//The Content Type supported by the parser
#endregion
#region Constructor and Singleton Instance members
/// <summary>
/// The constructor is private so that only the class itself can create an instance.
/// </summary>
private TextParser()
{
//Initialize the synchronization mechanism
mutex = new Mutex();
//Initialize the Regular Expressions
hrefRegex = new Regex(@"(http|https)://[\w]+(\.[\w]+)+([\w\-\.,@?^=%&:/~\+#]*[\w\-\@?^=%&/~\+#])?", RegexOptions.CultureInvariant | RegexOptions.Multiline | RegexOptions.IgnoreCase | RegexOptions.Compiled);
//use "(http|ftp|https)://[\w]+(\.[\w]+)+([\w\-\.,@?^=%&:/~\+#]*[\w\-\@?^=%&/~\+#])?" to enable ftp urls
sessionIDRegex = new Regex(@"([0-9a-fA-F]{40,64})|([\{|\(]?[0-9a-fA-F]{8}[-]?([0-9a-fA-F]{4}[-]?){3}[0-9a-fA-F]{12}[\)|\}]?)$", RegexOptions.CultureInvariant | RegexOptions.Multiline | RegexOptions.IgnoreCase | RegexOptions.Compiled); //@"^([0-9a-f]{32})|(\{?[0-9a-f]{8}-([0-9a-f]{4}-){3}-[0-9a-f]{12}\}?)$"
spacesRegex = new Regex(@"\s+", RegexOptions.CultureInvariant | RegexOptions.Multiline | RegexOptions.IgnoreCase | RegexOptions.Compiled);
//Initialize the filters
robotsFilter = RobotsFilter.Instance();
domainFilter = DomainFilter.Instance();
//Get a reference to the global variables and application settings
globals = Globals.Instance();
}
/// <summary>
/// Provides a global access point for the single instance of the <see cref="RobotsFilter"/>
/// class.
/// </summary>
/// <returns>A reference to the single instance of <see cref="RobotsFilter"/>.</returns>
public static RobotsFilter Instance()
{
if (instance==null)
{
//Make sure the call is thread-safe. We cannot use the private mutex since
//it hasn't yet been initialized - it gets initialized in the constructor.
Mutex imutex=new Mutex();
imutex.WaitOne();
if( instance == null )
{
instance = new RobotsFilter();
}
imutex.Close();
}
return instance;
}
/// <summary>
/// Provides a global access point for the single instance of the <see cref="RobotsFilter"/>
/// class.
/// </summary>
/// <returns>A reference to the single instance of <see cref="RobotsFilter"/>.</returns>
public static RobotsFilter Instance()
{
if (instance == null)
{
//Make sure the call is thread-safe. We cannot use the private mutex since
//it hasn't yet been initialized - it gets initialized in the constructor.
Mutex imutex = new Mutex();
imutex.WaitOne();
if (instance == null)
{
instance = new RobotsFilter();
}
imutex.Close();
}
return(instance);
}
/// <summary>
/// Constructs a new istance of the <see cref="Crawler"/> class and initializes its
/// properties with the default values. There should be only one instance of Crawler
/// </summary>
public Crawler()
{
//first of all get a reference to the global variables because they are needed
//in order to initialize some variables.
globals = Globals.Instance();
mustStop = false;
stopping = false;
state = CrawlerState.Stopped;
stats = new long[10] {0,0,0,0,0,0,0,0,0,0};
numThreads = (int)globals.Settings.ConnectionSpeed;
runningThreads = 0;
sendResultsThread = null;
synchronizeThread = null;
crawlingThreads = null;
syncBackOff = new ExponentialBackoff(BackoffSpeed.Declining);
downloadBackOff = new ExponentialBackoff(BackoffSpeed.Fast);
urlsToCrawl = new Queue();
resultFileNames = new Queue();
crawledUrls = new ArrayList();
queueSize = 0;
dataFileName = String.Empty;
defaultEncoding = Encoding.GetEncoding("ISO-8859-7");
htmlParser = HtmlParser.Instance();
textParser = TextParser.Instance();
pdfParser = PdfParser.Instance();
swfParser = SwfParser.Instance();
robotsFilter = RobotsFilter.Instance();
domainFilter = DomainFilter.Instance();
hostRequestFilter =HostRequestFilter.Instance();
hostBanFilter = HostBanFilter.Instance();
proxy = WebServiceProxy.Instance();
}
/// <summary>
/// Constructs a new istance of the <see cref="Crawler"/> class and initializes its
/// properties with the default values. The constructor is private so that only the
/// class itself can create an instance.
/// </summary>
private Crawler()
{
//first of all get a reference to the global variables because they are needed
//in order to initialize some variables.
globals = Globals.Instance();
mustStop = false;
stopping = false;
state = CrawlerState.Stopped;
stats = new long[10] {0,0,0,0,0,0,0,0,0,0};
numThreads = (int)globals.Settings.ConnectionSpeed;
runningThreads = 0;
//sendResultsThread = null;
synchronizeThread = null;
crawlingThreads = null;
syncBackOff = new Backoff(BackoffSpeed.Declining, 30000);
downloadBackOff = new Backoff(BackoffSpeed.Fast);
urlsToCrawl = new Queue();
resultFileNames = new Queue();
crawledUrls = new ArrayList();
queueSize = 0;
dataFileName = String.Empty;
defaultEncoding = Encoding.GetEncoding("ISO-8859-7");
defaultGreekEncoding = Encoding.GetEncoding(1253);
contentRegex = new Regex("<meta\\s*http-equiv=([^>])*charset\\s*=\\s*([^>])*(utf-7|utf-8|utf-16|windows-1253)([^>])*>",RegexOptions.CultureInvariant|RegexOptions.Multiline|RegexOptions.IgnoreCase|RegexOptions.Compiled);
htmlParser = HtmlParser.Instance();
textParser = TextParser.Instance();
pdfParser = PdfParser.Instance();
swfParser = SwfParser.Instance();
nullParser = NullParser.Instance();
robotsFilter = RobotsFilter.Instance();
robotsFilter.LoadEntries();
domainFilter = DomainFilter.Instance();
hostRequestFilter = HostRequestFilter.Instance();
hostBanFilter = HostBanFilter.Instance();
//proxy = WebServiceProxy.Instance();
proxy = CrawlWaveServerProxy.Instance(globals);
}
private Regex spacesRegex; //Regular Expression for compacting white space characters
#endregion Fields
#region Constructors
/// <summary>
/// The constructor is private so that only the class itself can create an instance.
/// </summary>
private TextParser()
{
//Initialize the synchronization mechanism
mutex=new Mutex();
//Initialize the Regular Expressions
hrefRegex=new Regex(@"(http|https)://[\w]+(\.[\w]+)+([\w\-\.,@?^=%&:/~\+#]*[\w\-\@?^=%&/~\+#])?", RegexOptions.CultureInvariant|RegexOptions.Multiline|RegexOptions.IgnoreCase|RegexOptions.Compiled);
//use "(http|ftp|https)://[\w]+(\.[\w]+)+([\w\-\.,@?^=%&:/~\+#]*[\w\-\@?^=%&/~\+#])?" to enable ftp urls
sessionIDRegex = new Regex(@"([0-9a-fA-F]{40,64})|([\{|\(]?[0-9a-fA-F]{8}[-]?([0-9a-fA-F]{4}[-]?){3}[0-9a-fA-F]{12}[\)|\}]?)$", RegexOptions.CultureInvariant|RegexOptions.Multiline|RegexOptions.IgnoreCase|RegexOptions.Compiled); //@"^([0-9a-f]{32})|(\{?[0-9a-f]{8}-([0-9a-f]{4}-){3}-[0-9a-f]{12}\}?)$"
spacesRegex = new Regex(@"\s+",RegexOptions.CultureInvariant|RegexOptions.Multiline|RegexOptions.IgnoreCase|RegexOptions.Compiled);
//Initialize the filters
robotsFilter = RobotsFilter.Instance();
domainFilter = DomainFilter.Instance();
//Get a reference to the global variables and application settings
globals = Globals.Instance();
}
private Regex stylesRegex; //Regular Expression for stylesheets
#endregion Fields
#region Constructors
/// <summary>
/// The constructor is private so that only the class itself can create an instance.
/// </summary>
private HtmlParser()
{
//Initialize the synchronization mechanism
mutex=new Mutex();
//Initialize the Regular Expressions
ahrefRegex = new Regex("href\\s*=\\s*(?:\"(?<1>[^\"]*)\"|(?<1>\\S+))", RegexOptions.CultureInvariant|RegexOptions.Multiline|RegexOptions.IgnoreCase|RegexOptions.Compiled);
baseRegex = new Regex("base\\s*href=\\s*(?:\"(?<1>[^\"]*)\"|(?<1>\\S+))",RegexOptions.CultureInvariant|RegexOptions.Multiline|RegexOptions.IgnoreCase|RegexOptions.Compiled);
charsetRegex = new Regex("<meta\\s*http-equiv=([^>])*charset\\s*=\\s*([^>])*(iso-8859-7|windows-1253)([^>])*>",RegexOptions.CultureInvariant|RegexOptions.Multiline|RegexOptions.IgnoreCase|RegexOptions.Compiled);
frameRegex = new Regex("frame\\s*.*src\\s*=\\s*(?:\"(?<1>[^\"]*)\"|(?<1>\\S+))", RegexOptions.CultureInvariant|RegexOptions.Multiline|RegexOptions.IgnoreCase|RegexOptions.Compiled);
flashRegex = new Regex("<embed\\s*([^>])*src\\s*=([^>])*type\\s*=([^>])*application/x-shockwave-flash([^>])*>", RegexOptions.CultureInvariant|RegexOptions.Multiline|RegexOptions.IgnoreCase|RegexOptions.Compiled);
refreshRegex = new Regex("<meta\\s*http-equiv=([^>])*refresh([^>])*content\\s*=\\s*\"[^>]*\">", RegexOptions.CultureInvariant|RegexOptions.Multiline|RegexOptions.IgnoreCase|RegexOptions.Compiled);
robotRegex = new Regex("<meta\\s*name\\s*=\\s*\"robots\"\\s*content\\s*=\\s*\"[^>]*\">", RegexOptions.CultureInvariant|RegexOptions.Multiline|RegexOptions.IgnoreCase|RegexOptions.Compiled);
scriptRegex = new Regex(@"(?i)<script([^>])*>(\w|\W)*</script([^>])*>",RegexOptions.CultureInvariant|RegexOptions.Multiline|RegexOptions.IgnoreCase|RegexOptions.Compiled); //@"(?i)<script([^>])*>(\w|\W)*</script([^>])*>" or @"<script[^>]*>(\w|\W)*?</script[^>]*>"
spacesRegex = new Regex(@"\s+",RegexOptions.CultureInvariant|RegexOptions.Multiline|RegexOptions.IgnoreCase|RegexOptions.Compiled);
stylesRegex = new Regex(@"<style([^>])*>(\w|\W)*</style([^>])*>",RegexOptions.CultureInvariant|RegexOptions.Multiline|RegexOptions.IgnoreCase|RegexOptions.Compiled);
stripTagRegex = new Regex("<[^>]*>", RegexOptions.CultureInvariant|RegexOptions.Multiline|RegexOptions.IgnoreCase|RegexOptions.Compiled);//<[^>]+> or >(?:(?<t>[^<]*))
sessionIDRegex = new Regex(@"([0-9a-fA-F]{40,64})|([\{|\(]?[0-9a-fA-F]{8}[-]?([0-9a-fA-F]{4}[-]?){3}[0-9a-fA-F]{12}[\)|\}]?)$", RegexOptions.CultureInvariant|RegexOptions.Multiline|RegexOptions.IgnoreCase|RegexOptions.Compiled); //@"^([0-9a-f]{32})|(\{?[0-9a-f]{8}-([0-9a-f]{4}-){3}-[0-9a-f]{12}\}?)$"
inlinedSessionIDRegex = new Regex(@"/(%28|\{)?(([0-9a-fA-F]{8}[-]?(([0-9a-fA-F]{4}[-]?){3}[0-9a-fA-F]{12}))|([0-9a-fA-F]{12,64}))(%29|\})?/", RegexOptions.CultureInvariant|RegexOptions.Multiline|RegexOptions.IgnoreCase|RegexOptions.Compiled);
//Initialize the filters
robotsFilter = RobotsFilter.Instance();
domainFilter = DomainFilter.Instance();
//Initialize the culture info to Greek (ISO)
culture = CultureInfo.CreateSpecificCulture("el-GR");
//Get a reference to the global variables and application settings
globals = Globals.Instance();
}
