A prematurely approved purple kata creates an unfortunate situation. Any chance you can work on the remaining issues, and also the open suggestions below?

@eurydice5717 if you know how to improve them without changing the essence of the task too much, I’m totally fine with that. Honestly, I agree that this kata turned out to have way too many issues. I kept trying to figure out how to solve them, but after 3 years, the best I could do was just close my eyes and pretend it didn’t exist. Then I just stopped liking the whole idea of the kata because I thought I had created something interesting, but it ended up being pretty mediocre. So, I came to terms with the fact that it couldn’t be fixed and wouldn’t pass review. That’s why I was surprised when I saw an email a couple of days ago saying someone had approved it (thank you for that btw)

Edit: Comment removed

for anyone solving this currently and confused by this edgecase: The solver appears to expect the output array to grow by at least 2 on each iteration, so if your result is an empty array after trimming leading and trailing zeroes you can create a new array of list.length + 2 and fill it with zeroes.

Thanks @geoffp - after attempting to read and reread this kata and going slightly mad thinking I couldn't understand it, you've explained it in such a nicer way.

what a saint

The examples should be an image or ascii/unicode visualisation instead, so that the solver does not need to wonder whether it contains hidden information in its text, and to make it easier to parse in general.

All the rest of the instructions below the rules table should be replaced with an explanation of what part of the data to return, that current text is difficult to read and says nothing.

The wikipedia mention should instead be a hyperlink for the first two words "Rule 30" so as to not make the solver question whether there is required information there that the instructions leave out. Also improves signal-to-noise ratio.

I have updated the encoding to reduce the nesting by another layer and included much harder problems. Up to 15 x 15 now!

oh

my solution accounts for big and little endian platforms (tested on PowerPC and x86_64) but it's hard to tackle the middle-endian shenanigans ^^

Raised as an issue.

At the same time, the input arg phi means the azimuth at sunrise

What is implicating this?

I'm still not sure, what is the problem.

As much as I understand, geoffp means that after passing through the overhead point (zenith), actual azimuth of the Sun changes by 180 degrees. For example, if the azimuth is 45deg before the noon, it becomes 225 deg after the noon. At the same time, the input arg phi means the azimuth at sunrise, and not at current clock.