We talked about graph databases in this and this blog post. As you might have read we’re big fans of a graph datbase called neo4j, and we’re using it together with JRuby. In this post we’ll share a little piece of code we created to make expressing graph traversals super easy and fun.
Jogger is a JRuby gem that enables lazy people to do very expressive graph traversals with the great pacer gem. If you don’t know what the pacer gem is, you should probably check pacer out first. (And don’t miss the pacer section at the end of the post.)
Remember the named_scopes from back in the days when you were using rails? Jogger gives you named traversals and is a little bit like named scopes. Jogger groups multiple pacer traversals together and give them a name. Pacer traversals are are like pipes. What are pipes? Pipes are great!!
The most important conceptual difference is, that the order in which named traversals are called matter, while it usually doesn’t matter in which order you call named scopes.
The former is really just a syntax thing, whereas the latter can help you a great deal modeling semantics of your business logic as parts of traversals. Let’s jog!
To demonstrate why point 1) in the list above can be useful, look at this traversal. It helps me find out what movies my female friends like the most, so I can impress them in a conversation:
Since I’m a very lazy person, I would prefer to write it a little shorter. Especially, since these multi step traversals are a pattern I found in our code at moviepilot.com a lot.
So here’s the Jogger way of expressing this:
See what I did there? Jogger keeps the current pacer traversal and forwards all method calls to that traversal, and then returns itself. So you could also write (in jogger as well as pacer):
Just saying, you can chain your methods, but I don’t like it cause I can only focus on 72 characters per line at max. If you want the current traversal, just call result on your Jogger instance.
So that traversal above, traversing from a node to all its friends, is pretty simple, but it could be simpler. Especially if it does things that you want to reuse in many other places. How cool would it be if I just had to write this:
No problem. Just define named traversals that aggregate different pipes and give them a name. You have to put your traversal into Jogger’s NamedTraverals module.
Your methods have to be able to take at least one parameter: the current traversal. It represents the current traversal state of your Jogger instance. Your traversal can then modify this traversal and must return it. It will become your Jogger instance’s new state.
These are silly examples, but if you look at your traversals I guarantee that you will find repeated patterns all over the place, and Jogger can help you stop repeating these and making the actual traversals much easier on the eyes.
Installation
First, you need to load pacer and whatever graph db connector you need (we use neo4j, by the way) and define your named traversals as above. Jogger doesn’t include these on purpose. Then, you have to
and
or for your Gemfile
That’s it!
I gave YARD a shot, so to open the documentation in your browser just do this in the jogger directory:
Or you can browse the documentation online.
You can implement feature #2 purely in pacer, if you like. For example, you could express
with
For this to work you need to tell pacer, what Movie actually means. In the simplest form, you could say:
That way pacer knows, whenever you’re talking about a Movie you want to filter those elements whose type==Movie. To do something like the named traversals described above, you could go ahead and define a route to all female likers of a movie:
You can then go ahead and fetch:
To wrap it up, using Jogger to do named traversals is for the super lazy. If you use pacer exclusively and have more complicated structures it would probably make more sense to create your named traversals in the design of your domain logic and do it purely with pacer. You can still use Jogger to traverse these routes. If you want to share common traversal patterns between different models it might be easier to do with Jogger.
In the spirit of jogging, here’s our complimentary gif as a thank you for sticking through:
This post is brought you by Moviepilot Labs’s own Jannis (@jannis).
