IRC logs for #openttd on OFTC at 2018-09-26
⏴ go to previous day
00:00:19 *** Extrems has joined #openttd
00:02:49 <m3henry> And It'll definitely be wanting c++11 support
00:02:54 <m3henry> So I'm happy about this :3
00:17:51 *** Thedarkb-T60 has joined #openttd
00:18:50 *** Gustavo6056 has joined #openttd
00:21:12 *** ChanServ sets mode: +v tokai
00:25:55 *** Gustavo6056 is now known as Gustavo6046
00:36:49 <DorpsGek> Samu: 0.944881889764
00:42:40 *** Extrems` has joined #openttd
00:44:42 *** Extrems` is now known as Extrems
01:16:22 <DorpsGek> Samu: 0.925042589438
01:17:54 <Samu> @calc (956+819+804+779+761)/(1031+870+853+825+799)
01:17:54 <DorpsGek> Samu: 0.940840566469
01:18:35 <DorpsGek> Samu: 0.926114649682
01:18:57 *** Gustavo6056 has joined #openttd
01:19:12 *** haudrauf has joined #openttd
01:20:28 <Samu> real gains not even 10%, sad
01:25:52 *** Gustavo6056 is now known as Gustavo6046
01:32:45 *** haudrauf has joined #openttd
01:34:09 *** snail_UES_ has joined #openttd
02:08:20 <Eddi|zuHause> c++11 is only 7 years old, how can you demand support for that?
02:19:04 *** Gustavo6056 has joined #openttd
02:26:07 *** Gustavo6056 is now known as Gustavo6046
03:19:14 *** Gustavo6056 has joined #openttd
03:20:36 *** WWacko1976-work has joined #openttd
03:24:48 *** Gustavo6056 is now known as Gustavo6046
03:58:44 <Samu> what's the fastest priority queue for graph aystar?
04:00:36 <Samu> and when i mean fastest, it's the one with less op codes needed
04:19:20 *** Gustavo6056 has joined #openttd
04:19:42 <Samu> fibonacci heap vs binary heap
04:20:18 <Samu> fibonacci was faster than binary, but the routes are different
04:20:47 <Samu> all costs are the same on both
04:26:37 *** Gustavo6056 is now known as Gustavo6046
05:19:26 *** Gustavo6056 has joined #openttd
05:26:42 *** Gustavo6056 is now known as Gustavo6046
05:43:02 *** haudrauf has joined #openttd
06:06:19 *** Gustavo6046 has joined #openttd
06:06:19 *** Extrems has joined #openttd
06:06:20 *** Tharbakim has joined #openttd
06:06:20 *** innocenat has joined #openttd
06:06:20 *** ccfreak2k has joined #openttd
06:06:20 *** rocky1138 has joined #openttd
06:06:20 *** reldred has joined #openttd
06:06:20 *** HeyCitizen has joined #openttd
06:06:20 *** Maarten has joined #openttd
06:06:20 *** lethosor_ has joined #openttd
06:06:20 *** TinoDidriksen has joined #openttd
06:06:20 *** wodencafe has joined #openttd
06:06:20 *** Antheus has joined #openttd
06:06:20 *** mindlesstux has joined #openttd
06:06:20 *** Smedles has joined #openttd
06:19:43 *** Gustavo6056 has joined #openttd
06:26:42 *** Gustavo6056 is now known as Gustavo6046
06:48:26 *** Supercheese has joined #openttd
07:06:39 *** DorpsGek has joined #openttd
07:06:39 *** ChanServ sets mode: +o DorpsGek
07:19:25 *** Gustavo6056 has joined #openttd
07:19:32 *** ChanServ sets mode: +v tokai
07:25:36 *** dihedral has joined #openttd
07:25:42 *** Gustavo6056 is now known as Gustavo6046
07:34:12 *** grossing has joined #openttd
08:17:24 *** luaduck has joined #openttd
08:19:53 *** Gustavo6056 has joined #openttd
08:24:17 *** Gustavo6056 is now known as Gustavo6046
08:25:43 *** mikegrb has joined #openttd
08:27:14 *** sla_ro|master has joined #openttd
09:19:55 *** Gustavo6056 has joined #openttd
09:23:52 *** Gustavo6056 is now known as Gustavo6046
10:19:52 *** Gustavo6056 has joined #openttd
10:27:27 *** Gustavo6056 is now known as Gustavo6046
11:14:29 <crem4> Did TTD take a sawmill sound from beginning of Billy Joel's "Allentown" song?
11:18:03 <peter1138> Stock audio, probably.
11:20:18 *** Gustavo6056 has joined #openttd
11:24:37 *** Gustavo6056 is now known as Gustavo6046
11:35:30 <LordAro> i hear AoE2's horse sound so often
11:53:50 <Samu> why does fibonacci heap and binary heap produce different results?
11:53:59 <Samu> is that supposed to happen?
11:55:30 <Samu> take a look at the 2 images and the roads built on both, they're different
12:07:10 <Samu> just tried priority queue, it's utterly slow
12:07:37 <Samu> produced the same result as fibonacci
12:07:44 <Samu> does that mean it's binary heap that's bugged?
12:07:54 <Samu> or it's not a bug at all?
12:20:35 *** Gustavo6056 has joined #openttd
12:26:59 *** m3henry has joined #openttd
12:27:37 *** Gustavo6056 is now known as Gustavo6046
12:52:38 *** Hazzard has joined #openttd
12:53:10 *** Webster has joined #openttd
12:53:55 *** Terkhen has joined #openttd
12:53:55 *** ChanServ sets mode: +o Terkhen
12:54:25 *** planetmaker has joined #openttd
12:54:25 *** ChanServ sets mode: +o planetmaker
12:55:25 *** XeryusTC has joined #openttd
12:55:55 *** fonsinchen has joined #openttd
13:16:40 *** peter1138 has joined #openttd
13:16:40 *** ChanServ sets mode: +o peter1138
13:20:35 *** Gustavo6056 has joined #openttd
13:24:44 *** snail_UES_ has joined #openttd
13:24:52 *** Gustavo6056 is now known as Gustavo6046
13:30:12 *** thavinci has joined #openttd
14:20:29 *** Gustavo6056 has joined #openttd
14:24:22 *** Gustavo6056 is now known as Gustavo6046
14:41:27 *** Stimrol has joined #openttd
15:02:58 *** Gustavo6046 has joined #openttd
15:05:13 *** techmagus has joined #openttd
15:07:37 *** tokai|noir has joined #openttd
15:07:37 *** ChanServ sets mode: +v tokai|noir
15:20:37 *** Gustavo6056 has joined #openttd
15:27:52 *** Gustavo6056 is now known as Gustavo6046
16:20:51 *** Gustavo6056 has joined #openttd
16:21:23 *** sla_ro|master has joined #openttd
16:23:47 <Samu> are heaps supposed to give different results?
16:25:57 *** Gustavo6056 is now known as Gustavo6046
16:46:16 *** andythenorth has joined #openttd
16:53:47 <Samu> andythenorth: do u understand fibonacci heap and binary heap used in aystar?
16:57:40 <LordAro> Samu: asking anyone who says anything is bordering on harrassment
16:58:29 <Samu> google doesn't help here
16:59:25 *** V453000 has joined #openttd
16:59:28 <Samu> it's very specific to my observed results
17:00:09 <Samu> aren't priority queues supposed to deliver the same results?
17:02:52 <andythenorth> I am quitting the channel for a bit
17:02:59 <andythenorth> getting asked random questions about stuff
17:03:03 *** andythenorth has left #openttd
17:03:43 <Eddi|zuHause> that's why i have an ignore list
17:04:13 <Eddi|zuHause> with about 2 entries on it
17:04:25 <Samu> don't have where else to ask abouth this kind of things
17:04:51 <nielsm> find a professor of computer science, or research things yourself
17:07:31 <LordAro> Samu: first step would be working out what a heap is
17:07:39 <LordAro> wikipedia's pretty good at this sort of stuff
17:07:40 <nielsm> asking why two variants of a complex algorithm produce different results is a really difficult question and will likely require hours of research
17:08:06 <nielsm> that is, for someone already well acquianted with the concepts involved
17:08:47 <nielsm> for someone who hasn't learned about the algorithms and data structures it's built upon it'll take days or weeks to get up to speed
17:10:25 *** Wormnest has joined #openttd
17:12:25 *** Alberth has joined #openttd
17:12:25 *** ChanServ sets mode: +o Alberth
17:20:53 *** Gustavo6056 has joined #openttd
17:22:05 *** rocky113844 has joined #openttd
17:24:57 *** Gustavo6056 has joined #openttd
17:25:44 *** WWacko1976-work has quit IRC
17:27:27 *** Gustavo6056 is now known as Gustavo6046
17:31:22 <Samu> i think im gonna investigate this on my own
17:31:50 <Samu> in reading valgrinds and stack and heaps, nothing to do with what I'm looking for
17:34:12 <Samu> i used Queue.PriorityQueue-2 and produced the same result as Queue.FibonacciHeap-3
17:34:39 <Samu> only Queue.BinaryHeap-1 is different
17:35:09 <Samu> but that's what Graph.AyStar-6 is using, I must understand why
17:35:31 <FLHerne> Samu: The problem is that most of what you write here is just meaningless to anyone else without telepathy
17:36:29 <FLHerne> The sort of implementation-detail questions you're asking only really make sense in the context of how you're thinking about a specific bit of code
17:36:46 <FLHerne> But no-one else can know what you're looking at, nor what you're thinking about it
17:37:16 <FLHerne> I mean, yes I know about different kinds of heap, and yes I know about A*
17:37:18 <crem4> If it's a heap with standard opeartions, and there are no equal elements, any heap implementation will have the same results.
17:37:39 <crem4> If there are equal elements, they can be returned from heap in arbitrary order.
17:37:56 <FLHerne> But understanding why /you/ care about using different kinds of heap, in A*, right now, and what your problems with that are would take a whole conversation that I don't have time for
17:38:54 <FLHerne> ^ditto for the images
17:39:05 <FLHerne> I can see you have images with slightly-different debug output
17:39:16 <Samu> the routes are different in themselves
17:39:26 <Samu> and the only thing I changed was the queue
17:39:31 <FLHerne> I don't know what produces that output, nor what it means, nor why it (presumably) concerns you
17:40:29 <FLHerne> Then if they have different costs, your code is wrong somehow :P
17:41:03 <Samu> nop, the costs are the same, only difference was queue class
17:41:35 <Samu> towns didn't grow, didn't build roads, i have that disabled
17:41:50 <FLHerne> Either the heap sorting itself, or you have some error in how you take items from the queue
17:42:04 <FLHerne> It /says right in your images/ that the route costs were different
17:42:24 <Samu> that's not the pathfinder cost
17:42:52 <FLHerne> Ok, so back to "you don't provide enough context for anyone to provide meaningful input"
17:43:08 <FLHerne> I said that, and tried to ignore it anyway, and it didn't work :P
17:43:34 <Samu> ok, gonna try get the pathfinder costs displayed
17:44:19 <FLHerne> Trying to get input on that sort of debugging over a text channel /is just a waste of people's time/
17:44:38 <FLHerne> You need to share more state with the other person than anyone can reasonably type
17:44:57 <FLHerne> (and end up spamming the channel, like you have been for weeks)
17:45:30 <FLHerne> General design decisions, choice of algorithm, maybe
17:46:12 <FLHerne> But "my code doesn't work, why?!" isn't something it's worth discussing except with someone else looking at the same code and ideally in the same room
17:51:41 <Eddi|zuHause> weeks? my ignore list entry is from years ago... nothing improved...
17:53:33 <Samu> [16:37] <crem4> If there are equal elements, they can be returned from heap in arbitrary order.
17:53:45 <Samu> should be something like this that's happening, have to investigate
17:55:16 *** Progman has joined #openttd
17:56:40 <FLHerne> Samu: If the final pathfinder costs /are/ identical, there's no problem
17:57:17 <FLHerne> A* is guaranteed to return /an/ optimal path; if there's more than one of those it can return any of them based on anything
17:58:07 <FLHerne> (which is why the ship pathfinder is so slow, the big equal-cost areas of water give it far too many options)
18:01:45 *** sla_ro|master has joined #openttd
18:03:11 <Samu> Queue.PriorityQueue-2: ticks: 28016, pathfinder cost: 9120
18:04:17 <Samu> Queue.FibonacciHeap-3: ticks: 524, pathfinder cost: 9120
18:04:48 <Alberth> if equal, prefer smallest estimate (ie generally closer to the end-point)
18:05:30 <Samu> Queue.BinaryHeap-1-1: ticks: 544, pathfinder cost: 9120
18:12:36 <FLHerne> Ok, so there's no problem
18:12:48 <FLHerne> All the routes are equally valid
18:13:07 <Samu> min_cost = min(_AIMap.DistanceManhattan(cur_tile, tile) * self._cost_tile, min_cost);
18:13:56 <Samu> foreach (tile in goal_tiles) {
18:14:12 <Samu> there's only 1 goal tile
18:14:44 <Alberth> ah, fair enough, that's the 'min' about
18:17:32 <Samu> gonna plant signs, i wanna see how different they behave
18:20:39 <nielsm> I'm quite sure you're into entirely acememic exercise territory now. if your goal is to understand the algorithms, go ahead, but if your goal is optimization you're better off just picking one based on measurements and try to find other ways to fast-eliminate potential route segments
18:20:50 <nielsm> (i.e. experiment with heuristics)
18:27:18 *** Thedarkb-T60 has joined #openttd
18:37:41 *** TheMask96 has joined #openttd
18:39:57 *** Alkel_U3 has joined #openttd
18:43:59 *** IgnoredAmbience has joined #openttd
18:53:53 <Samu> those are the open neighbours
18:56:06 <Samu> seems that fibonacci is better for my purposes :p
18:57:34 <Samu> out of curiosity, will check Queue.PriorityQueue-2
18:57:49 <Samu> how this one builds them signs
18:59:52 *** Thedarkb-T60 has joined #openttd
19:01:23 <Alberth> you tried it in the other direction as well?
19:12:14 <Samu> heh, it took so long that industries closed, it affected outcome
19:12:44 <Samu> but it seems to match fibonacci
19:12:58 <Samu> except that part closer to the industry that went missing
19:14:26 <Samu> steel mill gave place to a forest
19:14:35 <Wolf01> Does somebody remember a game like SEGA Golden Axe but with a warrior, a ranger and a wizard? I played it on a coin-op like 25 years ago and I forgot the name, every time I see Golden Axe on my library I remember that one
19:15:15 <peter1138> I remember Golden Axe.
19:16:51 *** greeter has joined #openttd
19:21:51 <Samu> in sum, what I learned, binary heap ordering of equal elements is different than the other two, and while it's not wrong, it's still as mismatch :(
19:23:09 <Samu> is it *ahem* easy to fix?
19:26:07 *** HerzogDeXtEr has joined #openttd
19:27:22 <Alberth> give less room, consider tiles closer to the target (lower estimate) better if the total cost is the same
19:30:15 <nielsm> for building roads, I think giving penalty to turns, slopes, and directions away from the final destination, would be a start
19:30:22 <nielsm> but I don't know what you have now
19:30:43 <Samu> i could copy paste my code
19:31:06 <nielsm> also perhaps outright rejecting tiles outside a rectangle bounded by the source and destination plus some border space
19:31:46 <nielsm> and if finding a valid route with that bounding box fails, increase the allowed border space and try again
19:32:24 <nielsm> source and destination being the full route endpoints, not any partial route
19:40:55 *** andythenorth has joined #openttd
19:43:00 <Samu> it's something to do with the estimator, right
19:43:17 <nielsm> the neighbors function I think
19:43:36 <Samu> make the estimator return some incredibly high cost if going outside that rectangle
19:43:49 <nielsm> that's also a possibility
19:44:11 <nielsm> adding to the cost the further away from the ideal bounding box you go
19:47:55 *** gelignite has joined #openttd
19:56:03 <Samu> oops, forget distancesquare
20:00:47 <andythenorth> hello planetmaker :D
20:04:17 *** Supercheese has joined #openttd
20:22:53 *** Thedarkb-T60 has joined #openttd
20:24:14 <Wolf01> <Wolf01> Does somebody remember a game like SEGA Golden Axe but with a warrior, a ranger and a wizard? I played it on a coin-op like 25 years ago and I forgot the name, every time I see Golden Axe on my library I remember that one <- ha! should be "the king of dragons"
20:24:37 <Eddi|zuHause> andythenorth: it's a valid feature request, but it's probably more involved than you think
20:25:10 <andythenorth> my friend had a game with a warrior a ranger and a wizard
20:25:17 <andythenorth> or maybe that was goldenaxe
20:25:57 <Wolf01> Is the one with the green and orange dragon
20:27:38 <andythenorth> Eddi|zuHause: the version I want is probably invalid
20:27:46 <andythenorth> all handled in newgrf, I abuse it for livery tricks
20:28:02 <andythenorth> won't work due to recursion failure?
20:28:41 <Wolf01> And now, for the nostalgia time, I should find out the sega game I played with my cousin, of which I recreated the gun with lego :P
20:28:43 <Eddi|zuHause> it'll work, but only on otherwise symmetric vehicles.
20:28:59 <andythenorth> I think that's close enough to "won't work"
20:32:39 <Eddi|zuHause> ok, well, it's a "bad" feature request in the sense that you're not requesting what you want, but a have a too specific implementation of what you want in mind which doesn't fit with the reality of the game
20:33:33 <andythenorth> I wonder what shift-click does in depot
20:34:32 <Eddi|zuHause> likely the same thing as everywhere else, estimate the cost of whatever a normal click would do (usually 0 won't show up)
20:36:19 *** Wacko1976 has joined #openttd
20:43:37 <andythenorth> do we support alt-click? :P
20:43:51 <andythenorth> all the ideas I've seen for liveries involve pissing around with dropdown menus
20:43:55 <andythenorth> which is terribly indirect
20:44:50 <Eddi|zuHause> alt+click is usually special on linuxoid windowing systems, and doesn't arrive at the program level
20:44:55 <andythenorth> ctrl-click is really direct
20:46:19 <Eddi|zuHause> you could add more options for the "can this vehicle flip" callback to determine what action is actually done on ctrl+click
20:46:39 <Eddi|zuHause> (nothing, flip, cycle livery group, ...?)
20:47:06 <andythenorth> that is an appealing proposal
20:47:25 <andythenorth> I would just modify it to delete 'flip'
20:47:34 <andythenorth> newgrf author has to handle flipping explicitly anyway
20:47:50 <andythenorth> the alleged support doesn't work for non 8/8 vehicles
20:52:00 <andythenorth> nielsm: fancy patching flip? :)
20:53:08 <Eddi|zuHause> andythenorth: that is exactly the point of the callback. the builtin flipping doesn't work for non-8/8 so the callback is there to tell the game "hey, it's safe to flip this non-8/8 because i implemented it myself"
21:04:37 <andythenorth> it has been removed in JGR PP
21:04:40 <andythenorth> so why is it needed?
21:07:53 <andythenorth> hmm the newgrf wiki does say that flipping isn't magic for shorter vehicles
21:08:35 * andythenorth checks the nml docs
21:09:07 <andythenorth> nml docs aren't as explicit
21:34:10 <Samu> the further is goes away from the distancemanhattan(source, goal), the higher the estimate cost
21:34:59 <Samu> that comment is false, don't read it
21:48:27 <Eddi|zuHause> it's not "magic", you just need different sprite offsets, because the offsets are relative to the start of the bounding box, not the center
21:57:50 *** ToBeCloud has joined #openttd
21:57:50 *** ToBeCloud is now known as ToBeFree
21:59:29 <andythenorth> well it could be made magic
22:00:04 <andythenorth> I implemented my own last year so potato / potato
22:00:58 <Alberth> why do you have both multiple sources and multiple goals, samu? seems a bit too much, doesn't it?
22:05:47 <andythenorth> Eddi|zuHause: so what does the spec look like for ctrl-click has multiple interpretations?
22:06:29 <Samu> because it's what graph.Aystar reads
22:06:58 <Samu> i haven't really touched graph.aystar code, except for the priority queue class
22:12:24 *** Progman has joined #openttd
22:18:37 <Eddi|zuHause> andythenorth: needs more looking up what the actual spec is currently before i can answer that
22:27:46 *** HeyCitizen_ has joined #openttd
22:41:26 *** HeyCitizen has joined #openttd
22:47:19 *** Thedarkb1-T60 has joined #openttd
22:49:06 *** Wacko1976_ has joined #openttd
22:51:31 <sushibear> Hello guys! Is OpenTTD being developed in C or C++?
22:53:13 <LordAro> though it was originally C, so there's still quite a lot of C-isms in the codebase
22:55:37 *** Thedarkb-T60 has joined #openttd
23:03:11 <andythenorth> Eddi|zuHause: it might be just another flag :P
23:03:23 <andythenorth> but that seems slightly wrong
23:04:11 <andythenorth> I suspect there's no clean way to change the semantics of bit 3
23:04:15 <andythenorth> and bit 3 can't be removed
23:04:35 <andythenorth> and adding 'just another bit' for livery counting conflicts with bit 3
23:16:30 <andythenorth> can't have 2 separate bits for the behaviour
23:16:43 <andythenorth> needs a bitmask eh? :P
23:17:11 <andythenorth> how about changing semantics of bit 3 to 'ctrl-click can be used in depot'
23:17:40 <andythenorth> and adding a new bit for what var ctrl-click changes: 0 = flip, 1 = counter
23:17:51 <andythenorth> backwards compatible :P
23:24:58 <planetmaker> sushibear, and there's places in the codebase where C probably will loom long (or forever). And there's some objective-C in the OSX support.
23:28:32 *** Wormnest has joined #openttd
23:56:59 <sushibear> Mah dudes, what is FHC?
23:59:35 <LordAro> "failheap-challenge.com"
continue to next day ⏵