public void WriteFile(WebPage page)
{
if (_fileToDomainMap.ContainsKey(page.Uri.AbsoluteUri))
{
return;
}
Guid guid = Guid.NewGuid();
string fileName = guid.ToString() + ".html";
page.SavePage(Path.Combine(_outputPath, fileName));
_fileToDomainMap.TryAdd(page.Uri.AbsoluteUri, fileName);
}
private bool EnsurePoliteVisit(WebPage webpage)
{
// Check robotstxt to see if we can visit the page
string hostname = webpage.Uri.DnsSafeHost;
RobotsTxt robotsTxt;
if (!_robotsTxts.ContainsKey(hostname.GetHashCode()))
{
robotsTxt = _parser.Parse(hostname);
if (robotsTxt == null) // we could not get the robotstxt therefore we cannot visit and we requeue the Uri
{
_urlFrontier.AddUri(webpage.Uri);
return false;
}
_robotsTxts.TryAdd(hostname.GetHashCode(), robotsTxt);
}
else
{
robotsTxt = _robotsTxts[hostname.GetHashCode()];
}
if (!robotsTxt.CanVisit("*", webpage.Uri)) // We are not allowed to visit this page
return false;
// We are allowed to visit, ensure we are waiting enough time before next request
int address = webpage.Address;
if (_visitedServers.ContainsKey(address))
{
DateTime now = DateTime.Now;
DateTime safeVisit = new DateTime(_visitedServers[address]);
int delay = (int) safeVisit.Subtract(now).TotalMilliseconds;
if (delay > 0)
{
Thread.Sleep(delay);
}
}
return true;
}
private void UpdateTimestamp(WebPage webpage)
{
DateTime nextValidVisitTime = DateTime.Now.AddSeconds(TIME_BETWEEN_VISITS);
if (_visitedServers.ContainsKey(webpage.Address))
{
_visitedServers[webpage.Address] = nextValidVisitTime.Ticks;
}
else
{
_visitedServers.TryAdd(webpage.Address, nextValidVisitTime.Ticks);
}
}
private void ProcessUri(ParallelOptions options, Uri uri)
{
if (uri == null)
return;
WebPage webpage = new WebPage(uri);
// Is it safe to visit the webpage?
if (!EnsurePoliteVisit(webpage)) return;
// Visit webpage
webpage.LoadPage();
// make sure to update or add time for visit to the dictionary
UpdateTimestamp(webpage);
// Add webpage anchors to the queue
if (!webpage.IsLoaded)
return;
if (options.CancellationToken.IsCancellationRequested)
{
Thread.CurrentThread.Abort();
}
_store.WriteFile(webpage);
_statistics.IncrementPagesCrawled();
// Add extracted anchors to the queue
_urlFrontier.AddUriRange(webpage.GetAnchors());
}
