MinchinWeb's MetaLibrary v.9 [2015-01-10]

[R2dical]

Can you show me what tile 0x7F6F looks like? (Or alternately upload your test map here.) Did this tile used to be land and then was terraformed to be water?

Looked through the autosaves, but unfortunately it was overwritten by subsequent test runs. I was tweaking my pathfinder and doing random starts. I can say that it is extremely unlikely that 0x7F6F (32623?) was changed since map gen. Industry changes off, no DoCommands for AI and I was only watching the log, no landscaping etc. I usually save the seed on crashes but didn't in these tests If it would help attached are the mapgen settings. I will continue testing with the new version, thnx for fix.

So extrapolating, a 6,000 tile path could take 14k-422k ticks (Lakes) + 7k-77k ticks (or more!) (pathfinding) ... in any case, many years (a game day is about 74 ticks, so a year is about 27k ticks). I keep working on the pathfinder so I hope to bring both those numbers down.

Yeah maybe good the crash interrupted this, I was sitting waiting patiently for it to finish... (FYI looks like that particular path, as returned by my pathfinder, was 600 not 6000 as even with step of 10 it returns a GetLength of ~ the actual...weird)

On a practical level, do you plan on running ships along 6,000 tile routes? Once your ships take more than ~4 months to travel the route (one way), it gets much more complicated to determine whether a route is making you money...

Well the thing is you(or your AI) don't know at runtime what distance the end path will be (or if one is even possible, i.e the same water body). Sure in most practical cases an AI will say ignore candidates for a shipping route > 500 tiles manhattan apart, but if your tiles are 499 apart there is always the (likely) possibility of a very circular path existing which ends up at say 5000 tiles. Sure, then you can limit pathfinder by a max_cost factor (or equivalent max_length, say 750), but then this will discount perfectly feasible paths on other searches between somewhat closer points. Also an AI may decide to limit ship route length by est travel time, so a function of ship speed, and end game long paths may be perfectly usable.

[MinchinWeb]

Version 8 is released! Available at the top or on Bananas.

This update brings a speed improvement to Lakes, allowing it to run 9-30x faster than v7.

[MinchinWeb]

Version 9 is released! The changes here are in the backend - updating the Python build script so that the GameScript version imports the correct libraries and make the build script work with Python 3.

Available at the top or on Bananas.

[Kogut]

main.nut has

Code: Select all

 *			Import("util.MinchinWeb", "MinchinWeb", 5);

Code: Select all

 *			import("util.MinchinWeb", "MinchinWeb", 9);

would be better (new version, I instead of i results in a crash).

[MinchinWeb]

@Kogut: main.nut would be part of an AI. Which AI are you looking at?

[Kogut]

@Kogut: main.nut would be part of an AI. Which AI are you looking at?

There is a main.nut in Metalibrary.

At least there is one in downloaded archive - I attached what gets downloaded from BaNaNaS.

http://minchin.ca/openttd-metalibrary/index.html#faq is also affected

[MinchinWeb]

@Kugot: I think the documentation you point to has been updated. Let me know if I've missed something on the site. As for main.nut, that code is actually the source of the documentation, and so isn't used directly. I've updated the files post-release, but haven't put out a release with just documentation fixes. Do you think it's worthwhile to put out a documentation-fixes-only release?