Sadly, PCO is getting out of the batman.js game.
From where I sit, I think it boils down to:
- The framework never hit critical mass (and lost what it had)
- The framework’s magic made some problems impossible to debug
- Client-side apps aren’t good for business
- Rails, Turbolinks & React will do just fine
Batman.js is a great framework with some amazing, well-tested features and I’m sorry to see it go this way for us!
No Critical Mass
Batman.js was an early entrant to the “Framework Wars”. It was production-ready by early 2013, which made it appealing at that point. However,
- there was very little “evangelism” by its creators (almost no talks, very little documentation, no “media presence”)
- early adopters were not “invited in” (unanswered github issues, for example)
When Shopify pulled out (around fall 2013?), a ton of knowledge and resources were removed with no viable replacement.
Batman.js is loaded with awesome APIs that were fun to use and satisfying when they worked. However, when it didn’t work, you were up a creek without a paddle. There was:
- no helpful error messages
- no support on github, stack overflow or IRC
- no documentation or information of any kind about the inner workings
- no debugging tools
I’ve seen a lot of batman.js users pull their hair out yelling, “Why doesn’t this work?!” Indeed, that’s what drove me to learn it from the source.
- Bug-tracking tools (namely Bugsnag for JS) are not as good (especially when you throw CoffeeScript & minification in the mix)
- Browser environments are outside your control (I got a lot of bugsnags for peoples’ browser extensions)
- State can get weird – after having the app open for an hour, data can just get messed up! (Maybe this doesn’t happen for better programmers.)
Not my problem:
No es mi problema:
You just don’t realize the luxury of reliable bugsnags until they’re gone! So many Check-Ins bugsnags leave no trace of what actually went wrong.
Other Options are OK, too
We have a solid data model, HTML templates and CSS to boot. It stinks to throw away all that code, but I forgot how amazingly fast it is to code Ruby on Rails.
JS MVC types will look down their noses at Turbolinks, but it works and it has a lot of eyes on it. Combined with
react-rails, it’s a really strong option!
In short, Rails-rendered HTML, Turbolinks &
form_for ... remote: true, and ReactJS for live-updates. There are a few things I am looking forward to:
- RUBY BUGSNAGS with stack traces, request environments and everything!
- Less state. Fewer things that can randomly affect other things in ways I didn’t think about.
Batman.Viewis an amazing thing, but it makes me appreciate the dead-simple API of a React component. It can only get data from one place, no other objects can mess with it, its lifecycle is very simple (and works as documented), etc etc. I write a lot more code to do a lot less with React, but I think it’s going to be a lot more stable, and that’s worth it.
We held on to Batman.js for a long time because it lended itself to live-updating everything. I think React is going to cover the a few key things there, and just changing pages will do the rest. So far I’ve hooked up Pusher to Flux-style stores, then wrapped anything live-updating in a React component that observes that store. It’s a lot more hands-on than Batman.js updates, but it has other advantages.