Walkthrough: Display statement completion

You can implement language-based statement completion by defining the identifiers for which you want to provide completion and then triggering a completion session. You can define statement completion in the context of a language service, define your own file name extension and content type and then display completion for just that type. Or, you can trigger completion for an existing content type—for example, "plaintext". This walkthrough shows how to trigger statement completion for the "plaintext" content type, which is the content type of text files. The "text" content type is the ancestor of all other content types, including code and XML files.

Statement completion is typically triggered by typing certain characters—for example, by typing the beginning of an identifier such as "using". It's typically dismissed by pressing the Spacebar, Tab, or Enter key to commit a selection. You can implement the IntelliSense features that trigger when typing a character by using a command handler for the keystrokes (the IOleCommandTarget interface) and a handler provider that implements the IVsTextViewCreationListener interface. To create the completion source, which is the list of identifiers that participate in completion, implement the ICompletionSource interface and a completion source provider (the ICompletionSourceProvider interface). The providers are Managed Extensibility Framework (MEF) component parts. They're responsible for exporting the source and controller classes and importing services and brokers—for example, the ITextStructureNavigatorSelectorService, which enables navigation in the text buffer, and the ICompletionBroker, which triggers the completion session.

This walkthrough shows how to implement statement completion for a hard-coded set of identifiers. In full implementations, the language service and the language documentation are responsible for providing that content.

Create a MEF Project

To create a MEF project

  1. Create a C# VSIX project. (In the New Project dialog, select Visual C# / Extensibility, then VSIX Project.) Name the solution CompletionTest.

  2. Add an Editor Classifier item template to the project. For more information, see Create an extension with an editor item template.

  3. Delete the existing class files.

  4. Add the following references to the project and make sure that CopyLocal is set to false:

    Microsoft.VisualStudio.Editor

    Microsoft.VisualStudio.Language.Intellisense

    Microsoft.VisualStudio.OLE.Interop

    Microsoft.VisualStudio.Shell.15.0

    Microsoft.VisualStudio.Shell.Immutable.10.0

    Microsoft.VisualStudio.TextManager.Interop

Implement the completion source

The completion source is responsible for collecting the set of identifiers and adding the content to the completion window when a user types a completion trigger, such as the first letters of an identifier. In this example, the identifiers and their descriptions are hard-coded in the AugmentCompletionSession method. In most real-world uses, you would use your language's parser to get the tokens to populate the completion list.

To implement the completion source

  1. Add a class file and name it TestCompletionSource.

  2. Add these imports:

    using System;
    using System.Collections.Generic;
    using System.Linq;
    using System.Text;
    using System.ComponentModel.Composition;
    using Microsoft.VisualStudio.Language.Intellisense;
    using Microsoft.VisualStudio.Text;
    using Microsoft.VisualStudio.Text.Operations;
    using Microsoft.VisualStudio.Utilities;
    
  3. Modify the class declaration for TestCompletionSource so that it implements ICompletionSource:

    internal class TestCompletionSource : ICompletionSource
    
  4. Add private fields for the source provider, the text buffer, and a list of Completion objects (which correspond to the identifiers that will participate in the completion session):

    private TestCompletionSourceProvider m_sourceProvider;
    private ITextBuffer m_textBuffer;
    private List<Completion> m_compList;
    
  5. Add a constructor that sets the source provider and buffer. The TestCompletionSourceProvider class is defined in later steps:

    public TestCompletionSource(TestCompletionSourceProvider sourceProvider, ITextBuffer textBuffer)
    {
        m_sourceProvider = sourceProvider;
        m_textBuffer = textBuffer;
    }
    
  6. Implement the AugmentCompletionSession method by adding a completion set that contains the completions you want to provide in the context. Each completion set contains a set of Completion completions, and corresponds to a tab of the completion window. (In Visual Basic projects, the completion window tabs are named Common and All.) The FindTokenSpanAtPosition method is defined in the next step.

    void ICompletionSource.AugmentCompletionSession(ICompletionSession session, IList<CompletionSet> completionSets)
    {
        List<string> strList = new List<string>();
        strList.Add("addition");
        strList.Add("adaptation");
        strList.Add("subtraction");
        strList.Add("summation");
        m_compList = new List<Completion>();
        foreach (string str in strList)
            m_compList.Add(new Completion(str, str, str, null, null));
    
        completionSets.Add(new CompletionSet(
            "Tokens",    //the non-localized title of the tab
            "Tokens",    //the display title of the tab
            FindTokenSpanAtPosition(session.GetTriggerPoint(m_textBuffer),
                session),
            m_compList,
            null));
    }
    
  7. The following method is used to find the current word from the position of the cursor:

    private ITrackingSpan FindTokenSpanAtPosition(ITrackingPoint point, ICompletionSession session)
    {
        SnapshotPoint currentPoint = (session.TextView.Caret.Position.BufferPosition) - 1;
        ITextStructureNavigator navigator = m_sourceProvider.NavigatorService.GetTextStructureNavigator(m_textBuffer);
        TextExtent extent = navigator.GetExtentOfWord(currentPoint);
        return currentPoint.Snapshot.CreateTrackingSpan(extent.Span, SpanTrackingMode.EdgeInclusive);
    }
    
  8. Implement the Dispose() method:

    private bool m_isDisposed;
    public void Dispose()
    {
        if (!m_isDisposed)
        {
            GC.SuppressFinalize(this);
            m_isDisposed = true;
        }
    }
    

Implement the completion source provider

The completion source provider is the MEF component part that instantiates the completion source.

To implement the completion source provider

  1. Add a class named TestCompletionSourceProvider that implements ICompletionSourceProvider. Export this class with a ContentTypeAttribute of "plaintext" and a NameAttribute of "test completion".

    [Export(typeof(ICompletionSourceProvider))]
    [ContentType("plaintext")]
    [Name("token completion")]
    internal class TestCompletionSourceProvider : ICompletionSourceProvider
    
  2. Import a ITextStructureNavigatorSelectorService, which finds the current word in the completion source.

    [Import]
    internal ITextStructureNavigatorSelectorService NavigatorService { get; set; }
    
  3. Implement the TryCreateCompletionSource method to instantiate the completion source.

    public ICompletionSource TryCreateCompletionSource(ITextBuffer textBuffer)
    {
        return new TestCompletionSource(this, textBuffer);
    }
    

Implement the completion command handler provider

The completion command handler provider is derived from a IVsTextViewCreationListener, which listens for a text view creation event and converts the view from an IVsTextView—which enables the addition of the command to the command chain of the Visual Studio shell—to an ITextView. Because this class is a MEF export, you can also use it to import the services that are required by the command handler itself.

To implement the completion command handler provider

  1. Add a file named TestCompletionCommandHandler.

  2. Add these using directives:

    using System;
    using System.ComponentModel.Composition;
    using System.Runtime.InteropServices;
    using Microsoft.VisualStudio;
    using Microsoft.VisualStudio.Editor;
    using Microsoft.VisualStudio.Language.Intellisense;
    using Microsoft.VisualStudio.OLE.Interop;
    using Microsoft.VisualStudio.Shell;
    using Microsoft.VisualStudio.Text;
    using Microsoft.VisualStudio.Text.Editor;
    using Microsoft.VisualStudio.TextManager.Interop;
    using Microsoft.VisualStudio.Utilities;
    
  3. Add a class named TestCompletionHandlerProvider that implements IVsTextViewCreationListener. Export this class with a NameAttribute of "token completion handler", a ContentTypeAttribute of "plaintext", and a TextViewRoleAttribute of Editable.

    [Export(typeof(IVsTextViewCreationListener))]
    [Name("token completion handler")]
    [ContentType("plaintext")]
    [TextViewRole(PredefinedTextViewRoles.Editable)]
    internal class TestCompletionHandlerProvider : IVsTextViewCreationListener
    
  4. Import the IVsEditorAdaptersFactoryService, which enables conversion from a IVsTextView to a ITextView, a ICompletionBroker, and a SVsServiceProvider that enables access to standard Visual Studio services.

    [Import]
    internal IVsEditorAdaptersFactoryService AdapterService = null;
    [Import]
    internal ICompletionBroker CompletionBroker { get; set; }
    [Import]
    internal SVsServiceProvider ServiceProvider { get; set; }
    
  5. Implement the VsTextViewCreated method to instantiate the command handler.

    public void VsTextViewCreated(IVsTextView textViewAdapter)
    {
        ITextView textView = AdapterService.GetWpfTextView(textViewAdapter);
        if (textView == null)
            return;
    
        Func<TestCompletionCommandHandler> createCommandHandler = delegate() { return new TestCompletionCommandHandler(textViewAdapter, textView, this); };
        textView.Properties.GetOrCreateSingletonProperty(createCommandHandler);
    }
    

Implement the completion command handler

Because statement completion is triggered by keystrokes, you must implement the IOleCommandTarget interface to receive and process the keystrokes that trigger, commit, and dismiss the completion session.

To implement the completion command handler

  1. Add a class named TestCompletionCommandHandler that implements IOleCommandTarget:

    internal class TestCompletionCommandHandler : IOleCommandTarget
    
  2. Add private fields for the next command handler (to which you pass the command), the text view, the command handler provider (which enables access to various services), and a completion session:

    private IOleCommandTarget m_nextCommandHandler;
    private ITextView m_textView;
    private TestCompletionHandlerProvider m_provider;
    private ICompletionSession m_session;
    
  3. Add a constructor that sets the text view and the provider fields, and adds the command to the command chain:

    internal TestCompletionCommandHandler(IVsTextView textViewAdapter, ITextView textView, TestCompletionHandlerProvider provider)
    {
        this.m_textView = textView;
        this.m_provider = provider;
    
        //add the command to the command chain
        textViewAdapter.AddCommandFilter(this, out m_nextCommandHandler);
    }
    
  4. Implement the QueryStatus method by passing the command along:

    public int QueryStatus(ref Guid pguidCmdGroup, uint cCmds, OLECMD[] prgCmds, IntPtr pCmdText)
    {
        return m_nextCommandHandler.QueryStatus(ref pguidCmdGroup, cCmds, prgCmds, pCmdText);
    }
    
  5. Implement the Exec method. When this method receives a keystroke, it must do one of these things:

    • Allow the character to be written to the buffer, and then trigger or filter completion. (Printing characters do this.)

    • Commit the completion, but do not allow the character to be written to the buffer. (Whitespace, Tab, and Enter do this when a completion session is displayed.)

    • Allow the command to be passed on to the next handler. (All other commands.)

      Because this method may display UI, call IsInAutomationFunction to make sure that it's not called in an automation context:

      public int Exec(ref Guid pguidCmdGroup, uint nCmdID, uint nCmdexecopt, IntPtr pvaIn, IntPtr pvaOut)
      {
          if (VsShellUtilities.IsInAutomationFunction(m_provider.ServiceProvider))
          {
              return m_nextCommandHandler.Exec(ref pguidCmdGroup, nCmdID, nCmdexecopt, pvaIn, pvaOut);
          }
          //make a copy of this so we can look at it after forwarding some commands
          uint commandID = nCmdID;
          char typedChar = char.MinValue;
          //make sure the input is a char before getting it
          if (pguidCmdGroup == VSConstants.VSStd2K && nCmdID == (uint)VSConstants.VSStd2KCmdID.TYPECHAR)
          {
              typedChar = (char)(ushort)Marshal.GetObjectForNativeVariant(pvaIn);
          }
      
          //check for a commit character
          if (nCmdID == (uint)VSConstants.VSStd2KCmdID.RETURN
              || nCmdID == (uint)VSConstants.VSStd2KCmdID.TAB
              || (char.IsWhiteSpace(typedChar) || char.IsPunctuation(typedChar)))
          {
              //check for a selection
              if (m_session != null && !m_session.IsDismissed)
              {
                  //if the selection is fully selected, commit the current session
                  if (m_session.SelectedCompletionSet.SelectionStatus.IsSelected)
                  {
                      m_session.Commit();
                      //also, don't add the character to the buffer
                      return VSConstants.S_OK;
                  }
                  else
                  {
                      //if there is no selection, dismiss the session
                      m_session.Dismiss();
                  }
              }
          }
      
          //pass along the command so the char is added to the buffer
          int retVal = m_nextCommandHandler.Exec(ref pguidCmdGroup, nCmdID, nCmdexecopt, pvaIn, pvaOut);
          bool handled = false;
          if (!typedChar.Equals(char.MinValue) && char.IsLetterOrDigit(typedChar))
          {
              if (m_session == null || m_session.IsDismissed) // If there is no active session, bring up completion
              {
                  this.TriggerCompletion();
                  m_session.Filter();
              }
              else    //the completion session is already active, so just filter
              {
                  m_session.Filter();
              }
              handled = true;
          }
          else if (commandID == (uint)VSConstants.VSStd2KCmdID.BACKSPACE   //redo the filter if there is a deletion
              || commandID == (uint)VSConstants.VSStd2KCmdID.DELETE)
          {
              if (m_session != null && !m_session.IsDismissed)
                  m_session.Filter();
              handled = true;
          }
          if (handled) return VSConstants.S_OK;
          return retVal;
      }
      

  6. This code is a private method that triggers the completion session:

    private bool TriggerCompletion()
    {
        //the caret must be in a non-projection location 
        SnapshotPoint? caretPoint =
        m_textView.Caret.Position.Point.GetPoint(
        textBuffer => (!textBuffer.ContentType.IsOfType("projection")), PositionAffinity.Predecessor);
        if (!caretPoint.HasValue)
        {
            return false;
        }
    
        m_session = m_provider.CompletionBroker.CreateCompletionSession
     (m_textView,
            caretPoint.Value.Snapshot.CreateTrackingPoint(caretPoint.Value.Position, PointTrackingMode.Positive),
            true);
    
        //subscribe to the Dismissed event on the session 
        m_session.Dismissed += this.OnSessionDismissed;
        m_session.Start();
    
        return true;
    }
    
  7. The next example is a private method that unsubscribes from the Dismissed event:

    private void OnSessionDismissed(object sender, EventArgs e)
    {
        m_session.Dismissed -= this.OnSessionDismissed;
        m_session = null;
    }
    

Build and test the code

To test this code, build the CompletionTest solution and run it in the experimental instance.

To build and test the CompletionTest solution

  1. Build the solution.

  2. When you run this project in the debugger, a second instance of Visual Studio is started.

  3. Create a text file and type some text that includes the word "add".

  4. As you type first "a" and then "d", a list that contains "addition" and "adaptation" should appear. Notice that addition is selected. When you type another "d", the list should contain only "addition", which is now selected. You can commit "addition" by pressing the Spacebar, Tab, or Enter key, or dismiss the list by typing Esc or any other key.