I just realized that building the test case for the "chain delete" code would involve something that would look very similar to the "move & merge chains" code that I still have to write, so I decided to go ahead and write that instead. This is one of those cases where the test case involves code of the same nature, complexity and size as the code required to operate it. To put it another way: The "move & merge chains" code will be the test case for the "delete chains" code.
I started with "delete chains" because it gave insight into which data structures were required to provide for efficient delta-updates, without the need to actually maintain those data structures yet. As the "delete chains" code evolved, so did the representation of those data structures.
Before making "move & merge chains", a decision has to be made on how to represent chains. I decided to maintain "empty" chains as well, and that requires code to efficiently determine whether a move splits an empty chain, and if so, to actually split it. There are a plethora of ways to do it. I read the available literature and decided upon one of the proposed implementations. The goal is speed. Everything else is secondary. I will have a special case for ladders but the rest of all tactical analysis will forever build upon the code I'm writing now.
