by Samuel Mullen
Posted on Jun 12, 2014
Ruby on Rails will be ten years old in December 2015, and a lot of things have changed since that initial release. Although I’m sure there are some applications which have been around since the initial release (anyone else besides Basecamp?), most “old” applications I run across as an independent are 2.x or higher.
As these opportunities have come my way, I’ve found I need to know how to evaluate the projects to make the best choice for myself and my clients. You may not be a freelancer, but you may still find this article useful if you are considering working on an open source Rails project or taking a full time Rails position and have an opportunity to look at the code base.
As you would evaluate anything, I start out looking at surface details and slowly work my way in to analyzing the logic of the application. I take notes along the way to aid me when I write up the formal evaluation I provide my customers.
Before getting yourself in too deep, rummaging around in files, looking at code, and breaking things (intentionally!), I’ve found one is better served by getting a feel for the application as a whole. You want to know the answers to the following questions:
Answering the first question is easy enough: if there’s a Gemfile, look for the
rails gem and what version it’s set to. If there isn’t a Gemfile, you can find
the version number by looking for the constant
To find the version of Ruby the application runs, you can sometimes find it in
the Gemfile, but you’ll usually find it in the
you can’t find the Ruby version, that should also tell you something.
While you’re in the Gemfile, it’s always a good idea to find out exactly what libraries are used by the application. Are they up to date? Is there a Gem for everything under the sun? Are there competing gems?
Gemfile hasn’t been created for the application, you’ll need to look in
config/environment.rb again. Be warned, just because a gem isn’t listed in the
environtment.rb file doesn’t mean it’s not used by the app - it’s one of the
joys of pre-Bundler apps - it may be required elsewhere in the application.
You’ll usually find out what’s missing by attempting to start the server.
Along the same vein as libraries, it’s also useful to discover which plugins, if
any, are used. You can find these under the
vendor/plugins directory in older
Knowing how frequently the application has been updated can give you an idea about how important a project is to a person or organization. To find this information - if the project uses an Source Code Management system (SCM) - look at the SCM’s log. You can do that with one of the commands below depending on the SCM used:
If the project isn’t under source control you may be limited to looking at the timestamps of the files, but this should also be a warning to you about the last developer(s).
While you’re looking at the commit log, see if you can recognize the names of any of the developers. Knowing who’s previously worked on a project can make all the difference when deciding whether or not to get involved.
The easiest way to find out if there are tests is just to look under the
features directories, depending on what framework the previous
developers chose (there may be more than one). Just because test files exist
doesn’t mean tests have been written, so you’ll want to find out if the tests
are more than just boilerplate files provided by the framework. You can find that
out in one of two ways: you can manually look in each of the files, or you can
wc command (word count), but by passing the
-l flag, it can count
lines in a file.
$ wc -l test/models/* ... 126 test/models/proposal_test.rb 103 test/models/registration_test.rb 47 test/models/search_test.rb 91 test/models/service_estimate_test.rb 4 test/models/service_template_test.rb 28 test/models/service_test.rb 51 test/models/user_registration_test.rb 106 test/models/user_test.rb 2239 total
The first column is the line count for the file. If all the files are between 10 and 15 lines, you can be sure they’re just boilerplates.
Finding the logic is similar to determining if tests have been written. Unlike what we search for with tests, however, here we’re trying to discover two things: 1) Where are the “god objects”?; 2) Is the application controller-heavy or model-heavy?
“God objects” are those classes which seem to get involved in every aspect of an application. I’ve written before about how to deal with them in Delegation Patterns in Ruby over at O’Reilly Media’s blog. The simplest way to discover them is to look for the largest files in the
/app/models directory. We can use the same trick we just used to look for tests, but we can pipe it to
sort to more easily find the larger files.
$ wc -l app/models/* | sort 7 app/models/weight.rb 8 app/models/department.rb 9 app/models/location.rb 11 app/models/position.rb 12 app/models/role.rb 13 app/models/comment.rb 13 app/models/salary.rb 15 app/models/rating.rb 19 app/models/question.rb 28 app/models/appraisal.rb 30 app/models/company.rb 69 app/models/notification.rb 96 app/models/category.rb 111 app/models/user.rb 441 total
You can see here that
User will likely be involved in a lot of
the logic of the app. This isn’t always the case, but we’re just getting an idea
about the application. If none of the models were more than 15 lines long, that
would tell us that all the logic was being kept in the controllers. Speaking of
You can use this same technique on the controllers directory. Here, though, what we’re looking for is how “fat” the controllers are. Controllers larger than 100 lines probably need to be refactored. Seeing a lot of fat controllers tells you something about the state of the project.
Oftentimes just scanning a repository will be all you need to determine if you want to work on it. But there are other times when you’re still on the fence. In those cases, it makes sense to be more invasive.
This may seem like an obvious thing to do, and it is, but just getting a Rails application to start can be very telling about the project itself. If it’s an older app, you’ll discover missing libraries. If it’s a newer project, what deprecations and errors are you presented with. How long does it take to piece everything together in order to start it? In any case, launching the program for the first time should give you some idea of its state.
We are Ruby developers after all; there should be tests, right? The state of the tests suite (passing, failing, or non-existent) will give you ample information about the project.
If there are no tests, or if the tests are so out of date as to be useless, reconsider taking on the project (or upping your meds). It is a serious undertaking to add tests to a project that has none. Although it is a simple matter to add a test for some method, oftentimes that method will be bloated with logic and need to be refactored. Is this really the kind of project on which you want to work?
Arguably the best tool available for determining how much of a steaming pile of filth you’re going to have to deal with is the gem, flog. Flog is a tool to measure code complexity. Although it has features similar to other code metrics tools, Flog is specific to Ruby and is able to understand all of her Foibles.
To use flog, install it and run it against a directory such as
Without arguments, Flog retrieves the scores for the largest 60%. The output looks like this:
1265.4: flog total 6.2: flog/method average 57.3: AppointmentCalendar#to_ics app/models/appointment_calendar.rb:8 34.2: OrganizationClauseManager#move app/models/organization_clause_manager.rb:17 27.4: Area#none 25.5: LineItem#none 25.5: AreaService#none 23.9: Lead#set_facility_variables app/models/lead.rb:176 22.7: Activity#all app/models/activity.rb:9 ...
The higher the score, the more complex the code. For the
you’ll want to see scores below 15. Scores above that may require investigation.
controllers directory, expect higher scores; good scores are less than
There may be very good reasons for having high scores, but beware of applications which have nothing but high scores.
There are a few other minor things to make note of; none of which should affect your appraisal.
I hate observers in Rails. I’ve written before about their evil twin callbacks, but observers are worse because they’re so easy to miss. Like I said in the aforementioned post, they’re “kind of like ninja callbacks”.
When evaluating a Rails application, make sure to look for
Depending on the gems used by an application, or by the “ingenuity” of the
previous developers, there may be some non-standard directories under the
folder. Here are some of the directories I run into most frequently:
decorators: usually created by the draper gem. These are classes which adhere to the decorator pattern.
presenters: oftentimes created by developers wanting to follow the decorator pattern, but without the overhead of adding yet another library.
uploaders: Created by the carrierwave gem.
overrides: Not limited to spree, but Spree makes heavy use of this directory.
Much of what’s been written above can be applied to any project. Almost everything from the first section can be directly applied - although the state of testing in other languages is…disappointing. With the exception of “Will the Application Start?” and “Do the Tests Run?”, the rest of the article is probably moot, but there may still be some pieces which can be adapted to suit your needs.
As an independent Rails developer I’m faced with the decision of whether or not to take on various projects. The decision’s a lot easier if I’m dealing with a new project, but if it’s a legacy application, I need to know what I’m getting myself into before I make a decision. Just because an application is in a horrible state doesn’t mean I’ll not take on the project; it’s more important for me to have all the information necessary to make the best decision for myself and my clients.
I live in the greater Kansas City area with my beautiful wife, our two great kids, and our dog. I've been programming using Open Source technologies since '97 and I'm currently an independent software developer specializing in Ruby on Rails and iOS. I am for hire.