I’ve recently started working with
CoffeeScript and wanted to do things right. This means writting some tests and generally trying to follow
Code needs to be modular – that’s pretty obvious for anyone who ever worked with something else then tiny “hello world”-like programs. Having 1k Lines of Code in a file is pain. With 5k LoC it seems almost impossible to work effectively. And even more – I definitely like having my tests in separate files than my production code. So I my project structure like this:
For now it seems enough. Spec directory holds test files (aka. specifications). In main directory there are classes and everything else. It will probably change when project will start getting bigger, but for now it’s good. Lets just take a look at files content:
A simple class and test. It doesn’t get any simpler than that, does it? BTW, if you don’t know how the specifications work – take a look at Jasmine tutorial – it’s all pretty easy and nice.
How do I run those tests? Well,
Jasmine says that I should create test running
html file and include required scripts in there. Simple, sure, but – boooring. Who likes going to the browser to see if test are passing? No – I wanted to be able to run them from command line. In future this will make it easy to include them in some
Continuous Integration and what else.
Jasmine-node comes to help! It allows running
Jasmine specifications using
nodeJs. Exactly what I was looking for.
So I installed it:
npm install -g jasmine-node
and tried running the tests:
jasmine-node --coffee ./spec
--coffee parameter indicates that specifications are written using
CoffeeScript, second parameter is simply path to directory containing specifications that needs to be run. After spending some time fighting with this tool (there is a bug that suppresses any error messages when running under
nodeJs 0.10 so it wasn’t notifying me on compilation errors – workaround – try first compiling
spec.coffee files manually if you see that
jasmine-node quit with no error or output at all) it run my tests. And they failed.
Reason for this is pretty obvious – there is no reference for
Ball class in specification. In
C# we would need to add
using statement, but what about
Ball would be defined in different block of code and would not be visible outside of it.
CoffeeScirpt proposes attaching the class to some global variable (like window) or exporting it using
nodeJs exporting mechanism. First solution will work when running in browser. Second when in
nodeJs context. But I want both – tests in
nodeJs and normal code in browser!
In this case there is no other way but to decide on runtime which export mechanism is to be used. And since
root = exports ? this
root.Ball = Ball
CoffeeScript existing operator – it checks if variable is defined (i.e. is not undefined and not null) and returns that variable if defined. If variable does not exist – it returns statement on the right.
this is object which is used to create class definition – in browser it will be
window instance – who would’ve guessed? In second line we assign our class to property on exporting point – how I love dynamic languages!
OK, exporting is covered. But how to import this in specifications? It’s simple since we assume our specifications will be run only from
nodeJs – then we can use
require method and get what we want. Below is final code:
Require returns exports object. To this object we have attached our
Ball class and now we can retrieve it and use it in our code. Re-running
jasmine-node tool proves that everything is OK now – test passes. Problem solved!
What’s left? Well – it won’t work if we will try using normal
jasmine html runner. But I think I have an idea on how to solve this as well – with a little hack, but it should be fine, I hope. We will see how it turned out in next blog post. Take care!