Setting `PackageProperties.Title` with 0x0001 Unicode Characters in OpenXML Results in an Error

[lindexi]

Describe the bug
When attempting to configure the PackageProperties.Title attribute of a PresentationDocument within OpenXML, inserting a string containing the Unicode character \u0001 (often known as a soft hyphen) generates an ArgumentException. This occurs despite the presence of other valid properties.

Screenshots
None.

To Reproduce

The mini repro code is:

using System;

using (var presentationDocument =
       DocumentFormat.OpenXml.Packaging.PresentationDocument.Open("Test.pptx", true))
{
    presentationDocument.PackageProperties.Title = "\u0001";
}

The Test.pptx is any normal pptx file.

You can find all my demo project in https://github.com/lindexi/lindexi_gd/tree/6baba8b5407c1d23119d3ac150b1ab5af4cd810c/Pptx/FukemqibairLaylalljerowhem

Observed behavior

Run my mini repro code, and you can find the exception:

 System.ArgumentException:“'�', hexadecimal value 0x01, is an invalid character.”
  	 System.Private.Xml.dll!System.Xml.XmlUtf8RawTextWriter.InvalidXmlChar(int ch, byte* pDst, bool entitize)	
  	 System.Private.Xml.dll!System.Xml.XmlUtf8RawTextWriter.WriteElementTextBlock(char* pSrc, char* pSrcEnd)	
  	 System.Private.Xml.dll!System.Xml.XmlUtf8RawTextWriter.WriteString(string text)	
  	 System.Private.Xml.dll!System.Xml.XmlWellFormedWriter.WriteString(string text)	
  	 System.IO.Packaging.dll!System.IO.Packaging.PartBasedPackageProperties.SerializeDirtyProperties()	
  	 System.IO.Packaging.dll!System.IO.Packaging.PartBasedPackageProperties.Flush()	
  	 System.IO.Packaging.dll!System.IO.Packaging.PartBasedPackageProperties.Close()	
  	 System.IO.Packaging.dll!System.IO.Packaging.Package.System.IDisposable.Dispose()	
  	 System.IO.Packaging.dll!System.IO.Packaging.Package.Close()	
  	 DocumentFormat.OpenXml.Framework.dll!DocumentFormat.OpenXml.Features.StreamPackageFeature.Dispose(bool disposing)	
  	 DocumentFormat.OpenXml.Framework.dll!DocumentFormat.OpenXml.Features.StreamPackageFeature.Dispose()	
  	 DocumentFormat.OpenXml.Framework.dll!DocumentFormat.OpenXml.Packaging.PackageFeatureCollection.DocumentFormat.OpenXml.Features.IContainerDisposableFeature.Dispose()	
  	 DocumentFormat.OpenXml.Framework.dll!DocumentFormat.OpenXml.Packaging.OpenXmlPackage.Dispose(bool disposing)	
  	 DocumentFormat.OpenXml.Framework.dll!DocumentFormat.OpenXml.Packaging.OpenXmlPackage.Dispose()	
 >	 FukemqibairLaylalljerowhem.dll!Program.<Main>$(string[] args)

Expected behavior

Work well without exception.

Desktop (please complete the following information):

• .NET Target: .NET 9
• DocumentFormat.OpenXml Version: 3.2.0

Additional context

This error throw in dotnet runtime library: https://github.com/dotnet/runtime/blob/c1fe87ad88532f0e80de3739fe7b215e6e1f8b90/src/libraries/System.Private.Xml/src/System/Xml/Core/XmlUtf8RawTextWriter.cs#L932

Reference #1532

[mikeebowen]

Hi @lindexi,

This does seem like it should work. I did some digging today and I believe the issue might be in System.IO.Packaging. I will investigate further and see what I can find.

[mkaszewiak]

Hi @lindexi \u0001 is a Start of Heading which is a control code character https://en.wikipedia.org/wiki/List_of_Unicode_characters#Control_codes control codes don't have a visual representation.
Soft hyphen is \u00AD https://en.wikipedia.org/wiki/Soft_hyphen#:~:text=In%20this%20context%2C%20the%20soft,the%20text%20is%20re%2Dflowed.
I tested it and it works for the title

[lindexi]

@mkaszewiak Thank you. And I think we can intercept this issue, at least, we can make it do not throw the exception on dispose. Because the we can not guess which property be set the wrong value by this exception.

[lindexi]

This is the code I'm using to filter characters that cannot be included in an XML file:

internal static class XmlSafeTextContentHelper
{
    public static string ToSafeXmlText(string text, char replaceChar = '_')
    {
         var index = -1;
        for (var i = 0; i < text.Length; i++)
        {
            var c = text[i];
            if (!XmlConvert.IsXmlChar(c))
            {
                index = i;
                break;
            }
        }

        var canFindInvalidChar = index >= 0;
        if (canFindInvalidChar)
        {
            var stringBuilder = new StringBuilder(text.Length);
            stringBuilder.Append(text, 0, index);
            for (var i = index; i < text.Length; i++)
            {
                var c = text[i];
                if (XmlConvert.IsXmlChar(c))
                {
                    stringBuilder.Append(c);
                }
                else
                {
                    stringBuilder.Append(replaceChar);
                }
            }

            return stringBuilder.ToString();
        }
        else
        {
            return text;
        }
    }
}

[mkaszewiak]

@lindexi I wonder if below solution would also work for you?

    var text = "\u0001SomeTextABC\u0001";
    var validText = "";
    for (var i = 0; i < text.Length; i++)
    {
        if (!XmlConvert.IsXmlChar(text[i]))
        {
            validText += "_";
            continue;
        }
        validText += text[i];
    }

[lindexi]

@mkaszewiak Yeah, our approach yields the same result as yours, but my code offers better performance and lower resource consumption. For most normal strings, there is no extra memory allocation or string concatenation overhead. However, for strings containing abnormal characters, our method utilizes a StringBuilder to minimize the overhead of object allocation caused by string concatenations.

[mkaszewiak]

Hi @lindexi thanks for the explanation, Sorry but sometimes I am trying to be to helpful therefore I shared my solution which may be suitable for your needs.
I am pretty sure you can change validText type to StringBuilder and call .Append() instead but you know better your requirements therefore I believe your solution works better for your needs.
I hope everything else from OpenXML side of things is working for you? Thanks

[lindexi]

@mkaszewiak Thank you. I expect there to be a built-in mechanism in OpenXML to help us to filter the xml string.

And the #1532 (comment) may help.

But I can not sure the Office application can consume the document well. If the Office application can not consume the document, I think we should make the OpenXML SDK to save the document after xml string filter.