What I like about these projects is that you can make them as simple or as complicated as you want.

Or more. Actually some of them seem inverted. Simple compiler takes 3 months but a GameBoy emulator takes 2 weeks? That’s not my experience at all.

I was wondering if that was time spent toying with the thing, not necessarily time to a complete or even functional result.

Yes, exactly. The point is to make a toy that proves the concept and gives you insight into the overall design, not to build something with polish!

IMO the difficulties are a bit off, if a 2D physics engine is 5/10, then orbital simulation cannot be 10/10 (even with the difficult stuff like J2 perturbation) ;)

The 10/10 was a copy & paste typo, sorry! Amended now.

Also what's up with this 1 week to make a hash map? In C it's only slightly harder than creating a growable array.

And a POSIX shell is only 3 days. It's an interesting list but the difficulty/time elements are all over the map.

I wrote a shell that's designed to be good enough for the things likely to be passed to system() and popen(). It probably took me a week. To make it a non-toy POSIX shell, I could see potentially taking me a year. The shell is nuts. There's a reason why Thompson got replaced with Bourne and Bourne got replaced with Almquist, and most people use Bash, and then a lot of people started using Zsh. It's simple enough to attract greenfield implementors, and complicated enough for them all to get improved upon.

For some, it depends on the definition of “toy”.