/** Crawl Delay ***********************************************************/
public async Task <int> GetCrawlDelay(string Url)
{
int Delay = 0;
Robots robot;
if (!MacroscopePreferencesManager.GetFollowRobotsProtocol())
{
return(Delay);
}
robot = await this.FetchRobot(Url : Url);
if (robot != null)
{
long CrawlDelayTime = robot.CrawlDelay(userAgent: this.UserAgentName());
if (CrawlDelayTime == 0)
{
CrawlDelayTime = robot.CrawlDelay("*");
}
if (CrawlDelayTime > 0)
{
Delay = (int)(CrawlDelayTime / 1000);
}
DebugMsg(string.Format("ROBOTS CrawlDelayTime: {0}", CrawlDelayTime));
DebugMsg(string.Format("ROBOTS Delay: {0}", Delay));
}
return(Delay);
}
/** Sitemaps **************************************************************/
public async Task <List <string> > GetSitemapsAsList(string Url)
{
List <string> SitemapsList = new List <string>();
if (MacroscopePreferencesManager.GetFollowRobotsProtocol())
{
Robots robot = await this.FetchRobot(Url : Url);
try
{
if ((robot != null) && (robot.Sitemaps != null))
{
foreach (Sitemap SitemapEntry in robot.Sitemaps)
{
string SitemapUrl = SitemapEntry.Url.ToString();
string SitemapUrlAbs = MacroscopeHttpUrlUtils.MakeUrlAbsolute(BaseUrl: Url, Url: SitemapUrl);
SitemapsList.Add(SitemapUrlAbs);
this.DebugMsg(string.Format("ROBOTS SitemapUrl: {0}", SitemapUrl));
}
}
}
catch (Exception ex)
{
this.DebugMsg(ex.Message);
}
}
return(SitemapsList);
}
/**************************************************************************/
public MacroscopeJobWorker(MacroscopeJobMaster JobMaster)
{
this.SuppressDebugMsg = true;
this.JobMaster = JobMaster;
this.DocCollection = this.JobMaster.GetDocCollection();
this.AllowedHosts = this.JobMaster.GetAllowedHosts();
this.IncludeExcludeUrls = this.JobMaster.GetIncludeExcludeUrls();
if (MacroscopePreferencesManager.GetCrawlDelay() > 0)
{
this.CrawlDelay = MacroscopePreferencesManager.GetCrawlDelay();
}
if (MacroscopePreferencesManager.GetFollowRobotsProtocol())
{
if (this.JobMaster.GetCrawlDelay() > 0)
{
this.CrawlDelay = this.JobMaster.GetCrawlDelay();
}
}
}
/** -------------------------------------------------------------------- **/
public async Task <bool> ApplyRobotRule(string Url)
{
bool Allowed = true;
if (MacroscopePreferencesManager.GetFollowRobotsProtocol())
{
Allowed = await this.CheckRobotRule(Url : Url);
}
return(Allowed);
}
/** Generate Robot URL ****************************************************/
public static string GenerateRobotUrl(string Url)
{
string RobotUrl = null;
if (MacroscopePreferencesManager.GetFollowRobotsProtocol())
{
DebugMsgStatic(string.Format("ROBOTS Disabled: {0}", Url));
Uri BaseUri = null;
string BaseUriPort = "";
Uri RobotsUri = null;
string RobotsTxtUrl = null;
try
{
BaseUri = new Uri(Url, UriKind.Absolute);
if (BaseUri.Port > 0)
{
BaseUriPort = string.Format(":{0}", BaseUri.Port);
}
RobotsUri = new Uri(
string.Format(
"{0}://{1}{2}{3}",
BaseUri.Scheme,
BaseUri.Host,
BaseUriPort,
"/robots.txt"
),
UriKind.Absolute
);
RobotsTxtUrl = RobotsUri.ToString();
}
catch (InvalidOperationException ex)
{
DebugMsgStatic(string.Format("GenerateRobotUrl: {0}", ex.Message));
}
catch (UriFormatException ex)
{
DebugMsgStatic(string.Format("GenerateRobotUrl: {0}", ex.Message));
}
if (!string.IsNullOrEmpty(RobotsTxtUrl))
{
RobotUrl = RobotsTxtUrl;
}
}
return(RobotUrl);
}
/** ROBOT RULES ***********************************************************/
public Boolean ApplyRobotRule(string Url)
{
Boolean Allowed = false;
if (!MacroscopePreferencesManager.GetFollowRobotsProtocol())
{
DebugMsg(string.Format("ROBOTS Disabled: {0}", Url));
return(true);
}
else
{
Robots robot = this.FetchRobot(Url: Url);
Uri BaseUri = null;
try
{
BaseUri = new Uri(Url, UriKind.Absolute);
}
catch (UriFormatException ex)
{
DebugMsg(string.Format("ApplyRobotRule: {0}", ex.Message));
}
catch (Exception ex)
{
DebugMsg(string.Format("ApplyRobotRule: {0}", ex.Message));
}
if ((robot != null) && (BaseUri != null))
{
if (robot.IsPathAllowed("*", BaseUri.AbsolutePath))
{
Allowed = true;
}
else
{
DebugMsg(string.Format("ROBOTS Disallowed: {0}", Url));
DebugMsg(string.Format("ROBOTS AbsolutePath: {0}", BaseUri.AbsolutePath));
}
}
}
return(Allowed);
}
/**************************************************************************/
public void SetPrefsFormControlFields()
{
{ // Configure Display Options
this.checkBoxPauseDisplayDuringScan.Checked = MacroscopePreferencesManager.GetPauseDisplayDuringScan();
this.checkBoxShowProgressDialogues.Checked = MacroscopePreferencesManager.GetShowProgressDialogues();
}
{ //Configure Form Fields
/** Spidering Control ---------------------------------------------- **/
this.numericUpDownDepth.Minimum = -1;
this.numericUpDownDepth.Maximum = 10000;
this.numericUpDownPageLimit.Minimum = -1;
this.numericUpDownPageLimit.Maximum = 10000;
this.numericUpDownCrawlDelay.Minimum = 0;
this.numericUpDownCrawlDelay.Maximum = 60;
this.numericUpDownMaxRetries.Minimum = 0;
this.numericUpDownMaxRetries.Maximum = 10;
}
{
/** WebProxy Options ----------------------------------------------- **/
this.comboBoxProxyType.SelectedIndex = (int)MacroscopePreferencesManager.GetProxyType();
/** Server Certificate Options --------------------------------------- **/
this.checkBoxServerCertificateValidation.Checked = MacroscopePreferencesManager.GetServerCertificateValidation();
/** Spidering Control ---------------------------------------------- **/
this.numericUpDownMaxThreads.Value = MacroscopePreferencesManager.GetMaxThreads();
this.numericUpDownDepth.Value = MacroscopePreferencesManager.GetDepth();
this.numericUpDownPageLimit.Value = MacroscopePreferencesManager.GetPageLimit();
this.numericUpDownCrawlDelay.Value = MacroscopePreferencesManager.GetCrawlDelay();
this.numericUpDownRequestTimeout.Value = (Decimal)MacroscopePreferencesManager.GetRequestTimeout();
this.numericUpDownMaxRetries.Value = (Decimal)MacroscopePreferencesManager.GetMaxRetries();
this.checkBoxCrawlStrictUrlCheck.Checked = MacroscopePreferencesManager.GetCrawlStrictUrlCheck();
this.checkBoxCheckExternalLinks.Checked = MacroscopePreferencesManager.GetCheckExternalLinks();
this.checkBoxFetchExternalLinks.Checked = MacroscopePreferencesManager.GetFetchExternalLinks();
this.checkBoxFollowRobotsProtocol.Checked = MacroscopePreferencesManager.GetFollowRobotsProtocol();
this.checkBoxFollowSitemapLinks.Checked = MacroscopePreferencesManager.GetFollowSitemapLinks();
this.checkBoxProbeHumansText.Checked = MacroscopePreferencesManager.GetProbeHumansText();
this.checkBoxCheckRedirects.Checked = MacroscopePreferencesManager.GetCheckRedirects();
this.checkBoxFollowRedirects.Checked = MacroscopePreferencesManager.GetFollowRedirects();
this.checkBoxFollowNoFollow.Checked = MacroscopePreferencesManager.GetFollowNoFollow();
this.checkBoxIgnoreQueries.Checked = MacroscopePreferencesManager.GetIgnoreQueries();
this.checkBoxIgnoreHashFragments.Checked = MacroscopePreferencesManager.GetIgnoreHashFragments();
this.checkBoxFollowCanonicalLinks.Checked = MacroscopePreferencesManager.GetFollowCanonicalLinks();
this.checkBoxFollowAlternateLinks.Checked = MacroscopePreferencesManager.GetFollowAlternateLinks();
this.checkBoxFollowHrefLangLinks.Checked = MacroscopePreferencesManager.GetFollowHrefLangLinks();
this.checkBoxDowncaseLinks.Checked = MacroscopePreferencesManager.GetDowncaseLinks();
this.checkBoxFetchStylesheets.Checked = MacroscopePreferencesManager.GetFetchStylesheets();
this.checkBoxFetchJavascripts.Checked = MacroscopePreferencesManager.GetFetchJavascripts();
this.checkBoxFetchImages.Checked = MacroscopePreferencesManager.GetFetchImages();
this.checkBoxFetchAudio.Checked = MacroscopePreferencesManager.GetFetchAudio();
this.checkBoxFetchVideo.Checked = MacroscopePreferencesManager.GetFetchVideo();
this.checkBoxFetchXml.Checked = MacroscopePreferencesManager.GetFetchXml();
this.checkBoxFetchBinaries.Checked = MacroscopePreferencesManager.GetFetchBinaries();
this.checkBoxScanSitesInList.Checked = MacroscopePreferencesManager.GetScanSitesInList();
this.checkBoxProbeParentFolderUrls.Checked = MacroscopePreferencesManager.GetProbeParentFolderUrls();
this.checkBoxProbeHead404sWithGet.Checked = MacroscopePreferencesManager.GetProbeHead404sWithGet();
/** Analysis Options ----------------------------------------------- **/
this.checkBoxResolveAddresses.Checked = MacroscopePreferencesManager.GetResolveAddresses();
this.checkBoxCheckHreflangs.Checked = MacroscopePreferencesManager.GetCheckHreflangs();
this.checkBoxDetectLanguage.Checked = MacroscopePreferencesManager.GetDetectLanguage();
this.checkBoxProcessStylesheets.Checked = MacroscopePreferencesManager.GetProcessStylesheets();
this.checkBoxProcessJavascripts.Checked = MacroscopePreferencesManager.GetProcessJavascripts();
this.checkBoxProcessImages.Checked = MacroscopePreferencesManager.GetProcessImages();
this.checkBoxProcessPdfs.Checked = MacroscopePreferencesManager.GetProcessPdfs();
this.checkBoxProcessAudio.Checked = MacroscopePreferencesManager.GetProcessAudio();
this.checkBoxProcessVideo.Checked = MacroscopePreferencesManager.GetProcessVideo();
this.checkBoxProcessXml.Checked = MacroscopePreferencesManager.GetProcessXml();
this.checkBoxProcessBinaries.Checked = MacroscopePreferencesManager.GetProcessBinaries();
this.numericUpDownRedirectChainsMaxHops.Value = MacroscopePreferencesManager.GetRedirectChainsMaxHops();
this.checkBoxWarnAboutInsecureLinks.Checked = MacroscopePreferencesManager.GetWarnAboutInsecureLinks();
this.checkBoxEnableTextIndexing.Checked = MacroscopePreferencesManager.GetEnableTextIndexing();
this.checkBoxCaseSensitiveTextIndexing.Checked = MacroscopePreferencesManager.GetCaseSensitiveTextIndexing();
this.checkBoxDisregardHtml5ElementNav.Checked = MacroscopePreferencesManager.GetDisregardHtml5ElementNav();
this.checkBoxDisregardHtml5ElementHeader.Checked = MacroscopePreferencesManager.GetDisregardHtml5ElementHeader();
this.checkBoxDisregardHtml5ElementFooter.Checked = MacroscopePreferencesManager.GetDisregardHtml5ElementFooter();
this.checkBoxDetectQrCodeInImage.Checked = MacroscopePreferencesManager.GetDetectQrCodeInImage();
/** SEO Options ---------------------------------------------------- **/
this.numericUpDownTitleMinLen.Value = MacroscopePreferencesManager.GetTitleMinLen();
this.numericUpDownTitleMaxLen.Value = MacroscopePreferencesManager.GetTitleMaxLen();
this.numericUpDownTitleMinWords.Value = MacroscopePreferencesManager.GetTitleMinWords();
this.numericUpDownTitleMaxWords.Value = MacroscopePreferencesManager.GetTitleMaxWords();
this.numericUpDownTitleMaxPixelWidth.Value = MacroscopePreferencesManager.GetTitleMaxPixelWidth();
this.numericUpDownDescriptionMinLen.Value = MacroscopePreferencesManager.GetDescriptionMinLen();
this.numericUpDownDescriptionMaxLen.Value = MacroscopePreferencesManager.GetDescriptionMaxLen();
this.numericUpDownDescriptionMinWords.Value = MacroscopePreferencesManager.GetDescriptionMinWords();
this.numericUpDownDescriptionMaxWords.Value = MacroscopePreferencesManager.GetDescriptionMaxWords();
this.numericUpDownMaxHeadingDepth.Value = MacroscopePreferencesManager.GetMaxHeadingDepth();
this.checkBoxAnalyzeKeywordsInText.Checked = MacroscopePreferencesManager.GetAnalyzeKeywordsInText();
this.checkBoxAnalyzeTextReadability.Checked = MacroscopePreferencesManager.GetAnalyzeTextReadability();
this.comboBoxAnalyzeTextReadabilityEnglishAlgorithm.SelectedIndex = (int)MacroscopePreferencesManager.GetAnalyzeTextReadabilityEnglishAlgorithm();
this.checkBoxEnableLevenshteinDeduplication.Checked = MacroscopePreferencesManager.GetEnableLevenshteinDeduplication();
this.comboBoxLevenshteinAnalysisLevel.SelectedIndex = (int)MacroscopePreferencesManager.GetLevenshteinAnalysisLevel();
this.numericUpDownMaxLevenshteinSizeDifference.Value = MacroscopePreferencesManager.GetMaxLevenshteinSizeDifference();
this.numericUpDownMaxLevenshteinDistance.Value = MacroscopePreferencesManager.GetMaxLevenshteinDistance();
this.checkBoxAnalyzeClickPaths.Checked = MacroscopePreferencesManager.GetAnalyzeClickPaths();
// TODO: Finish implementing click path analysis:
#if DEBUG
this.groupBoxPageNavigationAnalysis.Visible = true;
#else
this.groupBoxPageNavigationAnalysis.Visible = false;
#endif
/** Custom Filter Options ------------------------------------------ **/
this.checkBoxCustomFiltersEnable.Checked = MacroscopePreferencesManager.GetCustomFiltersEnable();
this.numericUpDownCustomFiltersMaxItems.Value = MacroscopePreferencesManager.GetCustomFiltersMaxItems();
this.checkBoxCustomFiltersApplyToHtml.Checked = MacroscopePreferencesManager.GetCustomFiltersApplyToHtml();
this.checkBoxCustomFiltersApplyToCss.Checked = MacroscopePreferencesManager.GetCustomFiltersApplyToCss();
this.checkBoxCustomFiltersApplyToJavascripts.Checked = MacroscopePreferencesManager.GetCustomFiltersApplyToJavascripts();
this.checkBoxCustomFiltersApplyToText.Checked = MacroscopePreferencesManager.GetCustomFiltersApplyToText();
this.checkBoxCustomFiltersApplyToXml.Checked = MacroscopePreferencesManager.GetCustomFiltersApplyToXml();
/** Extractor Options ---------------------------------------------- **/
this.checkBoxDataExtractorsEnable.Checked = MacroscopePreferencesManager.GetDataExtractorsEnable();
this.checkBoxDataExtractorsCleanWhiteSpace.Checked = MacroscopePreferencesManager.GetDataExtractorsCleanWhiteSpace();
this.numericUpDownDataExtractorsMaxItemsCssSelectors.Value = MacroscopePreferencesManager.GetDataExtractorsMaxItemsCssSelectors();
this.numericUpDownDataExtractorsMaxItemsRegexes.Value = MacroscopePreferencesManager.GetDataExtractorsMaxItemsRegexes();
this.numericUpDownDataExtractorsMaxItemsXpaths.Value = MacroscopePreferencesManager.GetDataExtractorsMaxItemsXpaths();
this.checkBoxDataExtractorsApplyToHtml.Checked = MacroscopePreferencesManager.GetDataExtractorsApplyToHtml();
this.checkBoxDataExtractorsApplyToCss.Checked = MacroscopePreferencesManager.GetDataExtractorsApplyToCss();
this.checkBoxDataExtractorsApplyToJavascripts.Checked = MacroscopePreferencesManager.GetDataExtractorsApplyToJavascripts();
this.checkBoxDataExtractorsApplyToText.Checked = MacroscopePreferencesManager.GetDataExtractorsApplyToText();
this.checkBoxDataExtractorsApplyToPdf.Checked = MacroscopePreferencesManager.GetDataExtractorsApplyToPdf();
this.checkBoxDataExtractorsApplyToXml.Checked = MacroscopePreferencesManager.GetDataExtractorsApplyToXml();
/** Export Options ------------------------------------------------- **/
this.checkBoxSitemapIncludeLinkedPdfs.Checked = MacroscopePreferencesManager.GetSitemapIncludeLinkedPdfs();
/** Ignore Errors Settings ----------------------------------------- **/
this.checkBoxIgnoreErrors410.Checked = MacroscopePreferencesManager.GetIgnoreErrors410();
this.checkBoxIgnoreErrors451.Checked = MacroscopePreferencesManager.GetIgnoreErrors451();
}
}
/** Fetch Robot ***********************************************************/
public Robots FetchRobot(string Url)
{
Robots robot = null;
if (!MacroscopePreferencesManager.GetFollowRobotsProtocol())
{
DebugMsg(string.Format("ROBOTS Disabled: {0}", Url));
return(robot);
}
Uri BaseUri = null;
Uri RobotsUri = null;
string RobotsTxtUrl = null;
try
{
BaseUri = new Uri(Url, UriKind.Absolute);
string BaseUriPort = "";
if (BaseUri.Port > 0)
{
BaseUriPort = string.Format(":{0}", BaseUri.Port);
}
RobotsUri = new Uri(
string.Format(
"{0}://{1}{2}{3}",
BaseUri.Scheme,
BaseUri.Host,
BaseUriPort,
"/robots.txt"
),
UriKind.Absolute
);
RobotsTxtUrl = RobotsUri.ToString();
}
catch (InvalidOperationException ex)
{
DebugMsg(string.Format("FetchRobot: {0}", ex.Message));
}
catch (UriFormatException ex)
{
DebugMsg(string.Format("FetchRobot: {0}", ex.Message));
}
/*
* lock( this.BadRobots )
* {
* if( !this.BadRobots.ContainsKey( RobotsUri ) )
* {
* return( robot );
* }
* }
*/
if (!string.IsNullOrEmpty(RobotsTxtUrl))
{
lock (this.RobotSquad)
{
if (this.RobotSquad.ContainsKey(RobotsTxtUrl))
{
robot = this.RobotSquad[RobotsTxtUrl];
}
else
{
String RobotsText = this.FetchRobotTextFile(RobotsUri: RobotsUri);
if (RobotsText != null)
{
robot = new Robots(content: RobotsText);
this.RobotSquad.Add(RobotsTxtUrl, robot);
}
}
}
}
return(robot);
}
