00:04:52 GabeW has quit ("Client Exiting") 00:22:55 Ouch. 00:23:44 Hm. If I go with your interpretation of query...my life becomes vastly easier. 00:29:05 Sean, are you around? 00:29:12 * bijan askes again. 00:29:40 yes, I am 00:29:54 I'm getting bored driving up and down the same old strip 00:30:29 Hmm? 00:30:56 you asked if I was around... 00:31:21 driving up and down? 00:31:27 Anyway, if I have the following: 00:31:28 :bob :parent :jane. 00:31:28 :jane :parent :janeII. 00:31:28 this log:forAll :x, :y. :z. 00:31:28 {:x :parent :y} log:implies {:y :child [:parent :y].}. 00:31:30 {:x :parent :y. :y :parent :z} log:implies {[a :gParent]. [a :gChild].}. 00:31:58 would you expect a filter to generate 6 triples with 4 bnodes? 00:33:02 sorry had to pop out again - the midnight hour is my "busy hour" 00:33:12 heh, that didn't come out right, did it? 00:33:15 anyway... 00:33:16 Heh. 00:33:30 * bijan feels close. 00:33:48 I have some forms for builtins, and now for bnodes. 00:33:57 Lists will be pretty much all that's left. 00:34:04 That and implementing all the builtins :( 00:34:07 O.K., for the first rule we get: :jane :child [ :parent :jane ] . :janeII :child [ :parent :janeII ] . 00:34:25 X = 'document#jane' 00:34:25 Y = 'document#child' 00:34:25 Z = anon_61 ; 00:34:26 X = anon_61 00:34:26 Y = 'document#parent' 00:34:26 Z = 'document#jane' ; 00:34:28 ugh, I should label those bNodes: call them _:x and _:y 00:34:28 X = 'document#janeII' 00:34:30 Y = 'document#child' 00:34:32 Z = anon_62 ; 00:34:34 X = anon_62 00:34:36 Y = 'document#parent' 00:34:38 Z = 'document#janeII' 00:34:40 Ooh, an bnode syntax. 00:34:48 But that's not a fundemental bit. 00:35:07 Of the inference engine. 00:35:43 then we have 2 bNodes for the second bit, surely 00:35:58 Yep. 00:35:59 Ok. 00:35:59 so four bNodes, I agree 00:36:49 and six triples... 00:36:50 now, you think that the first :jane :child anon_61 would be blocked by :jane :child :bob. 00:36:56 no, wait, more than six 00:37:05 nope... six 00:37:17 blocked? 00:37:29 If I had :jane :child :bob in the target context. 00:37:40 And given the principle that we don't add duplicates if we can help it. 00:37:58 Should I nevertheless add a :jane :child anon_61? 00:38:15 you don't have :jane :child :bob in there at all 00:38:20 I know. 00:38:24 *Suppose*. 00:38:27 oh, sorry: if 00:38:28 got it 00:38:29 Contray to the expresssed facts ;) 00:38:36 That in the *target* context... 00:38:52 So, e.g., if I were applying rules and the kb had :jane :child :bob. 00:38:55 I think you should add it 00:39:00 Me too. 00:39:05 00:39:32 Could you check cwm on this? 00:39:36 I could 00:39:42 Would you? 00:39:42 :) 00:39:48 I would 00:39:55 just the rule that you pasted in above, yes? 00:40:04 Yes. 00:40:57 :bob :parent :jane . 00:40:58 00:40:58 :jane :child [ 00:40:58 :parent :jane ]; 00:40:58 :parent :janeII . 00:40:58 00:41:01 :janeII :child [ 00:41:02 :parent :janeII ] . 00:41:27 6 triples, two bNodes 00:41:45 not what we expected... ugh 00:41:46 This is with :jane :child :bob? 00:41:58 input:- 00:41:59 @prefix : <#> . 00:41:59 @prefix log: . 00:41:59 :bob :parent :jane. 00:41:59 :jane :parent :janeII. 00:41:59 this log:forAll :x, :y. :z. 00:42:01 {:x :parent :y} log:implies {:y :child [:parent :y].}. 00:42:03 {:x :parent :y. :y :parent :z} log:implies {[a :gParent]. [a :gChild].}. 00:42:09 oh, heh. you borked that 00:42:15 You need a :jane :child :bob. 00:42:19 In the input. 00:42:30 that's better, now it gives:- 00:42:31 [ a :gChild ]. 00:42:31 00:42:31 [ a :gParent ]. 00:42:32 too 00:42:38 you want to add that triple? O.K. 00:42:43 Wait. 00:42:45 I'm confused. 00:42:54 you had: this log:forAll :x, :y. :z. 00:43:04 so it didn't recognize the :z as being universally quantified 00:43:13 Oops. 00:43:26 I added :jane :child :bob. to the document 00:43:31 and now it gives... [big paste] 00:43:40 [ a :gParent ]. 00:43:41 00:43:41 [ a :gChild ]. 00:43:41 00:43:41 :bob :parent :jane . 00:43:46 00:43:46 :jane :child [ 00:43:46 :parent :jane ], 00:43:47 :bob; 00:43:49 :parent :janeII . 00:43:51 00:43:53 :janeII :child [ 00:43:55 :parent :janeII ] . 00:43:57 [EOF] 00:44:25 Ok, so we're right. 00:44:28 yep 00:44:47 Ugh, that complicates things a bit. 00:51:28 Ok, y'know something really cool? 00:51:38 what? 00:51:53 [ a :CoolThingKnownBySean ] . 00:52:01 By compiling n3 to prolog I get an *exact* specification of the semantics of the rules. 00:52:11 In fairly transparent terms. 00:52:24 that's a good point. DanC tried ot express N3 in terms of KIF once before... 00:52:38 Yes, but in prolog, it becomes executable :) 00:52:52 For example: 00:52:54 n3implies(1, id(user, user32, user33), _G412):- 00:52:54 rdf(_G413, 'document#parent', _G415, user), 00:52:54 \+ rdf(_G415, 'document#child', Anon_4, _G412), 00:52:54 \+ rdf(Anon_4, 'document#parent', _G415, _G412), 00:52:54 catch(\+ atom_prefix(Anon_4, anon_),_, fail), %this will *NOT* work! Literals. 00:52:54 write_ln(ho), 00:52:57 logimplies, 00:52:58 rdf_gen_id('anon_', New_anon_1), 00:53:00 rdf_assert(_G415, 'document#child', New_anon_1, _G412), 00:53:02 rdf_assert(New_anon_1, 'document#parent', _G415, _G412). 00:53:18 Ignore the head, it's just infrastructure. 00:53:26 The first line checks that the premise holds. 00:53:46 2-3 check that the conclusion doesn't 00:53:56 2-5 00:54:03 6 is debugging :) 00:54:11 the logimplies is just a marker. 00:54:18 The last three lines add the triples. 00:54:32 excellent 00:54:40 But it's also pretty logical. 00:54:56 (the catch is wart. atom_prefix throws exceptions...) 00:55:17 what's the thing about literals for? 00:55:26 Oh, because of atom_prefix 00:55:29 It will work *now*. 00:55:35 Since I'm catching the exceptions. 00:55:39 * sbp eats some lovely garlic bread meanwhile 00:55:45 ah, got it 00:55:54 atom_prefix(literal(foo), X) will hork and exception rather than failing. 00:56:27 Alas, they arent' working :) 00:56:49 Duh. 00:56:56 Becasue I want an exception to == true :) 00:57:12 there we go. 00:57:48 Ugh. 00:58:00 Well, it will be interesting to see what optimization opportunities there are. 00:58:05 On real N3 code. 00:58:14 Right now I just query the crap out of everything. 00:58:26 yeah, CWM has a real problem with that 00:58:49 I still don't have good performance testers. 00:59:01 Oh, and between each application, i recompile *all* the rules. 00:59:22 As compiling gets more compilicated, and iterations get larger, it's going to make sense not to do that. 00:59:31 I.e., only compile new rules. 01:01:42 sbp has changed the topic to: Coming soon: "Swhack - the movie" 01:02:17 we could get Lucy Lawless to play xena 01:03:04 Ooo, scopt of the negations is wrong wrong wrong. 01:03:33 negations? in N3? 01:04:17 heh... just wait 'till you have to implement log:notIncludes. You're going to explode in a kind of critical frenzy. I can't wait 01:04:30 I'm NOT looking forward to it. 01:04:48 CWM's implementation of it *really* confuses me. It makes no sense 01:05:17 Well, I don't see that I need to implement all the log built-ins. 01:05:22 They seem sorta shady to me. 01:05:37 yep, but they're the meat on the N3 bone 01:05:59 Eh. 01:06:07 At SOME point someone else has to join in and help :) 01:06:13 Mustn't they? 01:06:23 well, most of the things that are done with N3 require builtins... 01:06:26 someone else?! 01:06:35 Sure, but not *all* the log built-ins 01:06:44 not just all, but more! 01:06:56 Er... 01:06:58 what's the point of making a CWM clone if you can't extend it :-) 01:07:00 * bijan doesn't parse that. 01:07:02 Oh sure! 01:07:06 that's why I added the crypto stuff to Eep 01:07:17 But *I* don't have to implement *all* the log built-ins. 01:07:22 sure you do 01:07:26 Nope. 01:07:32 Part of my point is to have a proof of concept. 01:07:33 otherwise, it's not much of a clone, is it? A sub-clone, perhaps :-) 01:07:51 And mostly on the inference engine. 01:08:04 I've still not seen a supposedly tough n3 file. 01:08:09 I.e., one that cwm grinds on forever. 01:08:17 er... the pool stuff is goo 01:08:25 Hmm. 01:08:43 [but it requires built-ins] 01:08:52 Sure. 01:09:05 I'll, perhaps, implement all the builtins necessary to run soemthing liek that. 01:09:31 in http://www.w3.org/2001/08swws67/ 01:09:39 the poolGame* stuff 01:09:45 If I had a good example of an N3 program that took CWM 1/2 hour to run on a fast machine, I'd focus on getting it running. 01:10:22 Good god. 01:10:26 3 files? 01:10:32 yeah? 01:10:45 what were you expecting? 01:11:01 I dotn' know, I just want to make sure I understand it. 01:11:06 oh! 01:11:10 rules.n3! 01:11:15 Hmm/ 01:11:23 that doesn't have any builtins, but when you apply it to data, it'll grind for quite a while 01:11:32 Pointer? 01:11:48 Ah. you think on that? 01:11:57 http://infomesh.net/2001/05/rdflint/rules.n3 01:12:11 Yes, got it. 01:13:25 try it on http://www.w3.org/2001/03/earl/0.95.n3 01:13:35 I'll run it locally, to see how long it takes... 01:14:35 * sbp sets it off 01:15:00 newhak (djflsdkjfl@ACA5AC4E.ipt.aol.com) has joined #swhack 01:15:02 it's getting files from the Web too... you should add that functionality (if you haven't got it already) 01:15:06 Hi there newhak 01:15:11 Argh. No I don't 01:15:27 this channel is publically logged etc., to make an off-log comment, prefix your junk with "# " 01:15:30 hey there newhak. 01:15:34 got any good 0 day? 01:15:48 Which is getting files from the web? 01:15:54 CWM 01:15:58 hi 01:16:07 ooh, CWM barfed:- 01:16:09 RuntimeError: ('Should only see first and rest in list mode', ((1, 'file:/home/#0_work'), (0, 'http://www.w3.org/1999/02/22-rdf-syntax-ns#type'), (0, 'http://infomesh.net/2001/05/rdflint/rules.n3#_g35'), (0, 'http://www.w3.org/2000/01/rdf-schema#Resource'))) 01:16:16 I mean, which file is making it get stuff from the web? 01:16:26 file? it's from command line... 01:16:29 eek, if cwm can't run them ... 01:16:38 CWM used to be able to 01:16:41 newhak has left #swhack 01:16:50 I want to load up a KB 01:16:51 damn 01:16:52 Think. 01:16:54 Time 01:16:55 i really wanted 0 day. 01:17:13 If possible, solely from local data. 01:17:24 I don't want to test my network latency or bandwidth :) 01:18:25 Why would rules.n3 and the eral thing fetch stuff fromthe net? 01:18:26 Tim's validation thing can also take a while, but that uses mega-builtins 01:18:37 rules.n3 doesn't take anything from the Web 01:18:48 * bijan is lost as why you said it would :) 01:19:23 I didn't say that it would. I meant that I just put the URIS in the command line, and CWM slurped the files from the Web. It can also take them using log:semantics 01:19:32 Ah, yes. 01:19:39 That's not a benchmarking feature though. 01:19:43 it is a feature I need to add. 01:19:43 nope 01:19:54 It's not? 01:19:58 Whew, then. 01:21:43 Oh bugger, the earl things has triple quoted stuff. 01:22:01 Definitely somethign to add soon :) 01:22:40 it's not a benchmarking feature, that is... it *is* a feature you need to add :-) 01:22:45 * sbp is running more CWM tests 01:22:54 ugh, it barfed again 01:23:12 Heheheh. 01:23:42 the debug information certainly isn't very helpful, either 01:24:36 heh. 01:24:44 Right now I have no good error messages. 01:24:50 I hope eventually to add some. 01:25:02 But being able to inspect and query the dbs is pretty helpful. 01:25:24 Weird. 01:27:02 Oops. lists. 01:27:05 no can do. 01:27:51 argh, it fucked up again?! what the smeg is wrong with it? it's like it can't handle any damn lists at all 01:28:11 where the hell is AaronSw? 01:28:13 Well, I *know* I can't :) 01:28:55 * sbp tries an old version of the CWM thing 01:30:55 Pff:- 01:30:57 $ u timenow; python cwm/cwm.py http://infomesh.net/2001/05/rdflint/rules.n3 http://www.w3.org/2001/03/earl/0.95.n3 --think > earl`u timenow`.n3; u timenow 01:31:00 20020227-013010 01:31:02 20020227-013029 01:31:04 works almost instantly, given that it has to get the files off of the Web 01:31:10 it's so much better than the new CWM. The new one seems quite broken 01:31:13 So not a good test. 01:31:38 Heh. /me is proud to have a CWM that is almost merely as broken as teh current CWM! 01:32:12 it *is* a good test, it's just that the latest CWM is buggy 01:32:22 well, it's not a good test in that it runs fast... 01:32:25 Yes. 01:32:28 heh, let's try the pool stuff 01:32:30 Sorry, that's all I meant. 01:33:39 cool 01:33:58 * sbp tries feeding it all of the pool stuff, and getting it to filter, using the *-filter* file 01:34:54 Morbus has quit ("http://www.disobey.com/") 01:35:06 a minute and a half-ish 01:35:07 $ u timenow; python cwm/cwm.py http://www.w3.org/2001/08swws67/poolGame-kb.n3 http://www.w3.org/2001/08swws67/poolGame.n3 --think --filter=http://www.w3.org/2001/08swws67/poolGame-filter.n3 > pool`u timenow`.n3; u timenow 01:35:08 20020227-013328 01:35:08 20020227-013452 01:35:23 I can't read your timestamps :) 01:35:38 130 some seconds? 01:36:04 HH:MM:SS 01:36:20 1 minute, 34 seconds 01:36:29 90-ish seconds... 01:36:38 that's pretty efficient, actually 01:36:42 That's a reasonable test. 01:36:46 No, you mean that's pretty quick. 01:36:56 We have no idea if it's efficient. 01:37:03 no, I mean efficient compared to the new CWM 01:37:08 Ah! 01:37:10 Yes. well. 01:37:13 which would take a year if it could do it at all! :-) 01:37:18 New CWM sucks. :) 01:37:29 good ol' 1.82 01:43:05 what is he on now? 01:43:08 * AaronSw waves 01:43:58 Morbus, I took the day off. 01:45:30 * AaronSw plays "Spanish Lady" again, in preparation for the RDF Core singalong 01:46:08 ooh, very neat: http://www.w3.org/2000/10/swap/pim/usPlace2LatLong.n3 01:47:28 erk: 80-ish seconds 01:48:58 lol: http://www.ftrain.com/robot_exclusion_protocol.html 01:49:24 .google aaron 01:49:24 been there, done that. Stop reading our logs! they're private 01:49:25 aaron: http://www.rall.com 01:49:33 heh. that's baffling, that 01:50:10 aha: "These terms only appear in links pointing to this page: aaron" - http://www.google.com/search?q=cache:krhpLrSaHI8C:www.rall.com/+Aaron&hl=en 01:50:18 that is unbelievably weird 01:50:28 .google link:www.rall.com aaron 01:50:29 no results found. 01:51:50 * AaronSw reports this using the quality_form 01:53:33 bitsko (~ken@kmacleod.static.iaxs.net) has joined #swhack 01:53:51 * sbp waves 01:53:56 Ken returns. 01:54:28 hey. know anybody who has an RSS generating toolkit (more than just one website) that'd be willing to put out an update that includes admin:generatorAgent? 01:54:40 Rael? 01:54:51 heh, I was just about to say that! 01:55:04 he also has privs for Perl's XML::RSS 01:55:04 You mean something used by more than one *person*? 01:55:13 Or a big old site? 01:55:59 generatorAgent is particularly for the software used to generate the RSS, different person that the website owner 01:56:38 right, but an aggregator could collect 0.91 files, and genreate rss1.0 01:56:44 Aaron! Arron might do that. 01:56:48 he's a tool guy. 01:57:25 you'll never get his attention that way. 01:57:30 You need to say, "Aaron, someone using a namespace with a hash on the end!" 01:57:37 heh 01:58:00 sorry, i'm reading the logs, but i hear lines with "aaron" in them and no periods with characters following them 01:58:05 btw, I just chatted with Jeff Barr and explained the idea and he's cool with it 01:58:14 and my name is not Arron 01:58:25 Sorry, a simple arror on my part. 01:58:42 heh. You make me sic 01:59:13 This is a comma typo banter on swhack? 01:59:25 Aaron: I need some RSS 1.0 toolkit authors to incorporate admin:generatorAgent 01:59:45 oh yeah, i sort of started that mess like that didn't i... 02:00:52 * bitsko holds breath: this may be the first widely used community-driven extension to RSS (of any flavour) 02:01:06 Oookay, I've got a working and perhaps feasible target for compiling bnoded rules. 02:01:24 bnoded? 02:01:29 sbp, have you met the unix time command? 02:01:34 * bitsko wonders what Bijan is cooking... 02:01:35 he means "bNoded" 02:01:41 $ time python somelongfile.py 02:01:46 unix time: that sucks 02:01:54 {[: a :child]} log:implies {[ a :parent]} 02:02:01 Ken: CWMclone! 02:02:10 wow! the new CWM had been going for 20 minutes 02:02:14 * sbp stops it 02:02:18 See: http://www.unc.edu/~bparsia/sw/cwmclone/cwmclone.html 02:02:38 Hmm. To which I've just added some grovely stuff about sean. 02:02:55 ooh, grovely stuff! 02:03:14 why does unix time suck? 02:03:37 * bijan is reminded of the excellent "Long Painful History of Time" 02:03:38 On the web. 02:03:46 * sbp would slave for Bijan if he hadn't mentioned Eep crapping out :-) 02:03:50 * bitsko wonders if he means Unix time(3) or time(1) 02:03:56 No you wouldn't :) 02:03:57 or is that time(2)? 02:04:00 .google "Long Painful History of Time" 02:04:00 "Long Painful History of Time": http://www.stud.ifi.uio.no/~enag/lugm-time.html 02:04:07 ah: http://www.stud.ifi.uio.no/~enag/lugm-time.html 02:04:08 i mean the one that times programs for you 02:04:10 Pff 02:04:14 instead of this crazy u timenow stuff 02:04:25 Ooh, Naggum.no! 02:04:36 I know him. 02:04:40 Personally? 02:04:51 No. 02:05:02 I should say that I know of him. 02:05:13 bitsko, so what exactly do I add to my generator? 02:05:13 Hmm. Actually, sean, you could do worse than try to implement builtins. As a prolog learning task. 02:05:21 I knew (of) him before he shot SGML in the back ;) 02:06:06 Morbus (~Morbus@s108.terminal3.totalnetnh.net) has joined #swhack 02:06:14 Naggum was a huge fan/supporter of SGML when it should have been doing RDF things, but HyTime didn't come soon enough before he bailed. 02:06:18 hello 02:06:22 hey j 02:06:29 AaronSw: I was looking for the plex docs today 02:06:34 and I found a newer version 02:06:35 it seems 02:06:37 Er... 02:06:39