Malformed mailto Hyperlink causes Exception on .NET 4.5+

[koyote]

Hi,

This issue is similar to #7 but for recent versions of the .NET framework.

If you create a blank document and set some text to the following field code (note the extra space):
{ HYPERLINK "mailto:email@address%20.com" }

Running the following on .NET 4.0 will work fine:
var wpdoc = WordprocessingDocument.Open(@"TestDoc.docx", false)
Running the same line on .NET 4.5+ will throw the following exception:

DocumentFormat.OpenXml.Packaging.OpenXmlPackageException: Invalid Hyperlink: Malformed URI is embedded as a hyperlink in the document.
   at DocumentFormat.OpenXml.Packaging.OpenXmlPackage.Load() in OpenXmlPackage.cs: line 490
   at DocumentFormat.OpenXml.Packaging.OpenXmlPackage.OpenCore(String path, Boolean readWriteMode) in OpenXmlPackage.cs: line 402
   at DocumentFormat.OpenXml.Packaging.WordprocessingDocument.Open(String path, Boolean isEditable, OpenSettings openSettings) in PackageDocument.cs: line 297
   at  DocumentFormat.OpenXml.Packaging.WordprocessingDocument.Open(String path, Boolean isEditable) in PackageDocument.cs: line 256

The inner-exception being this:

   at System.Uri.CreateThis(String uri, Boolean dontEscape, UriKind uriKind)
   at System.Uri..ctor(String uriString, UriKind uriKind)
   at MS.Internal.IO.Packaging.InternalRelationshipCollection.ProcessRelationshipAttributes(XmlCompatibilityReader reader)
   at MS.Internal.IO.Packaging.InternalRelationshipCollection.ParseRelationshipPart(PackagePart part)
   at MS.Internal.IO.Packaging.InternalRelationshipCollection..ctor(Package package, PackagePart part)
   at System.IO.Packaging.PackagePart.EnsureRelationships()
   at System.IO.Packaging.PackagePart.GetRelationshipsHelper(String filterString)
   at System.IO.Packaging.PackagePart.GetRelationships()
   at DocumentFormat.OpenXml.Packaging.PackagePartRelationshipPropertyCollection..ctor(PackagePart packagePart)
   at DocumentFormat.OpenXml.Packaging.OpenXmlPart.Load(OpenXmlPackage openXmlPackage, OpenXmlPart parent, Uri uriTarget, String id, Dictionary`2 loadedParts)
   at DocumentFormat.OpenXml.Packaging.OpenXmlPartContainer.LoadReferencedPartsAndRelationships(OpenXmlPackage openXmlPackage, OpenXmlPart sourcePart, RelationshipCollection relationshipCollection, Dictionary`2 loadedParts)
   at DocumentFormat.OpenXml.Packaging.OpenXmlPackage.Load()

I assume this is due to the following change introduced in .Net 4.5 (https://msdn.microsoft.com/en-us/library/hh367887%28v=vs.110%29.aspx):
An invalid mailto: URL throws an exception in the Uri class constructor.

Word seems to be perfectly happy with the malformed mailto hyperlink so I assume this would count as a valid document.
I am therefore not sure how this could be fixed elegantly.

Thanks,

[EricWhiteDev]

Hi,

As you determined, this is an issue in the underlying System.IO.Packaging.

Currently, we are assessing whether we can patch System.IO.Packaging to fix this. One issue is that the System.IO.Packaging returns a Uri object when you request a list of external relationships, and if the Uri is invalid in the document, then we can't return one.

Perhaps the good semantics would be:

• Don't throw an exception upon opening.
• If the program using the SDK (and therefore System.IO.Packaging) requests an invalid Uri, then throw an exception at that point
• If the program using the SDK never requests the invalid link, then upon serializing, the Open XML SDK and System.IO.Packaging would serialize the bad Uri as is.

Currently, the recommended workaround is:

http://openxmldeveloper.org/blog/b/openxmldeveloper/archive/2014/08/19/handling-invalid-hyperlinks-openxmlpackageexception-in-the-open-xml-sdk.aspx

[EricWhiteDev]

I'm assigning this to myself, and to version 4.0. This is an issue intrinsic to the design of System.IO.Packaging, and we need to be careful about how we fix this. In the meantime, the best approach is the workaround in the above comment.

[EricWhiteDev]

Closing this suggestion. This is most likely not going to be implemented, so should close this issue to reflect that.

[houssemzaier]

Hello
I have the sale problem when I want to read an Excel file containing a non valid URI I don't have any solution to this problem yet... any help please ?

[davideloper]

I guess I have the same problem with xlsx: "PackageRelationship target must be relative URI if TargetMode is Internal"
This error is only presented on mono, on windows everything looks fine.
I am using OpenXmlSdk 2.6 and System.IO.Packaging from .NETStandard.
The error fires on SpreadsheetDocument.Open(filePath ...).
Any possible solutions?

[igitur]

For those still struggling with this issue, please see @EricWhiteDev 's article on how to handle this: http://ericwhite.com/blog/handling-invalid-hyperlinks-openxmlpackageexception-in-the-open-xml-sdk/

The link Eric posted above seems obsolete.

@EricWhiteDev: Is there absolutely no way that this workaround will be implemented going forward?

[silviomarquesferreira]

Hi Guys, I'm facing the same problem with a docx document.
Is there some new implementation inside system.io.packaging?

[AmeetShinde]

Hello guys, we are facing similar issue and the workaround is not acceptable in this case (no write access to document and making in-memory copy of the document is too costly). Can we get this fixed in System.IO.Packaging library itself?

[robertmuehsig]

He @EricWhiteDev - is this problem solved when using a newer release of the OpenXML SDK or is this "by design" broken? @twsouthwick

[twsouthwick]

This is a "by design" issue by the fact that System.IO.Packaging.Package throws the exception and there's no way to catch it. That said, I have a design for a fix I'm going to propose on CoreFx that would allow to hook into the package reading to handle it manually. It hasn't been high on my priority list, but if there's interest I'll get that proposal submitted so there'll be a viable workaround.

[robertmuehsig]

Uh - this is bad, because we process Office documents and our main message is "use whatever feature you like in Word and we can handle it" (at least we shouldn't fail on it) and we have a customer issue because she or he inserted a link with a custom URL schema, like "myapp:foobar&param=123".

Word itself is doing fine, but our application needs to insert some CustomXML into the document, but because of this issue it will fail.

Now our application looks quite bad, because "Word can handle it" and we fail pretty early (and the workaround seems quite perf heavy :-/)