public CrawlResult CrawlPage(CrawlPlan plan)
{
var crawlResult = new CrawlResult(plan.AbsoluteUri);
string driverPageSource;
//using (IWebDriver driver = new ChromeDriver(new ChromeOptions() { }))
//{
//WebDriverWait wait = new WebDriverWait(driver, TimeSpan.FromSeconds(10));
try
{
//todo: disable redirect
driver.Navigate().GoToUrl(plan.AbsoluteUri);
if (_pageLoadWait > 0)
{
Thread.Sleep(TimeSpan.FromSeconds(_pageLoadWait));
}
driverPageSource = driver.PageSource;
}
catch (Exception e)
{
Console.WriteLine(e);
crawlResult.BrowserFailedAt = DateTime.UtcNow;
crawlResult.BrowserFailedException = e.ToString();
//throw;
return(crawlResult);
}
//string html = driver.ExecuteJavaScript<string>("return document.documentElement.outerHTML;");
//driver.FindElement(By.Name("q")).SendKeys("cheese" + Keys.Enter);
//IWebElement firstResult = wait.Until(ExpectedConditions.ElementExists(By.TagName("h1")));
//Console.WriteLine(firstResult.GetAttribute("textContent"));
//}
//crawlResult.Doc = doc;
crawlResult.BrowserCrawledAt = DateTime.UtcNow;
crawlResult.BrowserContent = driverPageSource;
var linkSets = new List <List <string> >();
//find links
var linkElements = driver.FindElements(By.XPath("//a[@href]"));
if (linkElements != null && linkElements.Count > 0)
{
linkSets.Add(linkElements.Select(o => o.GetAttribute("href")).ToList());
}
//------------------------------------------
//Some Webpages DO NOT show all elements unless interacted with e.g. hovering)
//try hovering and findind more contents
//------------------------------------------
var hoverableList = new string[] {
//"//p[.='Creators']", "//p[.='Learn more']", "//p[.='Crypto Community']"
};
if (hoverableList != null && hoverableList.Length > 0)
{
var action = new Actions(driver);
foreach (var hoverableXPath in hoverableList)
{
var hoverableElement = driver.FindElement(By.XPath(hoverableXPath));
if (hoverableElement != null)
{
action.MoveToElement(hoverableElement).Perform();
linkElements = driver.FindElements(By.XPath("//a[@href]"));
if (linkElements != null && linkElements.Count > 0)
{
linkSets.Add(linkElements.Select(o => o.GetAttribute("href")).ToList());
}
}
}
}
var linksToSave = new List <CrawledLink>();
foreach (var linkSet in linkSets)
{
//Console.WriteLine($"\tfound {links.Count} child links.");
foreach (var href in linkSet)
{
//var href = link.GetAttribute("href");
//if(href.Contains("fortnite-stats/")) Debugger.Break();
var decoded = HttpUtility.HtmlDecode(href);
//if (decoded != href)
// Debugger.Break();
if (decoded == "" || decoded.StartsWith("javascript:") || decoded.StartsWith("mailto:") || decoded.StartsWith("skype:"))
{
continue;
}
try
{
var childUri = Util.GetUriObjectFromUriString(decoded, plan.AbsoluteUri);
//no duplicated links
if (linksToSave.All(o => o.AbsoluteUri != childUri.AbsoluteUri))
{
linksToSave.Add(new CrawledLink()
{
AbsoluteUri = childUri.AbsoluteUri,
IsBrowserRequired = true,
});
}
}
catch (UriFormatException e) //for mal-formated uris, just add them into the list without using System.Uri
{
if (linksToSave.All(o => o.AbsoluteUri != decoded))
{
linksToSave.Add(new CrawledLink()
{
AbsoluteUri = decoded,
IsBrowserRequired = true,
});
}
Console.WriteLine(e);
}
}
}
crawlResult.LinkAbsoluteUris = linksToSave;
return(crawlResult);
}
public CrawlResult CrawlPage(CrawlPlan plan)
{
var crawlResult = new CrawlResult(plan.AbsoluteUri);
var web = new HtmlWeb();
web.CaptureRedirect = true;
web.PreRequest = request => WebPreRequest(request);
web.PostResponse = (request, response) => HtmlWeb_PostResponse(request, response);
var stopWatch = new Stopwatch();
stopWatch.Start();
//Console.WriteLine($"loading...\t{plan.AbsoluteUri}");
HtmlDocument doc;
try
{
doc = web.Load(plan.AbsoluteUri);
}
catch (Exception e)
{
Console.WriteLine(e);
//if (e is HtmlWebException && e.Message.StartsWith("Unsupported uri scheme:"))
//{
//}
//else
//{
crawlResult.FailedAt = DateTime.UtcNow;
crawlResult.FailException = e.ToString();
//}
return(crawlResult);
}
stopWatch.Stop();
var statusCode = (int)web.StatusCode;
var statusCodeString = web.StatusCode.ToString();
var timeTaken = stopWatch.Elapsed.TotalSeconds;
crawlResult.CrawledAt = DateTime.UtcNow;
crawlResult.StatusCodeStr = statusCodeString;
crawlResult.StatusCode = statusCode;
crawlResult.TimeTaken = (decimal?)timeTaken;
//not 200 OK
if (web.StatusCode != HttpStatusCode.OK)
{
//Console.WriteLine($"\tstatus code = {statusCode}");
//3xx redirect
if (statusCode / 100 == 3)
{
try
{
//Console.WriteLine($"\tfound redirect {statusCode} {_redirectLocation}");
var locationUri = Util.GetUriObjectFromUriString(_redirectLocation, plan.AbsoluteUri);
//Console.WriteLine($"\tfound new location {locationUri.AbsoluteUri}");
crawlResult.LocationAbsoluteUri = locationUri.AbsoluteUri;
}
catch (Exception e)
{
Console.WriteLine(e);
}
}
return(crawlResult);
}
//not a document
if (doc.ParsedText == null)
{
return(crawlResult);
}
//crawlResult.Doc = doc;
crawlResult.ContentLength = doc.ParsedText.Length;
crawlResult.Content = doc.ParsedText;
//find links
var links = doc.DocumentNode.SelectNodes("//a[@href]");
if (links != null)
{
//Console.WriteLine($"\tfound {links.Count} child links.");
var linksToSave = new List <CrawledLink>();
foreach (var link in links)
{
var href = link.Attributes["href"].Value;
//if(href.Contains("fortnite-stats/")) Debugger.Break();
var decoded = HttpUtility.HtmlDecode(href);
//if (decoded != href)
// Debugger.Break();
if (decoded == "" || decoded.StartsWith("javascript:") || decoded.StartsWith("mailto:") || decoded.StartsWith("skype:"))
{
continue;
}
try
{
var childUri = Util.GetUriObjectFromUriString(decoded, plan.AbsoluteUri);
//no duplicated links
if (linksToSave.All(o => o.AbsoluteUri != childUri.AbsoluteUri))
{
linksToSave.Add(new CrawledLink()
{
AbsoluteUri = childUri.AbsoluteUri,
});
}
}
catch (UriFormatException e)//for mal-formated uris, just add them into the list without using System.Uri
{
if (linksToSave.All(o => o.AbsoluteUri != decoded))
{
linksToSave.Add(new CrawledLink()
{
AbsoluteUri = decoded,
});
}
Console.WriteLine(e);
}
}
crawlResult.LinkAbsoluteUris = linksToSave;
}
//find canonical
var canonicalLinks = doc.DocumentNode.SelectNodes("//link[@rel='canonical']");
if (canonicalLinks != null && canonicalLinks.Count > 0)
{
var canonicalLinkValue = canonicalLinks[0].Attributes["href"].Value;
//Console.WriteLine($"\tfound canonical");
crawlResult.Canonical = canonicalLinkValue;
}
return(crawlResult);
}
