public M4Spider()
{
_maxOutboundLinks = 99;
_numOutboundLinks = 0;
_avoidHttps = true;
_insideLinks = false;
_verifyBackLinks = true;
_unSpidered = new List <string>();
_spideredURLs = new List <string>();
_obLinks = new List <string>();
_avoidPatterns = new List <string>();
_proxyDomain = "";
_proxyLogin = "";
_proxyPassword = "";
_proxyPort = 0;
_backLinkDN = "";
SR = new SpiderResults();
PrxShut = new ProxyShuttler();
AvoidPatterns.Add("google.");
AvoidPatterns.Add("yahoo.");
AvoidPatterns.Add("bing.");
AvoidPatterns.Add("altavista.");
AvoidPatterns.Add("princeton.edu");
AvoidPatterns.Add("amazon.com");
AvoidPatterns.Add("baidu.com");
AvoidPatterns.Add("planet-lab.edu");
}
private bool ParsePage(StreamReader strmRdr, string DN)
{
SR.Init(_backLinkDN);
if (strmRdr != null)
{
Uri thisUri;
try
{
thisUri = new Uri(DN);
}
catch (Exception URIex)
{
_errMsg = URIex.Message;
return(false);
}
SR.Host = thisUri.Host;
SR.Path = thisUri.PathAndQuery;
SR.Port = thisUri.Port.ToString();
SR.Protocol = thisUri.Scheme;
SR.DDN = _backLinkDN;
try
{
// Create an instance of StreamReader to read from CapturePage().
// The using statement also closes the StreamReader.
_errMsg = "";
using (strmRdr)
{
string HypLinkPattern = "href=\"(?<Link>http:.*?)\"";
string sHypLinkPattern = "href=\"(?<Link>https?:.*?)\"";
string HostPattern = @"^[a-z][a-z0-9+\-.]*://([a-z0-9\-._~%!$&'()*+,;=]+@)?(?<host>[a-z0-9\-._~%]+|\[[a-z0-9\-._~%!$&'()*+,;=:]+\])";
string VerifyPattern = @"^[a-z][a-z0-9+\-.]*://([a-z0-9\-._~%!$&'()*+,;=]+@)?(?<host>.*(domain))";
string KeywordPattern = "<meta name=\"keywords\" content=\"(?<keywords>.*?)\">";
string DescriptionPattern = "<meta name=\"description\" content=\"(?<description>.*?)\">";
string val;
string source, DN_host = "";
bool lb_avoid = false;
string totalSrc = "";
MatchCollection HypLinkMatch, KeywordMatch, DescripMatch;
Match DNmatch, hlDNmatch, DNmatch2;
//totalSrc = strmRdr.ReadToEnd();
//while ((source = strmRdr.ReadToEnd()) != null)//strmRdr.ReadLine()) != null)
if ((source = strmRdr.ReadToEnd()) != null)
{
if (null != strmRdr)
{
strmRdr.Close();
}
_errMsg = "";
KeywordMatch = Regex.Matches(source, KeywordPattern, RegexOptions.IgnoreCase | RegexOptions.Singleline);
if (KeywordMatch.Count > 0)//.Success)
{
foreach (Match m1 in KeywordMatch)
{
val = m1.Groups["keywords"].ToString();
if (val != null)
{
SR.Keywords = val.ToString();
}
}
}
DescripMatch = Regex.Matches(source, DescriptionPattern, RegexOptions.Singleline);
if (DescripMatch.Count > 0)//.Success)
{
foreach (Match m2 in DescripMatch)
{
val = m2.Groups["description"].ToString();
if (val != null)
{
SR.Description = val.ToString();
}
}
}
if (!_avoidHttps)
{
HypLinkPattern = sHypLinkPattern;
}
HypLinkMatch = Regex.Matches(source, HypLinkPattern, RegexOptions.Singleline);
if (HypLinkMatch.Count > 0)//.Success)
{
DateTime startTime = DateTime.Now;
foreach (Match m in HypLinkMatch)
{
if (_numOutboundLinks >= _maxOutboundLinks)
{
_sr1 = SR;
_numUnspidered--;
_numSpidered++;
_errMsg = "";
return(true);
}
val = m.Groups["Link"].ToString();
if (val != null)
{
foreach (string item in AvoidPatterns)
{
lb_avoid = false;
if (val.ToLower().Contains(item.ToLower()))
{
lb_avoid = true;
break;
}
}
if (!lb_avoid)
{
hlDNmatch = Regex.Match(val, HostPattern);
if (hlDNmatch.Success)
{
DN_host = hlDNmatch.Groups["host"].ToString();
}
if (DN_host != null && DN_host.Trim() != "")
{
if (!_insideLinks)
{
DNmatch = Regex.Match(val, BackLinkDN);
if (!DNmatch.Success)
{
if (_verifyBackLinks) // verify the linked page has the domain linked back
{
StreamReader strmrdr2 = CapturePage(val);
if (strmrdr2 != null)
{
try
{
using (strmrdr2)
{
totalSrc = strmrdr2.ReadToEnd();
strmrdr2.Close();
VerifyPattern = VerifyPattern.Replace("(domain)", BackLinkDN);
DNmatch2 = Regex.Match(totalSrc, VerifyPattern, RegexOptions.IgnoreCase);// | RegexOptions.Singleline);
if (!DNmatch2.Success)
{
//foreach (string item in AvoidPatterns)
//{
// if (val.ToLower().Contains(item.ToLower()))
// {
// lb_avoid = true;
// break;
// }
//}
//if (!lb_avoid)
//{
SR.HyperLinks.Add(val.ToString());
_numOutboundLinks++;
_errMsg = "";
//}
//lb_avoid = false;
}
}
}
catch (Exception e3)
{
_errMsg = e3.Message;
_numFailed++;
}
}
else
{
_numFailed++;
}
}
else //then just add the link
{
SR.HyperLinks.Add(val.ToString());
_numOutboundLinks++;
_errMsg = "";
}
}
}
}
}
}
}
DateTime stopTime = DateTime.Now;
TimeSpan timeDif = stopTime - startTime;
float procTime = (float)timeDif.TotalSeconds;
bool blrtn = CycleTime("Backlinks in ParsePage", BackLinkDN, procTime);
}
if (_numOutboundLinks >= _maxOutboundLinks)
{
_sr1 = SR;
_numUnspidered--;
_numSpidered++;
_errMsg = "";
return(true);
}
}
}
_sr1 = SR;
_numUnspidered--;
_numSpidered++;
_errMsg = "";
return(true);
}
catch (Exception e1)
{
_errMsg = e1.Message;
_numFailed++;
return(false);
}
}
else
{
_numFailed++;
return(false);
}
}
