Gate#getBlockBase now uses the nearest matching block

[BlvckBytes]

As I've already stated in the CraftBook Discord channel, the Gate mechanic often selects blocks which completely go against the intuition of the user, simply due to its search-algorithm, which doesn't consider distance at all. Expanding radially outwards from the point of the sign to me seems like the only right way of doing it.

[me4502]

Thanks for this!

Have you done any testing to assess how this performs to the old method? A simple loop was chosen here for speed, as some servers have tens of thousands of gates controlled by redstone/etc. My assumption would be that this will have substantially worse cache locality / predictions, but will likely find the gate block earlier than the previous one for most cases.

[me4502]

Would you be able to please make sure that import style is retained & follows our style guide?

[me4502]

Could you please ensure that the if statements/etc here keep the braces and follow the style guide?

[BlvckBytes]

Have you done any testing to assess how this performs to the old method? A simple loop was chosen here for speed, as some servers have tens of thousands of gates controlled by redstone/etc. My assumption would be that this will have substantially worse cache locality / predictions, but will likely find the gate block earlier than the previous one for most cases.

While I didn't conduct any benchmarks, I have a hard time imagining that this implementation will affect performance in a way that is even noteworthy in real-world scenarios. We're not talking about pipes here, where thousands of blocks are enumerated (possibly) every tick, but rather concise gates with a modest search-radius that may be toggled every so often. Even if there are thousands of them, the overall CPU-time is so minuscule that I wouldn't worry about, personally speaking. This is definitely one of the cases where algorithmic correctness (finding the closest block, as the player expects) is by far outweighing the tiny performance-penalty we take for doing so. Also, are we even sure that cache-hits are increased by walking in an axis-aligned way? I, for one, am not sure if that makes all the difference in the first place.

The only concern I had was the needless allocations of hashing blocks (#getLocation()) in the visited-set, which could use something analogous to my compact-ids, but then again, does it really matter? This version is certainly easier to read and comprehend.

[me4502]

There's just this if statement missing braces if you'd be able to please fix it

[me4502]

"Does it matter?" mostly just depends on whether something is going to result in another issue from someone stating that Gates have gotten significantly slower. Overall I feel this is probably fine at least for now, I was mostly just asking if you'd done much performance testing around it.