Viewing Visio Document Changes in Git

Developing a Visio solution usually involves both .Net code and Visio ShapeSheet formulas. Good practice dictates that the source code is saved into a code repository, such as Git, where changes can be committed and commented. Visual Studio 2019 now includes native Git support, and can be linked to Azure DevOps easily. The code can be viewed by others and the changes made by commits can be reviewed. This is tried and tested for the .Net code, but any changes made to a Visio template, stencil or drawing document is a black box. If the Visio document is included in the Git project, then all that is visible is the fact that the file has been changed, but the detail of the actual changes are unknown. There may be some comments with the commit, but that is not a precise definition. So, what if there was a way to easily spot the changes?

Microsoft Visio, and the other Microsoft Office apps, use the Open Packaging Conventions (OPC) format, which is a compressed (.zip) file with folders, mainly containing .xml files. When you make a minor change to your template you normally only change a couple of these xml files. For example, if you change one master in your Visio template then only one contained xml file is materially changed, apart from the timestamp, and possibly the windows sizes and positions, in other xml files.

This also works for Open Office XML documents and AutoCAD web drawings. You can see a complete list on Wikipedia.org.

To get the contents of the file, simply rename it to <mytemplate>.zip and (in Windows) open it in File Explorer. You can read more about the Visio XML file format at your leisure.

But we’re interested in saving it, automatically, to a Git repository, and to be able to see the changes through each commit. In this example, there is a single Visio template called, Custom Flowchart.vstx, that contains 5 masters which are customised versions of the same-named built-in ones (with the Match master by name on drop option ticked, of course.)

In the following screenshot of Azure DevOps, notice that there is a Custom Flowchart.vstx folder that contains the sub-folders and files within them. These are all within an Expanded folder, and are, in fact, an automatically created unzipped copy of the Custom Flowchart.vstx file that is still visible in the Documents folder. In addition to being unzipped, the xml files are prettified so that each xml element is on a separate line, making it much easier to review changes. The xml text inside Visio xml documents is all in one line, making it impossible to spot the differences without extracting historical files to use within a viewer that can prettify automatically.

So, the process required is to copy, unzip and prettify each specified Visio, or other OPC, document. This means that there will initially be a large number of folders and files added to the repository, but then only the deltas will be required subsequently.

Even though VS 2019 now has a dedicated Git menu and windows, it does not expose the hidden .git/hooks folder that is automatically created in the solution folder. See this article for more ideas. There are a number of sample files in that folder, and now a two line pre-compile file containing the following lines:

#!/usr/bin/env sh
dotnet run --project prettyunzip --extract "Documents/Custom Flowchart.vstx" "Documents/Expanded"

This simply executes a small C# DotNet Core app, prettyunzip, with the arguments to extract a specified file into a specific folder. This was my first time writing a DotNet Core app, so I followed the walkthrough at Microsoft Docs. I then copied and adapted the example from Tom Laird-McConnell on GitHub. I added code to iterate a directory from Microsoft Docs, and finally include Bash command as shown on loune.net.

This is the code block that performs the extract, passes the folder through to the WalkDirectoryTree(…) method where it is prettified, and then the whole folder is added to Git.

var folder = (args.Length == 3) ? args[2] : Path.GetFileNameWithoutExtension(args[1]);

var fi = new FileInfo(args[1]);
var subfolder = fi.Name;
var fullfolder = Path.Combine(folder, subfolder);
if (fullfolder.Length > 0 && Directory.Exists(fullfolder))
{
  Directory.Delete(fullfolder, true);
}
Console.WriteLine($"Extracting {args[1]} to {fullfolder}");
ZipFile.ExtractToDirectory(args[1], fullfolder);
var rootDir = new DirectoryInfo(fullfolder);
WalkDirectoryTree(rootDir, "xml", true);
var result = Bash(@$"git add {folder}" );

The WalkDirectoryTree() has two additional parameters to the original:

• string extension – allows the “xml” extension to be used in the root.GetFiles(“*.{extension}”) call
• bool prettify – instructs to prettify if true as follows:

if (prettify)
{
  var doc = XDocument.Load(fi.FullName);
  doc.Save(fi.FullName);
}

That’s all that is required to prettify and xml file, since it is the default action of XDocument.Save(…)!

The last action is to execute the Bash command:

public static string Bash(string cmd)
{
  var escapedArgs = cmd.Replace("\"", "\\\"");
  var process = new Process()
  {
     StartInfo = new ProcessStartInfo
     {
       FileName = "C:/Program Files/Git/bin/bash",
       Arguments = $"-c \"{escapedArgs}\"",
       RedirectStandardOutput = true,
       UseShellExecute = false,
       CreateNoWindow = true,
     }
  };
  process.Start();
  string result = process.StandardOutput.ReadToEnd();
  process.WaitForExit();
  return result;
}

I am not concerning myself with editing the unzipped xml files in Git, because I always make Visio ShapeSheet edits in the Visio for the Desktop application. I merely want to have a record of changes made to the Visio ShapeSheet for transparency and understanding.

For now, I am running the prettyunzip project that is in the same solution as my Visio project, but I do have the option to install locally or globally.

I would like to thank my long-time friend and colleague, Nick Ajderian of Flikk.net, for help in working out this process, and, also a not-quite-as-long-time friend and fellow Visio MVP, John Goldsmith of visualSignals.co.uk, for pointing me towards DotNet Core.

Using Notepad++ to Edit Visio ShapeSheet formulas

Visio shapes get their smartness from the formulas that are entered into the ShapeSheet, but editing these formulas can be extremely tricky and prone to error because of the lack of a modern programmer’s interface. Formulas can be quite long (up to 64k characters) but even medium size ones like the one in User.GetWorkdays cell…

Adding a second Function header bar to Visio swimlanes

I was recently asked if a second function header bar can be added to the swimlanes in the cross-functional flowchart templates in Visio. Some swimlanes can get quite wide, so it can be useful to have a duplicate function header shape on the far-side too. It is quite simple to duplicate the existing function header…

More Parsing XML Data in Visio Shapes

My last article looked at parsing an XML string with a known structure and order of elements and attributes. This can be acceptable in some scenarios, but what if the elements or attributes are in a different order? Then it is necessary to use the element and attribute names rather than their index position. This…

Parsing XML data in Visio Shapes

One of my current projects uses XML data, and some of the values in the XML data control the display and content of Visio shapes. Therefore, I looked deeper into how the XML data can be used directly to control parts of the graphics. Although the external data linking feature in Visio Professional and Visio…

Highlighting dirty text in Visio shapes

I suppose I should explain what I mean by dirty text first 🙂 I often display the value of a Shape Data row in Visio text, but sometimes the solution requires that the value is also editable. This is ok if the client accepts that all editing is done with the Shape Data window or…

My new book on Visualizing Processes with Microsoft Visio has launched

Back in the early 1990s, there was an application called ABC Flowcharter that was the market leader for diagramming business flowcharts, but some of the brains behind Aldus PageMaker saw an opportunity to create something smarter, and left to write the Visio product, with the stated aim to overtake ABC Flowcharter within 2 years. They…