After expressing my feelings about the OSD700 course, and all of the experiences that I went through this term, I thought, "what a good way to end the term..." But then I remembered that I have to speak about what I managed to get in for release 3.0!
So, I have one PR that took a while, despite being a single line change, and I would like to go into extreme detail, because it ends up in a really interesting lesson for other people to learn.
So, back in release 2.9, I developed a simple "star field", which is a simple animation showing the GitHub contributors as profile pictures. I got the initial code and idea thanks to the Coding Train :D
One weird quirk that the star field had is that the profile pictures would "flicker". For some reason, a picture would get small, then appear super big, and then get small again, acting like normal. It is somewhat difficult to describe, but this would give a flickering effect that was very annoying and somewhat horrible to look at...
We did what a responsible developer would do: file an issue and leave it for later :)
After a while, the issue was picked up by @jerryhue. He mentioned something about rewriting the math logic so that it would fix the flickering issue.
I wasn't super sure what that would mean, because the logic itself was fine, it was just a weird flickering that was happening. I thought to myself, "how is that related to the flickering?"
After a loooooong while, the team decided to assign me the issue in a meeting. After the meeting ended, I asked @jerryhue why he couldn't solve it. He told me that it was difficult for him to rewrite the math logic since he was struggling to think of an easier approach to do the star field simulation.
I wanted to know why he was so obssesed on rewriting the logic. He thought that the current logic was fine, but that it could be improved. One reason for this improvement is to make the stars a little more "believable", and the zoom effect would look better.
I understood his reasons, so I decided to improve the code to achieve something like that.
Of course, I didn't manage to do much...
I was back at square one. I was feeling frustrated. "Why I cannot solve this?", "why is this fix so difficult to happen?" I was almost going to give up and give it to someone else, however, I decided to try one last time.
If I was going to debug it properly, I wanted to write it in the p5js debugger, since I just wanted to focus on that specific part, and I didn't want to wait for Nextjs to compile every time I made a small change.
When I was trying it out, I didn't want to include the images at first, because I had to make an array of images urls that p5.js and I was feeling a little lazy, so I just went with an array of circles that will be drawn on the screen.
The only change I did was the circles being drawn instead of GitHub profile images, everything in the code was the same, and to my surprise, the flickering stopped.
I tried to understand the situation: how come the flickering occurs only with images and not with regular circles? At that moment, I decided to load a single image that could be reused for several stars and understand why this was the case. As expected, the flickering started to happen when I drew images.
I was utterly confused, since the behaviour of an image was clearly different of that from a circle. So, now that I knew that it would only happen with images only, I had to understand when that flicker would occur.
I wrote some code to make note of the values, I noticed something very consistent: the flickering wouldn't start at the end of a star reaching the maximum level, but right at the very start, when the size of star was 0.
When I thought about this, I was still confused, why at that moment? To answer some of my questions I went to the documentation to read upon the behaviour of the
image function, the function that draws images on a canvas. Nothing much was mentioned when the width and height of an image was 0.
Since nothing was documented, I said to myself, "it wouldn't hurt to try in the live examples available", and so I did. I wrote the size of the image to be 0, and all my doubts started to disappear. If you try to write an image to a canva with dimensions 0, it would use the original dimensions of the loaded image!
So, that meant, if I wanted to avoid the flickering, I had to stop making the size to be zero, and instead to be something somewhat close enough. This line came out as a result:
this.z = p5.random(p5.width * 0.1, p5.width * 0.95);
Instead of getting a value that might be zero, I will instead a value that is never zero, and thus, get rid of the flickering once and for all!
To be honest, this helped to learn a huge lesson. Sometimes it is harder to create a solution with the tools that you have rather than create your own tools to create your solution, however, at the same time, it is valuable to not have to create anything from scratch or rewrite everything. Even though it would have been nice to rewrite the math logic so that it looks better, I prefer this solution much better!