public override byte[] GetFileSource(string fileAbsoluteUriOrID, IArachnodeDAO arachnodeDAO)
{
if (ApplicationSettings.DownloadedFilesDirectory == null)
{
throw new Exception("_applicationSettings.DownloadedFilesDirectory is null. This is usually the result of failing to initialize the Application configuration from the ArachnodeDAO.");
}
ArachnodeDataSet.FilesRow filesRow = arachnodeDAO.GetFile(fileAbsoluteUriOrID);
if (filesRow != null)
{
if (filesRow.Source.Length != 0)
{
return(filesRow.Source);
}
else
{
string discoveryPath = GetDiscoveryPath(ApplicationSettings.DownloadedFilesDirectory, filesRow.AbsoluteUri, filesRow.FullTextIndexType);
if (!File.Exists(discoveryPath))
{
throw new Exception("Could not find the File Source in the database or on disk.");
}
return(File.ReadAllBytes(discoveryPath));
}
}
return(null);
}
private void FileUtilities_OnFileProcessed(ArachnodeDataSet.FilesRow filesRow, string message)
{
BeginInvoke(new MethodInvoker(delegate
{
rtbPostProcessingStatus.Text = message + Environment.NewLine + rtbPostProcessingStatus.Text;
if (rtbPostProcessingStatus.Text.Length > 10000)
{
rtbPostProcessingStatus.Text = rtbPostProcessingStatus.Text.Substring(0, 10000);
}
}));
//Application.DoEvents();
//Thread.Sleep(100);
}
private void nudFileID_ValueChanged(object sender, EventArgs e)
{
_filesRow = _arachnodeDAO.GetFile(nudFileID.Value.ToString());
if (_filesRow != null)
{
_fileDiscoveryPath = _discoveryManager.GetDiscoveryPath(_applicationSettings.DownloadedFilesDirectory, _filesRow.AbsoluteUri, _filesRow.FullTextIndexType);
llFileDiscoveryPathDirectory.Visible = true;
llFileDiscoveryPathDirectory.Text = Path.GetDirectoryName(_fileDiscoveryPath);
if (cbAutoView.Checked)
{
btnViewFile_Click(sender, e);
}
}
else
{
llFileDiscoveryPathDirectory.Visible = false;
wbMain.DocumentText = "The File with the ID of " + nudFileID.Value + " does not exist.";
}
}
/// <summary>
/// Process a range of FileID after crawling. Useful if crawled Files were not processed at crawl time according to desired ApplicationSettings configuration.
/// Calling this method DOES change the 'LastDiscovered' fields where applicable.
/// This method is not when crawling, rather during post-processing.
/// </summary>
/// <param name = "fileIDLowerBound"></param>
/// <param name = "fileIDUpperBound"></param>
public static void ProcessFiles(Crawler <TArachnodeDAO> crawler, long fileIDLowerBound, long fileIDUpperBound)
{
//do not assign the application settings. doing so will override the ApplicationSetting you set before calling this method...
TArachnodeDAO arachnodeDAO = (TArachnodeDAO)Activator.CreateInstance(typeof(TArachnodeDAO), crawler.ApplicationSettings.ConnectionString, crawler.ApplicationSettings, crawler.WebSettings, false, false);
ConsoleManager <TArachnodeDAO> consoleManager = new ConsoleManager <TArachnodeDAO>(crawler.ApplicationSettings, crawler.WebSettings);
ActionManager <TArachnodeDAO> actionManager = new ActionManager <TArachnodeDAO>(crawler.ApplicationSettings, crawler.WebSettings, consoleManager);
CookieManager cookieManager = new CookieManager();;
MemoryManager <TArachnodeDAO> memoryManager = new MemoryManager <TArachnodeDAO>(crawler.ApplicationSettings, crawler.WebSettings);
RuleManager <TArachnodeDAO> ruleManager = new RuleManager <TArachnodeDAO>(crawler.ApplicationSettings, crawler.WebSettings, consoleManager);
CacheManager <TArachnodeDAO> cacheManager = new CacheManager <TArachnodeDAO>(crawler.ApplicationSettings, crawler.WebSettings);
CrawlerPeerManager <TArachnodeDAO> crawlerPeerManager = new CrawlerPeerManager <TArachnodeDAO>(crawler.ApplicationSettings, crawler.WebSettings, null, arachnodeDAO);
Cache <TArachnodeDAO> cache = new Cache <TArachnodeDAO>(crawler.ApplicationSettings, crawler.WebSettings, crawler, actionManager, cacheManager, crawlerPeerManager, memoryManager, ruleManager);
DiscoveryManager <TArachnodeDAO> discoveryManager = new DiscoveryManager <TArachnodeDAO>(crawler.ApplicationSettings, crawler.WebSettings, cache, actionManager, cacheManager, memoryManager, ruleManager);
//load the CrawlActions, CrawlRules and EngineActions...
ruleManager.ProcessCrawlRules(crawler);
actionManager.ProcessCrawlActions(crawler);
actionManager.ProcessEngineActions(crawler);
//these three methods are called in the Engine.
UserDefinedFunctions.RefreshAllowedExtensions(true);
UserDefinedFunctions.RefreshAllowedSchemes(true);
UserDefinedFunctions.RefreshDisallowed();
//instantiate a WebClient to access the ResponseHeaders...
WebClient <TArachnodeDAO> webClient = new WebClient <TArachnodeDAO>(crawler.ApplicationSettings, crawler.WebSettings, consoleManager, cookieManager, new ProxyManager <TArachnodeDAO>(crawler.ApplicationSettings, crawler.WebSettings, consoleManager));
webClient.GetHttpWebResponse("http://google.com", "GET", null, null, null, null);
FileManager <TArachnodeDAO> fileManager = new FileManager <TArachnodeDAO>(crawler.ApplicationSettings, crawler.WebSettings, discoveryManager, arachnodeDAO);
for (long i = fileIDLowerBound; i <= fileIDUpperBound; i++)
{
ArachnodeDataSet.FilesRow filesRow = null;
try
{
//get the File from the database. we need the source data as we don't store this in the index.
//even though most of the fields are available in the Document, the File is the authoritative source, so we'll use that for all of the fields.
filesRow = arachnodeDAO.GetFile(i.ToString());
if (filesRow != null)
{
if (filesRow.Source == null || filesRow.Source.Length == 0)
{
if (File.Exists(discoveryManager.GetDiscoveryPath(crawler.ApplicationSettings.DownloadedFilesDirectory, filesRow.AbsoluteUri, filesRow.FullTextIndexType)))
{
filesRow.Source = File.ReadAllBytes(discoveryManager.GetDiscoveryPath(crawler.ApplicationSettings.DownloadedFilesDirectory, filesRow.AbsoluteUri, filesRow.FullTextIndexType));
}
else
{
Console.WriteLine("FileID: " + i + " was NOT processed successfully.");
if (OnFileProcessed != null)
{
OnFileProcessed.BeginInvoke(filesRow, "FileID: " + i + " was NOT processed successfully.", null, null);
}
}
}
ProcessFile(crawler.ApplicationSettings, crawler.WebSettings, crawler, filesRow, webClient, cache, actionManager, consoleManager, crawlerPeerManager, discoveryManager, fileManager, memoryManager, ruleManager, arachnodeDAO);
Console.WriteLine("FileID: " + i + " was processed successfully.");
if (OnFileProcessed != null)
{
OnFileProcessed.BeginInvoke(filesRow, "FileID: " + i + " was processed successfully.", null, null);
}
}
}
catch (Exception exception)
{
Console.WriteLine("FileID: " + i + " was NOT processed successfully.");
Console.WriteLine(exception.Message);
if (OnFileProcessed != null)
{
OnFileProcessed.BeginInvoke(filesRow, "FileID: " + i + " was NOT processed successfully.", null, null);
OnFileProcessed.BeginInvoke(filesRow, exception.Message, null, null);
}
arachnodeDAO.InsertException(null, null, exception, false);
}
}
//stop the CrawlActions, CrawlRules and EngineActions...
ruleManager.Stop();
actionManager.Stop();
}
/// <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);
}
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];
}
}
