/// <summary>
/// Getting all unique links on the page
/// </summary>
/// <param name="page"></param>
/// <returns></returns>
protected List <string> GetUniqueLinks(HtmlPageInfo page)
{
if (page.Document == null)
{
return(new List <string>());
}
return(page.Document.All
// Select only the tag <a> (link)
.Where(tag => tag.LocalName == "a")
// Project the value of the href attribute (link)
.Select(tag => ToAbsoluteUrl(tag.GetAttribute("href")))
// Select only links of this domain (including relative links)
.Where(href => !string.IsNullOrEmpty(href) && (href.StartsWith(Settings?.UrlSchemeAndHost) || href.StartsWith("/")))
// Select links that have not been added before and remove the repetition
.Where(href => !_tempLinks.Contains(href)).Distinct().ToList());
}
/// <summary>
/// Start parse process
/// </summary>
/// <returns></returns>
public async Task Start()
{
HtmlPageInfo rootPage = await GetPageInfo(Settings?.Url?.AbsoluteUri);
if (rootPage == null)
{
OnCompleted?.Invoke(this);
return;
}
_tempLinks.Add(rootPage.Url);
Pages.Add(rootPage);
await FillChildPages(rootPage);
OnCompleted?.Invoke(this);
}
/// <summary>
/// Recursive parsing of link tree
/// </summary>
/// <param name="parrentPage"></param>
/// <returns></returns>
protected async Task FillChildPages(HtmlPageInfo parrentPage)
{
List <string> links = GetUniqueLinks(parrentPage);
_tempLinks.AddRange(links);
foreach (string link in links)
{
HtmlPageInfo childPage = await GetPageInfo(link);
if (childPage == null)
{
continue;
}
Pages.Add(childPage);
OnNewPage?.Invoke(this, childPage);
await FillChildPages(childPage);
}
}
