/**************************************************************************/ 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(); } }
/**************************************************************************/ public bool CanApplyCustomFiltersToDocument(MacroscopeDocument msDoc) { bool CanApply = true; if ( (msDoc == null) || (msDoc.GetIsRedirect()) || (msDoc.GetStatusCode() != HttpStatusCode.OK) || (!msDoc.GetIsInternal())) { CanApply = false; } else { if ( !(msDoc.IsDocumentType(Type: MacroscopeConstants.DocumentType.HTML) || msDoc.IsDocumentType(Type: MacroscopeConstants.DocumentType.CSS) || msDoc.IsDocumentType(Type: MacroscopeConstants.DocumentType.JAVASCRIPT) || msDoc.IsDocumentType(Type: MacroscopeConstants.DocumentType.TEXT) || msDoc.IsDocumentType(Type: MacroscopeConstants.DocumentType.XML)) ) { CanApply = false; } else { switch (msDoc.GetDocumentType()) { case MacroscopeConstants.DocumentType.HTML: if (!MacroscopePreferencesManager.GetCustomFiltersApplyToHtml()) { CanApply = false; } break; case MacroscopeConstants.DocumentType.CSS: if (!MacroscopePreferencesManager.GetCustomFiltersApplyToCss()) { CanApply = false; } break; case MacroscopeConstants.DocumentType.JAVASCRIPT: if (!MacroscopePreferencesManager.GetCustomFiltersApplyToJavascripts()) { CanApply = false; } break; case MacroscopeConstants.DocumentType.TEXT: if (!MacroscopePreferencesManager.GetCustomFiltersApplyToText()) { CanApply = false; } break; case MacroscopeConstants.DocumentType.XML: if (!MacroscopePreferencesManager.GetCustomFiltersApplyToXml()) { CanApply = false; } break; default: break; } } } return(CanApply); }
/** -------------------------------------------------------------------- **/ private async Task _ProcessCssPage() { MacroscopeHttpTwoClient Client = this.DocCollection.GetJobMaster().GetHttpClient(); MacroscopeHttpTwoClientResponse Response = null; string ResponseErrorCondition = null; DebugMsg(string.Format("ProcessCssPage: {0}", "")); try { Response = await Client.Get( this.GetUri(), this.ConfigureCssPageRequestHeadersCallback, this.PostProcessRequestHttpHeadersCallback ); } catch (MacroscopeDocumentException ex) { this.DebugMsg(string.Format("_ProcessCssPage :: MacroscopeDocumentException: {0}", ex.Message)); ResponseErrorCondition = ex.Message; this.SetStatusCode(HttpStatusCode.BadRequest); this.AddRemark("_ProcessCssPage", ex.Message); } catch (Exception ex) { this.DebugMsg(string.Format("_ProcessCssPage :: Exception: {0}", ex.Message)); ResponseErrorCondition = ex.Message; this.SetStatusCode(HttpStatusCode.BadRequest); this.AddRemark("_ProcessCssPage", ex.Message); } if (Response != null) { string RawData = ""; this.ProcessResponseHttpHeaders(Response: Response); /** Get Response Body ---------------------------------------------- **/ try { DebugMsg(string.Format("MIME TYPE: {0}", this.MimeType)); RawData = Response.GetContentAsString(); this.SetContentLength(Length: RawData.Length); // May need to find bytes length this.SetWasDownloaded(true); } catch (Exception ex) { DebugMsg(string.Format("Exception: {0}", ex.Message)); this.SetStatusCode(HttpStatusCode.Ambiguous); this.SetContentLength(Length: 0); } if (!string.IsNullOrEmpty(RawData)) { try { StylesheetParser CssParser = new StylesheetParser(); Stylesheet CssStylesheet = CssParser.Parse(RawData); this.ProcessCssOutlinks(CssStylesheet: CssStylesheet); } catch (Exception ex) { this.DebugMsg(string.Format("ProcessHtmlAttributeCssLinks: {0}", ex.Message)); this.AddRemark("ProcessHtmlAttributeCssLinks", ex.Message); } } else { DebugMsg(string.Format("ProcessCssPage: ERROR: {0}", this.GetUrl())); } /** Custom Filters ------------------------------------------------- **/ if (!string.IsNullOrEmpty(RawData)) { if ( MacroscopePreferencesManager.GetCustomFiltersEnable() && MacroscopePreferencesManager.GetCustomFiltersApplyToCss()) { MacroscopeCustomFilters CustomFilter = this.DocCollection.GetJobMaster().GetCustomFilter(); if ((CustomFilter != null) && (CustomFilter.IsEnabled())) { this.ProcessGenericCustomFiltered( CustomFilter: CustomFilter, GenericText: RawData ); } } } /** Data Extractors ------------------------------------------------ **/ if (!string.IsNullOrEmpty(RawData)) { if ( MacroscopePreferencesManager.GetDataExtractorsEnable() && MacroscopePreferencesManager.GetDataExtractorsApplyToCss()) { this.ProcessGenericDataExtractors(GenericText: RawData); } } /** Title ---------------------------------------------------------- **/ { MatchCollection reMatches = Regex.Matches(this.DocUrl, "/([^/]+)$"); string DocumentTitle = null; foreach (Match match in reMatches) { if (match.Groups[1].Value.Length > 0) { DocumentTitle = match.Groups[1].Value.ToString(); break; } } if (DocumentTitle != null) { this.SetTitle(DocumentTitle, MacroscopeConstants.TextProcessingMode.NO_PROCESSING); DebugMsg(string.Format("TITLE: {0}", this.GetTitle())); } else { DebugMsg(string.Format("TITLE: {0}", "MISSING")); } } } if (ResponseErrorCondition != null) { this.ProcessErrorCondition(ResponseErrorCondition); } }
/**************************************************************************/ public Boolean CanApplyCustomFiltersToDocument(MacroscopeDocument msDoc) { Boolean CanApply = true; if ( (msDoc == null) || (msDoc.GetIsRedirect()) || (msDoc.GetStatusCode() != HttpStatusCode.OK) || (!msDoc.GetIsInternal())) { CanApply = false; } else { if ( !(msDoc.GetIsHtml() || msDoc.GetIsCss() || msDoc.GetIsJavascript() || msDoc.GetIsText() || msDoc.GetIsXml())) { CanApply = false; } else { if ( msDoc.GetIsHtml() && (!MacroscopePreferencesManager.GetCustomFiltersApplyToHtml())) { CanApply = false; } else if ( msDoc.GetIsCss() && (!MacroscopePreferencesManager.GetCustomFiltersApplyToCss())) { CanApply = false; } else if ( msDoc.GetIsJavascript() && (!MacroscopePreferencesManager.GetCustomFiltersApplyToJavascripts())) { CanApply = false; } else if ( msDoc.GetIsText() && (!MacroscopePreferencesManager.GetCustomFiltersApplyToText())) { CanApply = false; } else if ( msDoc.GetIsXml() && (!MacroscopePreferencesManager.GetCustomFiltersApplyToXml())) { CanApply = false; } } } return(CanApply); }
/**************************************************************************/ private void ProcessCssPage() { HttpWebRequest req = null; HttpWebResponse res = null; string ResponseErrorCondition = null; Boolean IsAuthenticating = false; DebugMsg(string.Format("ProcessCssPage: {0}", "")); try { req = WebRequest.CreateHttp(this.DocUrl); req.Method = "GET"; req.Timeout = this.Timeout; req.KeepAlive = false; req.AutomaticDecompression = DecompressionMethods.GZip | DecompressionMethods.Deflate; this.PrepareRequestHttpHeaders(req: req); IsAuthenticating = this.AuthenticateRequest(req); MacroscopePreferencesManager.EnableHttpProxy(req); res = ( HttpWebResponse )req.GetResponse(); } catch (UriFormatException ex) { DebugMsg(string.Format("ProcessCssPage :: UriFormatException: {0}", ex.Message)); ResponseErrorCondition = ex.Message; } catch (TimeoutException ex) { DebugMsg(string.Format("ProcessCssPage :: TimeoutException: {0}", ex.Message)); ResponseErrorCondition = ex.Message; } catch (WebException ex) { DebugMsg(string.Format("ProcessCssPage :: WebException: {0}", ex.Message)); DebugMsg(string.Format("ProcessCssPage :: WebException: {0}", ex.Status)); DebugMsg(string.Format("ProcessCssPage :: WebException: {0}", ( int )ex.Status)); ResponseErrorCondition = ex.Status.ToString(); } if (res != null) { string RawData = ""; this.ProcessResponseHttpHeaders(req, res); if (IsAuthenticating) { this.VerifyOrPurgeCredential(); } // Get Response Body try { DebugMsg(string.Format("MIME TYPE: {0}", this.MimeType)); Stream ResponseStream = res.GetResponseStream(); StreamReader ResponseStreamReader; if (this.GetCharacterEncoding() != null) { ResponseStreamReader = new StreamReader(ResponseStream, this.GetCharacterEncoding()); } else { ResponseStreamReader = new StreamReader(ResponseStream); } RawData = ResponseStreamReader.ReadToEnd(); this.ContentLength = RawData.Length; // May need to find bytes length this.SetWasDownloaded(true); } catch (WebException ex) { DebugMsg(string.Format("WebException: {0}", ex.Message)); if (ex.Response != null) { this.SetStatusCode((( HttpWebResponse )ex.Response).StatusCode); } else { this.SetStatusCode(( HttpStatusCode )ex.Status); } RawData = ""; this.ContentLength = 0; } catch (Exception ex) { DebugMsg(string.Format("Exception: {0}", ex.Message)); this.SetStatusCode(HttpStatusCode.BadRequest); this.ContentLength = 0; } if (!string.IsNullOrEmpty(RawData)) { ExCSS.Parser ExCssParser = new ExCSS.Parser(); ExCSS.StyleSheet ExCssStylesheet = ExCssParser.Parse(RawData); this.ProcessCssOutlinks(ExCssStylesheet); } else { DebugMsg(string.Format("ProcessCssPage: ERROR: {0}", this.GetUrl())); } /** Custom Filters ------------------------------------------------- **/ if (!string.IsNullOrEmpty(RawData)) { if ( MacroscopePreferencesManager.GetCustomFiltersEnable() && MacroscopePreferencesManager.GetCustomFiltersApplyToCss()) { MacroscopeCustomFilters CustomFilter = this.DocCollection.GetJobMaster().GetCustomFilter(); if ((CustomFilter != null) && (CustomFilter.IsEnabled())) { this.ProcessGenericCustomFiltered( CustomFilter: CustomFilter, GenericText: RawData ); } } } /** Data Extractors ------------------------------------------------ **/ if (!string.IsNullOrEmpty(RawData)) { if ( MacroscopePreferencesManager.GetDataExtractorsEnable() && MacroscopePreferencesManager.GetDataExtractorsApplyToCss()) { this.ProcessGenericDataExtractors(GenericText: RawData); } } /** Title ---------------------------------------------------------- **/ { MatchCollection reMatches = Regex.Matches(this.DocUrl, "/([^/]+)$"); string DocumentTitle = null; foreach (Match match in reMatches) { if (match.Groups[1].Value.Length > 0) { DocumentTitle = match.Groups[1].Value.ToString(); break; } } if (DocumentTitle != null) { this.SetTitle(DocumentTitle, MacroscopeConstants.TextProcessingMode.NO_PROCESSING); DebugMsg(string.Format("TITLE: {0}", this.GetTitle())); } else { DebugMsg(string.Format("TITLE: {0}", "MISSING")); } } res.Close(); res.Dispose(); } if (ResponseErrorCondition != null) { this.ProcessErrorCondition(ResponseErrorCondition); } }