A Walnut's Thoughts

Partial borrows across function boundaries don't really work in Rust. Unfortunately, that's kind of a major issue. There are workarounds, some are outlined here, but all of them come with pretty major drawbacks.

Oh how naive I was when I thought I'd be moving on to pathfinding and abilities. I had hoped some of the game's mechanics would function better in isolation, so I could gradually add complexity on top of the tracer bullet in a modular way. Unfortunately, that doesn't seem possible without a huge refactoring burden every time I add the next layer. It'll be more worth my time to properly architect it right now, and build it from the ground up with all of the systems it'll need.

A while back, I made a small application meant to demonstrate some important principles about RTS design, namely that unit stats in a vacuum can be very misleading. As a quick example, stalkers have ~9.7 dps on paper. When fighting marines though, their effective DPS drops to ~8.4 (8.2 with combat shields) due to overkill. That's about the same DPS as a sentry. Unfortunately, there's only so much info that can be extracted from a Time To Kill calculation between two units.

I encountered a neat example recently while solving Advent of Code 2015, day 6. The problem can be boiled down to "There is a 2D array of values. Given a range and instruction, apply the instruction to all values in that range. How many values are 'on' at the end?". There are only 2 possible states for each value: on and off, and only 3 possible instructions: on, off, and toggle. I won't focus too much on parsing the input or the structure of the algorithm here, I just want to investigate the hot loop which applies the instruction to each value.