In building out a custom solution I get tired of writing procedural code to create a list, add the columns, setup the views, and add some initial data. Many moons ago I decided to follow a creational design pattern called the Builder pattern. This pattern separates the construction of a complex object from it’s representation so that the same process can create different representations. Basically, let’s build a bunch of custom lists using OO instead of repetitive procedural-like code. Here’s how.
First let’s take a look at the Builder pattern. Here it in UML form for visual dudes:
Pretty straight forward. The principle here is that all concrete objects will adhere to a contract (the Builder) and implement the specific way they need to build things while the Director handles the correct sequence of object creation.
This suites Lists pretty nicely in SharePoint. A list needs to be created via the SPWeb it lives in, it needs fields added to it, views setup, and items added to the finished list.
Okay, time for some code.
First here’s our abstract ListBuilder. All implementations of creating a new list will inherit from this. It provides abstract methods that make up all the parts of a list (fields, views, items, and the list itself).
Pretty basic stuff. We have methods to create and delete the list (which requires the SPWeb object it belongs to) and there’s methods for creating fields and views and adding some initial data (for example if you have a lookup list you’re creating and you want to seed it with some initial values).
There’s also a protected property called TheList which holds an instance of the list created. This is for adding fields and views later in the construction process. P.S. Personally I hate the property name “TheList” but “List” gets all confused with System.Generic.List types and SPList is already a SharePoint type so to avoid conflicts I picked this. I’m also lazy and didn’t want to call it SharePointList or something. Feel free to make it whatever you want.
Now that we have our abstract ListBuilder let’s create the ListDirector. This class will handle building the list by orchestrating the assembly of the pieces. It also provides outside access to create and delete the lists (which we’ll call from our feature activating/deactivating).
The ListDirector is created using the SPWeb object (so it can tell the list builder where to create/delete lists from) and knows the order to put the pieces of the list together. This way ListBuilder (and any classes inheriting from it) don’t need to know how to orchestrate the pieces and can just work independently. You can also create special ListDirectors that do things like update lists (avoid the creation) or append data.
Armed with our abstract class and our director we head down the path of actually creating a specific list. For this sample we’ll do something really simple. A lookup list for countries and contains a name we’ll use for picking items and some internal codes for each item.
First we’ll start with the bare bones builder:
Now let’s look at each method in detail. There’s not a lot here to see because our list builder should be lightweight. Of course this completely depends on your needs so you might have something with 30 fields but the ListBuilder class should simplify things and remove redundancies for you.
First we’ll create the list. To do this we need the SPWeb object which we’ll pass in.
Here we create a generic list, retrieve it, and update the metadata (in this case if it’s on the Quick Launch or not, but you could do things like setting the Title or other values. The call to Update() is only needed to update the metadata. If all you’re doing is adding the list then the first line is all you need.
Next we’ll add our additional fields:
Since we’re creating this list from a generic (custom) list we already have a title field. Here we add two new fields.
Next up is the views.
When a list is created a default view is created that contains the Title field. Here we grab that view then add our two custom fields to it. You could also do all kinds of things with the view like setting the sort order, adding filters or creating entirely new views.
Finally we want to seed our list with some initial data so every time our feature is activated we have some data for it.
Here we have a helper to add an item to the list. Adding an item requires a few steps (creating the SPItem via Add, setting the data, calling update) so it’s wrapped in a method in the class.
You might be balking at the use of hard coded values and field names and whatnot here. That’s fine. You could go ahead and assign those to constants inside the class. The point is that the field names are all contained within this one class so if you ever wanted to change them you only have to change the values here. Create constants for the field name might be a good idea too. You could also do something like read from a text or xml file in your AddInitialData method if you had a lot of data to seed (or even pull it from a resource file in the assembly).
Now we have all the parts to build up the list. Let’s call it in a feature. As we move up the ladder and away from our list builder things get dead simple. Here’s the code we have to activate the feature:
Like I said. Dead simple. We pass the SPWeb object to the director (who will internally pass it along to the ListBuilder as needed) and we call the CreateUsing method passing it a new CountryListBuilder.
If we want to add new lists all we have to do is this:
- Create a new ListBuilder class inheriting from ListBuilder
- Fill in the abstract methods
- Add a new CreateUsing call in our FeatureActivated method
Okay, so we have it build but what about getting rid of it? Just as easy.
First we need to finish our CountryListBuilder. There’s one more method we need to implement:
Now to finish off we just implement the FeatureDeactivating method:
Just reverse of activating we call the DeleteUsing method on the director passing it a CountryListBuilder object. We really don’t have to keep the variable around and there’s little cost to creating a new object to pass it. There’s nothing unique that we have to keep around from creation. The CountryListBuilder class is just there to provide us the interace to it’s methods.
You might be thinking why do all this? Why don’t I just create the ListBuilder class and call the methods directly. Like I said, the Director abstracts away all the co-ordination you need to build the list while the list knows how to build its parts, it doesn’t necessarily know how to construct itself. This is done for reason. Maybe you want to build the list up differently depending on the environment. Or maybe you have a Director to handle creating items from scratch and another director for updating existing data (say in the Feature Upgrading event with a production site). The same pattern and classes apply, there are just different implementations to create. Also with the abstraction the ListDirector and ListBuilder provides, the code you have to maintain (and debug) in the Feature events is next to nothing and any errors can happen and be handled down where they need to be, not have to bubble up to the feature (remember you have no UI in the feature events).
So that’s it. Now I just have to create a new class for each List I want to build, inherit from ListBuilder and just add one line of code to my Feature activating/deactiving. Pretty clean and pretty quick to get up and running.
This is just *one* way of doing this. You could use other patterns that might be more appropriate (the Abstract Factory or perhaps Prototype comes to mind) but this is the one I went with and it works pretty good for me so far. Fields are more complex beasts so my CreateFields() method is sort of doing double duty by adding fields and also updating existing ones (or new ones) by setting the field lengths. This could be split out to say two methods (CreateFields and UpdateFields) or even a FieldBuilder class could be created and handled by the ListBuilder (who would act as the Director) but maybe that’s over complicating things.
I’m sure there are a lot of things that could be better in this code. Interface vs. abstract class; passing in the SPWeb as a constructor to the ListBuilder; etc. And yes, there’s no exception handling, logging, or backout in case things fail here. What if someone renamed the list? What if the list doesn’t exist when the feature is deactivated? Hey guys, it’s a blog post, not production code. Feel free however to comment and suggest improvements and alternate ideas.
Hope this helps! Happy coding.