About the use of the id. Coming from a .NET background I find it weird and confusing to define your class and your instance(it’s what you are doing when adding an id to the config). I did it anyway, because I am lazy and the use of Ext.getCmp() was so convenient. But at the end of the day, I ran into multiple and frustrating run time issues because of this poor design. Can you develop a bit more on that and tell us more about the Ext.ComponentQuery method ?

I haven’t read “Test Driven”, but maybe I’ll check it out based on your recommendation.

You’re also right in that this isn’t a new topic… but I do think many software developers are oblivious to the idea. Developers are often uninvolved in the “requirement” process, or at least until much later in those discussions. I feel that developers in particular would benefit from reading this book for that reason alone.

I will agree that I thought the author should have “just told me how to do it already” early in the book… the use of so many anecdotes probably contributed to some of the dryness I mentioned. That being said, Adzic’s clear vocabulary is certainly re-enforced by the examples.

Testing the logic for a method is simple (expect a specific output given an input) and should be run in complete isolation from the running application. Testing functional aspects of your application require the application be running… and these tests often don’t have a clearly defined “pass” or “fail” scenario.

In other words, I tend to see logical tests as objective, whereas functional tests are (to a certain degree) subjective.

I’ve been working a solid example based on Sencha Touch to demonstrate this… but I have not yet had the time to complete the functional tests. If you’re curious, check out my Prize Patrol app. It’s not really all that impressive, but you should get the idea.

How do you know you’ve tested enough? My advice is to test the “important” or “troublesome” parts of your logic, and apply functional tests as needed to prevent regression bugs. It’s damn near impossible to get 100% code coverage… I’m simply advocating a middle ground where you can be confident to a certain degree.

In your case (like many developers) you will probably have far more functional tests than logic tests – but it also depends on your code. In order to write Jasmine tests for ExtJS components you have to have “testable” methods that aren’t tightly coupled to the DOM or other dependencies.

The trouble with highly interactive UI’s built with ExtJS (or other frameworks) is that Selenium can have a hard time mapping the dynamic DOM elements to the expected test result. I have had some better success using Phantom.js to drive those functional tests as you can create screenshots – and compare the expected screen to the result screen for pixel-perfect tests.

The obvious flaw in that approach is that Phantom.js isn’t a true browser… it is based on WebKit, so it’s a good bet for covering Chrome/Safari and mobile browsers. It’s useless for Firefox, IE and others.

In my earlier post (linked above) I say that flaw “might be okay” depending on your application, but tools like Selenium do offer you the ability to test across browsers using the actual browser.

One note… as the latest OSX is now 10.7.3, I had to additionally install this driver (don’t know if the above link was necessary)

http://displaylink.com/support/mac_downloads.php (Version 1.7)

So the real magic is here:

return {
name : config.name,
age : config.age,

get : function() {
return getPrivateValue(); },
set : function(newValue) { setPrivateValue(newValue); }
};

Not in the var…. above.

Doh! thanks!!