private void Callback(Page page)
{
Interlocked.Increment(ref _numberOfPagesCrawled);
try
{
page.Examiner = _examinerFactory.CreateExaminer(_log, page);
}
catch (Exception exception)
{
_log.Log(new LogMessage(LogType.Error, "Uncaught exception when creating examiner", exception, page.Uri));
return;
}
foreach (var inspector in _inspectors)
{
try
{
inspector.InspectPage(page);
}
catch (Exception exception)
{
_log.Log(new InspectorLogMessage(inspector, new LogMessage(LogType.Error, "Uncaught exception when calling InspectPage", exception, page.Uri)));
}
}
}
private void CheckExternalLinks(Page page)
{
var externalLinks = page.GetExternalLinks();
foreach (var externalLink in externalLinks)
{
try
{
HttpStatusCode httpStatusCode;
if (_checkedExternalLinks.ContainsKey(externalLink))
httpStatusCode = _checkedExternalLinks[externalLink];
else
{
httpStatusCode = GetHttpStatusCode(externalLink);
_checkedExternalLinks.GetOrAdd(externalLink, httpStatusCode);
}
if (httpStatusCode != HttpStatusCode.OK)
Error(externalLink, page.Uri, httpStatusCode);
}
catch (Exception exception)
{
Error("Could not get page", exception, externalLink);
}
}
}
public override void InspectPage(Page page)
{
foreach (var link in page.Examiner.Links)
{
if (_targets.Exists(t => t == link) && IsInfoEnabled)
Info($"{link} referenced from {page.Uri}");
}
}
public override void InspectPage(Page page)
{
foreach (var link in page.Examiner.Links)
{
if (_schemes.Exists(s => s == link.Scheme) && IsInfoEnabled)
Info($"{link} found on {page.Uri}");
}
}
public override void InspectPage(Page page)
{
var externalLinks = page.GetExternalLinks();
foreach (var externalLink in externalLinks)
{
if(!_urisToIgnore.Contains(externalLink) && IsInfoEnabled)
Info($"{externalLink} found on {page.Uri}");
}
}
public override void InspectPage(Page page)
{
if (page.ContentByteSize > _byteSize)
{
Error(GetMessage(page));
return;
}
if (IsInfoEnabled)
Info(GetMessage(page));
}
public override void InspectPage(Page page)
{
if (page.Response.StatusCode == HttpStatusCode.OK)
{
if (_checkExternalLinks)
CheckExternalLinks(page);
return;
}
Error(page.Uri, page.Referrer, page.Response.StatusCode);
}
public override void InspectPage(Page page)
{
var timeSpan = page.RequestCompletedOnUtc - page.RequestStartedOnUtc;
var loadTime = (int) timeSpan.TotalMilliseconds;
if (loadTime > _milliseconds)
{
Error(GetMessage(page.Uri, loadTime));
return;
}
if (IsInfoEnabled)
Info(GetMessage(page.Uri, loadTime));
}
private Page Convert(CrawledPage crawledPage)
{
var requestCompleted = crawledPage.RequestCompleted;
if (crawledPage.DownloadContentCompleted.HasValue)
requestCompleted = crawledPage.DownloadContentCompleted.Value;
var byteSize = 0;
if (crawledPage.Content.Bytes != null)
byteSize = crawledPage.Content.Bytes.Length;
var page = new Page(
crawledPage.Uri,
crawledPage.ParentUri,
crawledPage.RequestStarted.ToUniversalTime(),
requestCompleted.ToUniversalTime(),
Convert(crawledPage.HttpWebRequest),
Convert(crawledPage.HttpWebResponse),
crawledPage.Content.Text,
byteSize);
return page;
}
public IExaminer CreateExaminer(ILog log, Page page)
{
return new AngleSharpExaminer(log, page.Uri, page.Content);
}
private string GetMessage(Page page)
{
return $"{page.Uri} is {page.ContentByteSize} bytes";
}
public abstract void InspectPage(Page page);