void ReadToEndTag(HtmlReader reader, string endTag)
{
while (!reader.EOF)
{
var n = reader.NextNode();
if (n.Value == endTag && n.Type == HtmlTokenType.EndTag)
{
return;
}
}
}
string ReadTitle(HtmlReader reader, string endTag)
{
string title = "";
while (!reader.EOF)
{
var node = reader.NextNode();
if (node.Type == HtmlTokenType.Text)
{
title += node.Value;
}
if (node.Type == HtmlTokenType.EndTag && node.Value == "title")
{
return(SanitiseAttribute(title, 128));
}
}
return(SanitiseAttribute(title, 128));
}
public override async Task ProcessResponseStream(Stream dataStream)
{
if (dataStream == null)
{
return;
}
var ms = new MemoryStream();
dataStream.CopyTo(ms);
ms.Seek(0, SeekOrigin.Begin);
if (ContentType.IsUnknownOrNull(ContentType))
{
ContentType = ContentType.Guess(ms);
}
ms.Seek(0, SeekOrigin.Begin);
if (ContentType.IsUnknownOrNull(ContentType) || ContentType.MediaType == "text")
{
using (var p = new HtmlReader(ms))
{
while (!p.EOF)
{
var node = p.NextNode();
if (node.Type == HtmlTokenType.StartTag)
{
if (node.Value == "script")
{
string src = p.GetAttribute("src");
if (!string.IsNullOrWhiteSpace(src))
{
Console.WriteLine(src);
}
}
}
}
}
}
}
public Uri NextUri()
{
Uri uri = null;
while (!reader.EOF && uri == null)
{
var node = reader.NextNode();
if (node.Value == "a")
{
var href = reader.GetAttribute("href");
if (!string.IsNullOrWhiteSpace(href))
{
var links = Utility.GetLinks(CurrentSourceUri, href);
if (links.Any())
{
uri = links.First();
}
}
}
else if (node.Value == "script")
{
var href = reader.GetAttribute("src");
if (!string.IsNullOrWhiteSpace(href))
{
var links = Utility.GetLinks(CurrentSourceUri, href);
if (links.Any())
{
uri = links.First();
}
}
}
else if (node.Value == "link")
{
var href = reader.GetAttribute("href");
if (!string.IsNullOrWhiteSpace(href))
{
var links = Utility.GetLinks(CurrentSourceUri, href);
if (links.Any())
{
uri = links.First();
}
}
}
else if (node.Value == "base")
{
var href = reader.GetAttribute("href");
if (!string.IsNullOrWhiteSpace(href))
{
// sometimes the links are bogus?!
var l = Utility.GetLinks(null, href);
if (l.Any())
{
CurrentSourceUri = l.First();
}
}
}
}
return(uri);
}
public WorkspaceResult Build(Stream stream, string requestString, string responseHeaders, out string evalText)
{
WorkspaceResult result = new WorkspaceResult();
result.SourceServerId = FetchoConfiguration.Current.CurrentServerNode.ServerId;
if (!stream.CanSeek)
{
throw new FetchoException("WorkspaceResultBuilder needs a seekable stream");
}
ProcessHeaders(result, requestString, responseHeaders);
result.DataHash = MD5Hash.Compute(stream).ToString();
result.PageSize = stream.Length;
stream.Seek(0, SeekOrigin.Begin);
ContentType contentType = GetContentType(result);
int titleness = 4;
if (contentType != null)
{
if (contentType.SubType.Contains("html"))
{
using (var reader = new HtmlReader(stream))
{
while (!reader.EOF)
{
var node = reader.NextNode();
if (node.Type == HtmlTokenType.Text)
{
evaluationText.Append(node.Value);
evaluationText.Append(' ');
}
if (node.Value == "script")
{
ReadToEndTag(reader, "script");
}
else if (node.Value == "style")
{
ReadToEndTag(reader, "style");
}
else if (node.Value == "title" && titleness > 1 && !result.PropertyCache.ContainsKey("title"))
{
string title = ReadTitle(reader, "title");
result.PropertyCache.Add("title", title);
titleness = 1;
}
else if (node.Value == "h1" && titleness > 2 && !result.PropertyCache.ContainsKey("title"))
{
string title = ReadTitle(reader, "h1");
result.PropertyCache.Add("title", title);
titleness = 2;
}
else if (node.Value == "h2" && titleness > 3 && !result.PropertyCache.ContainsKey("title"))
{
string title = ReadTitle(reader, "h2");
result.PropertyCache.Add("title", title);
titleness = 3;
}
else if (node.Value == "meta")
{
ProcessMetaTag(reader, result);
}
}
}
}
else if (contentType.IsTextType || ContentType.IsJavascriptContentType(contentType))
{
// leave the stream open so other tasks can reset it and use it
using (var reader = new StreamReader(stream, Encoding.Default, true, 1024, true))
{
evaluationText.Append(reader.ReadToEnd());
}
}
}
result.UriHash = MD5Hash.Compute(result.RequestProperties.SafeGet("uri") ?? "").ToString();
result.RefererUri = result.RequestProperties.SafeGet("referer");
result.Uri = result.RequestProperties.SafeGet("uri") ?? "";
result.Title = result.PropertyCache.SafeGet("title")?.ToString();
result.Description = result.PropertyCache.SafeGet("description")?.ToString();
result.Created = DateTime.UtcNow;
result.Updated = DateTime.UtcNow;
evalText = evaluationText.ToString();
return(result);
}