Open in new window / Try shogun cloud
--- Log opened Sun Jul 22 00:00:17 2012
blackburnsonney2k: here comes nice pic for the website https://dl.dropbox.com/u/10139213/shogun/sonne.jpg00:05
blackburnwiking: ^ :D00:05
gsomix_blackburn, lol :300:05
wikingblackburn: lol man wtf is this? :)00:09
blackburnwiking: secret http://sonnenburgs.de/soeren/category/ photo00:10
gsomix_good night guys00:26
-!- gsomix_ [~gsomix@80.234.25.187] has quit [Ping timeout: 264 seconds]00:31
-!- zxtx [~zv@ool-457e7550.dyn.optonline.net] has quit [Ping timeout: 246 seconds]00:48
blackburnwiking: here it goes00:57
CIA-18shogun: Sergey Lisitsyn master * r0e9c608 / (11 files in 4 dirs): Introduced multiple output multiclass labels - http://git.io/x84IUg00:57
blackburnhog is next on the list :)00:57
wikingcool00:59
wikingi'm just about to00:59
wikingfinish the unittester for SGVector00:59
wikingand then i'll do a pr01:00
wikingbut not for actually merging01:00
wikingbut more for discussion01:00
wikinghahahahah01:09
wikingblackburn: i think i even found a bug01:09
wikingwhile unit testing :)))01:09
blackburnnice01:09
blackburnwiking: I just checked classes01:19
blackburnhog won't work there for sure01:19
wikingwhy?01:43
wikingok here goes my stupid example01:43
wikinglet the flame begin :DD01:43
wikingthere01:54
wikingpr: https://github.com/shogun-toolbox/shogun/pull/66701:54
blackburnwiking: well as I can see they are not centered or anything that would make hog work02:04
wikingblackburn: can u merge this plz: https://github.com/shogun-toolbox/shogun/pull/65902:13
CIA-18shogun: Sergey Lisitsyn master * r532a5dd / (5 files in 3 dirs): Merge pull request #659 from vigsterkr/latent - http://git.io/VkPGqA02:14
CIA-18shogun: Viktor Gal master * r5b8915d / (5 files in 3 dirs): Switch from function ptrs to CLatentModel - http://git.io/LUhFlg02:14
CIA-18shogun: Viktor Gal master * r924dc20 / (2 files in 2 dirs): Fix spacing when casting - http://git.io/Kxx_7w02:14
wikingthnx02:14
wikingok fuck i had trouble rebasing02:14
wikingnow it works02:14
wikingthnx a lot02:14
wikingblackburn: centered? in what sense?02:19
wikingok oh shit02:19
wikingi should make your passwd work finally02:19
wikingjust a sed02:19
wikingsec02:19
blackburnI mean we don't know bounding window so HOG will just produce some noise02:20
wiking?02:20
blackburnwiking: HOG of what do we want to compute?02:21
blackburnI actually feel lost without any knowledge how to approach the problem :D02:28
droopy:*)02:28
wikingblackburn: we do a slidng window02:36
wikingfor hog02:36
wikingdroopy: HOG IT02:36
blackburnwiking: how do you want to match it ?02:36
wikingblackburn: well for starters02:37
wikingwe can do the same way as the guys02:37
wikingin Large-scale Image Classification:02:37
wikingFast Feature Extraction and SVM Training02:37
wikingsee page 2 figure 202:38
blackburnthey state liblinear didn't manage to converge after two weeks02:40
blackburnomfg :D02:40
shogun-buildbotbuild #190 of deb3 - modular_interfaces is complete: Failure [failed test libshogun]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/190  blamelist: Viktor Gal <viktor.gal@maeth.com>02:40
wikingmmm what's tiiiis02:41
blackburnmy fault probably02:41
wikingblackburn: aoh yeah i see02:41
wikingtransfer_multitasklogisticregression02:42
blackburnehm02:44
blackburnstrange it fails only under valgrind02:44
-!- blackburn [~blackburn@109.226.92.17] has quit [Quit: Leaving.]03:06
shogun-buildbotbuild #191 of deb3 - modular_interfaces is complete: Success [build successful]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/19103:16
-!- Netsplit *.net <-> *.split quits: @sonney2k, naywhayare, shogun-buildbot07:33
-!- Netsplit over, joins: @sonney2k07:39
-!- Netsplit *.net <-> *.split quits: @sonney2k07:40
-!- Netsplit over, joins: @sonney2k07:40
-!- Netsplit over, joins: naywhayare07:41
-!- Netsplit over, joins: shogun-buildbot07:41
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has joined #shogun08:08
-!- gsomix_ [~gsomix@80.234.25.187] has joined #shogun08:28
-!- gsomix_ [~gsomix@80.234.25.187] has quit [Client Quit]08:29
-!- gsomix_ [~gsomix@80.234.25.187] has joined #shogun08:29
gsomix_good morning09:17
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has quit [Ping timeout: 245 seconds]09:30
-!- pluskid [~pluskid@VPN-18-101-16-244.MIT.EDU] has joined #shogun09:51
-!- pluskid [~pluskid@VPN-18-101-16-244.MIT.EDU] has quit [Ping timeout: 240 seconds]10:16
-!- pluskid [~pluskid@202.130.113.141] has joined #shogun10:38
-!- rieck [~rieck@134.76.96.43] has quit [Quit: ZNC - http://znc.sourceforge.net]10:39
-!- rieck [~rieck@134.76.96.40] has joined #shogun10:41
droopyrieck10:41
-!- rieck [~rieck@134.76.96.40] has quit [Client Quit]10:42
-!- rieck [~rieck@134.76.96.43] has joined #shogun10:43
-!- rieck [~rieck@134.76.96.43] has left #shogun []10:43
-!- rieck [~rieck@134.76.96.43] has joined #shogun10:43
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has joined #shogun11:47
rieckno action = sunday12:22
-!- gsomix__ [~gsomix@85.112.35.219] has joined #shogun12:29
-!- heiko [~heiko@host86-181-80-49.range86-181.btcentralplus.com] has joined #shogun12:31
-!- gsomix_ [~gsomix@80.234.25.187] has quit [Ping timeout: 252 seconds]12:32
-!- rieck [~rieck@134.76.96.43] has left #shogun []12:33
wikingmorning12:38
* wiking loves mosh. 3g connection lost at least 3 times, terminal connection remains \o/12:40
-!- rieck [~rieck@134.76.96.43] has joined #shogun12:41
gsomix__wiking, mosh? http://en.wikipedia.org/wiki/File:Moshpit2.jpg12:42
-!- gsomix__ is now known as gsomix12:43
rieckmorning.12:43
wikinggsomix: http://mosh.mit.edu/12:45
gsomixwiking, huh, ok. :)12:45
rieckwiking: does it require a server and a client? or does it operate onto of ssh12:47
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has quit [Ping timeout: 264 seconds]12:52
-!- heiko [~heiko@host86-181-80-49.range86-181.btcentralplus.com] has quit [Quit: Leaving.]12:55
rieckhmmm. needs extra udp port for communication. that's a downside12:55
rieckalso crypt of mosh's state protocol is not really audited12:55
rieckstill, very interesting12:56
wikingrieck: well it'll use udp13:04
wikingrieck: but yeah you need to install mosh on both server and client13:05
rieckit's a good idea for sessions over wifi and mobile phones13:11
rieckbut for slow ssh forwarding sessions it will not really help13:12
wikingrieck: well it uses udp, so that should be 'faster' a bit13:13
wikingbut yeah for me it's more the case when i'm switching between vpn-s13:13
wikingor when my connection is bad13:13
rieckyep but you need to tunnel udp over ssh13:13
rieckif a direct connection is not possible13:14
wikingit's really nice that i dont have to reconnect all the time13:14
wikinganyhow, i gotta go... bbl13:14
-!- blackburn [~blackburn@109.226.92.17] has joined #shogun14:05
-!- heiko [~heiko@host86-181-80-49.range86-181.btcentralplus.com] has joined #shogun14:17
heikosonney2k, blackburn, What do you think of classes that generate example data14:22
heikoso far I did this in the target language/interface14:22
heikobut it would be nice to have it central14:22
heikoWould be just a few for me in order to demonstrate results14:22
heikoand since the example require different number of samples, it would be cool to have a class generating the data14:23
heiko(also for streaming)14:23
blackburnheiko: why to do that?14:27
heikoblackburn, I am asking14:27
blackburnyeah I am too :)14:28
heikoI would like to have some data generation methods that I can use from python and c++14:28
blackburnwhere?14:28
heikothese would come great for a demo of my new framwork14:28
heikocurrently I implemented this in python and c++ seperately14:28
blackburnso you want to add C++ class that generates data14:29
blackburnright?14:29
heikoyeah14:29
heikojust for examples14:29
blackburnheiko: have you seen secret soeren's photo?14:29
heikowhat? :)14:29
blackburnhttps://dl.dropbox.com/u/10139213/shogun/sonne.jpg14:30
heikooh blackburn, I think for now we should change the link of the \shogunclass to the nightly doxygen14:30
heikobut I forgot the link14:30
heikowhooo, where does that come from? :D14:30
blackburn:D14:30
blackburnI don't really remember nightly link as well14:31
rieckhttps://dl.dropbox.com/u/10139213/shogun/sonne.jpg hahaha! not bad14:33
droopyyep14:33
heikoblackburn, where does that photo come from? ;)14:33
blackburnheiko: from my gimp :)14:36
blackburnrieck: have you seen him in that 'suit'? :)14:37
rieckblackburn: yes, often he's wearing such things in the office.14:38
blackburnI thought so14:38
heikoblackburn, what about the data gen classes?, any opinion?14:53
blackburnheiko: no idea actually14:53
blackburnI'd like data get methods more14:53
blackburngen*14:54
heikowhat?14:54
blackburndata gen methods I mean14:54
heikoyeah but where?14:54
blackburnfeatures may be14:54
blackburnor may be one class generating all the stuff14:55
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has joined #shogun14:55
heikoone class for all14:56
heikook, Ill add such a thing in features14:56
heikomight be usefull for a lot of examples actually14:56
blackburnagree14:56
heikoone thing that is better for fixed datasets: You can assert results14:56
heikoBut we could do that via seeding14:58
heikoif data is generated randomly14:59
heikois seeding platform independent?14:59
heikoin shogun?14:59
blackburnI hope so14:59
blackburnheiko: I have some strange issue with subset15:32
blackburnheiko: should subset be sorted?15:33
heikoblackburn, no, no need for that15:33
heikowhats the issue?15:33
blackburnheiko: what is default subset?15:34
heikono subset15:34
heikoso empty stack15:34
blackburnoh lol :D15:34
heikothen the mapping method just returns original index15:34
blackburnsorry15:34
blackburnI just called it two times15:35
blackburnon features15:35
heikoyou dont add a [0:100] subset15:35
heikooh yeah :)15:35
blackburninstead of calling on features and labels15:35
heikoblackburn do you know the computational costs of computing eigenvalues in our SGVector method?15:35
heikocannot find that15:35
blackburneigenvalues of sgvector??15:35
blackburnheiko: what is lapack routine?15:35
heikosyev15:36
blackburnsyev is something n^3/3 or so15:36
blackburnheiko: IIRC it is cholesky + iterative refinement15:37
heikoso?15:37
blackburnheiko: so n^3/315:38
blackburn+ some n^215:38
heikokk15:38
blackburnheiko: actually all eigendecompositions are sticked to n^315:41
-!- blackburn [~blackburn@109.226.92.17] has left #shogun []15:41
heikoyeah, just to be sure15:42
-!- blackburn [~blackburn@109.226.92.17] has joined #shogun15:42
-!- blackburn [~blackburn@109.226.92.17] has left #shogun []15:42
-!- blackburn [~blackburn@109.226.92.17] has joined #shogun15:42
heikowhat if one doesnt want to compute *all* of them15:42
heikothen costs decrease massively right=?15:42
blackburnheiko: sorry alt+f4 spree15:42
blackburn:D15:42
blackburnall of eigenvalues?15:42
rieckdroopy: cholesky + iterative refinement?15:42
droopyrieck: i have a very good name of pirate15:42
blackburnheiko: arpack is probably the coolest way of computing a few eigenpairs of symmetrical matrix15:43
blackburnit is kind of lanczos and krylov subspace mix up so requires only mat-vec products15:45
heikommh15:45
heikobut we dont have a wrapper right?15:45
heikowould be cool to have15:45
blackburnfor arpack? sure we have15:45
blackburnI use it everywhere in converters15:45
blackburnheiko: which eigenvalues you want to compute?15:45
heikoof a kernel matrix15:46
blackburnI mean which, largest or smallest?15:46
blackburnthere are actually one way with lapack too15:46
blackburnhttp://www.netlib.org/lapack/double/dsyevr.f15:46
heikolargest15:47
heikomaybe we should add a method to SGMatrix for that15:47
heikowould be nice to have!15:47
-!- pluskid [~pluskid@202.130.113.141] has quit [Ping timeout: 260 seconds]15:47
heikousually one doesnt use all EV15:48
blackburnhah I just googled for arpack15:48
blackburnand found that one guy in blender made a C port for his purposes15:49
blackburnpeople are crazy15:49
blackburnheiko: simplest is to use dsyevr I think15:51
heikook, could you add that? or should I?15:51
blackburnadd what?15:51
heikoshould look similar to the compute_eigenvalues we have15:51
heikobut just with additional arugment for n largest EV15:51
heikoa wrapper15:51
heikosince I dont want to mess with messy LAPACK API ;)15:51
blackburnheiko: I could15:52
blackburnah one question before15:52
blackburnis it dense matrix?15:52
blackburnif so no need to even try arpack or anything15:53
blackburnheiko: I will add in a few minutes15:56
heikoblackburn, thanks!15:56
heikodense matrix15:56
heikoyes15:56
blackburnbut test it ;)15:56
heikoI will ;)15:56
heikoadd a unit test15:56
droopy;>15:56
blackburnheiko: ok api is15:59
blackburn596 void SGMatrix<T>::compute_few_eigenvectors(const double* matrix, double*& eigenvalues, double*& eigenvectors,15:59
blackburn597                                            int n, int m, int il, int iu)15:59
blackburnheiko: i let you allocate memory for eigenvalues and eigenvectors by yourself ;)16:00
heikoblackburn, why not adding a default parameter to existing method16:00
blackburnheiko: how?16:01
blackburnwhat are parameters?16:01
heikodont you have the SGMatrix.h file? :)16:02
blackburnheiko: what are default parameters I could add?16:02
heikoI would just add num_eigenvalues=-1 in the header16:02
heikooh, 0 or negative number16:02
heikoand then check in method to decide which lapack method to use16:02
blackburnbut you would need smallest or largest eigenvalues option then16:02
heikommh16:04
heikotrue16:04
heikodont know :)16:04
-!- pluskid [~pluskid@111.120.69.121] has joined #shogun16:07
blackburnheiko: ok ready, let me smoke test it16:10
blackburnI do not really understand why you don't love lapack's api :D16:13
wikingheiko: do u agree about mock?16:20
heikowiking, I think I dont know enough about that to give an opinion, but I agree that we need to test abstract classes16:20
heikoI just saw your comment, havent read that yet16:20
wikingheiko: okok no worries16:21
CIA-18shogun: Sergey Lisitsyn master * rf4b3852 / (src/shogun/lib/SGMatrix.cpp src/shogun/lib/SGMatrix.h): Added compute few eigenvectors routine - http://git.io/pZffOw16:21
wikingyou'll see that we'll need mocking fw as well16:21
blackburnheiko: test/taste it ;)16:21
wikingheiko: to be able to test more complex things we'll need it, but then we'll need to discuss where to put those code... anyhow i guess ppl r going to be more active - hopefully - tomorrow16:22
heikowiking, yes agreed in both points16:24
heikoblackburn, thanks!16:24
wikingbtw the api of SGVector is really getting a mess16:31
wikingwith all these extra stuff with cblas16:32
wikingthey should be included within a more general func calls like operator += and such16:32
-!- gsomix [~gsomix@85.112.35.219] has quit [Ping timeout: 248 seconds]16:39
-!- heiko [~heiko@host86-181-80-49.range86-181.btcentralplus.com] has quit [Quit: Leaving.]16:56
-!- gsomix [~gsomix@83.149.21.197] has joined #shogun16:56
-!- zxtx [~zv@ool-457e7550.dyn.optonline.net] has joined #shogun16:56
-!- gsomix [~gsomix@83.149.21.197] has quit [Client Quit]16:57
n4nd0wiking: thanks for the answer :)17:00
wikingnw17:01
blackburnwiking: don't use it17:01
wikingblackburn: what? :)17:01
wikingblackburn: ah the api that i dont like? :D17:01
blackburnwiking: sgvector += etc17:01
blackburnuse eigen3 inside17:01
wikingshiiiit17:03
wikingcan we do x-compile with shogun?17:03
wikingmm it seems not really :)17:03
blackburnwhat is x-compile?17:06
wikingcross-compile17:06
-!- pluskid [~pluskid@111.120.69.121] has quit [Quit: Leaving]17:06
wikingyou know when u compile the binary on a machine for another arch17:07
wikinglike compiling stuff for arm on x8617:07
wikingsince your x86 is much faster with compiling the binary than your ar17:07
blackburnah17:08
wikingmmm it's running17:08
wikingwonder what will be the end of this :DDD17:08
blackburnwell is it shogun problem?17:08
blackburnI mean if gcc can handle this17:08
blackburnwhy shogun can't17:08
wikingblackburn: it's more like a ./configure hack17:08
blackburnwell just manually edit17:08
blackburn.config17:08
blackburnand here you go17:08
blackburn:)17:08
wikingto be able to set up well your crosscompiler enviroment via ./configure17:08
wikingto try to link against the right libs... etc17:09
wikinganyhow now it compiles17:09
wikingso let's see where it might die17:09
blackburnwhat is your target?17:09
wikingarmhf17:09
wikingnamely raspberry pi17:09
blackburnomfg17:10
blackburn:D17:10
wikingjust wanna test how it would work out on it :)17:10
blackburnI am curious17:10
wikingi mean the other i've checked an iphones license stuff17:10
wikingand libsvm's license is in it :)17:10
wikingi wonder for they are using that lib in the phone :)17:10
blackburnlibsvm in iphone??17:11
wikingyep17:11
wiking:)17:11
blackburnthat's really interesting17:11
wikingdunno for what is it there17:12
wikingbut the license is there17:12
blackburnwiking: where did you find it?17:12
wikingsettings->general->about->license17:13
blackburnhmm okay no way for me :)17:13
droopyhrhr17:13
wikingok sorry it's instead of license it's Legal Notices17:14
blackburnI mean I am not apple user :)17:15
droopyyeah17:15
wikingthe 2011 license is in there17:15
wikingblackburn: yeah i've got that part ;)17:15
wikingbut yeah it's there and something is using it :D17:15
blackburnthat sounds like svm success story17:16
wikingbtw what happened with that opencl try back in march?17:16
blackburnwiking: http://www.youtube.com/watch?v=dqDqvKYDv9M&feature=player_detailpage17:17
wiking:D17:19
blackburnso it sank17:19
wikingi still haven't figured out wtf is happening with my NIS server but let see maybe i can get you the passwd working blackburn  ;)17:23
blackburnargh17:23
wikingmm i think i have a problem in pam.d17:37
wikingblackburn: try now17:52
wikingblackburn: should be able to login17:52
wikingand even to change the passwd17:53
wikingoh shit something agian17:53
wikingok i know17:53
blackburnwiking: let me find url again :D17:54
-!- blackburn [~blackburn@109.226.92.17] has left #shogun []18:23
-!- blackburn [~blackburn@109.226.92.17] has joined #shogun18:23
-!- heiko [~heiko@host86-181-80-49.range86-181.btcentralplus.com] has joined #shogun18:32
-!- heiko [~heiko@host86-181-80-49.range86-181.btcentralplus.com] has quit [Quit: Leaving.]18:47
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has quit [Quit: leaving]18:56
-!- heiko [~heiko@host86-181-80-49.range86-181.btcentralplus.com] has joined #shogun19:03
-!- gsomix [~gsomix@178.45.35.20] has joined #shogun19:35
gsomixgood evening19:35
CIA-18shogun: Sergey Lisitsyn master * r3d873b4 / src/shogun/lib/JLCoverTree.h : Made JLCoverTree parameters static - http://git.io/8ehorw19:41
CIA-18shogun: Sergey Lisitsyn master * raaa603d / (3 files in 2 dirs): A few fixes for malsar based algorithms - http://git.io/FufqeQ19:41
CIA-18shogun: Sergey Lisitsyn master * r9bef2a5 / (3 files): Added python examples for new classes - http://git.io/uTf5Qg19:43
CIA-18shogun: Heiko Strathmann master * r9f65805 / (4 files in 2 dirs): aded DataGenerator class - http://git.io/13T9EA20:01
CIA-18shogun: Heiko Strathmann master * rf1232bc / (2 files): first working method - http://git.io/0z_Z8A20:01
CIA-18shogun: Heiko Strathmann master * r3d23819 / (2 files): better way of generating mean samples - http://git.io/pYLqdg20:01
CIA-18shogun: Heiko Strathmann master * r1b70388 / examples/undocumented/libshogun/statistics_quadratic_time_mmd.cpp : make use of new data generator - http://git.io/G3ktEA20:01
CIA-18shogun: Heiko Strathmann master * r8e101b7 / (5 files in 3 dirs): Merge pull request #668 from karlnapf/master - http://git.io/JQVv1A20:01
-!- heiko [~heiko@host86-181-80-49.range86-181.btcentralplus.com] has quit [Quit: Leaving.]20:02
CIA-18shogun: Sergey Lisitsyn master * r0c3cb28 / (src/shogun/lib/SGMatrix.cpp src/shogun/lib/SGMatrix.h): Renamed conflicting matrix name in compute_few_eigenvectors - http://git.io/VVGpvw20:34
CIA-18shogun: Sergey Lisitsyn master * r82a598b / src/shogun/transfer/multitask/Task.cpp : Removed unused variable from CTask - http://git.io/-U8NAA20:34
CIA-18shogun: Sergey Lisitsyn master * rb99b671 / src/shogun/optimization/lbfgs/lbfgs.h : Removed explicit links in lbfgs - http://git.io/ZiavQw20:34
CIA-18shogun: Sergey Lisitsyn master * rb5c0c62 / src/shogun/mathematics/SparseInverseCovariance.h : Fixed sparse covariance estimation doc - http://git.io/HE4ReA20:34
CIA-18shogun: Sergey Lisitsyn master * r80530a6 / src/shogun/classifier/FeatureBlockLogisticRegression.h : Fixed doc of feature blocked logistic regression - http://git.io/2FfYjQ20:38
CIA-18shogun: Sergey Lisitsyn master * r8446c45 / (3 files in 3 dirs): Fixed a few undocumented warnings - http://git.io/CoXNoA20:38
-!- wiking [~wiking@info2k1.hu] has quit [Changing host]21:24
-!- wiking [~wiking@huwico/staff/wiking] has joined #shogun21:24
-!- gsomix_ [~gsomix@88.200.214.225] has joined #shogun22:31
-!- gsomix [~gsomix@178.45.35.20] has quit [Ping timeout: 252 seconds]22:33
@sonney2kgsomix_, sleeping already?23:03
gsomix_gsomix_, nope. hard working23:03
@sonney2kblackburn, didn't know you are an artist :D23:03
@sonney2kgsomix_, cool!23:03
blackburnsonney2k: you are23:03
blackburn:)23:03
-!- gsomix_ is now known as gsomix23:03
droopy;>23:03
gsomixdroopy, meow23:03
droopygsomix: what nick?23:03
@sonney2kblackburn, did you finally get the website to work?23:04
blackburnsonney2k: I didn't even try23:04
gsomixsonney2k, http://cat.sonnenburgs.de.meowbify.com/soeren/category/23:04
gsomixmeow23:04
* sonney2k wonders which internet meme will hit #shogun / the mailinglist next23:07
@sonney2kand who will be the next superhero to the rescue of course :D23:07
@sonney2kblackburn, I was like *megarotflbtc* seen what was going on between you / heiko23:07
blackburnsonney2k: in pics?23:08
blackburnyeah that was pretty funny :)23:08
wikinghttp://www.tugberkugurlu.com/Content/images/Uploadedbyauthors/wlw/1934ffeb460c_FB47/you-need-some-tests-yo.jpg23:10
@sonney2kwiking, btw about the tests - is it possible to keep these in a separate dir? I liked heikos idea of having a tests/shogun/<dir structure replicated>23:10
wiking;)23:10
droopy8)23:10
wikingsonney2k: yeah shouldn't be a problem at all...23:10
wikingsonney2k: it's just going to be a bit funky to keep somehow in sync the directory structure23:11
@sonney2kgsomix, miau is the German term for that :)23:11
gsomixsonney2k, huh, ok. :)23:11
wikingsonney2k: or u trust everybody that they are going to be concious enough about that? :)23:11
@sonney2kgsomix, but funny - didn't know that stuff exists :)23:12
@sonney2kwiking, well at least I know at least one other big java project that does exactly that23:12
@sonney2kand there are people doing the PR merges23:12
@sonney2kso renames should not be too hard to detect23:13
@sonney2kwiking, ^23:13
wikingsonney2k: okey23:13
wikingsonney2k: then i'll just git mv those stuff23:13
wikingsonney2k: src/tests/...23:13
wiking?23:13
wikingright?23:13
@sonney2kno23:13
wikingsonney2k: ./tests...?23:14
@sonney2knot sure - we currently have ../testsuite23:14
@sonney2kwe need some better name for that23:14
wikingwe can make ./unit-tests23:14
wikingand then it's solved23:14
@sonney2kor ../tests and then unit_test / regression_tests ?23:15
@sonney2kI mean move testsuite -> ../tests/regression_tests?23:15
@sonney2kI don't have a proper name for these btw23:15
wikingmmm23:15
@sonney2kbut I would prefer one directory for all tests (with subdirectories...)23:16
wikinganyhow i think having some funky git-hooks would be great after we create the unit-testing23:16
@sonney2k?23:16
wikingis there a way to add git-hooks in github?23:16
@sonney2kfor what?23:16
@sonney2kdoing what?23:16
wikingwell git is very powerful with hooks23:16
wikingso you can make a hook that it doesn't take in a commit23:17
wikingif u don't define for example corresponding unit tests for the new class23:17
@sonney2k*urgs* that is tough23:17
@sonney2kno idea if that is possible23:18
@sonney2kand if that is what we want23:18
wikingi mean u dont want to take commits without at least adding the correspoinding unit test.cc under unit tests dir23:18
@sonney2kbecause PR mergers check for tests once we have the infrastructure23:18
wikingok23:18
wikingi mean this would just take off some load from the admins23:18
-!- heiko [~heiko@host86-181-80-49.range86-181.btcentralplus.com] has joined #shogun23:19
wikingfunnily23:19
wikingi've seen somebody done this with python23:19
wikingon a project23:19
wikingthere's a pre-commit hook23:20
wikingand if the unit testing fails23:20
wikingit doesn't allow u to commit ;)23:20
@sonney2kwiking, well I don't mind but lets get the infrastructure set up fist23:20
@sonney2kfirst23:20
@sonney2kso ../tests/unit ?23:20
@sonney2kso ../tests/regression ?23:20
@sonney2kor what?23:21
wikingwhat's actually now in testsuites?23:21
@sonney2kwiking, the regression tests running large examples and checking if the output is still the same23:21
wikingheiko: how do u feel about that awesomeness? :D23:21
wikingsonney2k: ok well then yeah23:21
wikinggit add tests && git mv testsuits tests/regression && git add tests/unit23:22
@sonney2kI would expect some unit tests to be regression tests too but hey...23:22
wikingsonney2k: yeah especially if we'll use mocking23:22
@sonney2kblackburn, any comments?23:22
wikingheiko: http://bbgen.net/blog/2011/06/junit-git-hooks/23:22
wiking;)23:22
blackburnuhoh23:23
@sonney2kblackburn, ?23:23
blackburnsonney2k: haaard to think23:23
wikingblackburn: think harder :)23:23
droopy:)23:23
blackburnI hardly think hard23:23
wikingdroopy: get some vodka for blackburn23:23
droopywiking: +m please!?!23:23
wiking:DDDDDDD23:24
heikowhats going on here? ;D23:24
wikingheiko: discussing unit testing framework23:24
heikoalright, this site describes a nice thing that we should have!23:24
wikingheiko: i've just solved the thing for you: adding unit testing into git pre-commit hook so that u cannot commit unless the unit testing succesfully runs23:24
@sonney2kblackburn, get some vodka^H^H^H^Hfuel23:25
heikowiking, that is awesome!23:25
wikingheiko: yeah it'd be awesome i think as well23:25
wikingheiko: same can be done with valgrind23:25
blackburnI am just not drunk enough to discuss23:25
heikosonney2k, I saw a strange picture of you ;)23:25
wikingheiko: for mem leaks23:25
heikowiking wow23:26
heikoyou really get into that!23:26
blackburnpre-commit hook is something interesting23:26
blackburnproblem is23:26
heikohey blackburn, this time, I am a bit drunk too! :)23:26
blackburnhow can we23:26
blackburnwork on something when something is broken?23:27
blackburnheiko: I am not at all unfortunately ;(23:27
heikocant we just turn it off via a flag in this case?23:27
wikingheiko: heheh now the question is whether we can add hooks in github23:28
droopygreat wiking23:28
wikingheiko: so that when you fork git repo for yourself then you actually get the pre-commit hook as well23:28
heikowiking, I mean if its too hard, we can just rely on people performing tests before submitting patches23:28
heikoah ok23:28
heikothats better23:28
@sonney2kblackburn, it is not for *running* the test - just for checking that it is there23:29
@sonney2kso that is not a problem23:29
blackburnahh23:29
blackburnso we check23:29
blackburnwhether class has a test?23:29
@sonney2kwiking, ^ ?23:29
wikingsonney2k: yeah we can do that23:30
@sonney2kblackburn, is the directory structure good for you?23:30
blackburnyeah I am ok with it23:30
wikingsonney2k: and run unit testing as well as part of the pre-commit23:30
@sonney2kheiko, ?23:30
@sonney2kok too?23:30
wikingsonney2k: so that way buildbots will have less erros ;)23:30
heikoyes23:30
wikingsonney2k: ok i'll run that command now in my utests branch23:30
wikingsonney2k: and re-do the Makefile...23:31
wikingsonney2k: but i think you should really check my changes in Makefile.template and ./configure23:31
wikingsonney2k: just that u agree with them... i'm sure there's some funkyshit i shouldn't have done there23:31
@sonney2kwiking, actually I don't expect fewer buildbot failures - people just never build shogun for all interfaces / all configs23:31
@sonney2ke.g. heiko never builds with lapack disabled23:31
wikingsonney2k: :))) ok then i was wrong... anyhow23:32
wikingthen i'll do the quick changes now23:32
wikingdo and extra commit and let's see how we proceed23:33
wikingi'll add google mock checking as well23:33
wikingsince we'll most definitely will need that stuff23:33
heikowiking is the abstract class thing soved already?23:33
wikingfor making more complex running checks23:33
wikingheiko: google mock will solve it for us23:33
wikingheiko: and with this directory struct23:33
@sonney2kmock mock hurray!23:33
wikingheiko: it'll be not confusing at all23:33
heikonice!23:33
heikonow not even we are paid by google, we are even using their frameworks23:34
wiking:>23:35
wikinglol23:35
wikingmmm find script to regenerate the dir structure ... :>23:35
wikingsonney2k: do we want tests/unit/interface23:36
wikingafaik there won't be such23:36
wikingsonney2k: it should be just tests/unit/<directories under src/shogun>23:36
wikingor?23:36
@sonney2kwiking, this is what I understood is what you / heiko want23:36
wikingyep23:37
wikingdoing it23:37
@sonney2kkeeping the core sane and interface tests will be another chapter23:37
wikingsonney2k:  is it ok if i write a simple Makefile by hand in tests/unit/ ?23:41
wikingsonney2k: one thing23:42
wikingsonney2k: imo libshogun should not be installed anywhere for unit tests23:42
wikingsonney2k: i.e. we shouldn't do it as for the examples...23:43
-!- heiko [~heiko@host86-181-80-49.range86-181.btcentralplus.com] has quit [Ping timeout: 264 seconds]23:43
wikingsonney2k: and that'll be hard to require having libshogun locally built under ./src/shogun from ./tests/unit23:43
wikingany ideas how to solve this?23:43
wikingwithin a Makefile context?23:43
wikingso that if you do cd tests/unit && make then if ./src/shogun/libshogun.<dynamiclibrary> is not present then the building of the library is first called...23:44
wikingsonney2k: ^23:46
wikingmoreover i need some variables set in src/.config that is present only after ./configure has been exectuted... :S23:47
-!- heiko [~heiko@host86-181-80-49.range86-181.btcentralplus.com] has joined #shogun23:49
heikosonney2k, blackburn, what do you think of letting the buildbot compile the tutorial and produce a nightly latest version as for the doxygen?23:52
blackburnyeah we would need it at some point and may be even now23:53
heikothis way everybody could always read the current version without having to git pull and LaTeX around?23:53
wikingheiko: second that23:54
-!- heiko [~heiko@host86-181-80-49.range86-181.btcentralplus.com] has quit [Quit: Leaving.]23:57
-!- heiko [~heiko@host86-181-80-49.range86-181.btcentralplus.com] has joined #shogun23:58
--- Log closed Mon Jul 23 00:00:17 2012