Wednesday, September 19, 2012

Intro to ClojureScript - Part 3 - Using Shoreleave-pubsub

In this post we will move away from the function based event handling we put in place in my previous post to use Shoreleave’s pubsub library. Using pubsub will allow use to easily add and remove functionality over the lifetime of our app. Once we’ve replaced the current tightly coupled event handling we will add console logging and the ability to view our search history.

The pubsub library works like the name implies, topics are published and can be subscribed to. Topics can be strings, keywords, functions, atoms, web workers or local storage. In this post we will use keywords to indicate when a search is to be performed, when the results come back, and when the view history button is clicked. An atom will be used to track the current search and keep a list of previous searches.

Since ClojureScript is a bit of a mouthful I will use the abbreviation cljs for the remainder of this post. Remember I am learning cljs as I create these posts so any and all guidance is appreciated.

The Goal

By the end of this post we will have converted the event handling code over to use shorelave-pubsub and have added console logging and the ability to view our search history for the current session.

The Setup

If you don't already have cljs installed on your system go read Intro to ClojureScript - Getting Started to get your environment ready for cljs development. You will also need the code from my previous post. If you don't already have it you can grab it from my cljs-intro project on the part-2 branch. Once your environment is setup we can start making the necessary changes to add pubsub support.

Project.clj Changes

We only have one change to make in the project.clj file and that is to add the reference to the shoreleave-pubsub library:

[shoreleave/shoreleave-pubsub "0.2.2"]

Once the change is in place save the file and run lein deps.

Index.html Changes

We need to add the 'View Search History' button to our web page so users can view their previous searches. Add the new button next to the ‘Get Stats!’ button. The updated index.html should look like this:

After updating the HTML we are ready to get into the cljs changes. The first step is to create the new file that will contain the pubsub code.

The PubSub Code

A little background

Before we get into the pubsub code lets take a minute to go over shoreleave-pubsub. The shoreleave-pubsub library uses two protocols IMessageBrokerBus and IPublishable to provide the necessary functionality. The IMessageBrokerBus manages the bus, giving us the ability to publish, subscribe and unsubscribe to a given bus. The IPublishable protocol provides us with functions that define things that can be published. We will use the topicify and publishize functions to convert the keywords and atom to topics. For a more detailed look into what these two protocols provide read the Shoreleave pubsub docs.

src/cljs/pubsub.cljs

As you might expect the pubsub.cljs file is where we’ll put our pubsub related code will live. The file is broken down into three areas: config, helper functions and subscriptions.

config section

The first thing we need to do is require two namespaces shoreleave.pubsub.simple and shoreleave.pubsub.protocols. shoreleave.pubsubs.simple provides us with a synchronous bus which is fine for this example. The shoreleave.pubsubs.protocols library provides the functions we need to turn our keywords into topics and make our search-state atom ‘publishable’.

First we create the bus and then we ‘topicify’ the three keywords, :search, :results, and :history. Once the keywords have been 'topicify-ed' we can use them to publish messages. The search-topic will be used anytime a user clicks the ‘Get Stats!’ button. The results-topic will be used when that stats are returned. As you might expect the history-topic will be used to indicate that the user has clicked the ‘View Search History’ button.

Turning the search-state atom into something that can be used as a topic is slightly different. We need to call the publishize function passing in the atom and the bus it will be published on. Calling the publishize adds a watch function that publishes all changes to the atom’s subscribers. The published data contains a copy of the new and the old state of the atom. Subscribers can access the data by using the either the :new or :old keyword.

helper functions

The helper functions are there to keep the ‘guts’ of the pubsub code out of the search.cljs file. The publish-* functions are there so code in the search.cljs can publish messages to all subscribers. The subscribe-to function wraps the pubsub subscription process so subscriptions can be created outside of the pubsub.cljs.

subscriptions

The last two lines of the file create subscriptions to search-topic and search-state. These subscriptions, subscribe the anonymous functions to each topic. Now, anytime there is a message published on either one of these topics the published data will be written out to console.log.

src/cljs/search.cljs

The search.cljs file underwent quite a few changes during the conversion. Anything to do with event handling changed as did some of the functions called from within the event handlers. In addition to coding changes we had to add a new require statement for the new namespace cljs-intro.pubsub.

Subscriptions and Event handling

Since we are using pubsub to handle events and state changes we need subscribe to the topics we are interested in.

The subscription calls are using the helper function subscribe-to which takes the topic you want to subscribe to and the function that will be called when a new message is published.

Now when a handler is called instead of calling the function to handle the event, a message is publish for a specific topic. For instance, when a user clicks the ‘Get Stats!’ the handler publishes a message on the topic search-topic passing the value that was entered by the user to all subscribers. Here’s the difference between the two versions:

When we subscribed to the search-topic topic we established that the player-lookup function would be called. There was only two changes made, the first is to publish the results of the remote call when it returns instead of calling display-results directly. The published data is converted to a Clojure data structure prior to publishing freeing the subscribers of any unnecessary steps prior to using the data.

The display-results function is subscribed to the results-topic. The new display-results receives the converted data so we no longer have to convert it before we display it. I’ve also created a new function, update-results-div, to handle the updating of the search results.

The functionality we just walked through is the same as it was before we switched to using pubsub. You may be wondering whats the point? The conversion to pubsub allows us to open up the application so that adding new functionality can be done with little to no changes in the search.cljs code. As an example, we added console logging for the search string by adding a subscription to the search-topic and giving it a function that calls js/console.log. No changes were required to the core search code. The image above shows the search logging and the logging that occurs anytime the search-state atom changes. Thank you pubsub.

Viewing Search History

Another piece of functionality that was added is the ability to see the previous searches you’ve run.

When does search-state change? Every time a new search is submitted the current value of :lastname is conj’d into the :previous-searches sequence provided that :lastname is not nil or empty. When that happens a new message is published to notify all subscribers of the change. In our app we have subscriptions to search-state for managing the visibility of the ‘View Search History’ button and for logging.

The Code

When search-state is changed the search-state-change function is called. This function is where the visibility of the ‘View Search History’ button is determined. If the :previous-searches sequence contains at least one string then the domina set-styles! function is called to make the button visible.

The history button’s click event handler publishes a message on the history-topic. Our subscription for this topic calls cljs-intro.views/view-history function which takes the :previous-searches sequence as its only parameter. From that sequence an unordered HTML list is created and displayed using the new function update-results-div. Here’s what it looks like after a few searches and a click of the ‘View Search History’ button.

Summary

We now have an app that uses shoreleave-pubsub for handling events instead of the traditional, tightly coupled event handler functions. Using the library has allowed us to add new search related functionality without making any changes to the core part of our application. Now we have a more flexible application.

Whats Next?

In Intro to ClojureScript - Part 4 - Using Shoreleave’s Remoting we will convert the current remote callto the server to use shoreleave-remoting. Part 4 may not be up until two weeks due to an overbooked schedule. I will be speaking at Richmond Code Camp giving an ‘Intro to ClojureCLR’ talk and I need to get started on my talk. As soon as thats done Part 4 will be up.

Resources

shoreleave-pubsub docs. I spent a lot of time reading the docs for shoreleave-pubsub.  

Hockey Databank Database -- I created the database using data that originates from a Yahoo! group called hockey databank. After each season the group produces updated CSV files with the stats of the previous NHL season in addition to previous years data.

ClojureScript Experience Report - Resources Bits of information about Jason’s ClojureScript based application and his experience with ClojureScript.

ClojureScript: Up and Running - An early release book that discusses ClojureScript. I’ve read the released chapters and found it to be a good resource. Chapters 2 and 3 go over ClojureScript’s compilation, project structure, and other informative tidbits for both ClojureScript and Clojure. I’ve recommended the book to a colleague who is learning Clojure because of the way the authors describe data structures, immutability and sequences.

cljs-intro - My github repo that houses the code for this blog series. Each part of the series has its own branch. This post’s code is on the Part 3 branch.

2 comments:

  1. It's exciting to see Shoreleave popping up here! I have been enjoying this series very much. Thank you so much for exploring the library and taking the time to write this blog post.

    A few notes:
    You shouldn't have to topicify keywords, it'll happen automatically if you just use them in the bus.
    Keyword topics are great for cross-cutting broadcasts (system notifications, logging, etc)

    Functions themselves can publish their results directly to the bus if you publishize them (just like how you used the atom). That might simplify your code in the middle section.
    It's as if the function itself could have "watchers." You call it just like a standard function, and its results will be pushed to the bus and to any subscribers of that publishized function.

    Paul (OhPauleez)

    ReplyDelete
    Replies
    1. Paul, Thank you for the pointers. I really appreciate. I'll make sure and update the code for part 4 to include them. I think shoreleave is great and I plan on using it in future projects.

      Thanks for taking the time to read my post(s) and more importantly leaving you comment.

      Delete