Better Specs is a collection of best practices developers learned while testing apps that you can use to improve your coding skills, or simply for inspiration.
Better Specs came to life at Lelylan (open source IoT cloud platform) and checking out its test suite may be of inspiration.
Better Specs focus on Rails testing, but our goal is to create testing guidelines covering most languages and frameworks out there (e.g. Scala, Elixir, React). If you want to add your favourite language testing guidelines open an issue.
Be clear about what method you are describing. For instance, use the Ruby documentation convention of
. (or ::) when referring to a class method's name and # when
referring to an instance method's name.
A spec description should never be longer than 40 characters. If this happens you should split it using a context.
In the example we removed the description related to the status code, which has been replaced by the expectation is_expected. If you run this test typing rspec filename you will obtain a readable output.
The 'one expectation' tip is more broadly expressed as 'each test should make only one assertion'. This helps you on finding possible errors, going directly to the failing test, and to make your code readable. In isolated unit specs, you want each example to specify one (and only one) behavior. Multiple expectations in the same example are a signal that you may be specifying multiple behaviors.
Anyway, in tests that are not isolated (e.g. ones that integrate with a DB, an external webservice,
or end-to-end-tests), you take a massive performance hit to do the same setup over and over
again, just to set a different expectation in each test. In these sorts of slower tests, I
think it's fine to specify more than one isolated behavior.
Testing is a good practice, but if you do not test the edge cases, it will not be useful. Test valid, edge and invalid case. For example, consider the following action.
The error I usually see lies in testing only whether the resource has been removed. But there are at least two edge cases: when the resource is not found and when it's not
owned. As a rule of thumb think of all the possible inputs and test them.
When you have to assign a variable instead of using a before block to create an instance variable, use let. Using let the variable lazy loads only when it is used the first time in the test and get cached until that specific test is finished. A really good and deep description of what let does can be found in this
Use let to initialize actions that are lazy loaded to test your specs.
Use let! if you want to define the variable when the block is defined. This can be useful to populate your database to test queries or scopes. Here an example of what let actually is (learn more about rspec let).
If you have ever worked in a medium size project (but also in small ones), test suites can be heavy to run. To solve this problem, it's important not to load more data than needed. Also, if you think you need dozens of records, you are probably wrong.
This is an old topic, but it's still good to remember it. Do not use fixtures because they are difficult to control, use factories instead. Use them to reduce the verbosity
on creating new data (learn about Factory Bot).
One important note. When talking about unit tests the best practice would be to use neither fixtures or factories. Put as much of your domain logic in libraries that can be tested without needing complex, time consuming setup with either factories or fixtures. Read more in
Deeply test your models and your application behaviour (integration tests). Do not add useless complexity testing controllers.
When I first started testing my apps I was testing controllers, now I don't. Now I only create integration tests using RSpec and Capybara. Why? Because I
believe that you should test what you see and because testing controllers is an extra step you wont usually need. You'll find out that most of your tests go into the
models and that integration tests can be easily grouped into shared examples, building a clear and readable test suite.
This is an open debate in the Ruby community and both sides have good arguments supporting their idea. People supporting the need of testing controllers will tell
you that your integration tests don't cover all use cases and that they are slow. Both are wrong. You can easily cover all use cases (why shouldn't you?) and you can run single file specs using automated tools like Guard. In this way you will run only the specs you need to test blazing fast without stopping your flow.
Running all the test suite every time you change your app can be cumbersome. It takes a lot of time and it can break your flow. With Guard you can automate your test suite running only the tests related to the updated spec, model, controller or file you are working at.
Here you can see a sample Guardfile with some basic reloading rules.
Guard is a fine tool but as usual it doesn't fit all of your needs. Sometimes your TDD workflow works best with a keybinding that makes it easy to run just the examples you want when you want to. Then, you can use a rake task to run the entire suite before pushing code. Find an example vim keybinding here and learn more about guard-rspec.
When running a test on Rails the whole Rails app is loaded. This can take time and it can break your development flow. To solve this problem use solutions like Zeus,
Those solutions will preload all libraries you (usually) do not change and reload controllers, models, view, factories and all the files you change most often.
Here you can find a spec helper and a Guardfile configuration based on Spork.
With this configuration you will reload the whole app if a preloaded file (like initializers) change and you will run the single tests really, really fast.
The drawback of using Spork is that it aggressively monkey-patches your code and you could lose some hours trying to understand why a file is not reloaded. If you have some
code examples using Spin or any other solution let us know.
Here you can find a Guardfile configuration for using Zeus. The spec_helper does not need to be modified, however, you will have to run `zeus start` in a console to start the zeus server before running your tests. Although Zeus takes a less aggressive approach than Spork, one major drawback is the fairly strict usage requirements; Ruby 1.9.3+ (recommended using backported GC from Ruby 2.0)
as well as an operating system that supports FSEvents or inotify is required.
Many criticisms are moved to those solutions. Those libraries are a band aid on a problem that is better solved through better design, and being intentional about only loading the dependencies that you need. Learn more reading the related discussion.
“Feel free to submit a PR” are words often found in GitHub, but met with confusion and fear by many. Getting started with contributing open source is not always straightforward and can be tricky. If you are new to contribution, watch these videos and you’ll be equipped with the the tools, knowledge, and understanding you need to start on contributing to the world of open source projects. In particular, Better Specs needs your help on the following tasks.