I have set my sights on improving my Angular proficiency, however I am not ready to give up on all my other tools. I have a strong preference for Jasmine, and the recent release of 2.0 further cemented that comfort zone. Jasmine 2.0 made major improvements by removing window dependency and improving async support. Using Jasmine and Jasmine-Node together allows me to write tests in the same language for both server and client. I prefer to standardize as much as possible on a project, and having one language for tests on both application tiers is valuable to me in time and effort saved.
These are exciting times in the Enyo world; Enyo is soon to release version 2.4. This is an extremely important release that brings MV* features and a host of language improvements. I am going to discuss a feature new to Enyo called "Observables", and the power it has when combined with Enyo's inheritance model.
In the previous post I introduced the concept of using Rewire with Jasmine-Node to simplify Node.js unit testing. Rewire added a __set__ and a __get__ function to the module which let us grab a function in the module, and test it directly. We could test the function's output by controlling the objects that the function worked with, and checking them upon function completion. We didn't need to manipulate any internals of the function so our testing is still honoring the "Black Box" concept of unit testing. The idea is that the unit test only cares about what the testee can take in, return, and manipulate outside itself. How that operation is performed is not important to the unit test.
There was one test we didn't look at which I want to discuss now.
Node.js development can be a lot of fun. Unit testing can be fun too. Unit testing Node can be very much not fun. Unit testing Nodejs modules often feels like trying to build a neutered integration test with mocks, nocks, and intercepts. Since Node is a web server whose functionality revolves around responding to http requests, the most natural testing method would be to trip those URLs and check the response. However, that ends up being more of a test of the middleware, and doesn't allow for testing the function directly. If you want to ignore the http calls, the most rational way to test the functions inside the module would be to export them. Once exported the tests can easily be run against the exported functions. The problem with exposing all the functions via export means functionality that should be private is no longer private.
I am continuing along my path of "test driven development" discovery and am finding it fraught with challenges. All my progress has shown me how crucial TDD is for success, and I am constantly reminded of how effective it is. The latest challenge I had was to test the outgoing http call of my node server. I was testing a module that sent an http request in a private method. It was crucial that my testing confirm the outgoing URL. This is not as straight forward as it sounds.
I recently had the honor of one of my minor changes to the Jasmine-Node project being merged into the project master branch. I added a new option to the command line used to invoke the Jasmine-Node tests. Using a --configure option followed by two additional entries will add a variable to the process.env in the invoked node process. Adding this feature allowed me to move my tests with my node server through it's path from testing to production.