Schema Navigator

The W3C Schema XSLT 2 has been rendered here. You can see that the layout algorithm has organised the graph into two clusters. It happens that the each clusters correspond to a distinct namespace, which is indicated by the colour. The mauve elements belong to XSLT2 and the blue ones to W3C Schema.

You can see the namespace partitions the model very neatly. It is obvious that there are virtually no interrelationships between elements of the two domains. One that describes how a stylesheet works and the other how to describe an XML language. As XSLT2 may be schema aware it has the means of linkage. The point of contact as you can see in the centre of the diagram. It's the element import-schema, which can reference or physically contained the a schema.

Although the layout engine has no sense of the schema's type system or internal organisation it is fun to play around with different algorithms and see what is thrown up. After a bit of shuffling about I came across this. It is a very symmetrical arrangement which confines the 21 or so elements that derive from XSLT's 'sequence constructor' to a very small section of the graph (highlighted in green). It shows that the cluster has an almost identical relationship with the following elements that are arranged about it; apply-imports, apply-templates, sequence and call-template.

Looking closely at the XSLT element-constructor and its immediate relations pattern is discernible. Colour highlight indicates a relation; green self, blue child, red parent, and mauve parent-or-child. There are a large number of interdependencies and circularity confined to this part of the graph. You can compose XSLT instructions in may ways. So may be this is an indication of flexibility?

I don't know anything about this language but with the tool I can see it has about 180 elements and nearly 7000 links. That's almost that is almost as many links as it could have so it must be very circular. Maybe that indicates that it too is composable and flexible?.

Looking at relations it seems that the structure is fairly straightforward, there is a core of a elements surrounded by a sort of corona of others. All of the elements in the core can be parent to one and another, while the surrounding ones can only be children.

Here are a few more examples of languages which have this controlled composablility aspect. The first one is the XSL Formatting Object specification. The next picture came from the XHTML Schema. Although I think this digram more complicated than it needs to be. I have seen similar structures where this kind of complexity can be hidden with the careful use of substitution groups. For example there are a lot of inline elements that can occur in the same context and have similar content models.

XSL Formatting Objects and HTML seem to describe very similar structures. Both have tables and lists, and section like elements that can be composed into hierarchies, and in-line or spanning elements.

This diagram has a very different kind of feel. There are a lot of elements (500) and relatively few relations (700). So I guess, notwithstanding the number of elements, the number of possible structures is relatively constrained. So taking two points in the graph you would find relatively few paths connecting them as opposed to XHTML where there may be infinitely many.

Initially the pop-up dialogue will show some defaults for the namespace prefixes. Please override these as you see fit but they should be unique. If you click on a colour another dialogue comes up allowing you to choose from a palette. Subdued colours are better otherwise the final diagram may will look crowded. The application will remember your preferences for the future, but may still override them at start-up. Select okay.

To explore the document navigator needs a starting point. To the left along with their (new) prefix mappings will be a list of top-level elements. Choose the one which usually acts as a document or root element. At the moment the application is still experimental please leave abstract elements checked and select expand substitutions. Ignore coalesce names. Again select okay.

If all went well you should be looking at the diagram like the one above. To render a model like this should only take a second or two at most. For example XHTML takes a little under a second on a dual core 2.8 MacBook Pro and maths are now just under two. For very dense graphs some layouts take a little longer to settle than others. Alternative layouts may be selected from the menu or the operating system dependent shortcut keys.

One you have a graph you can interact with it. There are three operational tabbed panels to the right of the screen. By default the name choosers selected. If you pick an element in the main diagram with your mouse it will highlight along with any other elements that share its name, and the panel list will be updated. Likewise selecting an element from the list will highlight the corresponding node(s) in the graph. In the above example, which is displaying the W3C Schema, the element xs:element has been clicked, and oddly enough it seems there are three... Indeed there are. You can check yourself.

You may want to explore the type system behind the model. On the right-hand side select the model tab. Now you should see a list of global types have been declared in the schema. You can filter that list so that only complex or simple types of visible. You can also choose whether the application performs a deep compere when it inspects each element. This is in respect of the scheme is type system. So if type B inherits from a selecting type a will also highlight elements of type B. Again you can choose from either the graph or the panel to select the type

You can also examine relationships between elements. Select the relations type. The panel on the right has a number of components. First there was a drop-down box which selects the element to examine. Next is children and parents are listed. Items therein are selectable. The bottom half of the panel displays information about the element. Its main namespace content model, or rather its type. The number of inwardly directed edges and outward. Its attributes are also listed. As usual selecting an element highlighted in green. Its parents in red and its children the colour blue. Elements which can be both parent and child are highlighted in purple. When you travel from related elements your path is traced and highlighted in white when more related elements are selected for their relationships the elements on the current path are also highlighted. Selecting an unrelated element clears path.