Feature

When building applications, from content management to product catalogs, your project will typically be impacted by the method you choose for categorization. Designing code that is flexible enough to allow for ease of reuse, but simple enough to allow for ease of implementation, can often tie a developer in knots. The nested set model can help you fulfill both those goals.

This article will cover a simple implementation of this model, and draws extensively from two sources, Joe Celko's SQL for Smarties and Benjamin Elmore's Dynamic Publishing with ColdFusion MX. Both are recommended reading for a more in-depth look at the nested set model.

What Is the Nested Set Model?
In order to better understand the advantages of the nested set model in representing category trees within relational databases, we will first look at an alternate, commonly used method, the adjacency list model. The standard adjacency list model can be comprised of simply a categoryID, category, and parent field (see Figure 1).

Using the adjacency list model, a root node - a category that has no parent - is signified by a null value in the parent field (from here on out, node and category will often be used interchangeably). A leaf node is a category that has no children and can be determined through the use of a simple query. Determining the immediate parent or children of a given category also requires only a simple query. However, any function that would require traversing the tree beyond either the immediate parent or children of a given category would require complex recursive queries that are difficult to code, particularly when the depth of the tree is unknown. Similar problems occur when you wish to create functions for removing a sub tree. As adjacency list models are quite common, most experienced developers are well aware of their many deficiencies.

Rather than depend upon explicitly defined parents, the nested set model uses a set of left and right position integers to establish any given category's position within the tree structure. The left and right positions establish a range within which the left and right positions of any children will fall. The range between the left and right positions can easily be expanded to accommodate growth of the tree, and can be contracted just as easily when nodes are removed. In addition, determining the parents or children of a given node within a tree or even determining a given node's depth within the tree is a simple process regardless of the total depth of the tree.

In order to visualize this concept, let's take a root node (i.e., a node that has no parent) named "Macromedia," which is assigned the left and right positions of 1 and 20 respectively. A child of Macromedia called "ColdFusion" could have a left position of 2 and a right position of 9. Furthermore, a child of ColdFusion called "MX" could have a left position of 7 and a right position of 8, in which case it would be automatically recognizable as a leaf node given that the difference between the left and right positions is 1. See Figure 2 for a visual representation of a nested set model.

Building the CFC
Before I begin delving into the code of our component, I would first like to cover some assumptions I used in building it. Although the code is based upon examples from Dynamic Publishing with ColdFusion MX and SQL for Smarties, it has been modified in part to allow for compatibility with an Access database and also so that all public methods of the component return queries. While there are certainly benefits to using a more object-oriented approach within your components, and this component could be easily modified to meet that goal, I am choosing to return queries as I believe this makes the component easier to understand and implement for beginners and advanced users alike.

Before using this component, you first must create a table named "categories" in your database. If you are using Access, you would structure it as follows (see Figure 3):

1. CategoryID is your primary key and has a data type of text with a field size of 40.
2. Category has a data type of text with a field size of 255.
3. Lpos has a data type of number and a field size of long integer.
4. Rpos has a data type of number and a field size of long integer.

The first method of our component is called new (see Listing 1: categories.cfc)). This method is designed to simply create an empty query containing the columns defined within our table. A row is then inserted and a value set for categoryID, which is a ColdFusion UUID. When building components structured to return queries, I always add this method. The primary function of this method is to make it easier to build a form to update the database. Later we will cover building this form, at which time the purpose of this method will be more obvious.

The save method is fairly straightforward as it simply determines whether to call the Add or the Edit method depending on whether the categoryID sent already exists in the database.

The add, edit, and delete methods are fairly standard representations of their SQL counterparts (insert, update, and delete respectively), except for a few important points that we will cover. The first is that before a category is inserted, either the createRootNode or the createChildNode method must be called depending upon whether a parent category ID has been sent, keeping in mind that a root node is a node without a parent. We will discuss both methods in more detail later, but it is important to note that they both return a structure containing the left and right position integers.

Second, you should be aware that the edit method only allows changing of the text in the category field. This is because, as we discussed before, one of the few drawbacks of the nested set model is that it is extremely difficult to move a sub tree. Therefore, for simplicity's sake, this component simply disallows moving a category once it has been created.

Third, you will note a series of queries within the delete function called CloseGap 1, 2, 3, and 4. To understand the reason for these queries, you must first understand the logic behind including them, which is that, rather than delete a node (category) along with its entire subtree, I have chosen to promote the child nodes of any deleted node. For instance, let's take a category tree containing Grandpa Abe, which has a child, Homer; and Homer has three children, Bart, Lisa, and Maggie. Should something befall Homer, say an accident at the nuclear plant, Bart, Lisa, and Maggie would be adopted by Grandpa Abe. Thus the CloseGap queries are designed to do just that, close any gaps in the left and right position fields within the database, thereby promoting any children of the deleted node.

The get function within our component has been customized to implement a variety of different common queries required when integrating the category system into your site. This method takes four arguments: categoryID, children, parents, and immediate, all of which are optional. Due to the structure of the if statement, the Boolean argument children takes precedence over the Boolean argument parents. In addition, the Boolean argument immediate relates specifically to the children argument but is ignored otherwise. The method is as follows:

• If children is true and a categoryID is supplied, then the component will retrieve the children of a given node
  -If immediate is true, only the immediate children are returned (i.e., one level below the level of a given node), otherwise all children are returned
• If parents is true and a categoryID is supplied, then the component will return all parents of a given node
• If only a categoryID is supplied, then only the information for the specific node is returned
• If no arguments are supplied, then a raw query of all category data is returned

We will discuss later how this method is implemented within your application.

The getAllNodes method is a variation on the get method that returns all categories, with the difference being that getAllNodes returns a lvl column in the query that specifies the depth within the tree of any given node. For instance, given our example from earlier, the Bart category would have a value for lvl equal to 3, as it is three levels deep in the category tree. In addition, you can specify a lvl argument that will filter the returned results by a given depth. Again, using our previous example, specifying the value of 3 in the lvl argument would return only the categories of Bart, Lisa, and Maggie. Being able to retrieve the depth information for categories is a useful feature that you will find yourself using repeatedly throughout your application.

The remaining methods, createRootNode and createChildNode are both private methods, meaning they can only be called from inside the component. They are both designed to return the pos structure used within the add method as we saw earlier. The createChildNode method also updates the tree structure to make room for the node to be created.

Using the Component
Let's cover some basic implementations of the categories component by first looking at how to build a form to update the category tree. We will do this in the context of creating and organizing categories for Web site navigation. The form we will create is missing a lot of niceties, which I will leave for you to add at your leisure. For these examples to work, you must first create the database as explained earlier in this article and add that as a dsn within the ColdFusion administrator called CFDJ.

Our form is self-submitting and includes only two fields, one a text field to enter the name of the category and the other a select box to choose a parent category if applicable (see Listing 2: addCategories.cfm). The select box uses the lvl integer returned by the getAllNodes method to structure the list so that the depth within the category tree is visible (see Figure 4). The form is prefilled with either the value of a category determined by URL.categoryID or by the values returned by the new method, which is empty except for the categoryID. Note that when the form is prefilled with an existing category, the parent category select box is disabled because, as we discussed earlier, the component disallows changing the parent of an existing category.

When the form is submitted, we first run it through basic server-side error processing. Then we remove any empty values from the form structure so that they are not passed on to our component through the argumentCollection (this is simply a shortcut I prefer to use for submitting forms to a component). Finally, we call the save method, which, if you recall, determines whether the category should be added or edited.

Once submitted, you can view your category added to the tree within your select box, or view the test page, the basic code which I have also included (see Listing 3: index.cfm). The test page shows some standard uses for the different get methods within the component, such as getting the root level categories to propagate a category menu, getting the immediate children of a given category to propagate a submenu, and getting the parents of a given category to propagate breadcrumbs (see Figure 5). As you can see, all of these functions are exceedingly easy to use, as they require very little coding.

Conclusion
Using the nested set model, we have built a component and category structure that is flexible enough to use in a variety of applications with little to no modification. In addition, by abstracting this model into a component, we have made it portable enough to reuse relatively effortlessly.

Add Your Feedback

In order to post a comment you need to be registered and logged in.

Register | Log in

Please wait while we process your request...