Using Razor 2 to write RSS fails on the <link> element

Question

I upgraded to MVC 4 yesterday and just discovered a bug that the upgrade introduced.

I have a Razor view that is used to generate an RSS feed. It has some markup like this (simplified):

<item>
    <title>@post.BlogPost.Title</title> 
    <link>@Url.BlogPost(post.BlogPost, isAbsolute: true)</link>
</item>

In Razor version two, there's special support for HTML5 void elements. Such void elements are self closing, and do not have a closing tag.

Unfortunately, <link> is one such element.

This means the above Razor markup is no longer valid, and fails at runtime. Removing the closing </link> tag removes the parser error, but means that it's no longer valid RSS.

So, is there a way to get around this, or is Razor only really suitable for generating HTML5?

Answer 1

The short answer to this question seems to be that Razor, as over version 2, is tied to HTML to the exclusion of XML. I asked one of the dev's for confirmation, so hopefully he gets back.

I ended up changing my method to use Linq to XML and a custom ActionResult, bypassing Razor and indeed any view engine:

[HttpGet]
[OutputCache(Duration = 300)]
public ActionResult Feed()
{
    var result = new XmlActionResult(
        new XDocument(
            new XElement("rss",
                new XAttribute("version", "2.0"),
                new XElement("channel",
                    new XElement("title", "My Blog")
                    // snip
                )
            )
        )
    );

    result.MimeType = "application/rss+xml";

    return result;
}

This requires the following custom ActionResult:

public sealed class XmlActionResult : ActionResult
{
    private readonly XDocument _document;

    public Formatting Formatting { get; set; }
    public string MimeType { get; set; }

    public XmlActionResult([NotNull] XDocument document)
    {
        if (document == null)
            throw new ArgumentNullException("document");

        _document = document;

        // Default values
        MimeType = "text/xml";
        Formatting = Formatting.None;
    }

    public override void ExecuteResult(ControllerContext context)
    {
        context.HttpContext.Response.Clear();
        context.HttpContext.Response.ContentType = MimeType;

        using (var writer = new XmlTextWriter(context.HttpContext.Response.OutputStream, Encoding.UTF8) { Formatting = Formatting })
            _document.WriteTo(writer);
    }
}

Answer 2

I'd do it like this:

<item>
   <title>
      @post.BlogPost.Title
   </title>

   @Html.Raw("<link>")
      @Url.BlogPost(post.BlogPost, isAbsolute: true)
   @Html.Raw("</link>")
</item>

Generated source will look like this:

<item>
    <title>
        Google
    </title>

     <link>
         http://www.google.se
    </link>
</item>

Answer 3

For now I go with this workaround:

 @Html.Raw(string.Format(@"<param name=""{0}"">{1}</param>",Name, Value)) 

Answer 4

Since Alexander Taran has opened a bounty on this question in a search for a definitive answer to this, I thought I'd check out the Razor source code on CodePlex and provide some detail.

Firstly, take a look at HtmlMarkupParser. It contains this reference data:

//From http://dev.w3.org/html5/spec/Overview.html#elements-0
private ISet<string> _voidElements = new HashSet<string>(StringComparer.OrdinalIgnoreCase)
{
    "area", "base", "br", "col", "command", "embed", "hr", "img", "input", "keygen",
    "link", "meta", "param", "source", "track", "wbr"
};

This is exposed via HtmlMarkupParser.VoidElements, and the only usage of this property is in HtmlMarkupParser.RestOfTag(...). This is a parser that's walking through a sequence of tokens. The relevant snippet of code is:

if (VoidElements.Contains(tagName))
{
    // Technically, void elements like "meta" are not allowed to have end tags. Just in case they do,
    // we need to look ahead at the next set of tokens. If we see "<", "/", tag name, accept it and the ">" following it
    // Place a bookmark
    int bookmark = CurrentLocation.AbsoluteIndex;

    // Skip whitespace
    IEnumerable<HtmlSymbol> ws = ReadWhile(IsSpacingToken(includeNewLines: true));

    // Open Angle
    if (At(HtmlSymbolType.OpenAngle) && NextIs(HtmlSymbolType.Solidus))
    {
        HtmlSymbol openAngle = CurrentSymbol;
        NextToken();
        Assert(HtmlSymbolType.Solidus);
        HtmlSymbol solidus = CurrentSymbol;
        NextToken();
        if (At(HtmlSymbolType.Text) && String.Equals(CurrentSymbol.Content, tagName, StringComparison.OrdinalIgnoreCase))
        {
            // Accept up to here
            Accept(ws);
            Accept(openAngle);
            Accept(solidus);
            AcceptAndMoveNext();

            // Accept to '>', '<' or EOF
            AcceptUntil(HtmlSymbolType.CloseAngle, HtmlSymbolType.OpenAngle);
            // Accept the '>' if we saw it. And if we do see it, we're complete
            return Optional(HtmlSymbolType.CloseAngle);
        } // At(HtmlSymbolType.Text) && String.Equals(CurrentSymbol.Content, tagName, StringComparison.OrdinalIgnoreCase)
    } // At(HtmlSymbolType.OpenAngle) && NextIs(HtmlSymbolType.Solidus)

    // Go back to the bookmark and just finish this tag at the close angle
    Context.Source.Position = bookmark;
    NextToken();
}

This means that the following will be parsed successfully:

<link></link>

However the lookahead is limited, meaning that any extra tokens seen before the closing tag cause it to fail:

<link>Some other tokens</link>

It may be possible to extend the reach of the lookahead in this case. If anyone is keen, they can provide a pull request to the MVC team.

Answer 5

Html5 link is a special element used in header for stylesheets and the like.

Your Rss should not be Html5 but something like

<?xml version="1.0" encoding="UTF-8" ?>
<rss version="2.0">

you could have this in a layout controller that your rss feeds would use

<?xml version="1.0" encoding="UTF-8" ?>
<rss version="2.0">
    @RenderBody()
</rss>

An alternative way I've done it previosly is to have a completely empty view and then the controller below:

    [NHibernateActionFilter]
    public AtomActionResult Feed()
    {
        var dto = _service.GetThings(NHibernateSession);
        var items = Mapper.Map<List<ThingDto>, List<SyndicationItem>>(dto);
        var url = HttpContextWrapper.Request.UrlReferrer;
        var feed = new SyndicationFeed("MyTitle", "MyByline", url, items)
        {
            Copyright = new TextSyndicationContent("© 2012 SO"),
            Language = "en-IE"
        };
        return new AtomActionResult(feed);
    }

Of particular note is System.ServiceModel.Syndication.SyndicationFeed

And this is my custom result

 public class AtomActionResult : ActionResult
    {
        readonly SyndicationFeed _feed;

        public AtomActionResult() { }

        public AtomActionResult(SyndicationFeed feed)
        {
            _feed = feed;
        }

        public override void ExecuteResult(ControllerContext context)
        {
            //context.HttpContext.Response.ContentType = "application/atom+xml";
            //chrome does not yet support atom+xml 
            //http://code.google.com/p/chromium/issues/detail?id=104358
            context.HttpContext.Response.ContentType = "application/xml";
            var formatter = new Atom10FeedFormatter(_feed);
            using (XmlWriter writer = XmlWriter.Create(context.HttpContext.Response.Output))
            {
                formatter.WriteTo(writer);
            }
        }
    }

Answer 6

What you can do is this:

@("<link>" + Url.BlogPost(post.BlogPost, isAbsolute: true) + "</link>")

much simpler