/// <summary>
/// Processes a FilesRow after crawling.
/// </summary>
/// <param name = "filesRow">The files row.</param>
/// <param name="webClient"></param>
/// <param name="actionManager"></param>
/// <param name="consoleManager"></param>
/// <param name="discoveryManager"></param>
/// <param name = "fileManager">The file manager.</param>
/// <param name = "fileManager">The file manager.</param>
/// <param name="memoryManager"></param>
/// <param name="ruleManager"></param>
/// <param name = "arachnodeDAO">The arachnode DAO.</param>
/// <param name = "imageManager">The image manager.</param>
public static void ProcessFile(ApplicationSettings applicationSettings, WebSettings webSettings, Crawler <TArachnodeDAO> crawler, ArachnodeDataSet.FilesRow filesRow, WebClient <TArachnodeDAO> webClient, Cache <TArachnodeDAO> cache, ActionManager <TArachnodeDAO> actionManager, ConsoleManager <TArachnodeDAO> consoleManager, CrawlerPeerManager <TArachnodeDAO> crawlerPeerManager, DiscoveryManager <TArachnodeDAO> discoveryManager, FileManager <TArachnodeDAO> fileManager, MemoryManager <TArachnodeDAO> memoryManager, RuleManager <TArachnodeDAO> ruleManager, IArachnodeDAO arachnodeDAO)
{
CacheManager <TArachnodeDAO> cacheManager = new CacheManager <TArachnodeDAO>(applicationSettings, webSettings);
CookieManager cookieManager = new CookieManager();;
CrawlRequestManager <TArachnodeDAO> crawlRequestManager = new CrawlRequestManager <TArachnodeDAO>(applicationSettings, webSettings, cache, consoleManager, discoveryManager);
DataTypeManager <TArachnodeDAO> dataTypeManager = new DataTypeManager <TArachnodeDAO>(applicationSettings, webSettings);
EncodingManager <TArachnodeDAO> encodingManager = new EncodingManager <TArachnodeDAO>(applicationSettings, webSettings);
PolitenessManager <TArachnodeDAO> politenessManager = new PolitenessManager <TArachnodeDAO>(applicationSettings, webSettings, cache);
ProxyManager <TArachnodeDAO> proxyManager = new ProxyManager <TArachnodeDAO>(applicationSettings, webSettings, consoleManager);
HtmlManager <TArachnodeDAO> htmlManager = new HtmlManager <TArachnodeDAO>(applicationSettings, webSettings, discoveryManager);
Crawl <TArachnodeDAO> crawl = new Crawl <TArachnodeDAO>(applicationSettings, webSettings, crawler, actionManager, consoleManager, cookieManager, crawlRequestManager, dataTypeManager, discoveryManager, encodingManager, htmlManager, politenessManager, proxyManager, ruleManager, true);
//create a CrawlRequest as this is what the internals of SiteCrawler.dll expect to operate on...
CrawlRequest <TArachnodeDAO> crawlRequest = new CrawlRequest <TArachnodeDAO>(new Discovery <TArachnodeDAO>(filesRow.AbsoluteUri), 1, UriClassificationType.Host, UriClassificationType.Host, 0, RenderType.None, RenderType.None);
crawlRequest.Crawl = crawl;
crawlRequest.Discovery.DiscoveryType = DiscoveryType.File;
crawlRequest.Discovery.ID = filesRow.ID;
crawlRequest.Data = filesRow.Source;
crawlRequest.ProcessData = true;
crawlRequest.WebClient = webClient;
crawlRequest.WebClient.HttpWebResponse.Headers.Clear();
//parse the ResponseHeaders from the FilesRow.ResponseHeaders string...
foreach (string responseHeader in filesRow.ResponseHeaders.Split("\r\n".ToCharArray(), StringSplitOptions.RemoveEmptyEntries))
{
string[] responseHeaderSplit = responseHeader.Split(":".ToCharArray());
string name = responseHeaderSplit[0];
string value = UserDefinedFunctions.ExtractResponseHeader(filesRow.ResponseHeaders, name, true).Value;
crawlRequest.WebClient.HttpWebResponse.Headers.Add(name, value);
}
//refresh the DataTypes in the DataTypeManager... (if necessary)...
if (dataTypeManager.AllowedDataTypes.Count == 0)
{
dataTypeManager.RefreshDataTypes();
}
crawlRequest.DataType = dataTypeManager.DetermineDataType(crawlRequest);
if (applicationSettings.InsertFiles)
{
crawlRequest.Discovery.ID = arachnodeDAO.InsertFile(crawlRequest.Parent.Uri.AbsoluteUri, crawlRequest.Discovery.Uri.AbsoluteUri, crawlRequest.WebClient.HttpWebResponse.Headers.ToString(), applicationSettings.InsertFileSource ? crawlRequest.Data : new byte[] { }, crawlRequest.DataType.FullTextIndexType, applicationSettings.ClassifyAbsoluteUris);
}
crawlRequest.ManagedDiscovery = fileManager.ManageFile(crawlRequest, crawlRequest.Discovery.ID.Value, crawlRequest.Discovery.Uri.AbsoluteUri, crawlRequest.Data, crawlRequest.DataType.FullTextIndexType, applicationSettings.ExtractFileMetaData, applicationSettings.InsertFileMetaData, applicationSettings.SaveDiscoveredFilesToDisk);
actionManager.PerformCrawlActions(crawlRequest, CrawlActionType.PostRequest, arachnodeDAO);
discoveryManager.CloseAndDisposeManagedDiscovery(crawlRequest, arachnodeDAO);
}
/// <summary>
/// Processes a WebPagesRow after crawling.
/// </summary>
/// <param name = "webPagesRow">The web pages row.</param>
/// <param name="webClient"></param>
/// <param name="actionManager"></param>
/// <param name="consoleManager"></param>
/// <param name="discoveryManager"></param>
/// <param name="memoryManager"></param>
/// <param name="ruleManager"></param>
/// <param name = "webPageManager">The web page manager.</param>
/// <param name = "arachnodeDAO">The arachnode DAO.</param>
/// <param name = "fileManager">The file manager.</param>
/// <param name = "imageManager">The image manager.</param>
public static void ProcessWebPage(ApplicationSettings applicationSettings, WebSettings webSettings, Crawler <TArachnodeDAO> crawler, ArachnodeDataSet.WebPagesRow webPagesRow, WebClient <TArachnodeDAO> webClient, Cache <TArachnodeDAO> cache, ActionManager <TArachnodeDAO> actionManager, ConsoleManager <TArachnodeDAO> consoleManager, CrawlerPeerManager <TArachnodeDAO> crawlerPeerManager, DiscoveryManager <TArachnodeDAO> discoveryManager, MemoryManager <TArachnodeDAO> memoryManager, RuleManager <TArachnodeDAO> ruleManager, WebPageManager <TArachnodeDAO> webPageManager, IArachnodeDAO arachnodeDAO)
{
CacheManager <TArachnodeDAO> cacheManager = new CacheManager <TArachnodeDAO>(applicationSettings, webSettings);
CookieManager cookieManager = new CookieManager();
CrawlRequestManager <TArachnodeDAO> crawlRequestManager = new CrawlRequestManager <TArachnodeDAO>(applicationSettings, webSettings, cache, consoleManager, discoveryManager);
DataTypeManager <TArachnodeDAO> dataTypeManager = new DataTypeManager <TArachnodeDAO>(applicationSettings, webSettings);
EncodingManager <TArachnodeDAO> encodingManager = new EncodingManager <TArachnodeDAO>(applicationSettings, webSettings);
PolitenessManager <TArachnodeDAO> politenessManager = new PolitenessManager <TArachnodeDAO>(applicationSettings, webSettings, cache);
ProxyManager <TArachnodeDAO> proxyManager = new ProxyManager <TArachnodeDAO>(applicationSettings, webSettings, consoleManager);
HtmlManager <TArachnodeDAO> htmlManager = new HtmlManager <TArachnodeDAO>(applicationSettings, webSettings, discoveryManager);
Crawl <TArachnodeDAO> crawl = new Crawl <TArachnodeDAO>(applicationSettings, webSettings, crawler, actionManager, consoleManager, cookieManager, crawlRequestManager, dataTypeManager, discoveryManager, encodingManager, htmlManager, politenessManager, proxyManager, ruleManager, true);
//create a CrawlRequest as this is what the internals of SiteCrawler.dll expect to operate on...
CrawlRequest <TArachnodeDAO> crawlRequest = new CrawlRequest <TArachnodeDAO>(new Discovery <TArachnodeDAO>(webPagesRow.AbsoluteUri), webPagesRow.CrawlDepth, UriClassificationType.Host, UriClassificationType.Host, 0, RenderType.None, RenderType.None);
crawlRequest.Crawl = crawl;
crawlRequest.Discovery.DiscoveryType = DiscoveryType.WebPage;
crawlRequest.Discovery.ID = webPagesRow.ID;
crawlRequest.Data = webPagesRow.Source;
crawlRequest.CurrentDepth = webPagesRow.CrawlDepth;
crawlRequest.Encoding = Encoding.GetEncoding(webPagesRow.CodePage);
crawlRequest.ProcessData = true;
crawlRequest.WebClient = webClient;
crawlRequest.WebClient.HttpWebResponse.Headers.Clear();
//parse the ResponseHeaders from the WebPagesRow.ResponseHeaders string...
foreach (string responseHeader in webPagesRow.ResponseHeaders.Split("\r\n".ToCharArray(), StringSplitOptions.RemoveEmptyEntries))
{
string[] responseHeaderSplit = responseHeader.Split(":".ToCharArray());
string name = responseHeaderSplit[0];
string value = UserDefinedFunctions.ExtractResponseHeader(webPagesRow.ResponseHeaders, name, true).Value;
crawlRequest.WebClient.HttpWebResponse.Headers.Add(name, value);
}
//refresh the DataTypes in the DataTypeManager... (if necessary)...
if (dataTypeManager.AllowedDataTypes.Count == 0)
{
dataTypeManager.RefreshDataTypes();
}
crawlRequest.DataType = dataTypeManager.DetermineDataType(crawlRequest);
//now, process the bytes...
encodingManager.ProcessCrawlRequest(crawlRequest, arachnodeDAO);
if (applicationSettings.InsertWebPages)
{
crawlRequest.Discovery.ID = arachnodeDAO.InsertWebPage(crawlRequest.Discovery.Uri.AbsoluteUri, crawlRequest.WebClient.HttpWebResponse.Headers.ToString(), applicationSettings.InsertWebPageSource ? crawlRequest.Data : new byte[] { }, crawlRequest.Encoding.CodePage, crawlRequest.DataType.FullTextIndexType, crawlRequest.CurrentDepth, applicationSettings.ClassifyAbsoluteUris);
}
crawlRequest.ManagedDiscovery = webPageManager.ManageWebPage(crawlRequest.Discovery.ID.Value, crawlRequest.Discovery.Uri.AbsoluteUri, crawlRequest.Data, crawlRequest.Encoding, crawlRequest.DataType.FullTextIndexType, applicationSettings.ExtractWebPageMetaData, applicationSettings.InsertWebPageMetaData, applicationSettings.SaveDiscoveredWebPagesToDisk);
//assigning FileAndImageDiscoveries isn't applicable because Files and Images need to be crawled to be properly classified... without classification we don't know whether they belong in dbo.Files or dbo.Images...
crawlRequestManager.ProcessEmailAddresses(crawlRequest, arachnodeDAO);
crawlRequestManager.ProcessHyperLinks(crawlRequest, arachnodeDAO);
actionManager.PerformCrawlActions(crawlRequest, CrawlActionType.PostRequest, arachnodeDAO);
discoveryManager.CloseAndDisposeManagedDiscovery(crawlRequest, arachnodeDAO);
}
public override void ProcessCrawlRequest(CrawlRequest <TArachnodeDAO> crawlRequest, bool obeyCrawlRules, bool executeCrawlActions)
{
IssueWebRequest(crawlRequest, "GET");
crawlRequest.DataType = _dataTypeManager.DetermineDataType(crawlRequest);
if (obeyCrawlRules)
{
_ruleManager.IsDisallowed(crawlRequest, CrawlRuleType.PreGet, _arachnodeDAO);
}
if (executeCrawlActions)
{
_actionManager.PerformCrawlActions(crawlRequest, CrawlActionType.PreGet, _arachnodeDAO);
}
if (!crawlRequest.IsDisallowed)
{
try
{
if (crawlRequest.WebClient.HttpWebResponse != null)
{
crawlRequest.ProcessData = true;
bool isLastModifiedOutdated = true;
try
{
isLastModifiedOutdated = crawlRequest.WebClient.HttpWebResponse.LastModified != DateTime.Now;
}
catch (Exception exception)
{
_arachnodeDAO.InsertException(crawlRequest.Parent.Uri.AbsoluteUri, crawlRequest.Discovery.Uri.AbsoluteUri, exception, false);
}
if (isLastModifiedOutdated)
{
switch (crawlRequest.DataType.DiscoveryType)
{
case DiscoveryType.File:
if (ApplicationSettings.AssignFileAndImageDiscoveries) //ANODET: robots.txt
{
ArachnodeDataSet.FilesRow filesRow = _arachnodeDAO.GetFile(crawlRequest.Discovery.Uri.AbsoluteUri);
if (filesRow == null)
{
crawlRequest.ProcessData = true;
}
else
{
if (!filesRow.IsResponseHeadersNull())
{
DateTime lastModified;
SqlString lastModifiedValue = UserDefinedFunctions.ExtractResponseHeader(filesRow.ResponseHeaders, "Last-Modified: ", false);
if (!lastModifiedValue.IsNull && DateTime.TryParse(lastModifiedValue.Value, out lastModified))
{
//crawlRequest.WebClient.HttpWebResponse.LastModified will equal DateTime.Now (or close to it) if the 'Last-Modified' ResponseHeader is not present...
if ((crawlRequest.WebClient.HttpWebResponse).LastModified > lastModified)
{
crawlRequest.ProcessData = true;
}
else
{
crawlRequest.ProcessData = false;
}
}
else
{
crawlRequest.ProcessData = false;
}
}
else
{
crawlRequest.ProcessData = true;
}
if (!crawlRequest.ProcessData)
{
if (filesRow.Source.Length != 0)
{
crawlRequest.Data = filesRow.Source;
}
else
{
string discoveryPath = _discoveryManager.GetDiscoveryPath(ApplicationSettings.DownloadedFilesDirectory, crawlRequest.Discovery.Uri.AbsoluteUri, crawlRequest.DataType.FullTextIndexType);
if (File.Exists(discoveryPath))
{
crawlRequest.Data = File.ReadAllBytes(discoveryPath);
}
else
{
try
{
throw new Exception("The 'LastModified' HttpResponse Header indicated that the Data was not stale, but the Data (Source) could not be found in the Files database table or at _applicationSettings.DownloadedFilesDirectory. Therefore, the data was re-downloaded from the server. The File file may have been deleted from disk or the 'Source' column in the 'Files' table may have been cleared or a previous crawl may have crawled with both _applicationSettings.InsertFileSource = false and _applicationSettings.SaveDiscoveredFilesToDisk = false.");
}
catch (Exception exception)
{
_arachnodeDAO.InsertException(crawlRequest.Parent.Uri.AbsoluteUri, crawlRequest.Discovery.Uri.AbsoluteUri, exception, false);
}
crawlRequest.ProcessData = true;
}
}
}
}
}
else
{
crawlRequest.ProcessData = false;
}
break;
case DiscoveryType.Image:
if (ApplicationSettings.AssignFileAndImageDiscoveries)
{
ArachnodeDataSet.ImagesRow imagesRow = _arachnodeDAO.GetImage(crawlRequest.Discovery.Uri.AbsoluteUri);
if (imagesRow == null)
{
crawlRequest.ProcessData = true;
}
else
{
if (!imagesRow.IsResponseHeadersNull())
{
DateTime lastModified;
SqlString lastModifiedValue = UserDefinedFunctions.ExtractResponseHeader(imagesRow.ResponseHeaders, "Last-Modified: ", false);
if (!lastModifiedValue.IsNull && DateTime.TryParse(lastModifiedValue.Value, out lastModified))
{
//crawlRequest.WebClient.HttpWebResponse.LastModified will equal DateTime.Now (or close to it) if the 'Last-Modified' ResponseHeader is not present...
if (crawlRequest.WebClient.HttpWebResponse.LastModified > lastModified)
{
crawlRequest.ProcessData = true;
}
else
{
crawlRequest.ProcessData = false;
}
}
else
{
crawlRequest.ProcessData = false;
}
if (!crawlRequest.ProcessData)
{
if (imagesRow.Source.Length != 0)
{
crawlRequest.Data = imagesRow.Source;
}
else
{
string discoveryPath = _discoveryManager.GetDiscoveryPath(ApplicationSettings.DownloadedImagesDirectory, crawlRequest.Discovery.Uri.AbsoluteUri, crawlRequest.DataType.FullTextIndexType);
if (File.Exists(discoveryPath))
{
crawlRequest.Data = File.ReadAllBytes(discoveryPath);
}
else
{
try
{
throw new Exception("The 'LastModified' HttpResponse Header indicated that the Data was not stale, but the Data (Source) could not be found in the Images database table or at _applicationSettings.DownloadedImagesDirectory. Therefore, the data was downloaded from the server. The Image file may have been deleted from disk or the 'Source' column in the 'Images' table may have been cleared. A previous crawl may have crawled with both _applicationSettings.InsertImageSource = false and _applicationSettings.SaveDiscoveredImagesToDisk = false.");
}
catch (Exception exception)
{
_arachnodeDAO.InsertException(crawlRequest.Parent.Uri.AbsoluteUri, crawlRequest.Discovery.Uri.AbsoluteUri, exception, false);
}
crawlRequest.ProcessData = true;
}
}
}
}
else
{
crawlRequest.ProcessData = true;
}
}
}
else
{
crawlRequest.ProcessData = false;
}
break;
case DiscoveryType.WebPage:
ArachnodeDataSet.WebPagesRow webPagesRow = _arachnodeDAO.GetWebPage(crawlRequest.Discovery.Uri.AbsoluteUri);
if (webPagesRow == null)
{
crawlRequest.ProcessData = true;
}
else
{
if ((crawlRequest.WebClient.HttpWebResponse).LastModified > webPagesRow.LastDiscovered)
{
crawlRequest.ProcessData = true;
}
else
{
crawlRequest.ProcessData = false;
}
if (!crawlRequest.ProcessData)
{
if (webPagesRow.Source.Length != 0)
{
crawlRequest.Data = webPagesRow.Source;
}
else
{
string discoveryPath = _discoveryManager.GetDiscoveryPath(ApplicationSettings.DownloadedWebPagesDirectory, crawlRequest.Discovery.Uri.AbsoluteUri, crawlRequest.DataType.FullTextIndexType);
if (File.Exists(discoveryPath))
{
crawlRequest.Data = File.ReadAllBytes(discoveryPath);
}
else
{
try
{
throw new Exception("The 'LastModified' HttpResponse Header indicated that the Data was not stale, but the Data (Source) could not be found in the WebPages database table or at _applicationSettings.DownloadedWebPagesDirectory. Therefore, the data was re-downloaded from the server. The WebPage file may have been deleted from disk or the 'Source' column in the 'WebPages' table may have been cleared or a previous crawl may have crawled with both _applicationSettings.InsertWebPageSource = false and _applicationSettings.SaveDiscoveredWebPagesToDisk = false.");
}
catch (Exception exception)
{
_arachnodeDAO.InsertException(crawlRequest.Parent.Uri.AbsoluteUri, crawlRequest.Discovery.Uri.AbsoluteUri, exception, false);
}
crawlRequest.ProcessData = true;
}
}
}
}
break;
case DiscoveryType.None:
crawlRequest.ProcessData = true;
break;
}
}
}
}
catch (Exception exception)
{
_arachnodeDAO.InsertException(crawlRequest.Parent.Uri.AbsoluteUri, crawlRequest.Discovery.Uri.AbsoluteUri, exception, false);
}
if (crawlRequest.ProcessData)
{
if (crawlRequest.Data != null)
{
}
if (crawlRequest.RenderType == RenderType.None)
{
if (crawlRequest.Discovery.Uri.Scheme.ToLowerInvariant() != "ftp")
{
if (crawlRequest.WebClient.HttpWebResponse != null && crawlRequest.WebClient.HttpWebResponse.Method == "HEAD")
{
IssueWebRequest(crawlRequest, "GET");
}
if (crawlRequest.WebClient.HttpWebResponse != null)
{
crawlRequest.Data = crawlRequest.WebClient.DownloadHttpData(crawlRequest.Discovery.Uri.AbsoluteUri, crawlRequest.WebClient.HttpWebResponse.ContentEncoding.ToLowerInvariant() == "gzip", crawlRequest.WebClient.HttpWebResponse.ContentEncoding.ToLowerInvariant() == "deflate", crawlRequest.Crawl.Crawler.CookieContainer);
}
}
else
{
crawlRequest.Data = crawlRequest.WebClient.DownloadFtpData(crawlRequest.Discovery.Uri.AbsoluteUri);
}
}
else
{
RendererResponse rendererResponse = crawlRequest.Crawl.Crawler.Engine.Render(crawlRequest, RenderAction.Render, crawlRequest.RenderType);
if (rendererResponse != null)
{
if (rendererResponse.HTMLDocumentClass != null)
{
crawlRequest.Encoding = Encoding.GetEncoding(rendererResponse.HTMLDocumentClass.charset);
string outerHTML = rendererResponse.HTMLDocumentClass.documentElement.outerHTML;
crawlRequest.Data = crawlRequest.Encoding.GetBytes(outerHTML);
crawlRequest.DecodedHtml = HttpUtility.HtmlDecode(outerHTML);
crawlRequest.Html = outerHTML;
crawlRequest.HtmlDocument = rendererResponse.HTMLDocumentClass;
}
crawlRequest.RendererMessage = rendererResponse.RendererMessage;
}
}
}
}
else
{
if (crawlRequest.Data == null)
{
}
}
if (crawlRequest.Data == null)
{
crawlRequest.Data = new byte[0];
}
}
