public void Constructor_ValidUri_CreatesInstance()
{
CrawlDecision unitUnderTest = new CrawlDecision();
Assert.AreEqual(false, unitUnderTest.Allow);
Assert.AreEqual("", unitUnderTest.Reason);
Assert.IsFalse(unitUnderTest.ShouldHardStopCrawl);
Assert.IsFalse(unitUnderTest.ShouldStopCrawl);
}
//public static Dictionary<string, int> garmentTypes;
static void Main(string[] args)
{
Console.WriteLine("howdy!");
log4net.Config.XmlConfigurator.Configure();
//PrintDisclaimer();
//Uri uriToCrawl = GetSiteToCrawl(args);
//Uri uriToCrawl = new Uri("http://www.uniqlo.com/us/product/women-airism-tank-top-143149.html"); // out of stock
//Uri uriToCrawl = new Uri("http://www.uniqlo.com/us/product/women-extra-fine-cotton-long-sleeve-long-shirt-167548001.html");
//Uri uriToCrawl = new Uri("http://www.uniqlo.com/us/women.html"); // starting point for crawl - WOMEN
//Uri uriToCrawl = new Uri("http://www.uniqlo.com/us/men.html"); // starting point for crawl - MEN
//Uri uriToCrawl = new Uri("http://www.uniqlo.com/us/women/tops/t-shirts.html");
//Uri uriToCrawl = new Uri("http://www.hm.com/us/product/77536?article=77536-C");
//Uri uriToCrawl = new Uri("http://www.hm.com/us/");
//Uri uriToCrawl = new Uri("http://www.hm.com/us/products/men"); // starting point for crawl - men
//Uri uriToCrawl = new Uri("http://www.hm.com/us/products/men/tshirt");
//Uri uriToCrawl = new Uri("http://www.hm.com/us/products/ladies"); // starting point for crawl - women
//Uri uriToCrawl = new Uri("http://www.hm.com/us/product/30283?article=30283-J"); // multi-colors
//Uri uriToCrawl = new Uri("http://www.hm.com/us/product/04252?article=04252-A");
//Uri uriToCrawl = new Uri("http://www.zara.com/us/en/man/outerwear/view-all/navy-coat-c764502p3094502.html");
//Uri uriToCrawl = new Uri("http://www.zara.com/us/en/sale/man/coats-and-trench-coats/view-all/long-denim-parka-c794501p3276509.html");
//Uri uriToCrawl = new Uri("http://www.zara.com/us/en/sale/woman/t-shirts/view-all/crop-t-shirt-c732027p2874029.html"); // multiple colors
Uri uriToCrawl = new Uri("http://www.zara.com/us/en/collection-ss16/woman/outerwear/view-all-c719012.html"); // crawl start point
//Uri uriToCrawl = new Uri("http://www.zara.com/us/en/collection-ss16/man/jackets/bomber-jacket-c586542p3268146.html");
//Uri uriToCrawl = new Uri("http://www.zara.com/us/en/collection-ss16/woman/outerwear/view-all/wool-coat-with-lapels-c719012p3186217.html");
//Uri uriToCrawl = new Uri("http://www.zara.com/us/en/collection-ss16/woman/bags-c358019.html");
IWebCrawler crawler;
crawler = GetDefaultWebCrawler();
//Subscribe to any of these asynchronous events, there are also sychronous versions of each.
//This is where you process data about specific events of the crawl
crawler.PageCrawlStartingAsync += crawler_ProcessPageCrawlStarting;
crawler.PageCrawlCompletedAsync += crawler_ProcessPageCrawlCompleted;
crawler.PageCrawlDisallowedAsync += crawler_PageCrawlDisallowed;
crawler.PageLinksCrawlDisallowedAsync += crawler_PageLinksCrawlDisallowed;
crawler.ShouldCrawlPage((pageToCrawl, crawlContext) =>
{
CrawlDecision decision = new CrawlDecision { Allow = true, Reason = "OK" };
// H&M rules
if (pageToCrawl.Uri.AbsoluteUri.Contains("hm.com")) {
// only have rules for h&m
decision = new CrawlDecision { Allow = false, Reason = "Not good!" };
if (pageToCrawl.Uri.AbsoluteUri.Contains("http://www.hm.com/us/product/") ||
pageToCrawl.Uri.AbsoluteUri.Contains("http://www.hm.com/us/products/men") ||
pageToCrawl.Uri.AbsoluteUri.Contains("http://www.hm.com/us/products/ladies"))
return new CrawlDecision { Allow = true, Reason = "OK!" };
}
// UNIQLO rules
if (pageToCrawl.Uri.AbsoluteUri.Contains("uniqlo"))
{
// need some for uniqlo too
decision = new CrawlDecision { Allow = false, Reason = "Not good!" };
if (pageToCrawl.Uri.AbsoluteUri.Contains("http://www.uniqlo.com/us/women.html") ||
pageToCrawl.Uri.AbsoluteUri.Contains("http://www.uniqlo.com/us/women/") ||
pageToCrawl.Uri.AbsoluteUri.Contains("http://www.uniqlo.com/us/product/") ||
pageToCrawl.Uri.AbsoluteUri.Contains("http://www.uniqlo.com/us/men.html") ||
pageToCrawl.Uri.AbsoluteUri.Contains("http://www.uniqlo.com/us/men/"))
return new CrawlDecision { Allow = true, Reason = "OK!" };
}
// ZARA rules
if (pageToCrawl.Uri.AbsoluteUri.Contains("zara"))
{
// need some for zara too
decision = new CrawlDecision { Allow = false, Reason = "Not good!" };
if (pageToCrawl.Uri.AbsoluteUri.Contains("http://www.zara.com/us/en/collection-ss16/") ||
pageToCrawl.Uri.AbsoluteUri.Contains("http://www.zara.com/us/en/sale/woman") ||
pageToCrawl.Uri.AbsoluteUri.Contains("http://www.zara.com/us/en/sale/man/") ||
pageToCrawl.Uri.AbsoluteUri.Contains("http://www.zara.com/us/en/collection-ss16/man/"))
return new CrawlDecision { Allow = true, Reason = "OK!" };
}
return decision;
});
myConnection = new SqlConnection("user id=marakas;" +
"password=M@rakas69!;server=yzf0vdv9dr.database.windows.net;" +
"Trusted_Connection=False;Encrypt=True;" +
"database=superfashiondb_db; " +
"connection timeout=30");
try
{
myConnection.Open();
Console.WriteLine("DB OK!");
}
catch (Exception e)
{
Console.WriteLine(e.ToString());
}
//Start the crawl
//This is a synchronous call
CrawlResult result = crawler.Crawl(uriToCrawl);
// Keep the console window open in debug mode.
Console.WriteLine("Press any key to exit.");
Console.ReadKey();
}
protected virtual void SignalCrawlStopIfNeeded(CrawlDecision decision)
{
if (decision.ShouldHardStopCrawl)
{
_logger.InfoFormat("Decision marked crawl [Hard Stop] for site [{0}], [{1}]", _crawlContext.RootUri, decision.Reason);
_crawlContext.IsCrawlHardStopRequested = decision.ShouldHardStopCrawl;
}
else if (decision.ShouldStopCrawl)
{
_logger.InfoFormat("Decision marked crawl [Stop] for site [{0}], [{1}]", _crawlContext.RootUri, decision.Reason);
_crawlContext.IsCrawlStopRequested = decision.ShouldStopCrawl;
}
}
private void Crawl()
{
ThreadPool.QueueUserWorkItem((state) =>
{
PoliteWebCrawler crawler = new PoliteWebCrawler();
crawler.PageCrawlStartingAsync += crawler_ProcessPageCrawlStarting;
crawler.PageCrawlCompletedAsync += crawler_ProcessPageCrawlCompleted;
crawler.PageCrawlDisallowedAsync += crawler_PageCrawlDisallowed;
crawler.PageLinksCrawlDisallowedAsync += crawler_PageLinksCrawlDisallowed;
crawler.ShouldCrawlPage((crawledPage, crawledContext) =>
{
CrawlDecision decision = new CrawlDecision();
var uri = crawledPage.Uri.ToString();
if (crawledPage.IsRoot || uri.StartsWith("http://www.tingchina.com/"))
{
decision.Allow = true;
}
else
{
decision.Allow = false;
decision.Reason = "Just erge pages!";
}
return decision;
});
CrawlResult result = crawler.Crawl(new Uri("http://www.tingchina.com/"), cancellationTokenSource);
if (result.ErrorOccurred)
{
NextUrl = result.ErrorException.Message;
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.ReadLine();
}, null);
}
/// <summary>
/// Delegate passed to WebCrawler which calls functionaltiy to look at
/// blacklisted urls when deciding whether a page should be crawled or not.
/// If the <paramref name="pageToCrawl"/> Domain is blacklisted, then the
/// CrawlDecision.Allow is set to false. This delegate is called after the default
/// CrawlDecisionMaker.ShouldCrawlPage() method is called.
/// </summary>
/// <returns>CrawlDecision</returns>
/// <remarks>The default CrawlDecisionMaker.ShouldCrawlPage() method is called first, but then
/// this method will be called. No reason to override the default CrawlDecisionMaker, see
/// about line ~721 in WebCrawler.cs</remarks>
public CrawlDecision ShouldCrawlPage(PageToCrawl pageToCrawl, CrawlContext crawlContext)
{
CrawlDecision decision = null;
var domain = pageToCrawl.Uri.GetBaseDomain();
if (_repo.IsBlackListed(domain))
{
decision = new CrawlDecision
{
Allow = false,
Reason = string.Format("The domain {0} is blacklisted", domain)
};
}
else
{
decision = new CrawlDecision() { Allow = true };
}
return decision;
}
