Camelia, the Perl 6 bug

IRC log for #cdk, 2008-10-26

| Channels | #cdk index | Today | | Search | Google Search | Plain-Text | summary

All times shown according to UTC.

Time Nick Message
01:50 bag__ joined #cdk
05:58 jbrefort joined #cdk
09:01 CIA-36 cdk: egonw * r12809 /cdk/branches/cdk-1.2.x/src/ (4 files in 2 dirs): Fixed AtomTypeManipulator.configure() to not set values for which the AtomType source NULL; Added configureUnsetProperties() method for only setting NULL in the target Atom. (closes #1969156)
09:14 carsten joined #cdk
09:44 edrin joined #cdk
12:39 egonw joined #cdk
13:34 CIA-36 cdk: egonw * r12812 /cdk/branches/cdk-1.2.x/src​/test/org/openscience/cdk/ (39 files in 21 dirs): Removed all uses of the old CDKTestCase
13:34 CIA-36 cdk: egonw * r12813 /cdk/branches/cdk-1.2.x/src/test/or​g/openscience/cdk/CDKTestCase.java: Removed obsolete JUnit3 class
13:40 CIA-36 cdk: egonw * r12814 /cdk/branches/cdk-1.2.x/src​/test/org/openscience/cdk/ (294 files in 63 dirs): Renamed NewCDKTestCase to CDKTestCase
14:04 rguha joined #cdk
14:04 rguha hi
14:04 zarah hi rguha
14:05 rguha regardng the lp regressions - are you looking at it? If not I'll fix it this afternoon
14:05 rguha sorry, that was intended for egonw, not zarah :)
14:07 egonw hi rguha
14:07 rguha hi egonw
14:07 egonw no, not looking at it
14:07 egonw enough CDK for today
14:07 rguha ok
14:07 egonw need to finish a draft book chapter
14:08 rguha have fun
14:08 egonw hahaha :(
14:08 rguha :)
14:08 egonw really bad timing on my side
14:08 egonw not that I could have realistically anticipated that...
14:08 egonw but bad timing nevertheless
14:08 rguha bad timing for what?
14:08 egonw that book chapter
14:09 rguha aah
14:09 rguha what's it about?
14:10 egonw 3D representation
14:11 rguha btw, the ballester method for 3D sim, is quite crappy. I knew it wasn't going to be very good, but it turns out that it's hardly better than 2D similarity :(
14:11 egonw that sounds like a paper
14:12 rguha I was actually thinking of a blog post
14:12 rguha burnt out from writing :(
14:12 egonw yeah... writing sucks
14:12 rguha aah, well, better get back to writing applications
14:12 rguha ttyl
14:12 egonw there is so much more enjoying scientific communication
14:12 rguha I agree
14:13 egonw and -they- have to default to writing :(
14:13 egonw bummer
16:18 CIA-36 cdk: egonw * r12815 /cdk/branches/cdk-1.2.x/src/main/org/open​science/cdk/io/PCCompoundXMLReader.java: Fixed exception message: IMolecules is supported, not IChemFile
17:04 carsten joined #cdk
17:39 CIA-36 joined #cdk
19:47 jbrefort joined #cdk
19:54 CIA-36 cdk: rajarshi * r12816 /cdk/branches/cdk-1.2.x/src/test/org/opens​cience/cdk/qsar/DescriptorEngineTest.java: Removed unecessary print statements
20:07 CIA-36 cdk: rajarshi * r12817 /cdk/branches/cdk-1.2.x/src/test/org/ope​nscience/cdk/qsar/descriptors/molecular/ (6 files): Updated harness to provide class names of failing descriptors. Updated a bunch of descriptors so their setUp method is public as required by JUnit4
20:09 egonw joined #cdk
20:24 CIA-36 cdk: rajarshi * r12818 /cdk/branches/cdk-1.2.x/src/test/org/​openscience/cdk/qsar/descriptors/prot​ein/TaeAminoAcidDescriptorTest.java: Added copyright header. Fixes bug #2197592
20:27 rguha joined #cdk
20:27 rguha hi egonw
20:28 egonw hi rguha
20:28 egonw I'm running into the ReaderFactory bug again...
20:28 rguha what was that?
20:28 egonw nasty one...
20:28 egonw 2153298
20:28 rguha btw, the lp checker regression bug is found
20:28 egonw discovered that by adding unit tests...
20:29 egonw ah, happy to hear that :)
20:29 egonw then at least something works
20:29 rguha apparently it's not the fault of the recent commits
20:29 rguha adding explicit H's is not setting the atom type nameof the H;s
20:31 egonw so, what triggered the failing tests then?
20:32 egonw that's actually a good suggestion...
20:32 rguha I have a feeling that some tests were not being run
20:32 egonw I mean... the H-adder knows that the H's are simply H's, so just as well should set the atom type name
20:32 egonw ah
20:32 egonw possible indeed
20:32 rguha calling percieve... fixes it
20:32 egonw I also found a class that was never run
20:32 egonw we need a unit test for that too :)
20:33 rguha :)
20:33 egonw umm... better just have the H-adder set the atom type name
20:33 rguha I think after we convert everyhting to JUnit4 we'll be safe
20:33 egonw instead of new perception
20:33 rguha right
20:33 rguha one sec
20:33 egonw as we are not really changing the constituion
20:34 rguha right
20:34 rguha btw, did you see the mail regarding bond counts?
20:35 egonw yes
20:35 egonw was wondering about i too
20:35 CIA-36 cdk: rajarshi * r12819 /cdk/branches/cdk-1.2.x/src/main/o​rg/openscience/cdk/tools/manipulat​or/AtomContainerManipulator.java: Updated so that explicit H's have their atom type name set to "H" (since they can't be anything else!)
20:36 egonw not sure how the bond count is defined...
20:36 egonw I think it more sense to have heavy weight bond count only...
20:36 rguha it should ignore H's
20:36 egonw then it should
20:36 rguha since you can always work out how many H's need to be added
20:36 egonw and then the explicit H count counting should be fixed
20:36 rguha correct
20:38 egonw it's getting somewhere...
20:38 egonw I still have a few patches around...
20:38 rguha aah, 80 tests now fixed :)
20:38 egonw yeah :)
20:38 egonw but did not have time to merge that yet
20:39 egonw had a hard time getting the new CDK plugins working properly...
20:39 rguha ok
20:39 egonw Bioclipse/Spring is using a complex claasspath loading system... :(
20:40 rguha where does spring comeinto this?
20:40 egonw loading of the rcp plugins
20:41 egonw on top of osgi
20:41 egonw it's not fun at all
20:41 rguha aah
20:42 CIA-36 cdk: rajarshi * r12820 /cdk/branches/cdk-1.2.x/src/test/or​g/openscience/cdk/qsar/descriptors/​atomic/AtomicDescriptorTest.java: Updated to list the descriptor name when a failure occurs
20:47 CIA-36 cdk: rajarshi * r12821 /cdk/branches/cdk-1.2.x/src/main/or​g/openscience/cdk/qsar/descriptors/​molecular/BondCountDescriptor.java: Updated to explicitly ignore bonds to H's, so that we only count bonds to heavy atoms
20:51 rguha is there any shortcut method to strip H's?
20:51 egonw yes
20:52 egonw AtomContainerManipulator I guess, not sure though
20:52 rguha aah, right
20:52 rguha thanks
20:55 CIA-36 cdk: rajarshi * r12822 /cdk/branches/cdk-1.2.x/src/main/org/​openscience/cdk/qsar/descriptors/mole​cular/PetitjeanNumberDescriptor.java: Made sure that the calculation ignores H's, explicit and implicit
20:59 CIA-36 cdk: rajarshi * r12823 /cdk/branches/cdk-1.2.x/src/test/org/​openscience/cdk/qsar/descriptors/mole​cular/MolecularDescriptorTest.java: Added a more meaningful error message
21:05 egonw rguha: for human readabiliy you can better use exc.getMessage() than exc.toString()
21:05 rguha aah, OK
21:06 CIA-36 cdk: rajarshi * r12824 /cdk/branches/cdk-1.2.x/src/ (2 files in 2 dirs): Updated so that it works independent of H's. Also made sure to make a local copy of the molecule since it was doing aromaticity perception. also since we use a local copy no need to reset flags on exit
21:07 CIA-36 cdk: rajarshi * r12825 /cdk/branches/cdk-1.2.x/src/test/org/​openscience/cdk/qsar/descriptors/mole​cular/MolecularDescriptorTest.java: Updated to use more human readable exceptions
21:10 egonw ok, giving up on the ReaderFactory for today... :(
21:10 egonw bloody thing
21:10 rguha i might vive it a look and see if anything turen up
21:10 egonw will add that second mol desc test method, the one for hybrid bond info
21:10 egonw rguha: that would be appreciated
21:10 rguha ok
21:11 egonw the thing is this:
21:11 egonw 1. it needs to detect gziped files
21:11 egonw 2. support Reader and InputStream
21:11 egonw 3. and needs to detect the format of the file
21:11 rguha ok
21:11 egonw so, lot's of resetting...
21:12 rguha hmm, I'll try :)
21:12 egonw because the actually parsing must start at the start of the file again
21:12 egonw and that's where it now goes wrong...
21:12 rguha I thought JDK provides a way to detect gzipped files
21:12 egonw it might...
21:13 egonw the current code has stuff done by Michael Howard (Jmol fame)
21:13 rguha ok
21:13 CIA-36 cdk: rajarshi * r12826 /cdk/branches/cdk-1.2.x/src/main/org​/openscience/cdk/qsar/descriptors/mo​lecular/ZagrebIndexDescriptor.java: Updated so that it ignores explicit H's
21:13 rguha i love unit tests :) catching so many bugs
21:13 egonw yes, it's amazing... and scary
21:13 egonw CDK is full of bugs
21:14 rguha right!
21:14 egonw partly because of the flexible data structure
21:14 egonw might have been a bad design...
21:14 rguha also, mahy aren't bugs, so much as not documented
21:14 rguha many things are assumed
21:14 egonw yes, yes, indeed
21:15 egonw I have been fighting these things for many years now
21:15 egonw which is why I tend to be a bit harsh now and then...
21:15 rguha well, a few more bugs squashed :)
21:15 rguha understandable
21:15 egonw original authors about never actually get to fix things later...
21:15 egonw so, that has been mostly up to me as release manager ;(
21:15 rguha that's the problem with open source - maintenance
21:15 egonw well...
21:16 egonw proper code up front
21:16 rguha especially with something as niche as the CDK, the expertise is not always easily avaiklable
21:16 egonw that's the problem
21:16 egonw CDK has been way too relaxed
21:16 egonw fairly... we needed the critical mass of functionality
21:16 rguha I agree - but even the core devs can't be experts on every aspect of cheminfo
21:16 egonw nope
21:20 egonw have you met Tobias in person?
21:20 rguha no
21:20 egonw neither have I
21:20 rguha I owe him many beers :)
21:20 CIA-36 cdk: rajarshi * r12827 /cdk/branches/cdk-1.2.x/src/main/org/o​penscience/cdk/qsar/descriptors/molecu​lar/KappaShapeIndicesDescriptor.java: Made sure to ignore H's
21:20 egonw I hope Chris will invite him to the CDK workshop
21:20 egonw as speaker
21:20 rguha that'd be a good idea
21:21 rguha did you mention his name to Chris?
21:21 egonw feel free to relay that message to Christoph too :)
21:21 egonw yes, I did
21:21 rguha oh ok
21:21 egonw but if you do it too, it makes more impact :)
21:21 rguha right, I'll suggest it
21:21 egonw not sure how much funding Chris has for speakers though...
21:21 egonw and he has to pay you and me too :)
21:22 egonw actually, not sure about the exact plans...
21:22 rguha my presence is not fixed yet
21:22 rguha depends on what my job is next semester :(
21:23 egonw oh, still nothing fixed yet ?
21:23 egonw shame to hear that
21:23 rguha no, still searching
21:23 rguha yes, I'd love to be part of a CDK workshop/hack session
21:23 rguha we'd be able to get a tone of stuff done :)
21:23 egonw yes, think so
21:24 rguha ah well, I'll make use of the Internet :)
21:24 CIA-36 cdk: egonw * r12828 /cdk/branches/cdk-1.2.x/src/test/org/​openscience/cdk/qsar/descriptors/mole​cular/MolecularDescriptorTest.java: Added unit test for robustness against differences in representation of bond order info
21:25 rguha is the reader problem in trunk or 1.2.x?
21:26 egonw 1.2
21:26 egonw there should be unit tests failing because of it...
21:26 rguha can you point me to one you wer elooking at? there are lost of failures at http://cheminfo.informatics.indiana.edu/~rguh​a/code/java/nightly-1.2.x/test/result-io.html
21:26 egonw check the failing setReader_[Reader|InputStream]
21:26 egonw tests
21:27 zarah rguha's link is also http://tinyurl.com/53jjtq
21:27 rguha ok
21:27 egonw ok, got to hit my bed now...
21:27 egonw we'll visit Stockholm tomorrow
21:28 egonw won't be online
21:28 rguha good night
21:28 egonw I commited this second test...
21:28 egonw for the C=C vs C.sp2-C.sp2 representaion option stability for mol descs
21:28 rguha great
21:33 egonw ok, bye for now
21:52 rguha joined #cdk
22:11 CIA-36 cdk: rajarshi * r12829 /cdk/branches/cdk-1.2.x/src/main/org/op​enscience/cdk/io/CrystClustReader.java: Updated setReader(Reader) to check the input argument rather than the instane variable. Clears up the test failure
22:18 CIA-36 cdk: rajarshi * r12830 /cdk/branches/cdk-1.2.x/src/test/org/open​science/cdk/io/Gaussian98ReaderTest.java: Updated test to use the proper test file name!
23:33 edrin joined #cdk

| Channels | #cdk index | Today | | Search | Google Search | Plain-Text | summary