Differential D1160
Use correct distance for determining nearest dropsite Authored by temple on Dec 20 2017, 5:00 PM.
Details
The nearest dropsite is calculated using the center of the dropsite (in red).  With the patch it will instead use the nearest point on the dropsite (in green). (Strictly speaking, units have obstructions and ranges, so the new calculation isn't completely correct, but it's close enough.) We can quit the loop if we know that the distance to later dropsites will never be better than the current best. Since the range manager sorts dropsites by the distance to their centers, all we need to do is figure out how far away the nearest point on an obstruction can be from the center of the obstruction. Civic centers are at most 40m on a side, so a corner can't be farther than 40/sqrt(2) = 28m away from the center, but in the patch I use 40m to be safe. I don't think performance will be an issue, but I should probably test that.
Diff Detail
Event TimelineComment Actions Successful build - Chance fights ever on the side of the prudent. Updating workspaces... Build (release)... Build (debug)... Running release tests... Running cxxtest tests (308 tests)....................................................................................................................................................................................................................................................................................................................OK! Running debug tests... Running cxxtest tests (308 tests)....................................................................................................................................................................................................................................................................................................................OK! Checking XML files... Comment Actions A bit out of context, but as you did quite some patches with resourceDropsites lately, and are familiar with the hierarchicalPathfinder, some checks on dropsite accessibility for a next patch would be nice :) Comment Actions Seems like a good idea, agree with mimo that accessibility ought to be checked (perhaps another diff though indeed). I'll give this a look over the WE, thanks for the patch
Comment Actions I did think about how the walking distance to a dropsite might be different than the straight line distance, for example if a wall was in the way. But that seems too complicated to deal with. Looking at global regions in the hierarchical pathfinder (once D53 is done) might be possible, although it's not trivial since buildings themselves aren't in a region (see FindGoalRegions). Oh, I guess you might be able to use the terrain-only passability classes (which might be what you're more familiar with, with the AI), but then "islands" created by trees or walls wouldn't count.
That seems pretty crazy to me, that units could be gathering resources so far away from a dropsite that there's a nearer one on another island. Why not just build a new dropsite closer to the resource?
Comment Actions Indeed the idea by players and the design of the pathfinder is that units always chose the shortest path if they need a path.
Unlikely, but surely possible and if it happens, the player would certainly consider it a bug if the units clinch at the shoreline.
That means that the bug is recoverable, but it's still considered a bug IMO Comment Actions yes, i was thinking to the the terrain-only passability classes (as used by the AI). I agree that would not cover wall cases, but still is worth to have (but it was just a proposition of course if you were looking for new patches).
I had seen it several times in petra (and added a protection for it in the ai) when islands were very nearby. That's still a bug: the player should be allowed to gather far from its dropsite (even if not the best strategy) without having its units idling on the shoreline. There are cases where UnitAI make gatherers drift far away from the dropsite, or you may miss wood to build a new dropsite, or whatever could happen. Comment Actions (Although they don't actually go idle. If they did then it'd be easier spot them and take corrective action by building a new dropsite. I played a game last night where a similar thing happened, units getting stuck trying to gather from trees on a hill.  It would be nice if for a23 we could get a split of D13 to detect when this happens and tell units to abort the order. I've looked at this before, maybe I'll try to spend some more time on it.)
Sure. (I count.. 42 patches at the moment, plus others I've worked on but haven't submitted. We need more reviewers. :) ) Comment Actions Patch reads correct and ideal (and is deseparately needed).
Comment Actions and you didn't post a one-sentence ticket with a replayfile </3 refs #4902
(or chose ways to become more economical with the existing reviewers).
Comment Actions (thx linting)
Comment Actions Successful build - Chance fights ever on the side of the prudent. Updating workspaces... Build (release)... Build (debug)... Running release tests... Running cxxtest tests (308 tests)....................................................................................................................................................................................................................................................................................................................OK! Running debug tests... Running cxxtest tests (308 tests)....................................................................................................................................................................................................................................................................................................................OK! Checking XML files... Comment Actions Some naming refinements as inline comments, otherwise looks good. Thanks for the patch!
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||