public async Task ProcessAsync(ICrawler crawler, PropertyBag propertyBag)
{
if (propertyBag.StatusCode != HttpStatusCode.OK)
{
return;
}
if (!IsXmlContent(propertyBag.ContentType))
{
return;
}
using (var reader = propertyBag.GetResponse())
{
using (var sr = new StreamReader(reader))
{
var mydoc = XDocument.Load(sr);
if (mydoc.Root == null)
{
return;
}
var qualifiedName = XName.Get("loc", "http://www.sitemaps.org/schemas/sitemap/0.9");
var urlNodes =
from e in mydoc.Descendants(qualifiedName)
where !e.Value.IsNullOrEmpty() && e.Value.StartsWith("http://", StringComparison.OrdinalIgnoreCase)
select e.Value;
foreach (var url in urlNodes)
{
// add new crawler steps
var baseUrl = propertyBag.ResponseUri.GetLeftPart(UriPartial.Path);
var decodedLink = ExtendedHtmlUtility.HtmlEntityDecode(url);
var normalizedLink = NormalizeLink(baseUrl, decodedLink);
if (normalizedLink.IsNullOrEmpty())
{
continue;
}
await crawler.AddStepAsync(new Uri(normalizedLink), propertyBag.Step.Depth + 1,
propertyBag.Step, new Dictionary <string, object>
{
{ Resources.PropertyBagKeyOriginalUrl, url },
{ Resources.PropertyBagKeyOriginalReferrerUrl, propertyBag.ResponseUri }
});
}
}
}
}
public virtual async Task ProcessAsync(ICrawler crawler, PropertyBag propertyBag)
{
// Get text from previous pipeline step
var text = propertyBag.Text;
if (this.HasTextStripRules)
{
text = this.StripText(text);
}
if (text.IsNullOrEmpty())
{
return;
}
if (this.HasLinkStripRules)
{
text = this.StripLinks(text);
}
// Find links
var matches = s_LinkRegex.Value.Matches(text);
foreach (var match in matches.Cast <Match>().Where(m => m.Success))
{
var link = match.Value;
if (link.IsNullOrEmpty())
{
continue;
}
var baseUrl = propertyBag.ResponseUri.GetLeftPath();
var normalizedLink = link.NormalizeUrl(baseUrl);
if (normalizedLink.IsNullOrEmpty())
{
continue;
}
// Add new step to crawler
await crawler.AddStepAsync(new Uri(normalizedLink), propertyBag.Step.Depth + 1,
propertyBag.Step, new Dictionary <string, object>
{
{ Resources.PropertyBagKeyOriginalUrl, new Uri(link) },
{ Resources.PropertyBagKeyOriginalReferrerUrl, propertyBag.ResponseUri }
}).ConfigureAwait(false);
}
}
public virtual async Task ProcessAsync(ICrawler crawler, PropertyBag propertyBag)
{
AspectF.Define.
NotNull(crawler, "crawler").
NotNull(propertyBag, "propertyBag");
if (propertyBag.StatusCode != HttpStatusCode.OK)
{
return;
}
if (!IsHtmlContent(propertyBag.ContentType))
{
return;
}
var htmlDoc = new HtmlDocument
{
OptionAddDebuggingAttributes = false,
OptionAutoCloseOnEnd = true,
OptionFixNestedTags = true,
OptionReadEncoding = true
};
using (var reader = propertyBag.GetResponse())
{
var documentEncoding = htmlDoc.DetectEncoding(reader);
reader.Seek(0, SeekOrigin.Begin);
if (!documentEncoding.IsNull())
{
htmlDoc.Load(reader, documentEncoding, true);
}
else
{
htmlDoc.Load(reader, true);
}
}
var originalContent = htmlDoc.DocumentNode.OuterHtml;
if (this.HasTextStripRules || this.HasSubstitutionRules)
{
var content = this.StripText(originalContent);
content = this.Substitute(content, propertyBag.Step);
using (TextReader tr = new StringReader(content))
{
htmlDoc.Load(tr);
}
}
propertyBag["HtmlDoc"].Value = htmlDoc;
var nodes = htmlDoc.DocumentNode.SelectNodes("//title");
// Extract Title
if (!nodes.IsNull())
{
propertyBag.Title = string.Join(";", nodes.
Select(n => n.InnerText).
ToArray()).Trim();
}
// Extract Meta Data
nodes = htmlDoc.DocumentNode.SelectNodes("//meta[@content and @name]");
if (!nodes.IsNull())
{
propertyBag["Meta"].Value = (
from entry in nodes
let name = entry.Attributes["name"]
let content = entry.Attributes["content"]
where !name.IsNull() && !name.Value.IsNullOrEmpty() && !content.IsNull() && !content.Value.IsNullOrEmpty()
select name.Value + ": " + content.Value).ToArray();
}
// Extract text
propertyBag.Text = htmlDoc.ExtractText().Trim();
if (this.HasLinkStripRules || this.HasTextStripRules)
{
var content = this.StripLinks(originalContent);
using (TextReader tr = new StringReader(content))
{
htmlDoc.Load(tr);
}
}
var baseUrl = propertyBag.ResponseUri.GetLeftPath();
// Extract Head Base
nodes = htmlDoc.DocumentNode.SelectNodes("//head/base[@href]");
if (!nodes.IsNull())
{
baseUrl =
nodes.
Select(entry => new { entry, href = entry.Attributes["href"] }).
Where(@t => [email protected]() && [email protected]() &&
Uri.IsWellFormedUriString(@t.href.Value, UriKind.RelativeOrAbsolute)).
Select(@t => @t.href.Value).
AddToEnd(baseUrl).
FirstOrDefault();
}
// Extract Links
var links = htmlDoc.GetLinks();
foreach (var link in links.Links.Union(links.References))
{
if (link.IsNullOrEmpty())
{
continue;
}
var decodedLink = ExtendedHtmlUtility.HtmlEntityDecode(link);
var normalizedLink = this.NormalizeLink(baseUrl, decodedLink);
if (normalizedLink.IsNullOrEmpty())
{
continue;
}
await crawler.AddStepAsync(new Uri(normalizedLink), propertyBag.Step.Depth + 1,
propertyBag.Step, new Dictionary <string, object>
{
{ Resources.PropertyBagKeyOriginalUrl, link },
{ Resources.PropertyBagKeyOriginalReferrerUrl, propertyBag.ResponseUri }
}).ConfigureAwait(false);
}
}
/// <summary>
/// Add crawler step process synchronously.
/// </summary>
/// <param name="crawler">Crawler for which start crawling.</param>
/// <param name="uri">Url to crawl</param>
/// <param name="depth">Depth of the url</param>
public static void AddStep(this ICrawler crawler, Uri uri, int depth)
{
crawler.AddStepAsync(uri, depth).Wait();
}
/// <summary>
/// Queue a new step on the crawler queue
/// </summary>
/// <param name="uri">url to crawl</param>
/// <param name="depth">depth of the url</param>
public static async Task AddStepAsync(this ICrawler crawler, Uri uri, int depth)
{
await crawler.AddStepAsync(uri, depth, null, null).ConfigureAwait(false);
}
/// <summary>
/// Queue a new step on the crawler queue
/// </summary>
/// <param name="uri">url to crawl</param>
/// <param name="depth">depth of the url</param>
public static async Task AddStepAsync(this ICrawler crawler, Uri uri, int depth)
{
await crawler.AddStepAsync(uri, depth, null, null);
}
