Camelia, the Perl 6 bug

IRC log for #cdk, 2008-11-17

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

All times shown according to UTC.

Time Nick Message
01:52 Conrad joined #cdk
06:17 CIA-37 cdk: egonw * r13143 /cdk/branches/cdk-1.2.x/src/ (3 files in 3 dirs): Added tow Ga atom types
06:37 CIA-37 cdk: egonw * r13144 /cdk/branches/cdk-1.2.x/src/ (3 files in 3 dirs): Added a Ge atom type; completes atom types for all elements up to atomic number 36
06:38 jbrefort joined #cdk
08:19 egonw joined #cdk
08:33 egonw joined #cdk
08:34 sneumann joined #cdk
08:52 jonalv joined #cdk
09:11 thomasku joined #cdk
09:12 thomasku moin
09:12 zarah hi thomasku
09:18 egonw hi thomasku
09:18 thomasku hi egonw
09:37 sneumann Hi there
09:37 sneumann Together with shk3 I have started looking at a problem
09:38 sneumann with CDK fingerprinting, I get BitSets with length!=1024
09:38 sneumann but 1024,1020, 1000,996 etc, so several 1024,
09:38 sneumann but also some "almost" 1024 and alwas modulo 4
09:38 sneumann Anything known on this front ?
09:39 sneumann Otherwise we'll keep digging into that
10:04 egonw please do
10:04 egonw busy now, but will join in later
10:04 egonw sneumann: which version are you debugging?
10:05 sneumann 1.0.4, shk3 just sent me some patched version
10:05 sneumann my svn co is just running
10:05 sneumann apparently things work (with my molfiles) for him
10:05 sneumann and with my check code
10:17 egonw oh :(
10:17 egonw more 1.0.4 problems... :(
10:17 egonw I guess I have to hurry with 1.2.0 ...
10:17 egonw sneumann: any reason not to switch to 1.1.x ?
10:18 egonw the fingerprinter there is much faster
10:19 sneumann Is that the branch or trunk ?
10:19 egonw the cdk1.2.x branch
10:20 sneumann Binary jar anywhere ? Or build myself ?
10:22 egonw 1.1.1 is on SF
10:22 egonw and will create an 1.1.2 this week
10:22 egonw from SVN: ant clean dist-large
10:24 sneumann running, thanks
10:27 shk3 joined #cdk
10:28 sneumann Hi shk3,
10:28 sneumann just built 1.2.x, going for lunch now.
10:29 sneumann afterwards I'll find out new API for BitSet bs = new ExtendedFingerprinter().getFingerpr​int(molgo,arf.findAllRings(molgo));
10:41 gilleain joined #cdk
11:00 shk3 sneumann: did you read my latest email? I get the impression it's not a bug, it's a feature
11:12 sneumann Feature of BitStet, but if the downstream Stuff expects 1024 bits ?
11:13 sneumann i.e. if your (sub)structure search wants to compare 1024 bits vs. 996 bits ?
11:13 sneumann Should that be handled in my code, or in the CDK ?
11:19 egonw sneumann: compare 1024 bits with 996 bits?
11:19 egonw wouldn't that give binning issues?
11:19 egonw a 1024FP bit != a 996FP bit
11:23 sneumann Sorry, you dont have shk3's mail.
11:23 sneumann Die Apidoc zur size()-methode sagt: "Returns the number of bits
11:23 sneumann of space actually in use by this BitSet to represent bit values.
11:24 sneumann so if the highest 1-Bit is at position 996, BitSet will have length 996
11:25 shk3 but I get 1024 in any case, it seems
11:30 shk3 it looks like the code using cdk needs to handle that
11:30 egonw sneumann: oh... clearly it should say 1024 if it is a 1024 FP
11:30 egonw and not 996 if that just happens to be th highest used bit
11:31 shk3 well, I get the impression BitSet doesn't have a size in that sense
11:31 egonw guess not
11:31 shk3 BitSet seems to use the hightest set bit for size
11:31 shk3 that's the problem, it seems
11:31 egonw it only remembers which bits are set
11:31 egonw sounds like a CDK bug, then
11:31 shk3 but how should cdk enforce the bs to be of a certain size
11:31 shk3 if bs has no size?
11:32 egonw ah, it does not have to do that
11:32 egonw it just need to return the proper size() and not rely on the BS for that
11:32 sneumann BitSet bs = fingerprinter.getFingerprint(mol,rs, rslist);
11:32 sneumann System.out.println(molid + " " + bs.length());
11:33 sneumann So its not CDK's fault
11:33 sneumann but really a BitSet issue. So it would be the application to work with different lengths
11:33 egonw in that set up, yes
11:34 egonw what about using fingerprinter.length()
11:34 egonw or the appropriate method...
11:34 egonw checking...
11:34 egonw IFingerprint.getSize()
11:35 egonw System.out.println(molid + " " + fingerprint.getSize());
11:35 egonw see http://cheminfo.informatics.indiana.edu/~r​guha/code/java/nightly-1.2.x/api/org/opens​cience/cdk/fingerprint/IFingerprinter.html
11:35 zarah egonw's link is also http://tinyurl.com/6xc3ou
11:35 sneumann that'll gives always 1024
11:35 egonw yes
11:35 egonw sneumann: ok, then I do not understand your use case
11:35 egonw what is it you try to acchieve?
11:36 sneumann I adopted some of shk3's code from NMRshiftDB to calculate the fps,
11:36 sneumann to work with algorithms that expect fixed-width-bitstrings
11:37 egonw BitSet.length()
11:37 egonw BitSet.size()
11:37 egonw first returns the highest bit
11:37 egonw second returns the actual size
11:38 egonw which should not have changed in our use case, from the initial new BitSet(1024)
11:38 egonw will write a unit test for that...
11:38 egonw hang on
11:38 sneumann those who can read have a distinct advantage.
11:39 sneumann Changed to bs.size() and things are working
11:39 egonw hehehehe
11:39 sneumann hmpf
11:40 sneumann leaves the question why shk3 had the proper result with my code
11:40 sneumann or did you fix things unnoticedly ?!
11:40 egonw no, no code touched
11:42 sneumann I meant shk3, whether he fixed my stuff ...
11:44 egonw ah, ic
11:45 CIA-37 cdk: egonw * r13145 /cdk/branches/cdk-1.2.x/src/ (2 files in 2 dirs): Added unit test and explaining JavaDoc about BitSet.size()/.length()
11:49 shk3 hm, I think I didn't use your code
11:50 shk3 wrote a test case myself for PB000006
11:50 sneumann So you got it right from the start :-)
11:51 sneumann We'll come back to you once Björn had a look at the fingerprints
13:11 Julio joined #cdk
13:12 Julio left #cdk
13:19 Julio joined #cdk
13:20 Julio left #cdk
13:36 Julio joined #cdk
13:36 Julio left #cdk
13:39 Julio joined #cdk
13:39 Julio left #cdk
14:32 Julio joined #cdk
14:32 Julio left #cdk
16:04 shk3 sneumann: do you know how long bmc bioinformatics takes to come up with the final verson of a (our) paper?
16:20 sneumann We have a paper on alignment in the same issue, and that's already finished, IIRC
16:21 sneumann bye,
16:21 sneumann Steffen
16:38 shk3 hm, takes ages
17:31 jbrefort joined #cdk
17:34 carsten joined #cdk
20:33 edrin joined #cdk

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