// ------------------------------------------------------------------
/// <summary>
/// Constructor.
/// </summary>
/// <param name="settings">The settings.</param>
/// <param name="uriInfo">The URI info.</param>
/// <param name="textContent">Content of the text.</param>
public ResourceParser(
SpiderSettings settings,
UriResourceInformation uriInfo,
string textContent)
{
_settings = settings;
_uriInfo = uriInfo;
_textContent = textContent;
}
// ------------------------------------------------------------------
/// <summary>
/// Constructor.
/// </summary>
/// <param name="settings">The settings.</param>
/// <param name="uriInfo">The URI info.</param>
/// <param name="textContent">Content of the text.</param>
public ResourceParser(
SpiderSettings settings,
UriResourceInformation uriInfo,
string textContent )
{
_settings = settings;
_uriInfo = uriInfo;
_textContent = textContent;
}
/// <summary>
/// Constructor.
/// </summary>
/// <param name="copyFrom">The copy from.</param>
/// <param name="folderPath">The folder path.</param>
/// <param name="baseFolderPath">The base folder path.</param>
public DownloadedResourceInformation(
UriResourceInformation copyFrom,
DirectoryInfo folderPath,
DirectoryInfo baseFolderPath)
:
base(copyFrom)
{
_localFolderPath = folderPath;
_localBaseFolderPath = baseFolderPath;
}
/// <summary>
/// Initializes a new instance of the
/// <see cref="UriResourceInformation"/> class.
/// </summary>
/// <param name="copyFrom">The copy from.</param>
public UriResourceInformation(
UriResourceInformation copyFrom )
{
_options = copyFrom._options;
_originalUrl = copyFrom._originalUrl;
_relativeUri = copyFrom._relativeUri;
_baseUri = copyFrom._baseUri;
_absoluteUri = copyFrom._absoluteUri;
_linkType = copyFrom._linkType;
}
/// <summary>
/// Initializes a new instance of the
/// <see cref="UriResourceInformation"/> class.
/// </summary>
/// <param name="copyFrom">The copy from.</param>
public UriResourceInformation(
UriResourceInformation copyFrom)
{
_options = copyFrom._options;
_originalUrl = copyFrom._originalUrl;
_relativeUri = copyFrom._relativeUri;
_baseUri = copyFrom._baseUri;
_absoluteUri = copyFrom._absoluteUri;
_linkType = copyFrom._linkType;
}
/// <summary>
/// Stores a HTML resource to the local file system.
/// Does no hyperlink replacement.
/// </summary>
/// <returns>Return the info about the stored data.</returns>
public DownloadedResourceInformation StoreHtml(
string textContent,
Encoding encoding,
UriResourceInformation uriInfo)
{
DownloadedResourceInformation result =
new DownloadedResourceInformation(
uriInfo,
_settings.Options.DestinationFolderPath);
try
{
if (result.LocalFilePath.Exists)
{
result.LocalFilePath.Delete();
}
if (!result.LocalFilePath.Directory.Exists)
{
result.LocalFilePath.Directory.Create();
}
Trace.WriteLine(
string.Format(
@"Writing text content to file '{0}'.",
result.LocalFilePath));
using (FileStream s = new FileStream(
result.LocalFilePath.FullName,
FileMode.Create,
FileAccess.Write))
using (StreamWriter w = new StreamWriter(s, encoding))
{
w.Write(textContent);
}
}
catch (IOException x)
{
Trace.WriteLine(
string.Format(
@"Ignoring IO exception while storing HTML file: '{0}'.",
x.Message));
}
catch (UnauthorizedAccessException x)
{
Trace.WriteLine(
string.Format(
@"Ignoring exception while storing HTML file: '{0}'.",
x.Message));
}
return(result);
}
/// <summary>
/// Detects URLs in styles.
/// </summary>
/// <param name="baseUri">The base URI.</param>
/// <param name="attributeName">Name of the attribute.</param>
/// <param name="attributeValue">The attribute value.</param>
/// <returns></returns>
private List <UriResourceInformation> ExtractStyleUrls(
Uri baseUri,
string attributeName,
string attributeValue)
{
List <UriResourceInformation> result =
new List <UriResourceInformation>();
if (string.Compare(attributeName, @"style", true) == 0)
{
if (attributeValue != null &&
attributeValue.Trim().Length > 0)
{
MatchCollection matchs = Regex.Matches(
attributeValue,
@"url\s*\(\s*([^\)\s]+)\s*\)",
RegexOptions.Singleline | RegexOptions.IgnoreCase);
if (matchs.Count > 0)
{
foreach (Match match in matchs)
{
if (match != null && match.Success)
{
string url = match.Groups[1].Value;
UriResourceInformation ui =
new UriResourceInformation(
_settings.Options,
url,
new Uri(url, UriKind.RelativeOrAbsolute),
baseUri,
UriType.Resource);
bool isOnSameSite =
ui.IsOnSameSite(baseUri);
if ((isOnSameSite ||
!_settings.Options.StayOnSite) &&
ui.IsProcessableUri)
{
result.Add(ui);
}
}
}
}
}
}
return(result);
}
/// <summary>
/// Stores a binary resource to the local file system.
/// </summary>
/// <returns>Return the info about the stored data.</returns>
public DownloadedResourceInformation StoreBinary(
byte[] binaryContent,
UriResourceInformation uriInfo )
{
DownloadedResourceInformation result =
new DownloadedResourceInformation(
uriInfo,
_settings.Options.DestinationFolderPath );
try
{
if ( result.LocalFilePath.Exists )
{
result.LocalFilePath.Delete();
}
if ( binaryContent != null && binaryContent.Length > 0 )
{
Trace.WriteLine(
string.Format(
@"Writing binary content to file '{0}'.",
result.LocalFilePath ) );
using ( FileStream s = result.LocalFilePath.OpenWrite() )
{
s.Write( binaryContent, 0, binaryContent.Length );
}
}
}
catch ( IOException x )
{
Trace.WriteLine(
string.Format(
@"Ignoring IO exception while storing binary file: '{0}'.",
x.Message ) );
}
catch ( UnauthorizedAccessException x )
{
Trace.WriteLine(
string.Format(
@"Ignoring exception while storing binary file: '{0}'.",
x.Message ) );
}
return result;
}
/// <summary>
/// Stores a binary resource to the local file system.
/// </summary>
/// <returns>Return the info about the stored data.</returns>
public DownloadedResourceInformation StoreBinary(
byte[] binaryContent,
UriResourceInformation uriInfo)
{
DownloadedResourceInformation result =
new DownloadedResourceInformation(
uriInfo,
_settings.Options.DestinationFolderPath);
try
{
if (result.LocalFilePath.Exists)
{
result.LocalFilePath.Delete();
}
if (binaryContent != null && binaryContent.Length > 0)
{
Trace.WriteLine(
string.Format(
@"Writing binary content to file '{0}'.",
result.LocalFilePath));
using (FileStream s = result.LocalFilePath.OpenWrite())
{
s.Write(binaryContent, 0, binaryContent.Length);
}
}
}
catch (IOException x)
{
Trace.WriteLine(
string.Format(
@"Ignoring IO exception while storing binary file: '{0}'.",
x.Message));
}
catch (UnauthorizedAccessException x)
{
Trace.WriteLine(
string.Format(
@"Ignoring exception while storing binary file: '{0}'.",
x.Message));
}
return(result);
}
/// <summary>
/// Constructor.
/// </summary>
/// <param name="copyFrom">The copy from.</param>
/// <param name="baseFolderPath">The base folder path.</param>
public DownloadedResourceInformation(
UriResourceInformation copyFrom,
DirectoryInfo baseFolderPath)
:
base(copyFrom)
{
_localBaseFolderPath = baseFolderPath;
_localFilePath = new FileInfo(
Path.Combine(
baseFolderPath.FullName,
MakeLocalFileName(
copyFrom.AbsoluteUri,
copyFrom.BaseUri,
copyFrom.LinkType)));
_localFileName =
new FileInfo(_localFilePath.Name);
}
// ------------------------------------------------------------------
#endregion
#region Private methods.
// ------------------------------------------------------------------
/// <summary>
/// Does the extract links.
/// </summary>
/// <param name="xml">The XML.</param>
/// <param name="uriInfo">The URI info.</param>
/// <returns></returns>
private List <UriResourceInformation> DoExtractLinks(
XmlReader xml,
UriResourceInformation uriInfo)
{
List <UriResourceInformation> links =
new List <UriResourceInformation>();
while (xml.Read())
{
switch (xml.NodeType)
{
// Added 2006-03-27: Inside comments, too.
case XmlNodeType.Comment:
XmlReader childXml =
GetDocReader(xml.Value, uriInfo.BaseUri);
List <UriResourceInformation> childLinks =
DoExtractLinks(childXml, uriInfo);
links.AddRange(childLinks);
break;
// A node element.
case XmlNodeType.Element:
string[] linkAttributeNames;
UriType linkType;
// If this is a link element, store the URLs to modify.
if (IsLinkElement(
xml.Name,
out linkAttributeNames,
out linkType))
{
while (xml.MoveToNextAttribute())
{
links.AddRange(
ExtractStyleUrls(
uriInfo.BaseUriWithFolder,
xml.Name,
xml.Value));
foreach (string a in linkAttributeNames)
{
if (string.Compare(a, xml.Name, true) == 0)
{
string url = xml.Value;
UriResourceInformation ui =
new UriResourceInformation(
_settings.Options,
url,
new Uri(url, UriKind.RelativeOrAbsolute),
uriInfo.BaseUriWithFolder,
linkType);
bool isOnSameSite =
ui.IsOnSameSite(uriInfo.BaseUri);
if ((isOnSameSite ||
!_settings.Options.StayOnSite) &&
ui.IsProcessableUri)
{
links.Add(ui);
}
}
}
}
}
else
{
// Also, look for style attributes.
while (xml.MoveToNextAttribute())
{
links.AddRange(
ExtractStyleUrls(
uriInfo.BaseUriWithFolder,
xml.Name,
xml.Value));
}
}
break;
}
}
return(links);
}
/// <summary>
/// Stores a HTML resource to the local file system.
/// Does no hyperlink replacement.
/// </summary>
/// <returns>Return the info about the stored data.</returns>
public DownloadedResourceInformation StoreHtml(
string textContent,
Encoding encoding,
UriResourceInformation uriInfo )
{
DownloadedResourceInformation result =
new DownloadedResourceInformation(
uriInfo,
_settings.Options.DestinationFolderPath );
try
{
if ( result.LocalFilePath.Exists )
{
result.LocalFilePath.Delete();
}
if ( !result.LocalFilePath.Directory.Exists )
{
result.LocalFilePath.Directory.Create();
}
Trace.WriteLine(
string.Format(
@"Writing text content to file '{0}'.",
result.LocalFilePath ) );
using ( FileStream s = new FileStream(
result.LocalFilePath.FullName,
FileMode.Create,
FileAccess.Write ) )
using ( StreamWriter w = new StreamWriter( s, encoding ) )
{
w.Write( textContent );
}
}
catch ( IOException x )
{
Trace.WriteLine(
string.Format(
@"Ignoring IO exception while storing HTML file: '{0}'.",
x.Message ) );
}
catch ( UnauthorizedAccessException x )
{
Trace.WriteLine(
string.Format(
@"Ignoring exception while storing HTML file: '{0}'.",
x.Message ) );
}
return result;
}
/// <summary>
/// Replace URIs inside a given HTML document that was previously
/// downloaded with the local URIs.
/// </summary>
/// <returns>Returns the content text with the replaced links.</returns>
public string ReplaceLinks(
string textContent,
UriResourceInformation uriInfo)
{
ResourceParser parser = new ResourceParser(
_settings,
uriInfo,
textContent );
List<UriResourceInformation> linkInfos =
parser.ExtractLinks();
// For remembering duplicates.
Dictionary<string, string> replacedLinks =
new Dictionary<string, string>();
// --
foreach ( UriResourceInformation linkInfo in linkInfos )
{
if ( linkInfo.WantFollowUri || linkInfo.IsResourceUri )
{
DownloadedResourceInformation dlInfo =
new DownloadedResourceInformation(
linkInfo,
_settings.Options.DestinationFolderPath );
// /*
if ( !string.IsNullOrEmpty( linkInfo.OriginalUrl ) )
{
string textContentBefore = textContent;
string link =
Regex.Escape( linkInfo.OriginalUrl );
textContent = Regex.Replace(
textContent,
string.Format( @"""{0}""", link ),
string.Format( @"""Resources\{0}""", dlInfo.LocalFileName ),
RegexOptions.IgnoreCase | RegexOptions.Multiline );
textContent = Regex.Replace(
textContent,
string.Format( @"'{0}'", link ),
string.Format(@"'Resources\{0}'", dlInfo.LocalFileName),
RegexOptions.IgnoreCase | RegexOptions.Multiline );
// For style-"url(...)"-links.
textContent = Regex.Replace(
textContent,
string.Format( @"\(\s*{0}\s*\)", link ),
string.Format(@"(Resources\{0})", dlInfo.LocalFileName),
RegexOptions.IgnoreCase | RegexOptions.Multiline );
// Some checking.
// 2007-07-27, Uwe Keim.
if ( linkInfo.OriginalUrl != dlInfo.LocalFileName.Name &&
textContentBefore == textContent &&
!replacedLinks.ContainsKey( linkInfo.AbsoluteUri.AbsolutePath ) )
{
}
else
{
// Remember.
replacedLinks[linkInfo.AbsoluteUri.AbsolutePath] =
linkInfo.AbsoluteUri.AbsolutePath;
}
}
// */
}
}
// --
return textContent;
}
/// <summary>
/// Replace URIs inside a given HTML document that was previously
/// downloaded with the local URIs.
/// </summary>
/// <returns>Returns the content text with the replaced links.</returns>
public string ReplaceLinks(
string textContent,
UriResourceInformation uriInfo)
{
try
{
ResourceParser parser = new ResourceParser(
_settings,
uriInfo,
textContent);
List<UriResourceInformation> linkInfos =
parser.ExtractLinks();
// For remembering duplicates.
Dictionary<string, string> replacedLinks =
new Dictionary<string, string>();
// --
foreach (UriResourceInformation linkInfo in linkInfos)
{
if (linkInfo.WantFollowUri || linkInfo.IsResourceUri)
{
DownloadedResourceInformation dlInfo =
new DownloadedResourceInformation(
linkInfo,
_settings.Options.DestinationFolderPath);
// /*
if (!string.IsNullOrEmpty(linkInfo.OriginalUrl))
{
string textContentBefore = textContent;
string link =
Regex.Escape(linkInfo.OriginalUrl);
textContent = Regex.Replace(
textContent,
string.Format(@"""{0}""", link),
string.Format(@"""{0}""", dlInfo.LocalFileName),
RegexOptions.IgnoreCase | RegexOptions.Multiline);
textContent = Regex.Replace(
textContent,
string.Format(@"'{0}'", link),
string.Format(@"'{0}'", dlInfo.LocalFileName),
RegexOptions.IgnoreCase | RegexOptions.Multiline);
// For style-"url(...)"-links.
textContent = Regex.Replace(
textContent,
string.Format(@"\(\s*{0}\s*\)", link),
string.Format(@"({0})", dlInfo.LocalFileName),
RegexOptions.IgnoreCase | RegexOptions.Multiline);
// Some checking.
// 2007-07-27, Uwe Keim.
if (linkInfo.OriginalUrl != dlInfo.LocalFileName.Name &&
textContentBefore == textContent &&
!replacedLinks.ContainsKey(linkInfo.AbsoluteUri.AbsolutePath))
{
//throw new ApplicationException(
string.Format(
@"Failed to replace URI '{0}' with URI '{1}' in HTML text '{2}'.",
linkInfo.OriginalUrl,
dlInfo.LocalFileName,
textContent);
}
else
{
// Remember.
replacedLinks[linkInfo.AbsoluteUri.AbsolutePath] =
linkInfo.AbsoluteUri.AbsolutePath;
}
}
// */
}
}
}
catch (Exception ex)
{//Satwadhir written code }
// --
Console.WriteLine(ex.Message.ToString());
}
return textContent;
}
/// <summary>
/// Replace URIs inside a given HTML document that was previously
/// downloaded with the local URIs.
/// </summary>
/// <returns>Returns the content text with the replaced links.</returns>
public string ReplaceLinks(
string textContent,
UriResourceInformation uriInfo)
{
ResourceParser parser = new ResourceParser(
_settings,
uriInfo,
textContent);
List <UriResourceInformation> linkInfos =
parser.ExtractLinks();
// For remembering duplicates.
Dictionary <string, string> replacedLinks =
new Dictionary <string, string>();
// --
foreach (UriResourceInformation linkInfo in linkInfos)
{
if (linkInfo.WantFollowUri || linkInfo.IsResourceUri)
{
DownloadedResourceInformation dlInfo =
new DownloadedResourceInformation(
linkInfo,
_settings.Options.DestinationFolderPath);
// /*
if (!string.IsNullOrEmpty(linkInfo.OriginalUrl))
{
string textContentBefore = textContent;
string link =
Regex.Escape(linkInfo.OriginalUrl);
textContent = Regex.Replace(
textContent,
string.Format(@"""{0}""", link),
string.Format(@"""Resources\{0}""", dlInfo.LocalFileName),
RegexOptions.IgnoreCase | RegexOptions.Multiline);
textContent = Regex.Replace(
textContent,
string.Format(@"'{0}'", link),
string.Format(@"'Resources\{0}'", dlInfo.LocalFileName),
RegexOptions.IgnoreCase | RegexOptions.Multiline);
// For style-"url(...)"-links.
textContent = Regex.Replace(
textContent,
string.Format(@"\(\s*{0}\s*\)", link),
string.Format(@"(Resources\{0})", dlInfo.LocalFileName),
RegexOptions.IgnoreCase | RegexOptions.Multiline);
// Some checking.
// 2007-07-27, Uwe Keim.
if (linkInfo.OriginalUrl != dlInfo.LocalFileName.Name &&
textContentBefore == textContent &&
!replacedLinks.ContainsKey(linkInfo.AbsoluteUri.AbsolutePath))
{
}
else
{
// Remember.
replacedLinks[linkInfo.AbsoluteUri.AbsolutePath] =
linkInfo.AbsoluteUri.AbsolutePath;
}
}
// */
}
}
// --
return(textContent);
}
// ------------------------------------------------------------------
#endregion
#region Private methods.
// ------------------------------------------------------------------
/// <summary>
/// Does the extract links.
/// </summary>
/// <param name="xml">The XML.</param>
/// <param name="uriInfo">The URI info.</param>
/// <returns></returns>
private List<UriResourceInformation> DoExtractLinks(
XmlReader xml,
UriResourceInformation uriInfo )
{
List<UriResourceInformation> links =
new List<UriResourceInformation>();
while ( xml.Read() )
{
switch ( xml.NodeType )
{
// Added 2006-03-27: Inside comments, too.
case XmlNodeType.Comment:
XmlReader childXml =
GetDocReader( xml.Value, uriInfo.BaseUri );
List<UriResourceInformation> childLinks =
DoExtractLinks( childXml, uriInfo );
links.AddRange( childLinks );
break;
// A node element.
case XmlNodeType.Element:
string[] linkAttributeNames;
UriType linkType;
// If this is a link element, store the URLs to modify.
if ( IsLinkElement(
xml.Name,
out linkAttributeNames,
out linkType ) )
{
while ( xml.MoveToNextAttribute() )
{
links.AddRange(
ExtractStyleUrls(
uriInfo.BaseUriWithFolder,
xml.Name,
xml.Value ) );
foreach ( string a in linkAttributeNames )
{
if ( string.Compare( a, xml.Name, true ) == 0 )
{
string url = xml.Value;
UriResourceInformation ui =
new UriResourceInformation(
_settings.Options,
url,
new Uri( url, UriKind.RelativeOrAbsolute ),
uriInfo.BaseUriWithFolder,
linkType );
bool isOnSameSite =
ui.IsOnSameSite( uriInfo.BaseUri );
if ( (isOnSameSite ||
!_settings.Options.StayOnSite) &&
ui.IsProcessableUri )
{
links.Add( ui );
}
}
}
}
}
else
{
// Also, look for style attributes.
while ( xml.MoveToNextAttribute() )
{
links.AddRange(
ExtractStyleUrls(
uriInfo.BaseUriWithFolder,
xml.Name,
xml.Value ) );
}
}
break;
}
}
return links;
}
/// <summary>
/// Detects URLs in styles.
/// </summary>
/// <param name="baseUri">The base URI.</param>
/// <param name="attributeName">Name of the attribute.</param>
/// <param name="attributeValue">The attribute value.</param>
/// <returns></returns>
private List<UriResourceInformation> ExtractStyleUrls(
Uri baseUri,
string attributeName,
string attributeValue )
{
List<UriResourceInformation> result =
new List<UriResourceInformation>();
if ( string.Compare( attributeName, @"style", true ) == 0 )
{
if ( attributeValue != null &&
attributeValue.Trim().Length > 0 )
{
MatchCollection matchs = Regex.Matches(
attributeValue,
@"url\s*\(\s*([^\)\s]+)\s*\)",
RegexOptions.Singleline | RegexOptions.IgnoreCase );
if ( matchs.Count > 0 )
{
foreach ( Match match in matchs )
{
if ( match != null && match.Success )
{
string url = match.Groups[1].Value;
UriResourceInformation ui =
new UriResourceInformation(
_settings.Options,
url,
new Uri( url, UriKind.RelativeOrAbsolute ),
baseUri,
UriType.Resource );
bool isOnSameSite =
ui.IsOnSameSite( baseUri );
if ( (isOnSameSite ||
!_settings.Options.StayOnSite) &&
ui.IsProcessableUri )
{
result.Add( ui );
}
}
}
}
}
}
return result;
}
