public override void eventUniversal(crawlReportingStageEnum stage, directReporterBase __parent, crawlerDomainTask __task, modelSpiderSiteRecord wRecord)
{
switch (stage)
{
case crawlReportingStageEnum.DLCPreinitiation:
wRecord.context.OnTargetPageAttached += new modelSpiderSiteRecordEvent(onTargetPageAttached);
imbMCRepository mcRepo = mcm.activeRepository;
imbMCWebSite wRepo = mcRepo.GetWebSite(wRecord.domainInfo, true, loger);
if (!webSiteReposByDomain.ContainsKey(wRecord.domain))
{
webSiteReposByDomain.Add(wRecord.domain, wRepo);
}
else
{
loger.log("DLC sent to CrawlToMC plugin second time: " + wRecord.domain);
}
mcRepo.siteTable.AddOrUpdate(wRepo.entry);
wRepo.SaveDataStructure(mcRepo.folder, loger);
break;
}
}
/// <summary>
/// Gets all web pages registered in the <see cref="imbMCWebSite.pageTable" /> loaded, if <c>takeSettings</c> used then returns sampleTake
/// </summary>
/// <param name="site">The site repo to take pages for</param>
/// <param name="output">The log output.</param>
/// <param name="takeSettings">If specified it will return only fraction of web sites, according to the sampling settings.</param>
/// <returns>
/// All web pages in the site repo , or sampleTake if sampling settings specified
/// </returns>
public List <imbMCWebPage> GetAllWebPages(imbMCWebSite site, ILogBuilder output = null, samplingSettings takeSettings = null)
{
if (output == null)
{
output = aceLog.loger;
}
var all = site.pageTable.GetList();
if (takeSettings != null)
{
all = new sampleTake <imbMCWebPageEntry>(all, takeSettings);
}
List <imbMCWebPage> pages = new List <imbMCWebPage>();
foreach (var pe in all)
{
var repo = pe.HashCode.LoadDataStructure <imbMCWebPage>(site.folder, output);
if (repo != null)
{
pages.Add(repo);
}
}
return(pages);
}
/// <summary>
/// Gets the web page repository by resolved URL
/// </summary>
/// <param name="site">The site to query page from</param>
/// <param name="url">The fully resolved URL.</param>
/// <param name="autoCreate">if set to <c>true</c> [automatic create].</param>
/// <param name="output">The output.</param>
/// <returns></returns>
/// <exception cref="ArgumentException">Url must be in full and resolved form, and must come from the same root domain name (different TLD is allowed) - url</exception>
public imbMCWebPage GetWebPage(imbMCWebSite site, string url, bool autoCreate = false, ILogBuilder output = null)
{
if (output == null)
{
output = aceLog.loger;
}
if (!url.Contains(site.domainInfo.domainRootName))
{
if (doThrowOnDomainMismatch)
{
throw new ArgumentException($"Url [{url}] must be in full and resolved form, and must come from the same root domain name [{site.entry.domain}] (different TLD is allowed)", nameof(url));
}
}
string HashCode = md5.GetMd5Hash(url);
if (!autoCreate && (!site.pageTable.ContainsKey(HashCode)))
{
output.log($"Page repository {HashCode} (for: {url}) not found at {site.folder.path}");
return(null);
}
var entry = site.pageTable.GetOrCreate(HashCode);
imbMCWebPage repo = HashCode.LoadDataStructure <imbMCWebPage>(site.folder, output);
repo.deploy(entry);
site.pageTable.AddOrUpdate(entry);
return(repo);
}
public override void eventDLCFinished(directReporterBase __spider, crawlerDomainTask __task, modelSpiderSiteRecord __wRecord)
{
imbMCRepository mcRepo = mcm.activeRepository;
imbMCWebSite wRepo = webSiteReposByDomain[__wRecord.domain];
mcRepo.siteTable.AddOrUpdate(wRepo.entry);
wRepo.SaveDataStructure(mcRepo.folder, loger);
}
private void onTargetPageAttached(modelSpiderSiteRecord __wRecord, modelSpiderSiteRecordEventArgs __args)
{
imbMCRepository mcRepo = mcm.activeRepository;
imbMCWebSite wRepo = webSiteReposByDomain[__wRecord.domain];
ISpiderTarget target = __args.Target;
if (mcRepo.isTargetProper(target))
{
imbMCWebPage pRepo = mcRepo.BuildWebPage(target, wRepo, loger);
pRepo.indexEntry = imbWEMManager.index.pageIndexTable.GetPageForUrl(target.url);
pRepo.HtmlSourceCode = __args.sourceHtml;
pRepo.XmlSourceCode = __args.sourceXml;
pRepo.SaveDataStructure(wRepo.folder, loger);
}
}
/// <summary>
/// Builds the web page repository using <see cref="ISpiderTarget"/> crawl information
/// </summary>
/// <param name="target">Target information</param>
/// <param name="site">The site to build page for</param>
/// <param name="output">The output for console/log</param>
/// <returns>Built or updated web page repository</returns>
public imbMCWebPage BuildWebPage(ISpiderTarget target, imbMCWebSite site, ILogBuilder output = null)
{
imbMCWebPage page = GetWebPage(site, target.url, true, output);
ISpiderPage sPage = target.page;
page.entry.AnchorTextAll = sPage.captions.toCsvInLine(",");
page.entry.ClickDepth = sPage.iterationDiscovery;
page.entry.ResolvedRelativeURL = site.domainInfo.GetURLWithoutDomainName(target.url);
page.deploy(page.entry);
page.indexEntry = target.GetIndexPage();
page.TextContent = target.pageText;
page.name = target.page.name;
var htmlDoc = target.GetHtmlDocument();
if (htmlDoc != null)
{
page.HtmlSourceCode = htmlDoc.DocumentNode.OuterHtml; // ; = target.contentBlocks;
}
else
{
}
page.Blocks = new List <imbCommonModels.contentBlock.nodeBlock>();
page.TermTable = target.tokens.GetCompiledTable(output);
target.contentBlocks.ForEach(x => page.Blocks.Add(x));
site.pageTable.AddOrUpdate(page.entry);
page.SaveDataStructure(site.folder, output);
return(page);
}
/// <summary>
/// Builds or updates web site repositorium using crawling information.
/// </summary>
/// <param name="targetCollection">Collection of SpiderTargets, populated by DLC crawl</param>
/// <param name="domainInfo">DLC domain information</param>
/// <param name="output">The output.</param>
/// <returns>
/// Reference to created or updated web site repository
/// </returns>
/// <remarks>
/// This method uses completed DLC information to create <see cref="imbMCWebSite" /> repository and <see cref="imbMCWebPage" /> for all proper targets
/// </remarks>
public imbMCWebSite BuildWebSite(ISpiderTargetCollection targetCollection, domainAnalysis domainInfo, ILogBuilder output = null)
{
//Int32 siteCount = siteTable.Count;
int pageCount = 0;
imbMCWebSite repo = GetWebSite(domainInfo, true, output);
pageCount = repo.pageTable.Count;
if (pageCount == 0)
{
loger.log("Web site repository created [" + domainInfo.domainName + "]");
}
List <ISpiderTarget> crawledTargets = targetCollection.GetLoaded();
foreach (ISpiderTarget target in crawledTargets)
{
if (isTargetProper(target))
{
BuildWebPage(target, repo);
}
}
int nPageCount = repo.pageTable.Count - pageCount;
if (nPageCount > 0)
{
loger.log("Repository [" + domainInfo.domainName + "] expanded for [" + nPageCount + "] new pages, in total [" + (pageCount + nPageCount) + "] pages.");
}
siteTable.AddOrUpdate(repo.entry);
repo.SaveDataStructure(folder, output);
return(repo);
}
/// <summary>
/// Gets web site repositorium by clean domain name, like: "koplas.co.rs" for http://www.koplas.co.rs
/// </summary>
/// <param name="domainInfo">The domain information.</param>
/// <param name="autoCreate">if set to <c>true</c> it will automatically create new entry and new repository</param>
/// <param name="output">The log/diagnostic output</param>
/// <returns></returns>
public imbMCWebSite GetWebSite(domainAnalysis domainInfo, bool autoCreate = false, ILogBuilder output = null)
{
if (!autoCreate && (!siteTable.ContainsKey(domainInfo.domainRootName)))
{
return(null);
}
if (output == null)
{
output = aceLog.loger;
}
imbMCWebSiteEntry entry = siteTable.GetOrCreate(domainInfo.domainRootName);
entry.domainProperUrl = domainInfo.urlProper;
imbMCWebSite repo = entry.domain.LoadDataStructure <imbMCWebSite>(folder, output);
repo.domainInfo = domainInfo;
repo.deploy(entry);
entry.domainProperUrl = repo.domainInfo.urlProper;
siteTable.AddOrUpdate(entry);
if (repo.folder == null)
{
if (output != null)
{
output.log("Warning: folder instance is null in web site repo [" + repo.name + "]");
}
}
return(repo);
}
