Last Friday and Saturday, I attended the very first Pragmatic Rails Studio in Reston, VA. I highly recommend attending the studio to anyone interested in Rails development. Unfortunately, I wasn’t able to take more time off from work and attend the Thursday session on Ruby (though I already know enough Ruby to be dangerous). Dave and Mike really know their stuff and have “real world” examples of what was easy and where they ran into pitfalls in Rails. And Nicole took care of all of the little details which make a training session great (Thanks!)
Day 1
The pace of the first day was a bit slow for me as I’d already been playing around with Rails for a couple months and had written some small web-apps. But David and Mike had to lay the ground-work so that everyone was on the same page. Thankfully, for those of us who had already worked our way through Agile Web Development with Rails, they had updated the Depot application with some added features (such as Ajax, email, migrations, additional testing, etc). After getting the basics out of the way they went into detail about migrations — I had heard about migrations at OSCON from Lucas Carlson and John Butler, but at the time I didn’t bother investigating since it didn’t sound that useful to me — boy, was I wrong. Since learning migrations, I haven’t written another bit of SQL DDL — and I love it! After that they spoke a bit about scaffolding and models and then demo’ed script/console (basically irb with Rails loaded inside of it). Using the console, they showed how to debug your app, fix issues in the DB and simply browse the DB — after the training, the console became a bigger part of my development style and now I test things out before ‘coding and reloading’. After that we discussed model validation, the find method (and how the find_by_ methods are dynamically generated), SQL injection, rendering, testing and avoiding GETs with side-effects — all of this was review for me, but it had to be covered. Dave showed off some debugging techniques, where the logs were and that many issues can be resolved by removing or resetting the session. Finally, Dave and Mike added some Ajax to the Depot 2.0 app to update the cart in the sidebar without reloading the page and then added the yellow-fade technique so that the user’s attention is drawn to the cart change. There was one significant issue during the training which affected many folks: rendering the cart in the sidebar caused an infinite rendering loop so the page would load completely blank and there was no information about a stack overflow in the logs. So if you find that you’ve added a render call and now the page is completely blank you might have this problem be sure that the render command has set layout to false.
Day 2
Dave and Mike decided to change the pace so that instead of everyone following along coding, our job was to watch, listen and ask questions. This allowed us to cover ground much more quickly. First we discussed web forms and the two different ways in which they could be implemented with Rails: the two action update (two different methods in the controller are used, the scaffold works this way with edit and update) and the one action update (a single method in the controller handles both the GET request and the POST request). They said which one you choose depends on your circumstances, but with tools like Google Web Accelerator it might not be so bad for your actions to prevent GET requests from updating the application state. Then we talked about has_many and belongs_to and managing complex hierarchies (such as many-to-many associations and the acts_as tools). Since I had not messed with HABTM or the acts_as stuff, I found this interesting and am looking for ways to apply it in my next project. Following this Mike explained ActionMailer: how to configure Rails for sending and receiving email, how to generate a mailer and render it. Then we moved on to testing. First, functional testing (i.e. controller testing) was covered along with common assertions and the test environment, then we spoke about fixtures and how to embed Ruby code in them for generating many records, we discussed mock objects briefly and there was some discussion regarding how these mocks differ from Java’s mocks. Dave then talked about partial templates and their usefulness when rendering collections. A brief overview of the flash was next. Hooks, both in ActiveRecord and ActionController, were explained. Since I had never used the hooks in ActiveRecord, I found this discussion interesting while the overview of filters in ActionController was all review. The ActiveRecord hooks seem great for debugging, though I think they would slow down the application quite a bit if you took advantage of every single one — I guess if you need to massage the data going into/coming out of the DB then hooks could really be useful. Next routes were covered as an intro to web services. I had played around with routes briefly and found them to be really powerful. The overview of webservices was quick — I had already written a RESTful one and would be perfectly happy to spend the rest of my life without ever having to write a SOAP/XML-RPC web service, so I didn’t object. Mike then spoke about his experience working on VitalSource and how important interface design is for your web-app. He also urged everyone to get the web designer involved as early as possible in the project. He stressed the importance of creating many helpers to extract out the view for components. He recommended waiting to add Ajax until the end of the project and that it is better to use it subtly than to over do it. Following that we spoke more about SQL injection, CSS/XSS, not to trust parameters (especially the id), validate and sanitize uploaded files, don’t cache authenticated pages, use functional tests to seal potential holes (and recommended checking out OWASP). Deployment was next. First Mike discussed the architecture of your web-app (share nothing!) and that you should move to saving sessions in the DB since multiple instances of your app will be running on different servers. As far as the webserver, Dave and Mike were split on Apache and Lighty, respectively (Dave doesn’t think Lighty is quite there yet in terms of rock solid stability and Mike hasn’t had any issues with Lighty — he admits his Apache-fu is not that great). Mike talked about setting up your application to email you (or your cell phone) when things go wrong and Dave discussed the logs some more (and mentioned SyslogLogger). Some techniques for optimizing your app were discussed (such as preloading child rows, using find_by_sql, etc) and caching. Page caching was discussed first, followed by action caching and fragement caching — I had never used these techniques, but they seem straight-forward enough — Dave cautioned that caching only works in production mode, so if you enable caching and don’t see it working check that first. To end the studio, Mike gave a great demonstration of SwitchTower which showed how easy it is to deploy your app (whether you deploy to a single machine or a cluster) and how easy it is to rollback if you found an issue after deployment. Dave and Mike reminded us to roll our log files (I think there was an issue on TextDrive where someone’s exceeded their quota by far with huge log files), purge old sessions and schedule daily maintenance with cron.