2011-03-26T00:00:04 McLeopold: others agreed, i don't mind 2011-03-26T00:00:40 McLeopold: if we're going to go that route it would probably be better for a map to declare its "type" then the engine can decide if a particular food spawning methods fits with that map type 2011-03-26T00:00:40 i think we should just add a parameter to maps indicating what type it is 2011-03-26T00:00:42 they are very much separate concerns, although some combos do fit together nicely 2011-03-26T00:01:11 McLeopold: food generation seems to be closely related to map gen, because of symmetry types, so every different type of symmetry has another kind of food spawn system 2011-03-26T00:01:33 total random works with them all :) 2011-03-26T00:01:48 janzert: but total random is not fair 2011-03-26T00:02:03 especially if we decide we have to move toward symmetric everything 2011-03-26T00:02:25 I think map should pass its type 2011-03-26T00:02:28 but is it unfair in a way that actually ends up mattering in the long run? I know a lot of people here think so, I don't think it's really been determined 2011-03-26T00:02:52 at the same time I don't think the other methods are necessarily worse 2011-03-26T00:03:00 so I'm fine with them too 2011-03-26T00:03:05 janzert: I think later in the game it is not such an issue, but at the start it is 2011-03-26T00:03:05 janzert: depends on what people think 2011-03-26T00:03:20 sigh: start is pretty important 2011-03-26T00:03:34 amstan: wasn't saying otherwise 2011-03-26T00:04:17 and the latest method I've heard for spawning food should work on any map 2011-03-26T00:04:27 any symmetry or asymmetry 2011-03-26T00:04:31 the one antimatroid proposed? 2011-03-26T00:04:39 for asymettric? 2011-03-26T00:04:39 yes, I believe so 2011-03-26T00:04:54 janzert: which one? antimatroid: ping 2011-03-26T00:05:02 * sigh hates spelling symmetry or any variant thereof 2011-03-26T00:05:15 sigh: self-similar? 2011-03-26T00:05:52 variant was the wrong word, hate spelling words that use symmetry as the stem :P 2011-03-26T00:06:13 amstan: the one i came up with when you were here earlier 2011-03-26T00:06:17 amstan: take all the squares a player can reach first and check how long it takes to reach them from the starting position, then randomly pick a "same distance" square for each player to spawn food in 2011-03-26T00:07:14 janzert: how is the distribution chosen? 2011-03-26T00:07:32 i would do just random 2011-03-26T00:07:35 uniform I presume 2011-03-26T00:07:40 it's still not "symmetric" after the first turn 2011-03-26T00:07:44 uniform over what is the question :P 2011-03-26T00:08:01 eg. uniform over distances is not uniform over squares 2011-03-26T00:08:04 over distances and set of squares within the distance would be good enough 2011-03-26T00:08:10 I believe 2011-03-26T00:08:22 that gives preference to squares close to home 2011-03-26T00:08:23 yeah, i don't think that matters too much 2011-03-26T00:08:43 I think that's a nice side benefit ;) 2011-03-26T00:08:47 ok :) 2011-03-26T00:08:49 gives a "base" 2011-03-26T00:08:57 but shh don't tell jeff 2011-03-26T00:09:03 :P 2011-03-26T00:09:15 there is one other corner problem 2011-03-26T00:09:33 that I haven't seen mentioned so far 2011-03-26T00:10:01 ... 2011-03-26T00:10:09 in an asymmetric map one player is likely to have squares it can reach first but are farther than any other player has 2011-03-26T00:10:28 in a naive implementation presumably these squares will never spawn food 2011-03-26T00:11:00 we could get the map analyzer to try and minimise that, but yeah, good point 2011-03-26T00:11:00 there are ways it could be worked around but none very pretty that I've thought of so far 2011-03-26T00:11:04 there are multiple other things that can cause unfairness, but yes 2011-03-26T00:11:31 related to the same sort of issues 2011-03-26T00:11:51 just because the squares are the same distance, doesn't mean they are just as easy to control 2011-03-26T00:11:52 sigh: i don't know what you're going to do about squares that players can reach equally 2011-03-26T00:11:57 i'll let you worry about that :P 2011-03-26T00:12:07 :( 2011-03-26T00:12:32 sigh: probably but it seems I'm less concerned about "fairness" by trying to exactly equalize all game conditions than anyone else 2011-03-26T00:12:37 imo, trying too hard to make asymmetric maps completely fair 2011-03-26T00:12:45 and we should just go with the sections method 2011-03-26T00:12:52 so long as the unequal conditions aren't systematically biased toward any player 2011-03-26T00:12:55 trying too hard to make asymmetric maps completely fair is impossible 2011-03-26T00:13:14 erm? 2011-03-26T00:13:39 that sentence kind of implies there is nothing you could do that is too much in order to make asymmetric maps fair 2011-03-26T00:14:06 I think we are getting too complicated for little benefit 2011-03-26T00:14:29 and sweeping the edge cases under the rug for who ever implements it to clean up :P 2011-03-26T00:16:14 McLeopold: when you have the time it would be great if you had a look at parse_orders, because it has changed what is being communicated to the engine: https://github.com/aichallenge/aichallenge/blob/epsilon/ants/ants.py#L313 2011-03-26T00:53:43 antimatroid: I think I found a case where the vision algorithm depends on the order of the initial ants 2011-03-26T00:53:52 sigh: personally, I'm against using exceptions unless you are building a library for distribution. 2011-03-26T00:54:43 sigh: ? 2011-03-26T00:54:47 McLeopold: in this case the exceptions made the code a lot shorter, and let me specify the format of the invalid line in one place 2011-03-26T00:54:56 but I'm more concerned about the functionality 2011-03-26T00:57:06 antimatroid: just investigating the cause of my non-determinism in the engine... it seems that get_vision returns a slightly different result depending on what order you add ants to the initial queue... currently investigating 2011-03-26T00:57:48 where is get_vision? 2011-03-26T00:58:06 in ants.py 2011-03-26T00:58:09 it will be if you give a square to an ant the first time it's seen, and don't check the distance to a marked square you hit 2011-03-26T00:58:36 so say you have a...b, you will hit the middle square with one of them first, you need to count the second because it's at the same distance 2011-03-26T00:59:24 the vision algorithm only looks at the ants of one player at a time 2011-03-26T00:59:34 did you mean a...a? 2011-03-26T01:00:08 ah, then it shouldn't matter :\ 2011-03-26T01:01:00 sigh: you're using the exceptions as a structured way to shortcut the parsing and return a value 2011-03-26T01:01:37 It works, and would be fine, but I think there are better more semantic ways to accomplish that. It's just a matter of style. 2011-03-26T01:02:12 what do you think is better? 2011-03-26T01:02:24 perhaps a function that takes a single line and returns an error 2011-03-26T01:02:27 sigh: can you pastebin the example that comes out differently? 2011-03-26T01:02:32 you can have function in functions 2011-03-26T01:02:47 antimatroid: working on a reproducible case 2011-03-26T01:02:55 technically, these are not exceptions in the sense that you don't know what to do 2011-03-26T01:03:15 McLeopold: ok I'll do that... I don't mind either way, just didn't want to repeat code 2011-03-26T01:03:17 nor should you worry about which level is going to catch them 2011-03-26T01:03:53 yeah, the code looks nice :), just use a more semantic structure 2011-03-26T01:03:59 McLeopold: but more importantly, do you agree with moving all the validation into parse_orders? 2011-03-26T01:05:31 oh....um.... 2011-03-26T01:05:49 no 2011-03-26T01:06:02 parse should parse into tokens and numbers 2011-03-26T01:06:09 something else should check game logic 2011-03-26T01:06:32 I want to hook test bots in at different levels so I don't have to deal with strings 2011-03-26T01:07:55 ok, then move it up a level... should do_moves care about validation 2011-03-26T01:08:17 because currently that function is the only way to communicate to the engine about invalid orders 2011-03-26T01:08:48 the engine should only get strings that it logs or prints, nothing that it could intepret 2011-03-26T01:09:09 the engine will check is_alive if it needs to know if a player is kicked 2011-03-26T01:09:22 what is the point of invalid_orders return value then? 2011-03-26T01:09:28 so the game class makes the decision to kick a player, not the engine 2011-03-26T01:09:35 for logging 2011-03-26T01:10:55 bedtime, night 2011-03-26T01:11:01 ok, night 2011-03-26T01:15:38 *** moongrass has quit IRC (Ping timeout: 276 seconds) 2011-03-26T01:15:44 *** moongrass has joined #aichallenge 2011-03-26T01:20:30 antimatroid: you there 2011-03-26T01:20:32 ? 2011-03-26T01:20:34 yeah 2011-03-26T01:21:10 map random4.txt 2011-03-26T01:21:40 viewradius2 96 2011-03-26T01:21:43 ants at: [(15, 12), (18, 13), (22, 20), (21, 23), (16, 11), (23, 23), (14, 12), (22, 22), (20, 19), (21, 18)] 2011-03-26T01:22:00 can you paste bin? 2011-03-26T01:22:04 i've got a friend here watching movies 2011-03-26T01:22:04 get a different result for ants at: [(15, 12), (16, 11), (22, 22), (21, 23), (23, 23), (14, 12), (20, 19), (18, 13), (21, 18), (22, 20)] 2011-03-26T01:22:15 or is it just a number? 2011-03-26T01:22:32 you should try getting it to fill in your squares and output the result 2011-03-26T01:22:45 it's slightly different viewable area 2011-03-26T01:22:53 printing it out is not very enlightening :P 2011-03-26T01:25:11 I will pastebin it the two outputs 2011-03-26T01:28:06 http://pastebin.com/VRm3qcMv and http://pastebin.com/mrrXSEJ0 2011-03-26T01:28:37 ^^ antimatroid 2011-03-26T01:29:09 a dot from line 24 moves to line 27 2011-03-26T01:29:12 :D 2011-03-26T01:29:12 so there's two that swap bottom left 2011-03-26T01:29:17 yes 2011-03-26T01:29:50 the only difference between the ant lists is the order 2011-03-26T01:31:00 yeah, i think the algorithm must be wrong 2011-03-26T01:31:23 actually, i can break it 2011-03-26T01:31:27 i'll paste bin you a broken example 2011-03-26T01:31:28 the implementation is wrong? 2011-03-26T01:31:33 no, the algorithm 2011-03-26T01:31:36 ah :( 2011-03-26T01:31:41 yeah i know :P 2011-03-26T01:31:46 yeah, I was afraid of that 2011-03-26T01:32:26 http://pastebin.com/P0TWwFcU 2011-03-26T01:32:41 it's going to go through that water from one of the positions 2011-03-26T01:32:55 one is going to need to store multiple locations to check if there is a collision 2011-03-26T01:32:58 the algorithm doesn't care about water 2011-03-26T01:33:05 hmmm, good point 2011-03-26T01:33:07 this is just vision 2011-03-26T01:33:19 can that type of situation happen without water? 2011-03-26T01:33:35 given that water is not looked at at all, yes 2011-03-26T01:33:36 i don't think it can 2011-03-26T01:33:48 I will do some more testing, and make some more robust tests 2011-03-26T01:33:56 i'll keep thinking 2011-03-26T01:34:05 hopefully get a smaller test case :) 2011-03-26T01:35:41 antimatroid: this is what was causing the engine to be non-deterministic, btw :( 2011-03-26T01:36:46 sigh: at least we know what's causing it 2011-03-26T01:37:09 yeah... well this was an important bug to catch 2011-03-26T01:43:52 i think i found the problem 2011-03-26T01:44:12 consider 2011-03-26T01:44:12 a. 2011-03-26T01:44:12 .b 2011-03-26T01:44:24 aha... just output a case: bad_vision [(25, 46), (26, 45)] 2011-03-26T01:44:42 ran a check of vision output against brute force just now 2011-03-26T01:44:57 so yeah 2011-03-26T01:45:08 if you're checking the distance to a square in the same row as a on the right, then dependent on who goes to that square first it'll be different 2011-03-26T01:45:16 gtg, will be back in an hour or so 2011-03-26T01:45:23 okay sure 2011-03-26T01:45:58 sigh: update the ant to the one with the closest distance each turn 2011-03-26T01:46:05 that should fix it i think 2011-03-26T01:46:25 * janzert will stop working on his diagram to show the problem now :) 2011-03-26T01:46:32 :) 2011-03-26T01:46:35 since everyone else has also figured it ouy 2011-03-26T01:46:37 out 2011-03-26T01:46:39 :) 2011-03-26T01:46:47 janzert: you knew and didn't tell us? :P 2011-03-26T01:47:05 janzert is all knowing 2011-03-26T01:47:06 well, I wasn't sure I could explain it in words so was making a diagram to illustrate 2011-03-26T01:47:16 :} 2011-03-26T01:47:53 ergh, my fix isn't helpful 2011-03-26T01:47:54 if either of you can fix it while I'm out that would be great :) 2011-03-26T01:47:59 cya 2011-03-26T01:48:02 cya 2011-03-26T01:48:08 *** sigh has quit IRC (Remote host closed the connection) 2011-03-26T01:48:18 btw, once your algorithm is truly order independent you should be able to get rid of the deque 2011-03-26T01:48:22 janzert: any idea of how to fix that without scanning out from every ant separately 2011-03-26T01:49:00 you can for this, but people would need it for say voronoi diagrams 2011-03-26T01:49:29 well, you could always add the square to be checked even if it's already marked as visible 2011-03-26T01:49:39 it might be a matter of adding all joint ants to the list to check 2011-03-26T01:49:45 but that will mean it incurs the same work as scanning out from each ant 2011-03-26T01:49:53 nah, you can do better 2011-03-26T01:50:13 sure it's just the simple fix 2011-03-26T01:50:14 you just have a list of ant locations, and add ants that draw to get to that square first 2011-03-26T01:51:15 the question is how do you tell if it is a draw without testing against every ant 2011-03-26T01:51:27 since if you do that you're back to the same workload again 2011-03-26T01:51:36 or more 2011-03-26T01:51:49 you make visibility an int map rather than bool 2011-03-26T01:51:54 and store the distance you hit it at 2011-03-26T01:52:19 if you do that with voronois, you can use those numbers to back track shortest paths 2011-03-26T01:53:57 sure, and if you hit a visible square that is a tie for distance retroactively add the current ant to the list of ants for that square 2011-03-26T01:54:11 we could always change the distance metric :P 2011-03-26T01:54:23 so store the ants to check as in the vision map instead of the square queue 2011-03-26T01:54:29 yep 2011-03-26T01:54:39 then just need a stack of locations to check from 2011-03-26T01:54:45 no, needs to be queue 2011-03-26T01:55:16 because we need to wait to add all draw ants before moving on from that square 2011-03-26T01:55:17 I think it can get away with a stack 2011-03-26T01:55:22 nah 2011-03-26T01:55:25 nah 2011-03-26T01:55:33 we'll just double check some squares 2011-03-26T01:55:45 or you could just wait with a queue :P 2011-03-26T01:57:07 I'm worried the queue could still have cases it breaks with, although I can't see that it would 2011-03-26T01:57:45 I guess it shouldn't 2011-03-26T01:58:39 yeah, I think that should work 2011-03-26T01:58:42 i think we're good after that, similarly when blocking with water 2011-03-26T01:58:57 water blocks vision? 2011-03-26T01:59:07 no never 2011-03-26T01:59:21 but people will need something like this for their bots 2011-03-26T01:59:26 and they will want to be blocked by water 2011-03-26T02:00:09 for a reachable in x moves map I suppose? 2011-03-26T02:01:12 i don't think that matters then 2011-03-26T02:01:39 because your metric is manhatten based in a way 2011-03-26T02:01:44 ok, for closest ant to square 2011-03-26T02:01:51 eh, even then 2011-03-26T02:02:05 ok, when does it matter? :P 2011-03-26T02:02:21 i'm going to fix it.. 2011-03-26T02:02:23 unless you already are? 2011-03-26T02:02:29 no, go ahead 2011-03-26T02:03:15 it only matters when you're using euclidean metric 2011-03-26T02:03:26 people will need to worry about if for their own vision funtions 2011-03-26T02:03:39 yes, I mean when in the game do you care about euclidean and water blocking it? 2011-03-26T02:11:09 btw, the distances in the map should be the true euclidean distance not the manhatten integer distance, I think either actually gives a correct outcome but the euclidean will weed out extra ants much quicker 2011-03-26T02:12:35 only mentioning it because you initially were going to change the distances to an int map which won't work with euclidean distances well ;) 2011-03-26T02:13:20 *** chris__0076 has joined #aichallenge 2011-03-26T02:13:42 well yeah, i can use doubles 2011-03-26T02:13:53 one could just use the metric squares and have integers, but meh 2011-03-26T02:13:56 it's less clear 2011-03-26T02:14:11 yep, just use exactly whatever the distance function hands to you :) 2011-03-26T02:15:51 *** Chris_0076 has quit IRC (Ping timeout: 246 seconds) 2011-03-26T02:17:08 *** Chris_0076 has joined #aichallenge 2011-03-26T02:19:18 *** chris__0076 has quit IRC (Ping timeout: 246 seconds) 2011-03-26T02:30:36 janzert: how do i append a list to another list? 2011-03-26T02:30:50 l1 += l2? 2011-03-26T02:35:12 yes 2011-03-26T02:35:36 although you probably figured that out by now :) 2011-03-26T02:36:10 i was waiting anyway, i have a pastebin for you 2011-03-26T02:36:26 running an interactive python interpreter is really handy for checking things like that btw 2011-03-26T02:36:53 I do it semi frequently when writing python code 2011-03-26T02:37:32 http://pastebin.com/wc8C4FHx 2011-03-26T02:38:02 a game runs 2011-03-26T02:38:09 the ant count at the end is now different :\ 2011-03-26T02:38:22 I imagine 2011-03-26T02:39:04 why? 2011-03-26T02:39:11 it's like a hardly broken change to vision 2011-03-26T02:39:29 turns out it varied before : 2011-03-26T02:39:29 but I think not a too uncommon one 2011-03-26T02:39:34 that would be food spawning 2011-03-26T02:39:57 ahh, so not a deterministic run 2011-03-26T02:40:50 i'm pretty sure that's right, i'm going to push, sigh can yell at me later if it's not 2011-03-26T02:41:13 the inner loop doesn't ever weed out ants 2011-03-26T02:41:46 not sure if it actually makes a noticeable difference though 2011-03-26T02:42:19 weed out ants? 2011-03-26T02:42:25 oh, i thought about that 2011-03-26T02:42:31 but then it's very edge cases that it will miss 2011-03-26T02:42:44 this way you break the loop earlier, although i'm not adding them all, that's a point 2011-03-26T02:42:48 i might just weed them all 2011-03-26T02:43:11 oh i was adding them all, that was my += question :P 2011-03-26T02:43:28 i think it'd be faster to add them all and keep moving, cause usually you will match on the first ant location? 2011-03-26T02:44:31 the thing is with the case where the problem was found a..|.b. 2011-03-26T02:44:55 the distance is only tied at the first square, the next square out breaks the tie and would eliminate the correct ant 2011-03-26T02:45:09 so you never have to test against that ant again if you weed it out 2011-03-26T02:45:18 aichallenge: Nick Ham epsilon * rc2b92d7 / ants/ants.py : fixed get_vision - http://bit.ly/dQOjEW 2011-03-26T02:45:43 janzert: but does that always work? 2011-03-26T02:46:09 this will work for the moment 2011-03-26T02:46:31 but yeah I guess it doesn't matter cause even if it's first in the list it will only test negative on one square where the other tests positive 2011-03-26T02:46:57 *** sigh has joined #aichallenge 2011-03-26T02:47:00 yeah, it does at least look like it should return the correct result 2011-03-26T02:47:04 sigh: i fixed it for you :) 2011-03-26T02:47:12 umm actually 2011-03-26T02:47:16 :( 2011-03-26T02:47:28 you're now returning a different data structure than before 2011-03-26T02:47:43 heh 2011-03-26T02:47:49 crap i am 2011-03-26T02:47:52 you need to collapse the vision map back down 2011-03-26T02:47:56 yep 2011-03-26T02:48:04 sigh: you could probably do that neater than i? 2011-03-26T02:48:11 it shouldn't be more than a line or two 2011-03-26T02:48:16 where is code? 2011-03-26T02:48:24 get_vision in ants.py 2011-03-26T02:48:38 vision = [[v for v, a in col] for row in vision] 2011-03-26T02:48:41 vision is now a matrix of [bool,list] instead of [b] 2011-03-26T02:48:52 dw, i've got it :) 2011-03-26T02:49:27 oops 2011-03-26T02:49:33 *** closedbracket has quit IRC (Remote host closed the connection) 2011-03-26T02:49:41 what does it mean? 2011-03-26T02:49:46 yeah 2011-03-26T02:49:51 you wanted col == row yeah? 2011-03-26T02:50:04 yes, I think so 2011-03-26T02:50:12 although now I want to test it before say so :) 2011-03-26T02:50:49 antimatroid: what is the meaning of the [bool, list] ? 2011-03-26T02:51:10 bool is the square has been painted outwards from already, list is the ants that need to paint outwards 2011-03-26T02:51:55 hmmm... so you are treating each ant independently, then? 2011-03-26T02:52:23 they don't get in each other's ways, so to speak? 2011-03-26T02:52:24 yes, vision = [[v for v, a in row] for row in vision] should do what you need 2011-03-26T02:52:46 in this case, there is no point doing a bfs 2011-03-26T02:52:58 should just use the same thing that nearby ants does 2011-03-26T02:53:01 it's much faster 2011-03-26T02:53:19 aichallenge: Nick Ham epsilon * r971f736 / ants/ants.py : corrected get_vision to return the right type - http://bit.ly/fY4jLh 2011-03-26T02:53:25 sigh: you need bfs 2011-03-26T02:53:28 sigh: not quite independently only when there are "draws" in the distance between ants 2011-03-26T02:53:43 janzert: ah 2011-03-26T02:53:52 anyway, i need to go get some beer, some friends joint 21sts tonight 2011-03-26T02:53:56 :) 2011-03-26T02:54:37 umm, hate to say this 2011-03-26T02:55:03 :P 2011-03-26T02:55:08 antimatroid did you check the output of this? 2011-03-26T02:55:09 i broke it again? 2011-03-26T02:55:22 i checked that a game ran 2011-03-26T02:56:01 nevermind I thought the test "if not vision[n_row][n_col]:" could never be satisfied but I was reading it wrong 2011-03-26T02:56:01 i was hoping sigh would now check, he seemed to have it set up to, i have no idea how to easily :P 2011-03-26T02:56:08 phew :P 2011-03-26T02:56:24 yeah, hang on 2011-03-26T02:56:24 sigh: please check anyway, i did say before i was gambling that you could shout at me later if it's wrong :) 2011-03-26T02:56:39 I do worry that it may not be correct though 2011-03-26T02:56:48 why? 2011-03-26T02:56:49 just not completely broken ;) 2011-03-26T02:56:50 antimatroid: next time if you are trying experimental stuff do it in a new branch 2011-03-26T02:57:04 sigh: okay sure, i was expecting you back though 2011-03-26T02:57:32 yeah, sure... but the main branch is not the place for untested code :P 2011-03-26T02:58:00 yeah :P 2011-03-26T02:59:41 I hadn't satisfied myself that vision[c_row][c_col] = True won't ever happen before a "drawing" ant makes it there, but at least with 3am thinking I do now think it is fine. 2011-03-26T02:59:59 of course the 2:50am thinking is what got me worried about it :P 2011-03-26T03:00:47 * janzert should really stop trying to inspect code for logic errors after 1am 2011-03-26T03:00:50 hmm... it's throwing up differences with my naive vision calculation 2011-03-26T03:01:14 lemme check 2011-03-26T03:01:23 erm crap, anyway, back in like 10 2011-03-26T03:03:22 sigh: does it work if you take out the "if not vision[n_row][n_col]:" check? ;) 2011-03-26T03:04:13 well, now it throws errors 2011-03-26T03:04:24 foo.append += bar will tend to do that though 2011-03-26T03:04:33 heh 2011-03-26T03:04:36 hmm 2011-03-26T03:06:40 ok, yeah I'm back to thinking that a vision square can and will be set to true before all the drawing ants get there 2011-03-26T03:06:55 ok, here's what I'm going to do... I'm going to replace get_vision with a brute force version until we get an implementation that works 2011-03-26T03:07:07 :) 2011-03-26T03:07:09 I don't want to keep the repo in a broken state 2011-03-26T03:08:30 for various definitions of broken 2011-03-26T03:08:41 well at least it will be less broken 2011-03-26T03:15:04 hmm... brute force is way too slow for me to want to check in... it takes a 10 sec game to 2 minutes 2011-03-26T03:15:50 I'd rather have the slightly broken old version 2011-03-26T03:16:06 did you have a script to brute force check for differences against the naive implementation? 2011-03-26T03:16:47 janzert: not yet, atm I'm just comparing them inside the game 2011-03-26T03:16:53 ok 2011-03-26T03:17:40 janzert: I'm think of creating a branch for the current version of the repo, then reverting antimatroid's changes in the main repo to back to how it was 2011-03-26T03:17:47 does that sound reasonable 2011-03-26T03:18:03 as it is, the code in aichallenge/epsilon will not run 2011-03-26T03:18:28 ok it will, but it claims each ant can only see itself 2011-03-26T03:18:56 if you want, I think just dropping the extra check and removing .append may be all that's needed 2011-03-26T03:19:13 no, there were more errors 2011-03-26T03:19:57 if I knew how to test it, I might give fixing it a go :) 2011-03-26T03:20:28 but reverting it is fine with me 2011-03-26T03:20:35 i'm back 2011-03-26T03:20:46 or see if antimatroid shows back up soon.... 2011-03-26T03:20:51 ^^ 2011-03-26T03:20:57 :) 2011-03-26T03:21:08 i don't mind if you revert it, but we could spend 5 minutes trying to work out why it's broken 2011-03-26T03:21:40 would it have a problem with renaming vision? 2011-03-26T03:22:22 I'm trying to figure out why it is broken 2011-03-26T03:22:43 now it is stuck in an infinite loop and my computer is slow 2011-03-26T03:22:54 sgh: line 180 2011-03-26T03:22:59 get rid of .append 2011-03-26T03:23:12 that is a rather silly mistake :\ 2011-03-26T03:23:43 sigh ^* 2011-03-26T03:23:56 that's what I did, but now I have an infinite loop 2011-03-26T03:24:38 which is being kind enough to eat up all my ram and my laptop swap like crazy 2011-03-26T03:25:17 is that for running a game or your other stuff? 2011-03-26T03:25:37 it just ran a game for me 2011-03-26T03:27:12 antimatroid: as an aside here's how I would have probably done it without the deque 2011-03-26T03:27:16 I bet you just played a game where each any can only see itself 2011-03-26T03:27:41 completely untested of course 2011-03-26T03:28:26 ok, I'm sorry... but how does this thing know when to stop? 2011-03-26T03:29:02 heh, yes I just realized that is why the 'if not vision..." check has to be there 2011-03-26T03:30:06 sigh: when there's no squares left to check 2011-03-26T03:30:27 antimatroid: there's always squares to check 2011-03-26T03:30:34 no there's not 2011-03-26T03:30:41 i mark them as visited once i check them 2011-03-26T03:30:55 oh wait 2011-03-26T03:31:00 haha 2011-03-26T03:31:03 i completely misinterpreted vision 2011-03-26T03:31:09 *** closedbracket has joined #aichallenge 2011-03-26T03:32:03 give me 5 and i'll have a pastebin replacement for you 2011-03-26T03:32:12 (sorry :P) 2011-03-26T03:32:53 and I just realized I didn't paste the pastebin link of my non-deque using version :P 2011-03-26T03:32:56 http://pastebin.com/Q6zij8ZD 2011-03-26T03:33:59 firefox is still swapping back in :( 2011-03-26T03:34:53 sigh: i'm not convinced that isn't what i wanted to do 2011-03-26T03:35:14 what did you want to do? 2011-03-26T03:35:34 return you vision squares 2011-03-26T03:35:58 and it should terminate, because it only adds non-checked visible squares to the deque 2011-03-26T03:36:21 it doesn't do that 2011-03-26T03:36:49 it does, it marks a location as valid when it takes it off the deque 2011-03-26T03:36:53 doesn't the "if not vision[n_row][n_col]:" do that? 2011-03-26T03:36:58 yeah 2011-03-26T03:37:08 that will always evaluate to True 2011-03-26T03:37:21 put a [0] there :P 2011-03-26T03:37:33 vision[n_row][n_col][0] 2011-03-26T03:37:41 so many bugs, it'd hard to pick 2011-03-26T03:38:48 you run it first 2011-03-26T03:39:31 If I have to debug infinite loops, let me turn on my desktop where they run faster 2011-03-26T03:39:33 also is that check really in the right spot to not erroneously weed out draws? or should it be moved down like http://pastebin.com/zb2jKTks 2011-03-26T03:41:10 I'm now back to thinking it needs to be moved, of course I've changed my mind on that every 5 minutes or so for the last half hour 2011-03-26T03:44:20 *** antimatroid1 has joined #aichallenge 2011-03-26T03:44:24 *** antimatroid