private void ReplaceChildElementByText(INode parent, IElement child, IHtmlDocument document)
{
if (!string.IsNullOrEmpty(child.TextContent))
{
// Add a text content in place of the element
var newElement = document.CreateTextNode(child.TextContent);
parent.ReplaceChild(newElement, child);
}
else
{
// If no content then drop the element
parent.RemoveChild(child);
}
}
/// <summary>
/// Converts each<a> and<img> uri in the given element, and its descendants, to an absolute URI,
/// ignoring #ref URIs.
/// </summary>
/// <param name="articleContent">The node in which to fix all relative uri</param>
public static void FixRelativeUris(IElement articleContent, Uri uri, IHtmlDocument doc)
{
var scheme = uri.Scheme;
var prePath = uri.GetBase();
var pathBase = uri.Scheme + "://" + uri.Host + uri.AbsolutePath.Substring(0, uri.AbsolutePath.LastIndexOf('/') + 1);
var links = NodeUtility.GetAllNodesWithTag(articleContent, new string[] { "a" });
NodeUtility.ForEachNode(links, (link) =>
{
var href = (link as IElement).GetAttribute("href");
if (!String.IsNullOrWhiteSpace(href))
{
// Replace links with javascript: URIs with text content, since
// they won't work after scripts have been removed from the page.
if (href.IndexOf("javascript:") == 0)
{
var text = doc.CreateTextNode(link.TextContent);
link.Parent.ReplaceChild(text, link);
}
else
{
(link as IElement).SetAttribute("href", uri.ToAbsoluteURI(href));
}
}
});
var imgs = NodeUtility.GetAllNodesWithTag(articleContent, new string[] { "img" });
NodeUtility.ForEachNode(imgs, (img) =>
{
var src = (img as IElement).GetAttribute("src");
if (!String.IsNullOrWhiteSpace(src))
{
(img as IElement).SetAttribute("src", uri.ToAbsoluteURI(src));
}
});
}
/// <summary>
/// Converts each <a> and <img> uri in the given element, and its descendants, to an absolute URI,
/// ignoring #ref URIs.
/// </summary>
/// <param name="articleContent">The node in which to fix all relative uri</param>
/// <param name="uri">The base uri</param>
/// <param name="doc">The document to operate on</param>
internal static void FixRelativeUris(IElement articleContent, Uri uri, IHtmlDocument doc)
{
var scheme = uri.Scheme;
var prePath = uri.GetBase();
var pathBase = uri.Scheme + "://" + uri.Host + uri.AbsolutePath.Substring(0, uri.AbsolutePath.LastIndexOf('/') + 1);
var links = NodeUtility.GetAllNodesWithTag(articleContent, new string[] { "a" });
NodeUtility.ForEachNode(links, (link) =>
{
var href = (link as IElement).GetAttribute("href");
if (!string.IsNullOrWhiteSpace(href))
{
// Remove links with javascript: URIs, since
// they won't work after scripts have been removed from the page.
if (href.IndexOf("javascript:") == 0)
{
// if the link only contains simple text content, it can be converted to a text node
if (link.ChildNodes.Length == 1 && link.ChildNodes[0].NodeType == NodeType.Text)
{
var text = doc.CreateTextNode(link.TextContent);
link.Parent.ReplaceChild(text, link);
}
else
{
// if the link has multiple children, they should all be preserved
var container = doc.CreateElement("span");
while (link.ChildNodes.Length > 0)
{
container.AppendChild(link.ChildNodes[0]);
}
link.Parent.ReplaceChild(container, link);
}
}
else
{
(link as IElement).SetAttribute("href", uri.ToAbsoluteURI(href));
}
}
});
var medias = NodeUtility.GetAllNodesWithTag(articleContent, new string[] { "img", "picture", "figure", "video", "audio", "source" });
NodeUtility.ForEachNode(medias, (media_node) => {
if (media_node is IElement)
{
var media = media_node as IElement;
var src = media.GetAttribute("src");
var poster = media.GetAttribute("poster");
var srcset = media.GetAttribute("srcset");
if (src != null)
{
media.SetAttribute("src", uri.ToAbsoluteURI(src));
}
if (poster != null)
{
media.SetAttribute("poster", uri.ToAbsoluteURI(poster));
}
if (srcset != null)
{
var newSrcset = RE_SrcSetUrl.Replace(srcset, (input) =>
{
return(uri.ToAbsoluteURI(input.Groups[1].Value) + (input.Groups[2]?.Value ?? "") + input.Groups[3].Value);
});
media.SetAttribute("srcset", newSrcset);
}
}
});
}
