = 0,
MaxPagesToCrawlPerDomain = 0,
MaxCrawlDepth = int.MaxValue
};
var crawler = new PoliteWebCrawler(crawlConfig);
crawler.PageCrawlCompletedAsync += (sender, args) =>
{
var page = args.CrawledPage;
if (page.WebException != null && page.HttpWebResponse == null)
{
_log.Log(new LogMessage(LogType.Error, "Could not get page", page.WebException, page.Uri));
return;
}
callback(Convert(args.CrawledPage));
};
crawler.Crawl(uri);
}
static void Main(string[] args)
{
CrawlConfiguration config = new CrawlConfiguration();
config.MaxConcurrentThreads = 1; // Web Extractor is not currently thread-safe.
// Create the PhantomJS instance. This will spawn a new PhantomJS process using phantomjs.exe.
// Make sure to dispose this instance or you will have a zombie process!
IWebDriver driver = CreatePhantomJsDriver(config);
// Create the content extractor that uses PhantomJS.
IWebContentExtractor extractor = new JavaScriptContentExtractor(driver);
// Create a PageRequester that will use the extractor.
IPageRequester requester = new PageRequester(config, extractor);
using (IWebCrawler crawler = new PoliteWebCrawler(config, null, null, null, requester, null, null, null, null)) {
crawler.PageCrawlCompleted += OnPageCrawlCompleted;
CrawlResult result = crawler.Crawl(new Uri("http://wvtesting2.com/"));
if (result.ErrorOccurred)
Console.WriteLine("Crawl of {0} completed with error: {1}", result.RootUri.AbsoluteUri, result.ErrorException.Message);
else
Console.WriteLine("Crawl of {0} completed without error.", result.RootUri.AbsoluteUri);
}
Console.Read();
}
static void Main(string[] args) {
var proxies = new WebProxy[] { new WebProxy("http://192.168.1.1:3128"), new WebProxy("http://192.168.1.2:3128") };
var crawlConfig = new CrawlConfiguration();
var domainLimiter = new MultiProxyDomainRateLimiter(crawlConfig.MinCrawlDelayPerDomainMilliSeconds);
var pageRequester = new MultiProxyPageRequester(crawlConfig, new WebContentExtractor(), proxies);
var crawler = new MultiProxyPoliteWebCrawler(crawlConfig, null, null, null, pageRequester, null, null, domainLimiter, null);
crawler.Crawl(new Uri("http://localhost/"));
}
public PageRequester(CrawlConfiguration config)
{
if (config == null)
throw new ArgumentNullException("config");
_userAgentString = config.UserAgentString.Replace("@ABOTASSEMBLYVERSION@", Assembly.GetAssembly(this.GetType()).GetName().Version.ToString());
_config = config;
if (_config.HttpServicePointConnectionLimit > 0)
ServicePointManager.DefaultConnectionLimit = _config.HttpServicePointConnectionLimit;
}
public PageRequester(CrawlConfiguration config, IWebContentExtractor contentExtractor)
{
if (config == null)
throw new ArgumentNullException("config");
_config = config;
if (_config.HttpServicePointConnectionLimit > 0)
ServicePointManager.DefaultConnectionLimit = _config.HttpServicePointConnectionLimit;
_extractor = contentExtractor ?? new WebContentExtractor();
}
public CrawlConfiguration Convert()
{
AutoMapper.Mapper.CreateMap<CrawlBehaviorElement, CrawlConfiguration>();
AutoMapper.Mapper.CreateMap<PolitenessElement, CrawlConfiguration>();
CrawlConfiguration config = new CrawlConfiguration();
AutoMapper.Mapper.Map<CrawlBehaviorElement, CrawlConfiguration>(CrawlBehavior, config);
AutoMapper.Mapper.Map<PolitenessElement, CrawlConfiguration>(Politeness, config);
foreach (ExtensionValueElement element in ExtensionValues)
config.ConfigurationExtensions.Add(element.Key, element.Value);
return config;
}
public PoliteWebCrawler(
CrawlConfiguration crawlConfiguration,
ICrawlDecisionMaker crawlDecisionMaker,
IThreadManager threadManager,
IScheduler scheduler,
IPageRequester pageRequester,
IHyperLinkParser hyperLinkParser,
IMemoryManager memoryManager,
IDomainRateLimiter domainRateLimiter,
IRobotsDotTextFinder robotsDotTextFinder)
: base(crawlConfiguration, crawlDecisionMaker, threadManager, scheduler, pageRequester, hyperLinkParser, memoryManager)
{
_domainRateLimiter = domainRateLimiter ?? new DomainRateLimiter(_crawlContext.CrawlConfiguration.MinCrawlDelayPerDomainMilliSeconds);
_robotsDotTextFinder = robotsDotTextFinder ?? new RobotsDotTextFinder(new PageRequester(_crawlContext.CrawlConfiguration));
}
public void Crawl_CrawlTimeoutIs1Sec_TimesOut()
{
CrawlConfiguration configuration = new CrawlConfiguration();
configuration.CrawlTimeoutSeconds = 2;
int pagesCrawledCount = 0;
PoliteWebCrawler crawler = new PoliteWebCrawler(configuration, null, null, null, null, null, null, null, null);
crawler.PageCrawlCompletedAsync += (a, b) => pagesCrawledCount++;
CrawlResult result = crawler.Crawl(new Uri("http://localhost.fiddler:1111/"));
Assert.IsFalse(result.ErrorOccurred);
Assert.IsTrue(result.Elapsed.TotalSeconds < 8, "Took more than 8 seconds");
Assert.IsTrue(pagesCrawledCount < 2, "Crawled more than 2 pages");
}
public void Crawl_CrawlTimeoutIs1Sec_TimesOut()
{
CrawlConfiguration configuration = new CrawlConfiguration ();
configuration.CrawlTimeoutSeconds = 1;
int pagesCrawledCount = 0;
PoliteWebCrawler crawler = new PoliteWebCrawler (configuration, null, null, null, null, null, null, null, null);
crawler.PageCrawlCompletedAsync += (a, b) => pagesCrawledCount++;
CrawlResult result = crawler.Crawl (new Uri ("http://wvtesting2.com/"));
Assert.IsFalse (result.ErrorOccurred);
Assert.IsTrue (result.Elapsed.TotalSeconds < 5);
Assert.IsTrue (pagesCrawledCount > 0);
}
public void Crawl_MaxPagesTo25_OnlyCrawls25Pages()
{
new PageRequester(new CrawlConfiguration { UserAgentString = "aaa" }).MakeRequest(new Uri("http://localhost.fiddler:1111/PageGenerator/ClearCounters"));
CrawlConfiguration configuration = new CrawlConfiguration();
configuration.MaxPagesToCrawl = 25;
int pagesCrawledCount = 0;
PoliteWebCrawler crawler = new PoliteWebCrawler(configuration, null, null, null, null, null, null, null, null);
crawler.PageCrawlCompletedAsync += (a, b) => pagesCrawledCount++;
crawler.Crawl(new Uri("http://localhost.fiddler:1111/"));
Assert.AreEqual(25, pagesCrawledCount);
}
public PageRequester(CrawlConfiguration config, IWebContentExtractor contentExtractor)
{
if (config == null)
throw new ArgumentNullException("config");
_config = config;
if (_config.HttpServicePointConnectionLimit > 0)
ServicePointManager.DefaultConnectionLimit = _config.HttpServicePointConnectionLimit;
if (!_config.IsSslCertificateValidationEnabled)
ServicePointManager.ServerCertificateValidationCallback +=
(sender, certificate, chain, sslPolicyErrors) => true;
_extractor = contentExtractor ?? new WebContentExtractor();
}
public void SetUp()
{
_fakeHyperLinkParser = new Mock<IHyperLinkParser>();
_fakeHttpRequester = new Mock<IPageRequester>();
_fakeCrawlDecisionMaker = new Mock<ICrawlDecisionMaker>();
_fakeDomainRateLimiter = new Mock<IDomainRateLimiter>();
_fakeMemoryManager = new Mock<IMemoryManager>();
_fakeRobotsDotTextFinder = new Mock<IRobotsDotTextFinder>();
_fakeRobotsDotText = new Mock<IRobotsDotText>();
_dummyScheduler = new Scheduler();
_dummyThreadManager = new ManualThreadManager(1);
_dummyConfiguration = new CrawlConfiguration();
_dummyConfiguration.ConfigurationExtensions.Add("somekey", "someval");
_rootUri = new Uri("http://a.com/");
}
public void Crawl_MaxPagesTo5_WithCrawlDelay_OnlyCrawls5Pages()
{
new PageRequester(new CrawlConfiguration{ UserAgentString = "aaa" }).MakeRequest(new Uri("http://localhost.fiddler:1111/PageGenerator/ClearCounters"));
CrawlConfiguration configuration = new CrawlConfiguration();
configuration.MinCrawlDelayPerDomainMilliSeconds = 1000; //adding delay since it increases the chance of issues with abot crawling more than MaxPagesToCrawl.
configuration.MaxPagesToCrawl = 5;
int pagesCrawledCount = 0;
PoliteWebCrawler crawler = new PoliteWebCrawler(configuration, null, null, null, null, null, null, null, null);
crawler.PageCrawlCompletedAsync += (a, b) => pagesCrawledCount++;
crawler.Crawl(new Uri("http://localhost.fiddler:1111/"));
Assert.AreEqual(5, pagesCrawledCount);
}
private static IWebCrawler GetManuallyConfiguredWebCrawler()
{
var crawlConfiguration = new CrawlConfiguration();
crawlConfiguration.CrawlTimeoutSeconds = 0;
crawlConfiguration.DownloadableContentTypes = "text/html, text/plain";
crawlConfiguration.IsExternalPageCrawlingEnabled = false;
crawlConfiguration.IsExternalPageLinksCrawlingEnabled = false;
crawlConfiguration.IsRespectRobotsDotTextEnabled = false;
crawlConfiguration.IsUriRecrawlingEnabled = false;
crawlConfiguration.MaxConcurrentThreads = 10;
crawlConfiguration.MaxPagesToCrawl = 10;
crawlConfiguration.MaxPagesToCrawlPerDomain = 0;
crawlConfiguration.MinCrawlDelayPerDomainMilliSeconds = 1000;
crawlConfiguration.ConfigurationExtensions.Add("Somekey1", "SomeValue1");
crawlConfiguration.ConfigurationExtensions.Add("Somekey2", "SomeValue2");
return new PoliteWebCrawler(crawlConfiguration, null, null, null, null, null, null, null, null);
}
public void SetUp()
{
_fakeHyperLinkParser = new Mock<IHyperLinkParser>();
_fakeHttpRequester = new Mock<IPageRequester>();
_fakeCrawlDecisionMaker = new Mock<ICrawlDecisionMaker>();
_fakeMemoryManager = new Mock<IMemoryManager>();
_fakeDomainRateLimiter = new Mock<IDomainRateLimiter>();
_fakeRobotsDotTextFinder = new Mock<IRobotsDotTextFinder>();
_dummyScheduler = new Scheduler();
_dummyThreadManager = new TaskThreadManager(10);
_dummyConfiguration = new CrawlConfiguration();
_dummyConfiguration.ConfigurationExtensions.Add("somekey", "someval");
_unitUnderTest = new PoliteWebCrawler(_dummyConfiguration, _fakeCrawlDecisionMaker.Object, _dummyThreadManager, _dummyScheduler, _fakeHttpRequester.Object, _fakeHyperLinkParser.Object, _fakeMemoryManager.Object, _fakeDomainRateLimiter.Object, _fakeRobotsDotTextFinder.Object);
_unitUnderTest.CrawlBag.SomeVal = "SomeVal";
_unitUnderTest.CrawlBag.SomeList = new List<string>() { "a", "b" };
_rootUri = new Uri("http://a.com/");
}
private static IWebDriver CreatePhantomJsDriver(CrawlConfiguration p_Config)
{
// Optional options passed to the PhantomJS process.
PhantomJSOptions options = new PhantomJSOptions();
options.AddAdditionalCapability("phantomjs.page.settings.userAgent", p_Config.UserAgentString);
options.AddAdditionalCapability("phantomjs.page.settings.javascriptCanCloseWindows", false);
options.AddAdditionalCapability("phantomjs.page.settings.javascriptCanOpenWindows", false);
options.AddAdditionalCapability("acceptSslCerts", !p_Config.IsSslCertificateValidationEnabled);
// Basic auth credentials.
options.AddAdditionalCapability("phantomjs.page.settings.userName", p_Config.LoginUser);
options.AddAdditionalCapability("phantomjs.page.settings.password", p_Config.LoginPassword);
// Create the service while hiding the prompt window.
PhantomJSDriverService service = PhantomJSDriverService.CreateDefaultService();
service.HideCommandPromptWindow = true;
IWebDriver driver = new PhantomJSDriver(service, options);
return driver;
}
public void Constructor_ValidUri_CreatesInstance()
{
CrawlConfiguration unitUnderTest = new CrawlConfiguration();
Assert.IsNotNull(unitUnderTest.ConfigurationExtensions);
Assert.AreEqual(0, unitUnderTest.ConfigurationExtensions.Count);
Assert.AreEqual(0, unitUnderTest.CrawlTimeoutSeconds);
Assert.AreEqual("text/html", unitUnderTest.DownloadableContentTypes);
Assert.AreEqual(false, unitUnderTest.IsExternalPageCrawlingEnabled);
Assert.AreEqual(false, unitUnderTest.IsExternalPageLinksCrawlingEnabled);
Assert.AreEqual(false, unitUnderTest.IsRespectRobotsDotTextEnabled);
Assert.AreEqual(false, unitUnderTest.IsRespectMetaRobotsNoFollowEnabled);
Assert.AreEqual(false, unitUnderTest.IsRespectAnchorRelNoFollowEnabled);
Assert.AreEqual(false, unitUnderTest.IsUriRecrawlingEnabled);
Assert.AreEqual(10, unitUnderTest.MaxConcurrentThreads);
Assert.AreEqual(5, unitUnderTest.MaxRobotsDotTextCrawlDelayInSeconds);
Assert.AreEqual(1000, unitUnderTest.MaxPagesToCrawl);
Assert.AreEqual(0, unitUnderTest.MaxPagesToCrawlPerDomain);
Assert.AreEqual(0, unitUnderTest.MinCrawlDelayPerDomainMilliSeconds);
Assert.AreEqual("Mozilla/5.0 (Windows NT 6.3; Trident/7.0; rv:11.0) like Gecko", unitUnderTest.UserAgentString);
Assert.AreEqual("abot", unitUnderTest.RobotsDotTextUserAgentString);
Assert.AreEqual(0, unitUnderTest.MaxPageSizeInBytes);
Assert.AreEqual(200, unitUnderTest.HttpServicePointConnectionLimit);
Assert.AreEqual(15, unitUnderTest.HttpRequestTimeoutInSeconds);
Assert.AreEqual(7, unitUnderTest.HttpRequestMaxAutoRedirects);
Assert.AreEqual(true, unitUnderTest.IsHttpRequestAutoRedirectsEnabled);
Assert.AreEqual(false, unitUnderTest.IsHttpRequestAutomaticDecompressionEnabled);
Assert.AreEqual(false, unitUnderTest.IsSendingCookiesEnabled);
Assert.AreEqual(true, unitUnderTest.IsSslCertificateValidationEnabled);
Assert.AreEqual(false, unitUnderTest.IsRespectUrlNamedAnchorOrHashbangEnabled);
Assert.AreEqual(0, unitUnderTest.MaxMemoryUsageCacheTimeInSeconds);
Assert.AreEqual(0, unitUnderTest.MaxMemoryUsageInMb);
Assert.AreEqual(0, unitUnderTest.MinAvailableMemoryRequiredInMb);
Assert.AreEqual(100, unitUnderTest.MaxCrawlDepth);
Assert.AreEqual(false, unitUnderTest.IsForcedLinkParsingEnabled);
Assert.AreEqual(0, unitUnderTest.MaxRetryCount);
Assert.AreEqual(0, unitUnderTest.MinRetryDelayInMilliseconds);
Assert.AreEqual(null, unitUnderTest.LoginUser);
Assert.AreEqual(null, unitUnderTest.LoginPassword);
Assert.AreEqual(false, unitUnderTest.IsAlwaysLogin);
}
public void Crawl_IsRateLimited()
{
new PageRequester(new CrawlConfiguration { UserAgentString = "aaa" }).MakeRequest(new Uri("http://localhost.fiddler:1111/PageGenerator/ClearCounters"));
CrawlConfiguration configuration = new CrawlConfiguration();
configuration.MaxPagesToCrawl = 3;
configuration.MinCrawlDelayPerDomainMilliSeconds = 1000; // 1 second * 2 pages = 2 (or more) seconds
int pagesCrawledCount = 0;
var crawler = new PoliteWebCrawler(configuration);
crawler.PageCrawlCompletedAsync += (a, b) => pagesCrawledCount++;
var uriToCrawl = new Uri("http://localhost.fiddler:1111/");
var start = DateTime.Now;
crawler.Crawl(uriToCrawl);
var elapsed = DateTime.Now - start;
Assert.GreaterOrEqual(elapsed.TotalMilliseconds, 2000);
Assert.AreEqual(3, pagesCrawledCount);
}
public CSQueryHyperlinkParser(CrawlConfiguration config, Func<string, string> cleanURLFunc)
: base(config, cleanURLFunc)
{
}
/// <summary>
/// Creates a crawler instance with custom settings or implementation. Passing in null for all params is the equivalent of the empty constructor.
/// </summary>
/// <param name="threadManager">Distributes http requests over multiple threads</param>
/// <param name="scheduler">Decides what link should be crawled next</param>
/// <param name="httpRequester">Makes the raw http requests</param>
/// <param name="hyperLinkParser">Parses a crawled page for it's hyperlinks</param>
/// <param name="crawlDecisionMaker">Decides whether or not to crawl a page or that page's links</param>
/// <param name="crawlConfiguration">Configurable crawl values</param>
public WebCrawler(
CrawlConfiguration crawlConfiguration,
ICrawlDecisionMaker crawlDecisionMaker,
IThreadManager threadManager,
IScheduler scheduler,
IPageRequester httpRequester,
IHyperLinkParser hyperLinkParser,
IMemoryManager memoryManager)
{
_crawlContext = new CrawlContext();
_crawlContext.CrawlConfiguration = crawlConfiguration ?? GetCrawlConfigurationFromConfigFile() ?? new CrawlConfiguration();
CrawlBag = _crawlContext.CrawlBag;
_threadManager = threadManager ?? new ManualThreadManager(_crawlContext.CrawlConfiguration.MaxConcurrentThreads);
_scheduler = scheduler ?? new FifoScheduler(_crawlContext.CrawlConfiguration.IsUriRecrawlingEnabled);
_httpRequester = httpRequester ?? new PageRequester(_crawlContext.CrawlConfiguration);
_crawlDecisionMaker = crawlDecisionMaker ?? new CrawlDecisionMaker();
if(_crawlContext.CrawlConfiguration.MaxMemoryUsageInMb > 0
|| _crawlContext.CrawlConfiguration.MinAvailableMemoryRequiredInMb > 0)
_memoryManager = memoryManager ?? new MemoryManager(new CachedMemoryMonitor(new GcMemoryMonitor(), _crawlContext.CrawlConfiguration.MaxMemoryUsageCacheTimeInSeconds));
_hyperLinkParser = hyperLinkParser ?? new HapHyperLinkParser();
_crawlContext.Scheduler = _scheduler;
}
public PoliteWebCrawler(CrawlConfiguration crawlConfiguration)
: this(crawlConfiguration, null, null, null, null, null, null, null, null)
{
}
public PageRequesterWrapper(CrawlConfiguration config)
: base(config)
{
}
public void ShouldCrawlPage_OverMaxPagesToCrawlPerDomain_IsRetry_ReturnsTrue()
{
Uri uri = new Uri("http://a.com/");
CrawlConfiguration config = new CrawlConfiguration
{
MaxPagesToCrawlPerDomain = 100
};
ConcurrentDictionary<string, int> countByDomain = new ConcurrentDictionary<string, int>();
countByDomain.TryAdd(uri.Authority, 100);
CrawlContext crawlContext = new CrawlContext
{
CrawlConfiguration = config,
CrawlStartDate = DateTime.Now,
CrawlCountByDomain = countByDomain
};
CrawlDecision result = _unitUnderTest.ShouldCrawlPage(
new PageToCrawl(new Uri(uri.AbsoluteUri + "anotherpage"))
{
IsRetry = true,
IsInternal = true
},
crawlContext);
Assert.IsTrue(result.Allow);
Assert.IsFalse(result.ShouldHardStopCrawl);
Assert.IsFalse(result.ShouldStopCrawl);
}
public void Dispose()
{
if (_extractor != null)
{
_extractor.Dispose();
}
_cookieContainer = null;
_config = null;
}
protected virtual void PrintConfigValues(CrawlConfiguration config)
{
_logger.Info("Configuration Values:");
string indentString = new string(' ', 2);
foreach (PropertyInfo property in config.GetType().GetProperties())
{
if (property.Name != "ConfigurationExtensions")
_logger.InfoFormat("{0}{1}: {2}", indentString, property.Name, property.GetValue(config, null));
}
foreach (string key in config.ConfigurationExtensions.Keys)
{
_logger.InfoFormat("{0}{1}: {2}", indentString, key, config.ConfigurationExtensions[key]);
}
}
public PageRequester(CrawlConfiguration config)
: this(config, null)
{
}
