I’m bruised after fighting my own Ruby/Mac noobness in the last couple of days, but I think I’ve learned a bunch of things after this.
I was a nobody…
When attending to Houston TechFest last week, my buddies told me that if I’m not using RVM I’m a nobody. 🙂
I hadn’t had a reason to use RVM, to be honest. I’ve been mostly working on projects running Ruby 1.8.7, and haven’t really had time to look into the new features/fixes introduced to the language. Ben mentioned improved performance, which is a very good reason, but I felt I wasn’t ready to migrate just yet (I had heard of people having some issues, so I thought I was going to save that “fun” for later).
Well, little did I know what was coming up the very next day…
It all started with some weird errors in Heroku
I pushed some changes to Heroku, and when I went about testing the application, it started to crash in points unrelated to my changes.
After analyzing the logs, the core problem seemed to be related to the allowed quota on my database being reached (I’m running the free account right now). I thought, “well, let me just reset the database, since this is all junk data anyway“. Apparently, db:reset isn’t available for MongoDB on Heroku. I tried to clean up my local database and push it up, but something went wrong there.
I then decided to just ‘destroy’ the app on Heroku, and recreate it. That’s easy enough…
Off to new errors…
Once the application was redeployed to Heroku, I started to see new errors in the log: they were Ruby syntax errors that I didn’t use to have before. After giving it some thought, I’ve guessed that Heroku’s default for new apps is currently Ruby 1.9.2, whereas my app used to be running on 1.8.7. My guess turned out to be correct as confirmed by Heroku’s tech support a couple of hours later.
Time to upgrade Ruby version? RVM!
My first reaction was to upgrade to Ruby 1.9.2, as that should address my Heroku problem (later, their tech support told me I could specify what version I wanted my app to run on in their stack… but at that point, it was too late, as I had gone quite far down into the upgrade path).
Installing RVM was easy by simply following the installation instructions on the website. After that, however, my environment didn’t have bundler, so I installed it with gem install bundler.
Next, time for bundle install. I started to get errors related to certain gems, which were actually easy to fix by upgrading gems, changing versions, etc. But there was one that seems to always be the worst nightmare: RMagick. I remember last time Ben and I had to hack something to make it work. This time I found out this Magick-Installer script, which worked like a charm (it took like an hour to download and install everything, though).
Installing stuff from Bash
I’m still pretty much a Bash noob. When I looked at the section of how to install Magick-Installer script, it was kind of cryptic to me:
“Download and run this simple script and watch your ImageMagick support go from 0 to 1 without MacPorts.”
Download and run. Hmm… *how* do I do that? I’m kind of used to download a zip file and double-click on whatever comes out of it, but I couldn’t find anything similar there. But then, I recalled the instructions I used to install RVM just a little earlier, so after clicking a link to the Magick-installer.sh file on Github and clicking on the “Raw” link in there, I found out the direct link to the actual script file. Then it was easy:
bash < <(curl -s https://raw.github.com/maddox/magick-installer/master/magick-installer.sh)
Use curl to download the file, and bash to run it. I shall not forget that.
Yeah, yeah, you bash badasses out there are laughing at me right now… well, I’ll give you one more reason to laugh: I’ve run the command above inside of one of my project’s folder, instead of running it inside of a temp folder, as I came to find out when I looked at the status of my Hg repository and it showed me hundreds of files had been added to it…
Mercurial cleaning up my mess
…well, good thing I’m using a version control system, right? Once I figured out the mess I created, reverting it was just a matter of running hg revert –all. No more harm done.
Me messing Mercurial up
At one point, I committed a “gemfile” to my repository, without realizing it wasn’t the same “Gemfile” I already had in there; no idea where the new, all lowercase file came from. Whenever it was time for me to merge my branch into the trunk I get a nasty surprise: some “case-folding collision” on “gemfile”.
I’ve tried what seemed like a million things, including the steps outlined here, but none of it worked. Eventually I realized I had a good copy of my project on Heroku, so I pulled it down, dropped into the project’s folder where the Hg repository is, and managed to merge it that way, and I was back in business.
Swapping Gems due to versioning issues
I was using HTTParty in order to make a post call to a service that returns an XML fragment. For whatever reason, after the upgrade, HTTParty (or rather Nokogiri, which is used internally to parse out the responses) couldn’t handle the XML fragment in the response I get from the service.
Fortunately, my dependency on HTTParty was encapsulated in a single spot, so I was able to quickly swap it and use Patron instead. Looked like it was going to work fine, but when I deployed to Heroku I got errors when compiling the native extensions. I look it up, and I find out I’m supposed to use an older version of Patron. I try out the older version, and it doesn’t work on my local environment (WebMock had trouble mocking calls to Patron).
I then look *that* up, and somebody recommends RestClient, which worked like a charm.
Tests helped. A lot!!
I have a bunch of Cucumber and RSpec tests. They were VERY helpful pointing out things I needed to fix after upgrading to Ruby 1.9.2.
Besides certain syntax errors I had here and there, I got quite a bunch of errors related to Mongoid relations; I was using “reference_many” and “referenced_in”, instead of “has_many” and “belongs_to”, respectively.
After fixing most of those issues, I got stuck in a case where I had some metaprogramming magic going due to a misunderstood requirement. I had gotten clarification on the requirement with the client a month ago and added an item to my tracker in order to come back and clean that up later. Well now was the time to do it.
Thanks to my tests, I was able to fix my upgrading issues *and* clean up my code, refactoring parts that were written when I knew much less about both Ruby and the domain I’m working on.
Everyday I can feel my noobness on Ruby, Rails, Bash, Mac, etc. I’m working by myself, so there’s nobody sitting by my side I can turn to when I get stuck with stupid stuff. I have a couple of buddies who I ping on Skype or Twitter, and who are *very* helpful, but I try to keep that to a minimum, since everybody’s always busy with their own stuff.
Considering all of that, I’m feeling very good after the struggles I’ve had in the last couple of days, as I managed to get stuff done, and learn a bunch of stuff while doing it.
The single most important factor that allowed me to be brave and knock down every single hurdle in front of me? TESTS, TESTS, and TESTS!!