Background ==================== This demo application presumes that you have an interest in XML technologies and might want to leverage them for a fast, but rich and dynamic, website. In this demo application, we build up, bit-by-bit, the functionality. Thus, you don't have to know squatola about XML to follow along. In fact, the real purpose of this demo app is to teach its author how to use the stack (repoze.bfg, Paster, eggs, etc.) In summary: - Represent a hierarchical site as hierarchical XML - Inject ``repoze.bfg`` semantics into elements using :term:`lxml` - Support flexible-but-fast rendering with XSLT .. warning:: If you dislike XML and related technologies such as XPath and XSLT, you'll thoroughly detest this sample application. Just to be stupendously clear, ``repoze.bfg`` is in no way dependent on XML. On the other hand, ``repoze.bfg`` happens to make XML publishing kinda fun. What It Does ------------------- Imagine you have a website that looks like this:: / folder1/ doc1 doc2 image1 folder2/ doc2 Meaning, a heterogenous, nested folder structure, just like your hard drive. (Unless you're one of those folks that uses your Windows Desktop as a flat filing system.) How might I get that information into a website? Using ``repoze.bfg``, of course. More specifically, with an XML file that models that hierarchy: .. literalinclude:: step00/simplemodel.xml :language: xml How It Works ------------------- To coerce ``repoze.bfg`` into publishing this model, I just need to sprinkle in some Python behavior. For example, ``repoze.bfg`` uses ``__getitem__`` to traverse the model. I need my XML data to support this method. Moreover, I want some specific behavior: run an XPath express on the node to get the child with the ``@name`` attribute matching the URL hop. Fortunately :term:`lxml` makes this easy. I can inject my nodes with a class that I write, thus providing my own ``__getitem__`` behavior. That class can also assert that my XML nodes provide an interface. The interface then lets me glue back into the standard ``repoze.bfg`` machinery, such as associating views and permissions into the model. Neato torpedo. And stinking fast. Next up, I need to provide views for the elements in the model. I could, for example, use ZPT and manipulate the XML data using Python expressions against the :term:`lxml` API. Or, I could use XSLT. For the latter, I could register a different XSLT for every "view" on every interface. Or, I could write one big XSLT, and let its template matching machinery decide who to render in a certain context. And finally, I could pass in just a single node and render it, or pass in the entire tree with a parameter identifying the context node. In the course of this writeup, we'll build ``repoze.lxmlgraph`` step-by-step, starting with no XML. Each of those decisions will be analyzed an implemented. At the end, you'll see both the resulting demo application, plus the thought process that went along with it. What It Might Do -------------------- This demo application has the potential to show some other interesting investigations: #. **Authorization**. By hooking up support for an ``__acl__`` property, I can store ACL information on a single node, on an ancestor, on the ```` root, on the Python class, or any combination thereof. Additionally, I can wire up the ``__parent__`` attribute as a property that makes an :term:`lxml` ``node.getparent()`` call. #. **Multiple views**. Instead of just having a single default view on a node, I can allow other view names, all pointing at the same view function and XSLT. I simple grab that name and pass it in as a paramter to the XSLT, which will run a different rule for rendering. Adding a view would no longer required editing ZCML and adding a function. #. **Forms**. To edit data in the model, I need to render a form, then handle post data on the way back in. For the former, it's *really* easy in XSLT to make a very powerful, flexible, and extensisible form rendering system. For the latter, I'll have to learn more about POST handlers in ``repoze.bfg``.