/**************************************************************************/
private void BuildWorksheetPageDuplicatePages(
MacroscopeJobMaster JobMaster,
XLWorkbook wb,
string WorksheetLabel
)
{
var ws = wb.Worksheets.Add(WorksheetLabel);
int iRow = 1;
int iCol = 1;
int iColMax = 1;
decimal DocCount = 0;
decimal DocListCount = 0;
decimal CountOuter = 0;
decimal CountInner = 0;
MacroscopeDocumentCollection DocCollection = JobMaster.GetDocCollection();
MacroscopeAllowedHosts AllowedHosts = JobMaster.GetAllowedHosts();
Dictionary <string, bool> CrossCheckList;
CrossCheckList = MacroscopeLevenshteinAnalysis.GetCrossCheckList(
Capacity: DocCollection.CountDocuments()
);
DocCount = ( decimal )DocCollection.CountDocuments();
{
ws.Cell(iRow, iCol).Value = "Status Code";
iCol++;
ws.Cell(iRow, iCol).Value = "Status";
iCol++;
ws.Cell(iRow, iCol).Value = "Origin URL";
iCol++;
ws.Cell(iRow, iCol).Value = "Distance";
iCol++;
ws.Cell(iRow, iCol).Value = "Similar URL";
}
iColMax = iCol;
iRow++;
foreach (string UrlLeft in DocCollection.DocumentUrls())
{
MacroscopeDocument msDocLeft = DocCollection.GetDocumentByUrl(Url: UrlLeft);
MacroscopeLevenshteinAnalysis LevenshteinAnalysis = null;
CountOuter++;
CountInner = 0;
if (DocCount > 0)
{
this.ProgressForm.UpdatePercentages(
Title: null,
Message: null,
MajorPercentage: -1,
ProgressLabelMajor: string.Format("Documents Processed: {0}", CountOuter),
MinorPercentage: (( decimal )100 / DocCount) * CountOuter,
ProgressLabelMinor: UrlLeft,
SubMinorPercentage: 0,
ProgressLabelSubMinor: ""
);
}
if (msDocLeft.GetIsExternal())
{
continue;
}
if (!msDocLeft.IsDocumentType(Type: MacroscopeConstants.DocumentType.HTML))
{
continue;
}
LevenshteinAnalysis = new MacroscopeLevenshteinAnalysis(
msDoc: msDocLeft,
SizeDifference: MacroscopePreferencesManager.GetMaxLevenshteinSizeDifference(),
Threshold: MacroscopePreferencesManager.GetMaxLevenshteinDistance(),
CrossCheckList: CrossCheckList,
IPercentageDone: this
);
Dictionary <MacroscopeDocument, int> DocList;
DocList = LevenshteinAnalysis.AnalyzeDocCollection(
DocCollection: DocCollection
);
DocListCount = ( decimal )DocList.Count;
foreach (MacroscopeDocument msDocDuplicate in DocList.Keys)
{
int StatusCode = ( int )msDocLeft.GetStatusCode();
HttpStatusCode Status = msDocLeft.GetStatusCode();
string UrlDuplicate = msDocDuplicate.GetUrl();
int Distance = DocList[msDocDuplicate];
CountInner++;
iCol = 1;
if (DocCount > 0)
{
this.ProgressForm.UpdatePercentages(
Title: null,
Message: null,
MajorPercentage: -1,
ProgressLabelMajor: string.Format("Documents Processed: {0}", CountOuter),
MinorPercentage: (( decimal )100 / DocCount) * CountOuter,
ProgressLabelMinor: UrlLeft,
SubMinorPercentage: (( decimal )100 / DocListCount) * CountInner,
ProgressLabelSubMinor: UrlDuplicate
);
}
this.InsertAndFormatStatusCodeCell(ws, iRow, iCol, StatusCode);
iCol++;
this.InsertAndFormatStatusCodeCell(ws, iRow, iCol, Status);
iCol++;
this.InsertAndFormatUrlCell(ws, iRow, iCol, UrlLeft);
if (AllowedHosts.IsInternalUrl(Url: UrlLeft))
{
ws.Cell(iRow, iCol).Style.Font.SetFontColor(XLColor.Green);
}
else
{
ws.Cell(iRow, iCol).Style.Font.SetFontColor(XLColor.Gray);
}
iCol++;
this.InsertAndFormatContentCell(ws, iRow, iCol, Distance.ToString());
if (Distance <= MacroscopePreferencesManager.GetMaxLevenshteinDistance())
{
ws.Cell(iRow, iCol).Style.Font.SetFontColor(XLColor.Red);
}
else
{
ws.Cell(iRow, iCol).Style.Font.SetFontColor(XLColor.Green);
}
iCol++;
this.InsertAndFormatUrlCell(ws, iRow, iCol, UrlDuplicate);
if (AllowedHosts.IsInternalUrl(Url: UrlDuplicate))
{
ws.Cell(iRow, iCol).Style.Font.SetFontColor(XLColor.Green);
}
else
{
ws.Cell(iRow, iCol).Style.Font.SetFontColor(XLColor.Gray);
}
iRow++;
if (this.ProgressForm.Cancelled())
{
break;
}
}
if (this.ProgressForm.Cancelled())
{
break;
}
//Thread.Yield();
}
{
var rangeData = ws.Range(1, 1, iRow - 1, iColMax);
var excelTable = rangeData.CreateTable();
}
}
/**************************************************************************/
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();
}
}
/**************************************************************************/
private void BuildWorksheetPageDuplicatePages(
MacroscopeJobMaster JobMaster,
CsvWriter ws
)
{
decimal DocCount = 0;
decimal DocListCount = 0;
decimal CountOuter = 0;
decimal CountInner = 0;
MacroscopeDocumentCollection DocCollection = JobMaster.GetDocCollection();
MacroscopeAllowedHosts AllowedHosts = JobMaster.GetAllowedHosts();
Dictionary <string, bool> CrossCheckList;
CrossCheckList = MacroscopeLevenshteinAnalysis.GetCrossCheckList(
Capacity: DocCollection.CountDocuments()
);
DocCount = ( decimal )DocCollection.CountDocuments();
{
ws.WriteField("Status Code");
ws.WriteField("Status");
ws.WriteField("Origin URL");
ws.WriteField("Distance");
ws.WriteField("Similar URL");
ws.NextRecord();
}
foreach (string UrlLeft in DocCollection.DocumentUrls())
{
MacroscopeDocument msDocLeft = DocCollection.GetDocumentByUrl(Url: UrlLeft);
MacroscopeLevenshteinAnalysis LevenshteinAnalysis = null;
CountOuter++;
CountInner = 0;
if (DocCount > 0)
{
this.ProgressForm.UpdatePercentages(
Title: null,
Message: null,
MajorPercentage: -1,
ProgressLabelMajor: string.Format("Documents Processed: {0}", CountOuter),
MinorPercentage: (( decimal )100 / DocCount) * CountOuter,
ProgressLabelMinor: UrlLeft,
SubMinorPercentage: 0,
ProgressLabelSubMinor: ""
);
}
if (msDocLeft.GetIsExternal())
{
continue;
}
if (!msDocLeft.IsDocumentType(Type: MacroscopeConstants.DocumentType.HTML))
{
continue;
}
LevenshteinAnalysis = new MacroscopeLevenshteinAnalysis(
msDoc: msDocLeft,
SizeDifference: MacroscopePreferencesManager.GetMaxLevenshteinSizeDifference(),
Threshold: MacroscopePreferencesManager.GetMaxLevenshteinDistance(),
CrossCheckList: CrossCheckList,
IPercentageDone: this
);
Dictionary <MacroscopeDocument, int> DocList;
DocList = LevenshteinAnalysis.AnalyzeDocCollection(
DocCollection: DocCollection
);
DocListCount = ( decimal )DocList.Count;
foreach (MacroscopeDocument msDocDuplicate in DocList.Keys)
{
int StatusCode = ( int )msDocLeft.GetStatusCode();
HttpStatusCode Status = msDocLeft.GetStatusCode();
string UrlDuplicate = msDocDuplicate.GetUrl();
int Distance = DocList[msDocDuplicate];
CountInner++;
if (DocCount > 0)
{
this.ProgressForm.UpdatePercentages(
Title: null,
Message: null,
MajorPercentage: -1,
ProgressLabelMajor: string.Format("Documents Processed: {0}", CountOuter),
MinorPercentage: (( decimal )100 / DocCount) * CountOuter,
ProgressLabelMinor: UrlLeft,
SubMinorPercentage: (( decimal )100 / DocListCount) * CountInner,
ProgressLabelSubMinor: UrlDuplicate
);
}
this.InsertAndFormatStatusCodeCell(ws, StatusCode);
this.InsertAndFormatStatusCodeCell(ws, Status);
this.InsertAndFormatUrlCell(ws, UrlLeft);
this.InsertAndFormatContentCell(ws, Distance.ToString());
this.InsertAndFormatUrlCell(ws, UrlDuplicate);
ws.NextRecord();
if (this.ProgressForm.Cancelled())
{
break;
}
}
if (this.ProgressForm.Cancelled())
{
break;
}
//Thread.Yield();
}
}