XPS writer fails with .Net 4 / font rendering bug?

Question

We have a WPF application that allows users to enter and style text as they wish. Their text is then transformed to XPS to be inserted in a PDF using ABCPdf.

We have recently switched to .Net 4, and now the XPS generation sometimes generates badly-formed XPS.

The only time when a badly-formed XPS is generated is when the user uses a font that overlaps two characters (such as Sevillana, available on Google Fonts). In .Net 4, the characters overlap, whereas in .Net 3.5, they do not. From another SO question I gather that it's because .Net 4.0 has changed its font rendering engine.

.Net 3.5 :

.Net 4.0 : (note how the "d" and "A" overlap)

It looks like .Net 4 does not take the apostrophe's spacing into account, or even negates it. If I remove the apostrophe, the spacing widens. I have created a simple test project to demonstrate the issue : https://github.com/tbroust-trepia/wpf-4-font-rendering

Our method to save a FlowDocument to XPS is pretty much the same as in creating an XPS Document from a FlowDocument and attach it on the fly (I suspect that the original developer just copy/pasted that code) ; it just saves the XPS to a file instead of in a stream.

The MS tool IsXPS says that a specific node is invalid. As a matter of fact, in Documents\1\Pages\1.fpage, the value of Indices is ;,-16;,84;;;;,30. I can see that "0.-16" is not really a number.

So, I have many questions:

• Why is it happening ? I don't understand how the values are converted into these Indices. I guess that the transformation engine tried to set it to "-0.16" but messed up ?
• Can I modify/check the value of this Indices before it's saved ?
• If I cannot, how can I check afterwards ? I have found a sort of XPS validator here but I don't understand how I can check the Indices value from there. I could do it manually, reading and parsing the XML, but I'm really sure it's not a good idea.

Edit

I have opened a bug at Microsoft, but I don't expect much there. I think I will try to modify the generated XPS. It's dirty but it might work.

Answer 1

While waiting for MS to do whatever they want, I have found a way to fix the broken XPS.

Warning: dirty hack ahead

public class XpsFile
{
    /// <summary>
    /// Regex to validate XPS "indices" property (gotten from ABCPDF)
    /// </summary>
    private static readonly string IndicesRegex = @"(((\(([1-9][0-9]*)(:([1-9][0-9]*))?\))?([0-9]+))?(,(\+?(([0-9]+(\.[0-9]+)?)|(\.[0-9]+))((e|E)(\-|\+)?[0-9]+)?)?(,((\-|\+)?(([0-9]+(\.[0-9]+)?)|(\.[0-9]+))((e|E)(\-|\+)?[0-9]+)?)?(,((\-|\+)?(([0-9]+(\.[0-9]+)?)|(\.[0-9]+))((e|E)(\-|\+)?[0-9]+)?))?)?)?)(;((\(([1-9][0-9]*)(:([1-9][0-9]*))?\))?([0-9]+))?(,(\+?(([0-9]+(\.[0-9]+)?)|(\.[0-9]+))((e|E)(\-|\+)?[0-9]+)?)?(,((\-|\+)?(([0-9]+(\.[0-9]+)?)|(\.[0-9]+))((e|E)(\-|\+)?[0-9]+)?)?(,((\-|\+)?(([0-9]+(\.[0-9]+)?)|(\.[0-9]+))((e|E)(\-|\+)?[0-9]+)?))?)?)?)*";

    /// <summary>
    /// Fixes the XPS problems it encounters and knows about
    /// </summary>
    public static void FixXps(string filePath)
    {
        // first we'll load the XPS file
        using (var currentPackage = Package.Open(filePath, FileMode.Open))
        {
            var pageUri = new Uri("/Documents/1/Pages/1.fpage", UriKind.Relative);

            // check that the file we'll modify exists
            if (!currentPackage.PartExists(pageUri))
            {
                throw new Exception(string.Format("Unable to find first page in XPS {0} - unable to fix this XPS !", filePath));
            }

            // assume the broken part is in the first page
            var firstPage = currentPackage.GetPart(pageUri);
            var relationships = firstPage.GetRelationships();
            var pageContent = XDocument.Load(firstPage.GetStream());

            // then we'll look up each glyph and check if their "Indices" property is valid
            XNamespace ns = pageContent.Root.GetDefaultNamespace();
            var glyphs = (from g in pageContent.Descendants(ns + "Glyphs")
                          where g.Attribute("Indices") != null
                          select g).ToList();
            for (var i = 0; i < glyphs.Count(); i ++)
            {
                glyphs[i] = FixGlyph(glyphs[i]);
            }

            // remove the current (corrupted) file from the package
            currentPackage.DeletePart(pageUri);

            // add the new (shiny) file to the package
            var newPage = currentPackage.CreatePart(pageUri, "application/vnd.ms-package.xps-fixedpage+xml", CompressionOption.NotCompressed);
            using (var ms = new MemoryStream())
            {
                // we need to remove XML declaration, so we need to use the XmlWriter
                var settings = new XmlWriterSettings();
                settings.Indent = false;
                settings.NewLineChars = string.Empty;
                settings.NewLineHandling = NewLineHandling.Replace;
                settings.OmitXmlDeclaration = true;
                using (var xw = XmlWriter.Create(ms, settings))
                {
                    pageContent.WriteTo(xw);
                }

                ms.Seek(0, SeekOrigin.Begin);
                CopyStream(newPage.GetStream(), ms);
            }

            // now we need to re-create the relationships between the Page file and the fonts
            foreach (var relation in relationships)
            {
                newPage.CreateRelationship(relation.TargetUri, relation.TargetMode, relation.RelationshipType, relation.Id);
            }
        }
    }

    /// <summary>
    /// Tries to load the XPS, and returns false if it fails
    /// </summary>
    public static bool IsValidXps(string filePath)
    {
        try
        {
            using (var xpsOld = new XpsDocument(filePath, FileAccess.Read))
            {
                var unused = xpsOld.GetFixedDocumentSequence();
            }

            return true;
        }
        catch (System.Windows.Markup.XamlParseException)
        {
            return false;
        }
    }

    /// <summary>
    /// Writes the whole content of a stream into another
    /// </summary>
    /// <remarks>
    /// http://stackoverflow.com/a/18885954/2354542
    /// </remarks>
    private static void CopyStream(Stream target, Stream source)
    {
        const int bufSize = 0x1000;
        byte[] buf = new byte[bufSize];
        int bytesRead = 0;
        while ((bytesRead = source.Read(buf, 0, bufSize)) > 0)
        {
            target.Write(buf, 0, bytesRead);
        }
    }

    /// <summary>
    /// Fixes the glyph, if necessary
    /// </summary>
    private static XElement FixGlyph(XElement g)
    {
        var matchAttribute = Regex.Match(g.Attribute("Indices").Value, IndicesRegex);
        if (!matchAttribute.Success)
        {
            return g;
        }

        var hasProblem = false;
        foreach (var token in matchAttribute.Value.Split(";".ToCharArray()))
        {
            if (token == ",")
            {
                hasProblem = true;
                break;
            }
        }

        if (hasProblem)
        {
            // the Indices attribute is not well-formed: let's try to fix the one(s) that are wrong
            var fixedTokens = new List<string>();
            foreach (var token in g.Attribute("Indices").Value.Split(";".ToCharArray()))
            {
                var newToken = token;
                var matchToken = Regex.Match(token, @",(-\d+)");
                if (matchToken.Success) // negative number, yay ! it's not allowed :-(
                {
                    newToken = ",0"; // it should be zero, I believe
                }

                fixedTokens.Add(newToken);
            }

            g.Attribute("Indices").Value = string.Join(";", fixedTokens);
        }

        return g;
    }
}