Wednesday, April 21, 2010

Ruby: An Introduction to Behavioral Driven Development with RSpec and Cucumber: Version 2

Tonight, I'll be giving an "encore" presentation of my talk An Introduction to Behavioral Driven Development with RSpec and Cucumber to the San Francisco Ruby Meetup Group. This is an introduction to behavioral driven development in Rails using Cucumber, RSpec, Webrat, and factory_girl.

Happy testing! :-D

Thursday, April 15, 2010

Ruby: Modifying Class Attributes in Cucumber and RSpec

Imagine you're building Amazon.com, and you have a rating system. You might not want to show an average rating to the user until there's been some minimum number of ratings. Of course, you'll want to test that that feature works. However, in other tests, you might want to disable that feature by setting the minimum number of ratings required to 0.

More generally, I sometimes run into the situation where I have a constant in my model that I might want to change, but only for a specific test. Instead of using a constant, you can use a class attribute. However, since class attributes aren't part of a database transaction, they don't get reset after every test. That's a problem since one test that changes the class attribute might mysteriously break another test that relies on that class attribute. I had a problem where one feature file would pass when run alone, but it wouldn't pass when all the features files were run. Here's how to make it all work.

In your Rating model, create your class attribute as follows:
# Rating.min_num_ratings is the minimum number of ratings before
# we show the ratings average to the user. We don't want a single bad
# rating to forever scar a book's reputation.
#
# I'm setting this up as a class accessor so that you can set it to 0 in your
# tests. Beware that Cucumber and RSpec don't reset this after every test
# since this isn't part of a database transaction. Hence, if you change
# Rating.min_num_ratings, you must change it back after your test. In
# RSpec, you can do this with a begin/ensure block. In Cucumber, you can use
# the @afterwards_reset_min_num_ratings annotation.
DEFAULT_MIN_NUM_RATINGS = 3
cattr_accessor :min_num_ratings
self.min_num_ratings = DEFAULT_MIN_NUM_RATINGS
Your RSpec tests should look like:
it "should do something useful when the minimum number of ratings has been hit" do
Rating.min_num_ratings = 2
begin
# Test that it does something useful.
ensure
Rating.min_num_ratings = Rating::DEFAULT_MIN_NUM_RATINGS
end
end
Just to be sure I didn't forget to reset min_num_ratings anywhere, I have the following at the bottom of my .spec file:
it "should still have the default min_num_ratings" do
Rating.min_num_ratings.should == Rating::DEFAULT_MIN_NUM_RATINGS
end
Cucumber is a bit harder since you can't use a begin/ensure block. I started by adding a step:
# If you use this step, you *must* use the @afterwards_reset_min_num_ratings
# annotation on your scenario in order to reset Rating.min_num_ratings after
# your test.
Given /^the min_num_ratings has been set to (\d+)$/ do |min|
Rating.min_num_ratings = min.to_i
end
Then, I made use of the step and the annotation in my tests:
@afterwards_reset_min_num_ratings
Scenario: users can vote, and their vote affects the average
Given there is only one book
And the min_num_ratings has been set to 2
And a user has already given the book 2.0 stars
And I am logged in as a user
When I follow "Moby Dick"
And I choose by xpath "//input[@name='rating' and @value='1.0']"
And I press "Rate"
Then I should see "Moby Dick"
And I should see "Thanks for rating!"
And the CSS class for the flash message should be "success"
And I should see an average rating of 1.5 stars
Last of all, I created a new file, features/support/hooks.rb, with the following:
# This will only run after scenarios tagged with
# @afterwards_reset_min_num_ratings.
After("@afterwards_reset_min_num_ratings") do
Rating.min_num_ratings = Rating::DEFAULT_MIN_NUM_RATINGS
end
I know this is a bit esoteric, but this same approach can be used anytime you have just a few tests that require test-specific configuration.

Tuesday, April 13, 2010

ActionScript: Dynamic Streaming

There's an article here about dynamic streaming in Flash Media Server 3.5 using Flash 10's new QoS features. It's a three-part series, and I just spent several hours reading it.

It talks about a class that Adobe provides called "DynamicStream". It doesn't come with ActionScript 3.0--it's a separate download. I was hoping to integrate this class into JW Player in order to do more intelligent bitrate switching using Adobe's new QoS features.

Let me explain why I'm frustrated.

First of all, this article was a three-part series, each with several pages. Hence, it was very long--much longer than I think was necessary. Part 3 covered "Integrating dynamic streaming with existing video players". Since it was the final page in the entire series, I expected it to be very exciting and useful. Instead, it was very mediocre. I can summarize, "Hey, this stuff doesn't work in Flash 9, so you'll have to do something else to support older players. Here, read this long code sample to learn how to parse the Flash player version number." There's a lot of code like:
private function _verifyServerVersion( p_version:String ):void
{
var fmsVersion:Number = Number( p_version.split(",", 2).join(".") );

trace("fmsVersion: " + fmsVersion);

if ( fmsVersion >= _targetFMSVersion )
{
_isServerCapable = true;
}
else
{
_isServerCapable = false;
}
}
At the very least, that if/else could have been written:
_isServerCapable = (fmsVersion >= _targetFMSVersion);
However, I think it's a flaw in ActionScript that you have to parse the version numbers manually at all.

Next up, I looked at the DynamicStream class, and it had multiple different coding styles:
if(...) 
{
...
if(...){
...
if (...) {
...
if(...) { ... }
When I see stuff like that, I lose my confidence in the code. The file is 1049 lines long, by the way.

Having said all that, what really frustrates me is that the code doesn't take into consideration the widths for the various encodings. Currently, JW Player will pick the best encoding based on width and bitrate. That way, you can have different encodings for different size players. This is important for several reasons. First of all, there's no use pulling down an HD video for a 480x320 player until the user goes full-screen. It eats up bits that could be going to quality instead of size. Furthermore, it's better to avoid scaling on the client since it looks better to scale only once during encoding.

Hence, I'm more than a little disappointed by the DynamicStream class, and I'm stuck wondering how the heck I'm going to pull the useful bits out of this 1049 line file in order to integrate them into JW Player. Ugh, painful.

Oh, did I mention Silverlight takes care of bitrate switching entirely automatically? Double ugh.