public CrawledWindow(URLParentData parent)
{
InitializeComponent();
tb_title.Text = parent.title;
tb_url.Text = parent.url;
tb_rootURL.Text = parent.rootURL;
tb_crawledTime.Text = parent.urlCrawledTime.ToString();
TimeSpan ts = new TimeSpan(parent.urlCrawledTime.Ticks - parent.urlRegisteredTime.Ticks);
tb_duration.Text = ts.TotalSeconds + " s";
foreach (URLChildData child in parent.childrenURLs)
{
lb_children.Items.Add(child.url);
}
}
// Take input url as string
// 2019103002
public URLCrawler(URLParentData parentURLData)
{
this.parentURLData = parentURLData;
htmlWeb = new HtmlWeb();
htmlDoc = new HtmlDocument();
}
private void startSingleCrawler(URLChildData childToBecomeParent)
{
// transfer url from toBeCrawled to crawling
lock (dToBeCrawledURLs) // 2019103027
dToBeCrawledURLs.Remove(childToBecomeParent.url);
lock (dCrawlingURLs) // 2019103027
dCrawlingURLs.Add(childToBecomeParent.url, childToBecomeParent);
URLParentData parent = new URLParentData(childToBecomeParent);
URLCrawler crawler = new URLCrawler(parent);
// create a new task for crawler operation
var urlCrawlerTask = Task.Factory.StartNew(() =>
{
crawler.start();
}).ContinueWith(taskInfo =>
{
if (parent.state == CrawlerState.Crawled)
{
// transfer url from crawling hs to crawled hs
lock (dCrawlingURLs) // 2019103027
dCrawlingURLs.Remove(childToBecomeParent.url);
lock (dCrawledURLs) // 2019103027
dCrawledURLs.Add(parent.url, parent);
// loops over each child URL, checks if exists and adds it to tobecrawled hs
foreach (URLChildData child in parent.childrenURLs)
{
// 2019103026
bool externalURLactivated = lRootURLs.Where(rootData => rootData.rootURL == child.rootURL)
.Select(rootData => rootData.externalActivated)
.First();
// discards external URLs if externalURL checkbox is not activated
if (child.type == CrawledType.externalURL && !externalURLactivated)
{
continue;
}
if (!URLexist(child.url))
{
lock (dToBeCrawledURLs)
dToBeCrawledURLs.Add(child.url, child);
}
}
}
else
{
childToBecomeParent.errorCounter = parent.errorCounter;
// handle CrawlerState.Failed
if (childToBecomeParent.errorCounter >= 3)
{
childToBecomeParent.urlDisabledTime = DateTime.Now;
childToBecomeParent.errorCounter = 0;
lock (disabledURLs)
disabledURLs.Add(childToBecomeParent);
}
else
{
// transfer url from toBeCrawled to crawling
lock (dToBeCrawledURLs)
dToBeCrawledURLs.Add(childToBecomeParent.url, childToBecomeParent);
}
// remove url from crawling list
lock (dCrawlingURLs)
dCrawlingURLs.Remove(childToBecomeParent.url);
}
});
Tuple <string, Task, string> newTask = new Tuple <string, Task, string>(childToBecomeParent.rootURL, urlCrawlerTask, childToBecomeParent.url);
tl_activelyRunningTasks.Add(newTask);
}
