Realsolve Logo
Articles
 

Building an Eclipse Text Editor with JFace Text

How to create an enhanced Eclipse XML editor using JFace Text

By Phil Zoio, April 2006

Authors Note: I had originally intended to have this article published externally. Time constraints prevent me from putting in the necessary effort to achieve this, but it should still be a worthwhile read for the aspiring Eclipse plugin developer.

Introduction

JFace Text is a sophisticated framework which allows an Eclipse plug-in developer to build text editors with advanced features, such as syntax highlighting, basic content assistance and code formatting. In my view, an understanding of JFace Text is very important for Eclipse developers because most Eclipse plug-ins involve a text editor of some kind. It is not an easy API to understand, but is very powerful.

I wrote this article because found it difficult to find documentation on this impressive API. For example, I struggled to find material which explains the document partitioning process very well. If you don't know what this is right now, don't worry, since this is one of the major areas covered in this article.

The artide is built around the default example XML editor provided as an Eclipse plugin development template. Before we talk in more detail about the APIs, lets set the scene by describing the application we are going to use.

Part 1: Setting the Scene

The Application

The first likely step for anyone intending to write an Eclipse editor plugin, and a JFace Text based editor in particular, is to create a new Eclipse Plugin Development Environment with the XML editor example as a template.

You can do this as follows:

  1. File -> New -> Project ...
  2. Select Plug-in Project

Then follow the wizard screen until you get to the Templates selection as shown below:

Template Wizard to create a new default XML editor

Here you check the box to create the plug-in using one of the templates, and select the template Plug-in with an editor. Eclipse very kindly generates for you a working XML editor, albeit one which does very little.

The XML editor which Eclipse generates looks like this:

The basic autogenerated XML editor in action

Of course, you get all the source code, and the generated application is quite nicely written, so it makes an excellent starting point for learning the new technology. The XML editor is clearly more than just a text editor – it has syntax highlighting. What's missing though, are some really useful features:

So you want all of these features and more, and what you have is the basic text editor source, where do you go from here? That's what this article is all about. I am no expert. Hopefully, however, I have learnt enough to know what is easy to achieve, and give some helpful indications on how to accomplish the more difficult bits.

You can use this link to download the full source code on which this article is based.

Part 2: Getting to Know JFace Text

Text Editors

The JFace Text framework is ordinarily used with an Eclipse text editor, for which the abstract base class is a AbstractTextEditor (package and plugin name). Normally, it would be most convenient to subclass TextEditor, which is what the provided XML editor does in a class named XMLEditor.

The Eclipse provided XMLEditor class itself does does not do much – it simply uses the base class functionality, and delegates the job of adding additional features.

public class XMLEditor extends TextEditor
{

    private ColorManager colorManager;

    public XMLEditor()
    {
        super();
        colorManager = new ColorManager();
        setSourceViewerConfiguration(new XMLConfiguration(colorManager));
        setDocumentProvider(new XMLDocumentProvider());
    }

    public void dispose()
    {
        colorManager.dispose();
        super.dispose();
    }

}

Its first delegate is a SourceViewerConfiguration class instance, which is used to add additional features to the editor's user interface. The second is an instance of the IDocumentProvider interface, which encapsulates the mechanism for creating a JFace Text representation of the document being edited from its source (such as a file in the file system). Understanding how to use the SourceViewerConfiguration and the DocumentProvider is necessary, so we will cover both later in the article.

We will also see that when adding additional features to the XML editor, such as error marking and content outlining, we will also need to add some enhancements to the XML editor class.

JFace Text Documents

When working with JFace Text, the document you are editing works with an IDocument instance. The IDocument contains the text of your document, and can be used to query its structure. The IDocument can also be used to mark for positions in the document, which the framework can use to store document partition information and other metadata, such as highlighting ranges. We'll learn more about document partitioning later in the article. Positions are stored in documents as instances of the Position class or subclasses.

Your application probably won't often need to use anything other than the default provided IDocument implementations. All the provided IDocument implementations are also partition-aware, that is, can be divided into non-overlapping regions of text. Partitions play an important role in JFace text, so we will be covering them in some detail later in the article.

The IDocument interface is designed to work without any knowledge of how it is stored. That is, an IDocument instance has no knowledge of whether it has been loaded from the file system, from a database or from any other source. The job of creating the document instance and initiating the process that sets the documents initial state is the job of the IDocumentProvider. Eclipse provides excellent out of the box support form loading documents from the file system, through the FileDocumentProvider class.

We saw when introducting the XMLEditor class how the IDocumentProvider instance is made available to the editor. We cannot simply use the FileDocumentProvider as is, because the FileDocumentProvider has no knowledge of our plugin's partitioning scheme.

The XML editor's document provider remedies this by overriding only one method – createDocument() - with the following code:

protected IDocument createDocument(Object element) throws CoreException
{
    IDocument document = super.createDocument(element);
    ...
    return document;
}

Unless you already know JFace Text, there would be no point at this stage describing the code that goes in this method, except to say that in addition to the inherited behaviour, it configures the document's partitioning mechanism. At this stage, you may not be very clear on what partitioning is, or how it works. Since partitioning is so central to the way JFace Text works, we will now talk about it in a bit more detail.

Partitioning

When you open a document which uses JFace Text, the framework divides the document into partitions, that is, a region of text within the document. These partitions are non-overlapping. The partitions are categorised, so that each partition is associated with a particular "content type".

The easiest way to understand partitioning is by viewing an example. I have added a simple mechanism to easily print the document partitioning to the console, which I will describe later. For now, we are concerned about how the document is partitioned.

Below, we have an abridged version of the partitioning for the document shown above.

Partition type: __xml_pi, offset: 0, length: 21
Text:
<?xml version="1.0"?>
---------------------------

Partition type: __dftl_partition_content_type, offset: 21, length: 2
Text:

---------------------------

Partition type: __xml_doctype, offset: 23, length: 36
Text:
<!DOCTYPE world SYSTEM "cities.dtd">
---------------------------

Partition type: __dftl_partition_content_type, offset: 59, length: 2
Text:

---------------------------

Partition type: __xml_start_tag, offset: 61, length: 7
Text:
<world>
---------------------------

Partition type: __dftl_partition_content_type, offset: 68, length: 5
Text:
  
---------------------------

Partition type: __xml_start_tag, offset: 73, length: 27
Text:
<continent name = "Africa">
---------------------------

Partition type: __dftl_partition_content_type, offset: 100, length: 4
Text:

--------------------------

Partition type: __xml_start_tag, offset: 104, length: 13
Text:
<description>
---------------------------

... rest of document until end ...

Partition type: __xml_end_tag, offset: 1301, length: 8
Text:
</world>
---------------------------

Notice a few things about the document's partitioning:

By default JFace Text will partition the document when it is initially read, and will also repartion part of the document when modifications are made.

So now that we know what partitioning is, lets consider why it is important:

The default editor only defines two content types. The options for adding new features are as a result still quite limited. We will need to add some more content types to progress towards a more sophisticated editor. We tackle this task later in the article.

We are now in a position to go back to how the partitioning mechanism is set up in the XMLDocumentProvider's implementation of createDocument(), which we show below

protected IDocument createDocument(Object element) throws CoreException
{
    IDocument document = super.createDocument(element);
    if (document != null)
    {
        IDocumentPartitioner partitioner = new XMLPartitioner(
        new XMLPartitionScanner(), new String[]                            
          {
                  XMLPartitionScanner.XML_TAG, XMLPartitionScanner.XML_COMMENT
          });
        partitioner.connect(document);
        document.setDocumentPartitioner(partitioner);
    }
    return document;
}

We need to create an instance of IDocumentPartitioner. JFace Text has been neatly separated into parts of the framework which your application will need to provide implementations and parts that don't. The IDocumentPartitioner definitely fits into the latter category; normally you can simply get away with using the FastPartitioner in Eclipse 3.1 (or DefaultPartitioner in 3.0 or below). However, we have extended this for debugging purposes, simply so that we can get the quick and easy print out of the document's partitioning that we saw earlier in this section. The code to get this done is:

public void connect(IDocument document, boolean delayInitialise)
{
    super.connect(document, delayInitialise);
    printPartitions(document);
}

public void printPartitions(IDocument document)
{
    StringBuffer buffer = new StringBuffer();

    ITypedRegion[] partitions = computePartitioning(0, document.getLength());
    for (int i = 0; i < partitions.length; i++)
    {
        try
        {
            buffer.append("Partition type: " 
              + partitions[i].getType() 
              + ", offset: " + partitions[i].getOffset()
              + ", length: " + partitions[i].getLength());
            buffer.append("\n");
            buffer.append("Text:\n");
            buffer.append(document.get(partitions[i].getOffset(), 
             partitions[i].getLength()));
            buffer.append("\n---------------------------\n\n\n");
        }
        catch (BadLocationException e)
        {
            e.printStackTrace();
        }
    }
    System.out.print(buffer);
}

The printPartition() method is called once when the IPartitioner instance is connected to the document, which happens when you open a document in the editor.

Returning to our IDocumentProvider's createDocument() method, we see that the IDocumentPartitioner instance we create needs to be configured with an IPartitionTokenScanner instance, as well as a String array corresponding to the content types supported by the editor. This is how we set the partitioning scheme for the editor. Of course, this configuration needs to be backed by an underlying implementation – our plugin needs to implement functionality for document partitioning. We will learn how it does this when we discuss scanners, tokens and rules in the next section.

Scanners, Tokens and Rules

As the name suggests, the job of the IPartitionTokenScanner is to scan the document and find tokens corresponding to individual document partitions. A token is represented by an instance of IToken. You may be somewhat confused about what exactly a token is. A token does not represent a particular sequence of characters. Instead, it serves as an identifier for some characteristics that a sequence of text might embody. Some examples of what a token might represent include:

In keyword example, the token does not represent an instance of the keyword appearing in the text, but rather the keyword itself.

When partitioning a document, the IPartitionTokenScanner does its job by searching for tokens which correspond to partitions. We get some more clues to how it works by looking at the implementation of XMLPartitionScanner:

public class XMLPartitionScanner extends RuleBasedPartitionScanner
{
    public final static String XML_DEFAULT = "__xml_default";
    public final static String XML_COMMENT = "__xml_comment";
    public final static String XML_TAG = "__xml_tag";

    public XMLPartitionScanner()
    {

        IToken xmlComment = new Token(XML_COMMENT);
        IToken tag = new Token(XML_TAG);

        IPredicateRule[] rules = new IPredicateRule[2];

        rules[0] = new MultiLineRule("<!--", "-->", xmlComment);
        rules[1] = new TagRule(tag);

        setPredicateRules(rules);
    }
}

Each of the content types associated with the document editor is associated with a String contant. There are only three content types associated with the default XML editor – the default or unspecified content type, XML tags and XML comments. These are associated with the Strings __xml_default, __xml_comment and __xml_tag respectively. Notice that the model is rather simple – we don't have specific content types for XML text (textual content between element tags) or processing instructions. When the XMLPartitionScanner, each of the content types is associated with an instance of IToken. The IToken instance does very little – it simply serves as an identifier for the named content type.

Next, our IPartitionTokenScanner needs to be told how to recognise tokens when scanning the document. Notice that our XMLPartitionScanner extends RuleBasedPartitionScanner. As the name suggests, RuleBasedPartitionScanner contains functionality to scan a document for tokens using a configured set of rules. Partitioning according to rules is a very convenient way to partition a document which has a well defined structure. An XML document is a very good example.

In the default XML editor, the XMLPartitionScanner is configured to use only two rules, each corresponding to one of the editor's content types. The first, a MultiLineRule instance, is used to find XML comments. The second looks for XML tags using a customised rule implementation, TagRule.

Both the classes MultiLineRule and TagRule are implementations of IPredicateRule. The IPredicateRule has an important restriction which makes it suitable for partitioning by content type. When finding a match it is only allowed to return a single token. This token is called the success token. This restriction makes possible the one to one association between content type (e.g. __xml_comment), IPredicateRule implementation and IToken instance.

Partitioning using a rule-based scanner essentially works as follows:

The partitioning information stored in the document can be returned in full using the IDocumentPartitioner computePartitioning() method. Of course, the document partitioner needs a reference to the document, which will be available after the IDocumentPartitioner.connect(document) method has been called, which in the case of the XML example application takes place when the XMLDocumentProvider's createDocument() method is invoked.

When using a rule-based approach to document partitiioning, much of the time it is possible to use the Eclipse supplied IPredicateRule implementations, which include MultiLineRule, SingleLineRule and PatternRule, and others. The challenge is to decide how to configure the rules being used, and how to order them. For more demanding applications, it is also necessary to provide your own implementation(s) of IPredicateRule.

As an alternative to partitioning using rules, you could also create your own ITokenScanner implementation. The JDT in Eclipse 3.1 uses its own implementation of ITokenScanner, FastJavaPartitionScanner, to scan for partitions in Java class files. Obviously, there is a lot more work in creating a custom partition scanner than using the generic rule-based scanner, but the payoff may be worthwhile for some applications.

Text Viewers and Source Viewers

With our document suitably partitioned, we are now able to add some interesting features to our user interface. JFace Text uses the StyledText widget as the underlying control for presenting textual content. The ITextViewer interface provides an API abstraction for an IDocument instance to be used as the text model for the StyledText widget. This means that when using JFace Text the application code invariably does not need to interact with the StyledText widget.

ITextViewer also supports a range of features that you would not want to have to implement yourself, such as undo management. Read the JavaDocs for ITextViewer and see the TextViewer class's source for more details.

In order to further facilitate the task of creating editors for structured documents, JFace Text provides the ISourceViewer interface. ISourceViewer adds functionality to ITextViewer, such as visual annotations based on an annotation model, including the error markers we have described earlier, as well as range highlighting. We will return to both of these features later in the article.

As a developer of JFace Text applications, you need to understand the capabilities of text viewers and source viewers, and how they configured. However, it is unlikely that you will need to extend any of the provided ISourceViewer implementation classes in your application, at least not initially.

Configuring the SourceViewer

JFace provides a mechanism for configuration of the source viewer in the form of the SourceViewerConfiguration class. Your application will almost certainly need to extend this class to customise the behaviour of the source viewer. The default XML editor uses the XMLConfiguration, which of course extends SourceViewerConfiguration.

The application's SourceViewerConfiguration subclass plays an important role because it is used to add a range of features to your application, including text formatting, syntax highlighting, double click support, text hovering and content assistance. In fact, most of the value added features we discussed earlier in the article are introduced to the application through the SourceViewerConfiguration.

So powerful are the operations defined by the SourceViewerConfiguration, that your application code may not have to interface directly with the editor's source viewer. In the default XML editor, the only references in the plugin to ISourceViewer are in the SourceViewerConfiguration subclass itself!

Remember also that the SourceViewerConfiguration is added to the editor using the AbstractTextEditor.setSourceViewerConfiguration(SourceViewerConfiguration) method. In the case of XMLEditor, this method is called when the editor class is constructed, as we saw when introducing the text editor earlier in the article.

We won't describe in detail the methods of the SourceViewerConfiguration class at this point. Instead, we will introduce them when we discuss the relevant feature that the method adds to the application.

Part 3: Adding Features

The Enhanced Application

We mentioned earlier that the default generated XML editor provided by Eclipse did not include many of the features that you would expect from a sophisticated plugin. As part of writing this article, I have extended the basic XML application to add a number a number of new features. We'll call this the enhanced XML editor. The screenshot below shows what this enhanced editor looks like:

The enhanced XML editor in action

Looking at the document you will notice a few added features:

Experimenting with this application, you will find that this is certainly still not a fully-featured, commercial-grade XML editor. The scale of the task in producing such an application is way beyond the scope of this article. However, the modified editor serves as a useful guide to how the default editor can be enhanced with some useful features that would be present in real-world applications. Hopefully, also, the enhanced editor provides a some good examples of the value-added features which we have introduced, which will be discussed individually in more detail in the sections that follow.

Extending our Partitioning Model

In order to support all of the features we have just described, we need to extend our partitioning model. Specifically, we would like to create separate content types for the following:

The purpose of adding these content types is to exercise greater control over how regions of the document are managed in terms of formatting, syntax highlighting and content assistance.

A step-by-step example, lets consider what we would need to do to add a new content type to represent XML processing instructions.

Create a new a content type

We need to create a new content type which represents XML processing instructions. We can do this simply by adding a constant in the class XMLPartionScanner.

 
public final static String XML_PI = "__xml_pi";

Register the new content type with the document partitioner

The IDocumentPartitioner interface defines a method,
String[] getLegalContentTypes();
which returns the list of legal content types associated with the partitioner instance. According to the method's documentation "any result delivered by this partitioner may not contain a content type which would not be included in this method's result".

Our XMLPartitioner constructor provides an argument for populating this list. We simply need to add our new content type to the String[] array of legal content types, as shown below

protected IDocument createDocument(Object element) throws CoreException
{
    IDocument document = super.createDocument(element);
    if (document != null)
    {
        IDocumentPartitioner partitioner = new XMLPartitioner(
         new XMLPartitionScanner(), new String[]
        {
                XMLPartitionScanner.XML_START_TAG,
                XMLPartitionScanner.XML_PI,
                XMLPartitionScanner.XML_DOCTYPE,
                XMLPartitionScanner.XML_END_TAG,
                XMLPartitionScanner.XML_TEXT,
                XMLPartitionScanner.XML_CDATA,
                XMLPartitionScanner.XML_COMMENT
        });
        partitioner.connect(document);
        document.setDocumentPartitioner(partitioner);
    }
    return document;
}

Create a token and rule for identifying the content

We now need to make the IPartitionTokenScanner implementation aware of our new content type. Since we are using a RuleBasedPartitionScanner, we simply need to add a new token and rule pairing which can be used to identify processing instructions.

From the XML specification we know that processing instructions begin with the characters <? and end with the characters ?>, so we simply create a new token instance to represent the XML processing instruction content type, then use it to configure a new MultiLineRule instance which can be used to identify matching character sequences. This happens in the XMLPartitionScanner constructor.

public XMLPartitionScanner()
{

    IToken xmlComment = new Token(XML_COMMENT);
    IToken xmlPI = new Token(XML_PI);
    IToken startTag = new Token(XML_START_TAG);
    IToken endTag = new Token(XML_END_TAG);
    IToken docType = new Token(XML_DOCTYPE);
    IToken text = new Token(XML_TEXT);

    IPredicateRule[] rules = new IPredicateRule[7];

    rules[0] = new NonMatchingRule();
    rules[1] = new MultiLineRule("<!--", "-->", xmlComment);
    rules[2] = new MultiLineRule("<?", "?>", xmlPI);
    rules[3] = new MultiLineRule("</", ">", endTag);
    rules[4] = new StartTagRule(startTag);
    rules[5] = new MultiLineRule("<!DOCTYPE", ">", docType);
    rules[6] = new XMLTextPredicateRule(text);

    setPredicateRules(rules);
}

Of course this is a very simple case because we are simply reusing one of the standard JFace Text classes. For more complex requirements, it may be necessary to provide a custom implementation of IPredicateRule, or at least customise one of the existing classes.

Make the SourceViewerConfiguration aware of the new content type

The whole purpose of adding the new content type is to take advantage of functionality which is introduced to the application through the SourceViewerConfiguration subclass. SourceViewerConfiguration contains a method

public String[] getConfiguredContentTypes(ISourceViewer sourceViewer);

which includes the list of content types for which functionality is introduced. According to the source documentation, this list "tells the caller which content types must be configured for the given source viewer". We make our source viewer configuration aware of our new content type by adding it to the array of returned configured content types. This takes place in the overriding getConfiguredContentTypes() method in XMLConfiguration:

public String[] getConfiguredContentTypes(ISourceViewer sourceViewer)
{
    return new String[]
    {
            IDocument.DEFAULT_CONTENT_TYPE,
            XMLPartitionScanner.XML_COMMENT,
            XMLPartitionScanner.XML_PI,
            XMLPartitionScanner.XML_DOCTYPE,
            XMLPartitionScanner.XML_START_TAG,
            XMLPartitionScanner.XML_END_TAG,
            XMLPartitionScanner.XML_TEXT
    };
}

Add functionality for the new content type

The final step is to add functionality for the new content type, mostly be modifying methods in the application's SourceViewerConfiguration subclass. For example, IPresentationDamager and IPresentationRepairer instances need to be specified for the new content type to allow syntax highlighting, and an IFormattingStrategy instance may need to be specified for content-type specific formatting. Don't worry at this point if you don't know what these interfaces represent. We'll see examples of these and others in the following sections, when we discuss different types of features which can be added to the editor.

Syntax Highlighting

If you understand document partitioning then you are in a very good position to understand syntax highlighting in a JFace Text application. Essentially, syntax highlighting involves dividing a partition into tokens, each of which has its own text display attributes. The standard mechanism for doing this also involves the use of the already familiar tokens, scanners and rules. We'll explain in this section how these artifacts are applied to allow syntax highlighting.

The SourceViewerConfiguration method which deals with syntax highlighting is the method IPresentationReconciler getPresentationReconciler(ISourceViewer sourceViewer). The snippet below shows the implementation of this method in the default XML editor:

public IPresentationReconciler getPresentationReconciler(ISourceViewer sourceViewer)
{
    PresentationReconciler reconciler = new PresentationReconciler();

    DefaultDamagerRepairer dr = new DefaultDamagerRepairer(getXMLTagScanner());
    reconciler.setDamager(dr, XMLPartitionScanner.XML_TAG);
    reconciler.setRepairer(dr, XMLPartitionScanner.XML_TAG);

    dr = new DefaultDamagerRepairer(getXMLScanner());
    reconciler.setDamager(dr, IDocument.DEFAULT_CONTENT_TYPE);
    reconciler.setRepairer(dr, IDocument.DEFAULT_CONTENT_TYPE);
    
    ...

    return reconciler;
}

The IPresentationReconciler is responsible for tracking changes to the underlying IDocument. For each content (or partition) type, it holds a reference to an IPresentationDamager and IPresentationRepairer instance. Whenever a change is made to the document, a notification is sent to the IPresentationDamager for each content type in the affected area of the document. The IPresentationDamager in turn returns an IRegion indicating the area of the document which needs to be rebuilt as a result of the change. This information is then passed on to the IPresentationRepairer, which is responsible for reapplying the textual presentation for the affected region.

This all probably sounds quite complicated. Fortunately you probably won't ever need to implement any of these classes yourself. JFace Text provides a DefaultDamagerRepairer which does the job both of the IPresentationDamager and the IPresentationRepairer. It also provides a PresentationReconciler class which performs the co-ordination role we described earlier.

So what do we need to do? Of course, we need some way of letting the framework know how our application document is structured, on a per-content type basis. We also need to let the framework know what textual styles to apply for each of the elements in this structure.

The DefaultDamagerRepairer expects an ITokenScanner as a parameter in each of its constructors. The ITokenScanner is very similar to the partition token scanner we met earlier when discussing document partitioning. In both cases, the token is represented by an IToken instance. The key difference is in the granularity of the tokens. The IToken instance returned during partition scanning represents a region of the document belonging a particular content type. The IToken returned during syntax highlighting represents a sequence of characters which share a common textual format.

We can see this most easily when we look at the implementation of XMLTagScanner, which the DefaultDamagerRepairer uses for the XML_TAG content type. This of course represents the textual content between the XML < and > characters in an element tag.

public class XMLTagScanner extends RuleBasedScanner
{
    public XMLTagScanner(ColorManager manager)
    {
        Color color = manager.getColor(IXMLColorConstants.STRING);
        TextAttribute textAttribute = new TextAttribute(color);
        IToken string = new Token(textAttribute);
    
        IRule[] rules = new IRule[3];
    
        // Add rule for double quotes
        rules[0] = new SingleLineRule("\"", "\"", string, '\\');
        // Add a rule for single quotes
        rules[1] = new SingleLineRule("'", "'", string, '\\');
        // Add generic whitespace rule.
        rules[2] = new WhitespaceRule(new XMLWhitespaceDetector());
    
        setRules(rules);
    }
}

XMLTagScanner extends RuleBasedScanner, so it uses kind of rule processing mechanism for identifying tokens as we saw with the RuleBasedPartitionScanner. Viewing the edited document we can see a light green is used for the attribute text. This visual representation is created using the code

new TextAttribute(manager.getColor(IXMLColorConstants.STRING))

We also have two rules for returning this token: one which uses matching double quote characters, and another which uses matching single quote characters. There is also a rule to match white space.

You may be wondering why the rest of the XML tag content appears blue and not the black of the default text editor. This happens because the scanner can be configured to return a default token if none of the configured rules for the scanner find a match. We see this is the getXMLTagScanner() implementation in XMLConfiguration:

protected XMLTagScanner getXMLTagScanner()
{
    if (tagScanner == null)
    {
        tagScanner = new XMLTagScanner(colorManager);
        Color color = colorManager.getColor(IXMLColorConstants.TAG);
        TextAttribute textAttribute = new TextAttribute(color);
        Token token = new Token(textAttribute);
        tagScanner.setDefaultReturnToken(token);
    }
    return tagScanner;
}

This default return token is configured with a text attribute which represents the blue colour we see within the XML tags.

With our control over partitions through our document partitioner, and our ability to identify each String of text as an occurence of a particular token, we can gain complete control over the syntax highlighting of our structured document.

Content Formatting

Formatting is the ability to set the indentation and use of white space in the document to make the document more neatly structured and readable. Adding content formatting to your JFace Text application is a two step process:

Because content formatting rules can be applied per partition, we again see the importance of understanding and correctly applying document partitioning. A simple of example of a formatting strategy is TextFormattingStrategy, which we use to format character text nested within XML elements.

public class TextFormattingStrategy extends DefaultFormattingStrategy
{
    private static final String lineSeparator = System.getProperty("line.separator");

    public String format(String content, 
        boolean isLineStart, 
        String indentation, 
        int[] positions)
    {
        if (indentation.length() == 0)
            return content;
        return lineSeparator + content.trim() + lineSeparator + indentation;
    }
}

Here we subclass DefaultFormattingStrategy, which is essentially a no-op formatting implementation. In most cases, you'll need to override format(...). In this case, we simply trim the textual content, then surround it with line separators. For a real application, you'll probably want to use a preference page to allow the user to select formatting rules.

A more substantive implementation of IFormattingStrategy is found in the example application in the class XMLFormattingStrategy, which is used to format the XML elements. We won't go into details here on how this is done – interested readers can consult the example source code. It is worth saying that creating a good implementation can be quite tricky, because formatting needs to behave intelligently regardless of what portion of a document you select. It will probably take quite a bit of trial and error, or experience, to get it right.

Applying formatting rules is fairly straightforward. You simply need to provide an implementation of SourceViewerConfiguration.getContentFormatter(ISourceViewer sourceViewer). Below, we can see our implementation in XMLConfiguration:

public IContentFormatter getContentFormatter(ISourceViewer sourceViewer)
{
    ContentFormatter formatter = new ContentFormatter();
    XMLFormattingStrategy formattingStrategy = new XMLFormattingStrategy();
    DefaultFormattingStrategy defaultStrategy = new DefaultFormattingStrategy();
    TextFormattingStrategy textStrategy = new TextFormattingStrategy();
    DocTypeFormattingStrategy doctypeStrategy = new DocTypeFormattingStrategy();
    PIFormattingStrategy piStrategy = new PIFormattingStrategy();
    formatter.setFormattingStrategy(defaultStrategy, IDocument.DEFAULT_CONTENT_TYPE);
    formatter.setFormattingStrategy(textStrategy, XMLPartitionScanner.XML_TEXT);
    formatter.setFormattingStrategy(doctypeStrategy, XMLPartitionScanner.XML_DOCTYPE);
    formatter.setFormattingStrategy(piStrategy, XMLPartitionScanner.XML_PI);
    formatter.setFormattingStrategy(textStrategy, XMLPartitionScanner.XML_CDATA);
    formatter.setFormattingStrategy(formattingStrategy, XMLPartitionScanner.XML_START_TAG);
    formatter.setFormattingStrategy(formattingStrategy, XMLPartitionScanner.XML_END_TAG);

    return formatter;
}

We start by creating a ContentFormatter instance. Then for each of the partition types to which we want to apply specific formatting rules, we use the ContentFormatter.setFormattingStrategy() method to associate an IFormattingStrategy instance with the appropriate partition type.

Content Outline

Providing a content outline for a structured document editor is always a good idea for two reasons; it gives the user a simplified overview of the document's contents, and provides a means for navigating more easily within the document.

In our example application, we have created a content outline view, which is shown below:

The enhanced XML editor with content outline view

Lets consider some of the functionality that we need to provide when creating a functional content outline view:

Before any of this can be done we need to lay the necessary foundations by creating an association between our content outline implementation and the editor itself. Eclipse provides basic support for the content outline view in the form of the interface IContentOutlinePage. The easiest way to provide an implementation is to extend ContentOutlinePage, as we do with XMLContentOutlinePage.

The content outline page is exposed to the editor through its implementation of the method getAdapter(), an extensibility mechanism used widely in Eclipse. The Eclipse runtime requests a content outline page implementation by calling getAdapter() with an IContentOutlinePage Class instance as an argument. When this occurs, our application responds by instantiating and returning our implementation, as shown in the following source code from XMLEditor:

public Object getAdapter(Class required)
{
    if (IContentOutlinePage.class.equals(required))
    {
        if (outlinePage == null)
        {
            outlinePage = new XMLContentOutlinePage(this);
            ...
        }
        return outlinePage;
    }
    return super.getAdapter(required);
}

The job of our XMLContentOutlinePage is to instantiate the content and label providers which expose the model to the underlying framework, and to propagate updates. We'll learn more about these features in the next sections.

Intepreting the Document Structure

The intelligence for interpreting the structure of the document is in our example application is in our class OutlineContentHandler, which is nothing other than an implementation of the SAX ContentHandler interface. OutlineContentHandler's job is to build a tree representation of the XML document, in which information necessary for the content outline is captured. This information includes the names of elements, the names and values of attributes, and the position (line and column number) of elements in the tree, and is encapsulated in our classes XMLTree, XMLElement and XMLAttribute. Interested readers can view the source code of OutlineContentHandler to see how this is done for the example application.

Of course, we need a mechanism for the content of the XMLTree to be exposed to the user interface. The default content outline page, from which XMLContentOutlinePage inherits, wraps a JFace ITreeViewer instance. We take advantage of its presence by providing ITreeContentProvider and ITreeLabelProvider implementations. In our application, the classes which perform this role are OutlineContentProvider and OutlineLabelProvider. As readers familiar with ITreeViewer will know, ITreeContentProvider is responsible for building the structure of the tree's visual representation, while ITreeLabelProvider is responsible for providing text and image icons for individual tree nodes.

The rest of the OutlineContentProvider is responsible for populating the document model in response to to changes in the input, as shown below:

public void inputChanged(Viewer viewer, Object oldInput, Object newInput)
{

    ...

    input = (IEditorInput) newInput;

    if (newInput != null)
    {
        IDocument document = documentProvider.getDocument(newInput);
        if (document != null)
        {
            ...
            
            XMLElement rootElement = parseRootElement(document);
            if (rootElement != null)
            {
                root = rootElement;
            }
        }
    }
}

The OutlineContentHandler is called into action to parse the XML document and provide a XMLTree model object which can be used by the OutlineContentProvider to provide a representation of the document's structure.

Updating the outline view

There are essentially two strategies you can use for updating the content outline view. One is to update the view each time the edited document is modified (usually with a lag so that updating the content outline does not unnecessarily drain system resources). The second is to update the outline view each time the edited document is saved. For simplicity, we have taken the latter approach in our example application – you will only see the outline view updating when you save the document.

We mentioned that the content outline's structure is update each time OutlineContentProvider.inputChanged() is called. We can complete our picture by considering the sequences which lead to the invocation of this method. There are two situations when this sequence is initiated; when the document is loaded, and when the document is saved.

We noted earlier that when an editor for a document is loaded, Eclipse looks for a content outline page for the editor by calling the editor's getAdapter() method, passing an IContentOutlinePage Class literal as the argument. As the code segment below shows, our XMLContentOutlinePage instance is created at this point.

public Object getAdapter(Class required)
{
    if (IContentOutlinePage.class.equals(required))
    {
        if (outlinePage == null)
        {
            outlinePage = new XMLContentOutlinePage(this);
            if (getEditorInput() != null)
                outlinePage.setInput(getEditorInput());
        }
        return outlinePage;
    }
    return super.getAdapter(required);
}

In addition, we use this opportunity to call the outline page's setInput() method, passing in the editor's input. This results in a call to XMLContentOutlinePage.update(), which is shown below.

public void update()
{
    
    TreeViewer viewer = getTreeViewer();
    if (viewer != null)
    {
        Control control = viewer.getControl();
        if (control != null && !control.isDisposed())
        {
            control.setRedraw(false);
            viewer.setInput(input);
            viewer.expandAll();
            control.setRedraw(true);
        }
    }
}

The reference to IEditorInput is used to call setInput() on the enclosed ITreeViewer, which in turn calls inputChanged() in our OutlineContentProvider. Finally, the underlying GUI control is marked as eligible for redrawing.

Updating the content outline on saving is as simple as overriding the editorSaved() from AbstractTextEditor, and adding the following code to call XMLContentOutlinePage.update():

 
if (outlinePage != null)
outlinePage.update();

Navigating using outline selections

An necessary feature of our outline view is the ability to navigate the main document in the editor pane using selections in the outline view. We discussed earlier how the outline view allows you to visualise structure of the document, and how we had created an XMLTree instance represent this structure. To support navigation via selection, we need to add two capabilities. First, we need to maintain the information on how nodes in this XMLTree structure are mapped to locations within the editor. Second, we need to monitor selections in the outline view, and use the stored location information to select text within the editor.

Recording line and column information is fairly straightforward. We simply add a DocumentLocator to our SAX content handler implementation used to build the XMLTree structure.

private XMLElement parseRootElement(IDocument document)
{
    String text = document.get();
    XMLParser xmlParser = new XMLParser();
    OutlineContentHandler contentHandler = new OutlineContentHandler();
    contentHandler.setDocument(document);
    ...
    contentHandler.setDocumentLocator(new LocatorImpl());
    xmlParser.setContentHandler(contentHandler);
    xmlParser.doParse(text);
    XMLElement root = contentHandler.getRootElement();
    return root;
}

To apply changes we need to override ContentOutlinePage's default implementation of selectionChanged(), as shown in the implementation from XMLContentOutlinePage:

public void selectionChanged(SelectionChangedEvent event)
{
    super.selectionChanged(event);
    // find out which item in tree viewer we have selected, and set
    // highlight range accordingly

    ISelection selection = event.getSelection();
    if (selection.isEmpty())
        editor.resetHighlightRange();
    else
    {
        IStructuredSelection sel = (IStructuredSelection) selection;
        XMLElement element = (XMLElement) sel.getFirstElement();

        int start = element.getPosition().getOffset();
        int length = element.getPosition().getLength();
        try
        {
            editor.setHighlightRange(start, length, true);
        }
        catch (IllegalArgumentException x)
        {
            editor.resetHighlightRange();
        }
    }
}

We get a reference to the selected node in our XMLTree object model by calling
XMLElement element = (XMLElement) ((IStructuredSelection) selection).getFirstElement().
We then extract the stored location information from this object, and use this to update the highlighted range of the editor.

Error Marking

One very useful feature of an editor is the ability to identify and display errors. For example, an XML editor needs to be able to validate the XML document using the DTD or schema, and display any validation errors arising. Our example application has support for simple DTD valdidation added. The screenshot below shows the error marking at work.

Support for error marking in the enhanced XML editor

There are some close similarities in the way error marking is implemented to the way content outline is implemented. Firstly, identifying errors also requires parseing of the XML document being edited. This time, we use a SAX ErrorHandler to collect errors and their locations. Secondly, the validation and error marking occurs at roughly the same time as generation of the content outline: when the document is loaded, and each time the document is saved.

The validation and error marking is initiated in the validateAndMark() method in XMLEditor, shown below:

protected void validateAndMark()
{

    IDocument document = getInputDocument();
    MarkingErrorHandler markingErrorHandler = 
     new MarkingErrorHandler(getInputFile(), document);
    markingErrorHandler.removeExistingMarkers();

    XMLParser parser = new XMLParser();
    parser.setErrorHandler(markingErrorHandler);

    String text = document.get();
    parser.doParse(text);

}

A MarkingErrorHandler instance is created and passed a copy of the IFile instance representing the input file and a copy of the IDocument being edited. It needs the IFile to performing the marking (the Eclipse Marker API uses references to the underlying resource object). The IDocument reference is required to determine the correct start and end character locations of the markers to be inserted.

Any existing error markers are cleared before the document is parsed. The workhorse method in MarkingErrorHandler, called each time an error is encountered during document parsing, is handleError(), shown below:

protected void handleError(SAXParseException e, boolean isFatal)
{

    int lineNumber = e.getLineNumber();
    int columnNumber = e.getColumnNumber();

    Map map = new HashMap();
    MarkerUtilities.setLineNumber(map, lineNumber);
    MarkerUtilities.setMessage(map, e.getMessage());
    map.put(IMarker.MESSAGE, e.getMessage());
    map.put(IMarker.LOCATION, file.getFullPath().toString());

    Integer charStart = getCharStart(lineNumber, columnNumber);
    if (charStart != null)
        map.put(IMarker.CHAR_START, charStart);

    Integer charEnd = getCharEnd(lineNumber, columnNumber);
    if (charEnd != null)
        map.put(IMarker.CHAR_END, charEnd);

    map.put(IMarker.SEVERITY, new Integer(IMarker.SEVERITY_ERROR));

    try
    {
        MarkerUtilities.createMarker(file, map, ERROR_MARKER_ID);
    }
    catch (CoreException ee)
    {
        ee.printStackTrace();
    }
}

The implementation of this method is fairly straightforward. The error message, line number and column number are obtained, which are then used to create an error marker using the Eclipse marker API. The editor is of course relying on the quality of the error messages and location tracking of the underlying XML parser. Our example application uses Xerces, which luckily does a fairly good job for both of these.

Content Assistance

The final feature we'll cover in this article is content assist. This is the mechanism which allows your application to suggest text completions which can help users to edit their documents. For a commercial grade XML editor, you would certainly expect the content assist editor to suggest completions which are relevant given your cursor's position in the document and the document's DTD.

For our example application, we have extended the default XML editor by adding a simple content assist mechanism. The screenshot below shows our content assist support at work:

Support for content assist in the enhanced XML editor

Creating an intelligent content assist mechanism is a challenging task. In this context, to be intelligent, the content assist mechanism needs to have an understanding of the structure of the document, as well as the location within that structure of the text currently being edited.

As with many of the other features discusssed in this article, content assist is configured by overriding one of the methods in SourceViewerConfiguration – in this case getContentAssistant(). Our implementation is shown below:

public IContentAssistant getContentAssistant(ISourceViewer sourceViewer)
{

    ContentAssistant assistant = new ContentAssistant();

    IContentAssistProcessor tagContentAssistProcessor 
        = new TagContentAssistProcessor(getXMLTagScanner());
    assistant.setContentAssistProcessor(tagContentAssistProcessor,
            XMLPartitionScanner.XML_START_TAG);
    assistant.enableAutoActivation(true);
    assistant.setAutoActivationDelay(500);
    assistant.setProposalPopupOrientation(IContentAssistant.CONTEXT_INFO_BELOW);
    assistant.setContextInformationPopupOrientation(IContentAssistant.CONTEXT_INFO_BELOW);
    return assistant;

}

There is not much to explain here. We need to instantiate an IContentAssistant instance and configure its user interface options. Then we need to attach to the ContentAssistant an implementation of IContentAssistProcessor for each document partition for which we want this feature enabled. Notice here that we are only supporting content assistance within element declarations. The fact that content assistance works on a per-partition basis makes it easy to support this restriction. Of course, our document partitioning structure also needs to be designed with the requirements of this feature in mind.

The power in JFace Text is in providing the necessary UI support for content assist, allowing us to concentrate fully on application-specific tasks. While it is possible to subclass or even replace the provided ContentAssistant, this will usually not be necessary for most applications, and could be a relatively complex task when undertaken.

Instead, the starting point for the intelligence that we add to our content assist mechanism is typically the IContentAssistProcessor implementation. IContentAssistProcessor defines a number of methods, of which the most interesting for us is:
ICompletionProposal[] computeCompletionProposals(ITextViewer viewer, int offset);
since this method is the workhorse for providing suggestions for code completions. An abridged and simplified version of our implementation is shown below:

public ICompletionProposal[] computeCompletionProposals(ITextViewer viewer, int offset)
{

    IDocument document = viewer.getDocument();
    boolean isAttribute = isAttribute(offset, document);

    TextInfo currentText = currentText(document, offset);

    if (!isAttribute)
    {

        List allElements = dtdTree.getAllElements();

        ICompletionProposal[] result = new ICompletionProposal[allElements.size()];
        int i = 0;
        for (Iterator iter = allElements.iterator(); iter.hasNext();)
        {
            XMLElement element = (XMLElement) iter.next();
            String name = element.getName();

            String text = "" + name + ">" + "</" + name + ">";
            }

            result[i++] = new CompletionProposal(text, 
             currentText.documentOffset, 
             currentText.text.length(), 
             text.length());

        }
        return result;

    }
    else
    {

        List allAttributes = dtdTree.getAllAttributes();

        ICompletionProposal[] result = new ICompletionProposal[allAttributes.size()];
        int i = 0;
        for (Iterator iter = allAttributes.iterator(); iter.hasNext();)
        {
            String name = (String) iter.next();

            String text = name + "= \"\" ";
            
            result[i++] = new CompletionProposal(text, 
             currentText.documentOffset, 
             currentText.text.length(), 
             text.length());
        }
        return result;
    }

}

The logic is embarrassingly simple. The completion processor inspects the current element text to determine whether an element has already been entered. If so, it provides a list of completions corresponding with all the attributes known to the document. If not, it provides completions corresponding to all the known elements.

Of course, this is not a particularly intelligent mechanism A more advanced implementation would possibly scan the document, and identify where in the document's structure the current completion is being requested. It would then use the DTD to calculate completions which would be result in valid XML. In order to accomplish this, the editor would need to build an understanding of the structure implied by the DTD, by creating a DTD parser of sorts.

Nevertheless, the example hopefully gives a flavour for the tasks involved when creating a content assist processor.

Summary

Building a sophisticated text editor is one of the requirements that few Eclipse plug-ins can avoid. Fundamental to this task is JFace Text, one of the most powerful and important Eclipse APIs, but also one of the more complex APIs.

We've taken as our starting point the default XML editor which can be autogenerated using the Eclipse PDE wizards. We've then extended this to demonstrate how you might build some of the advanced features that a sophisticated text editor would required, including:

Many of these features require significant understanding of JFace Text. The article has discussed some of the key concepts, interfaces and classes at work, and how they fit together to allow the creation of a feature-rich Eclipse text editor.


Author's Note (October 9, 2008): Charlie Kelly has just provided me with the bundle manifest which he used to get the article's code working on Eclipse 3.4. Haven't tried it myself, but here it is:

Manifest-Version: 1.0
Bundle-ManifestVersion: 2
Bundle-Name: Ui Plug-in
Bundle-SymbolicName: com.matrixware.xml.ui
Bundle-Version: 1.0.0
Bundle-Activator: com.matrixware.xml.ui.XmlUiActivator
Bundle-Vendor: Matrixware.com
Require-Bundle: org.eclipse.core.runtime,
org.eclipse.jface.text;bundle-version="3.4.0",
org.eclipse.core.resources;bundle-version="3.4.0",
org.eclipse.ui.views;bundle-version="3.3.0",
org.eclipse.ui.editors;bundle-version="3.4.0",
org.eclipse.ui;bundle-version="3.4.0",
org.apache.xerces;bundle-version="2.9.0"
Bundle-RequiredExecutionEnvironment: J2SE-1.5
Bundle-ActivationPolicy: lazy
Import-Package: org.eclipse.ui
Export-Package: article,
article.editor,
article.editor.contentassist,
article.editor.format,
article.editor.markers,
article.editor.outline,
article.editor.rules,
article.editor.scanners,
article.editor.xml 



This is a working document. If there are any errors, or if you disagree with anything which I have said, or have any suggestions for improvements please email me at philzoio@realsolve.co.uk.