I eventually came up with a solution, at least for VS2010.
Whilst I have used this for coloring '#region' and '#endregion' tags, a similar solution ought to be applicable for any text content in a Visual Studio window.
It seems that this sort of problem can be resolved by creating a IViewTaggerProvider which will 'tag' parts of the source code with a 'classification'.
Visual Studio will provide a style for text tagged with that classification which can then be changed by the user to the desired style via Tools > Options... > Environment > Fonts and Colors.
The Tagger provider looks like:
[Export(typeof(IViewTaggerProvider))]
[ContentType("any")]
[TagType(typeof(ClassificationTag))]
public sealed class RegionTaggerProvider : IViewTaggerProvider
{
[Import]
public IClassificationTypeRegistryService Registry;
[Import]
internal ITextSearchService TextSearchService { get; set; }
public ITagger<T> CreateTagger<T>(ITextView textView, ITextBuffer buffer) where T : ITag
{
if (buffer != textView.TextBuffer)
return null;
var classType = Registry.GetClassificationType("region-foreground");
return new RegionTagger(textView, TextSearchService, classType) as ITagger<T>;
}
}
This creates an ITagger object, which, given a Visual Studio text view, will tag parts of the text with the given classification type. Note that this will work for all text views (i.e. source code editor, 'Find Results' windows etc.). It may be possible to change this by editing the ContentType attribute (to just C#?).
The classification type (in this case "region-foreground") is defined as:
public static class TypeExports
{
[Export(typeof(ClassificationTypeDefinition))]
[Name("region-foreground")]
public static ClassificationTypeDefinition OrdinaryClassificationType;
}
[Export(typeof(EditorFormatDefinition))]
[ClassificationType(ClassificationTypeNames = "region-foreground")]
[Name("region-foreground")]
[UserVisible(true)]
[Order(After = Priority.High)]
public sealed class RegionForeground : ClassificationFormatDefinition
{
public RegionForeground()
{
DisplayName = "Region Foreground";
ForegroundColor = Colors.Gray;
}
}
The Order attribute determines when the classification will be applied compared to other classifications which may also apply to a span of text.
The DisplayName will be used in the Tools > Options... dialog.
Once the classification is defined, an ITagger class can search a view's text and provide classifications for applicable sections of the text it finds.
Simply put, its job is to listen for the ViewLayoutChanged event of the text view, which is fired when the provided text view's content changes (e.g. because the the user has typed something).
It must then search the text for the area of text it is interested in (called a 'span'). Here, it returns spans of lines containing either #region or #endregion. I've kept this simple, but the TextSearchService used to find matches can also search using regular expressions.
Finally, a method is provided for Visual Studio to retrieve the tags of the text it has found, called GetTags(). For a given span collection, this will return text spans with classification tags, i.e. areas of those spans which should be classified in a certain way.
Its code is:
public sealed class RegionTagger : ITagger<ClassificationTag>
{
private readonly ITextView m_View;
private readonly ITextSearchService m_SearchService;
private readonly IClassificationType m_Type;
private NormalizedSnapshotSpanCollection m_CurrentSpans;
public event EventHandler<SnapshotSpanEventArgs> TagsChanged = delegate { };
public RegionTagger(ITextView view, ITextSearchService searchService, IClassificationType type)
{
m_View = view;
m_SearchService = searchService;
m_Type = type;
m_CurrentSpans = GetWordSpans(m_View.TextSnapshot);
m_View.GotAggregateFocus += SetupSelectionChangedListener;
}
private void SetupSelectionChangedListener(object sender, EventArgs e)
{
if (m_View != null)
{
m_View.LayoutChanged += ViewLayoutChanged;
m_View.GotAggregateFocus -= SetupSelectionChangedListener;
}
}
private void ViewLayoutChanged(object sender, TextViewLayoutChangedEventArgs e)
{
if (e.OldSnapshot != e.NewSnapshot)
{
m_CurrentSpans = GetWordSpans(e.NewSnapshot);
TagsChanged(this, new SnapshotSpanEventArgs(new SnapshotSpan(e.NewSnapshot, 0, e.NewSnapshot.Length)));
}
}
private NormalizedSnapshotSpanCollection GetWordSpans(ITextSnapshot snapshot)
{
var wordSpans = new List<SnapshotSpan>();
wordSpans.AddRange(FindAll(@"#region", snapshot).Select(regionLine => regionLine.Start.GetContainingLine().Extent));
wordSpans.AddRange(FindAll(@"#endregion", snapshot).Select(regionLine => regionLine.Start.GetContainingLine().Extent));
return new NormalizedSnapshotSpanCollection(wordSpans);
}
private IEnumerable<SnapshotSpan> FindAll(String searchPattern, ITextSnapshot textSnapshot)
{
if (textSnapshot == null)
return null;
return m_SearchService.FindAll(
new FindData(searchPattern, textSnapshot) {
FindOptions = FindOptions.WholeWord | FindOptions.MatchCase
});
}
public IEnumerable<ITagSpan<ClassificationTag>> GetTags(NormalizedSnapshotSpanCollection spans)
{
if (spans == null || spans.Count == 0 || m_CurrentSpans.Count == 0)
yield break;
ITextSnapshot snapshot = m_CurrentSpans[0].Snapshot;
spans = new NormalizedSnapshotSpanCollection(spans.Select(s => s.TranslateTo(snapshot, SpanTrackingMode.EdgeExclusive)));
foreach (var span in NormalizedSnapshotSpanCollection.Intersection(m_CurrentSpans, spans))
{
yield return new TagSpan<ClassificationTag>(span, new ClassificationTag(m_Type));
}
}
}
For brevity I have omitted namespaces and the using statements, which are typically of the form Microsoft.VisualStudio.Text.*. For these to be available, the Visual Studio 2010 SDK must first be downloaded.
I've been using this solution for the past few months without issue.
One limitation I noticed is colours are not 'blended', so a color with less than 100% opacity will not 'fade out' the existing colours in a span - which might be useful to preserve syntax highlighting.
I also have little idea of its efficiency, as it looks like it will repeatedly search a document on each keypress. I have not done the research to see if Visual Studio optimises this somehow. I do notice a slowdown of Visual Studio on large files (> ~1000 lines), but I also use Resharper, so I cannot attribute this to this plugin alone.
As this was coded mostly using guesswork, I welcome any comments or code changes which could clarify or simplify things or improve upon the code's performance.
