Camelia, the Perl 6 bug

IRC log for #cdk, 2012-08-13

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

All times shown according to UTC.

Time Nick Message
04:03 egonw__ joined #cdk
05:32 egonw__ joined #cdk
05:57 jbrefort joined #cdk
06:33 Gpox joined #cdk
06:35 konditorn joined #cdk
07:06 egonw__ joined #cdk
07:31 bag__ joined #cdk
08:06 jonalv joined #cdk
09:19 kiceek joined #cdk
11:18 kiceek joined #cdk
12:16 Alchimiste joined #cdk
12:16 Alchimiste moin
12:35 egonw_ joined #cdk
13:13 egonw__ joined #cdk
13:17 egonw__ hi Alchimiste
13:18 Alchimiste hi ebonw
13:18 Alchimiste sorry
13:18 Alchimiste hi egonw
13:18 egonw :)
13:20 kiceek joined #cdk
13:20 egonw hi kiceek
13:44 kiceek hi!
14:23 egonw kiceek: you are using the CDK?
14:24 kiceek partly, our system is based on chemaxon oracle cartridge, but if I need something quick and custom, I use cdk
14:25 kiceek mol file reading/processing etc
14:25 egonw OK, sounds good
14:25 egonw there is a CDK-based oracle cartridge too
14:26 egonw http://www.jcheminf.com/content/1/1/17
14:26 egonw I have no clue how they compare performance-wise
14:27 kiceek yes, i have seen this one. too bad our system is too much interwined with chemaxons tools that it would be total redesign to migrate
14:27 egonw understood
14:28 kiceek altough i'm trying to slowly replace some parts of it with opensource tools
14:28 * egonw mumbles vendor-lock-in...
14:28 kiceek :) aha
14:28 egonw would be nice if CDK and ChemAxon would interoperate...
14:28 kiceek well there are things that are not replacable, structure to name for example
14:28 egonw indeed, indeed
14:28 egonw hence, interoperate :)
14:29 egonw nothing wrong with ChemAxon...
14:29 kiceek we bought licence and now are generating all the names as fast as we can for our 23mil database before licence expires
14:29 egonw heheheh
14:29 egonw does the license allow you to use them beyond the license?
14:30 kiceek no, it's annual
14:30 egonw them -> the names
14:31 kiceek ah names, yes
14:31 kiceek it's just the functionality that get's disabled
14:31 kiceek and names are saved in our own fields, outside chemaxon generated molecular descriptor table
14:35 kiceek does cdk generate standard and non-standard inchi?
14:35 egonw it should be able to do both
14:38 kiceek api does not mention what kind of inchi is generated
14:39 egonw by default the StdInChI is created
14:40 egonw if you construct a InChIGenerator with additional options (like FixedH) then you get a non-standard InChI
14:40 * egonw reads from his API book :)
14:40 egonw latter is like:
14:40 egonw generator = factory.getInChIGenerator(
14:40 egonw mierezuur, "FixedH"
14:40 egonw );
14:41 egonw with just the mierezuur IAtomContainer, then you get a standard InChI
14:44 Gpox joined #cdk
14:44 kiceek hmm! should try tomorrow to replace our vendorlocked generator with cdk
14:45 kiceek and see how fast it is
14:46 egonw looking forward to the numbers!
14:46 kiceek i wonder if those optimisations and modifications they have made to c code really make anything faster or no
14:46 egonw lovely...
14:47 egonw taking advantage of open source, but not making your own stuff open source... happens too often :)
14:47 egonw but surely the send those patches back to the InChI team, right?
14:48 kiceek i hope so
14:48 egonw same here...
14:48 egonw because then the CDK will benefit from it too
14:48 egonw (and many other projects)
14:48 egonw oh, and make sure to use the SilentChemObjectBuilder
14:49 egonw (if you do not rely on change events to be passed around)
14:49 kiceek because jni-inchi crashes a lot when used in parallel
14:49 kiceek but chemaxon's generator not
14:49 egonw ouch
14:49 kiceek so it seems they have optimised somethin in that regard
14:49 kiceek optimized
14:51 kiceek ou they do provide modified jni-inchi source on their website
14:51 egonw oh... got a URL for me?
14:51 kiceek hmm...I should do diff on it
14:52 kiceek http://www.chemaxon.com/marvin​/help/formats/srcjniinchi.zip
14:52 egonw http://sourceforge.net/mailarchive/forum.php​?thread_name=4C7E33DC.2030502%40users.source​forge.net&forum_name=jni-inchi-discuss
14:53 egonw looking at the date, maybe these changes were already merge in...
14:53 egonw checking the repos log...
14:53 egonw (mmm... got 5 minutes left, before going home...)
14:54 kiceek looks like all the stuff is in .dll's
14:54 egonw at least is it post the 0.7 release
14:54 kiceek that is, this modified jni-inchi only downloads respective dll's for 32bit/64bit windows
14:55 kiceek i don't see any optimizations they have talked about...but maybe spokeperson just did not quite understand what he was talking about @ conference
14:56 kiceek anyway, will try to compare this with cdk
14:58 egonw the jar will indeed have .dll for windows...
14:58 egonw but all should be generated from the source...
14:59 egonw I will ask Sam if he intergrated any changes from ChemAxom...
14:59 egonw need to ask him about newer InChI option for tautomerization anyway...
14:59 egonw ttl
15:03 kiceek joined #cdk
15:19 egonw joined #cdk
15:26 egonw ok, just looked at the jni-inchi from trunk/
15:26 egonw I can compile it on my machine...
15:29 egonw kiceek:
15:29 egonw very few differences, in fact
15:29 egonw they only made a few things public
15:29 egonw moved around lib loading a big
15:29 egonw moved around lib loading a bit
15:30 egonw and removd logging
15:30 egonw the latter will help performance a bit, indeed
15:31 kiceek yet, they print their own logs to system.err
15:31 kiceek tb even warnings
15:31 kiceek and only way to disable it is to override system.err write() method
15:31 egonw :(
15:31 kiceek but they promised to fix it in some future version
15:31 kiceek it is very annoiying
15:32 kiceek annoying
17:21 kiceek joined #cdk
17:32 egonw joined #cdk

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