Hi Krab, You're absolutely right : there is bound to be some molecule that some descriptor will hang on. I was just distracted by the particular example you gave.
I'm pretty sure that the Descriptor architecture doesn't allow individual descriptors access to global state. However, there is the slight possibility of IO resources still being open if the thread is killed. For example, some of them seem read atom type information from config files. Sorry, but I don't know the qsar package very well, so I should really let someone else reply. gilleain On 11/14/11, Krab <sed...@flih.cz> wrote: > > Hi gilleain, > my problem is not specific to any particular molecule or descriptor. The > point is that for some molecules, the computation may hang. I cannot control > which smiles could be uploaded so there might be possibly any garbage. > > Arsenious acid with BCUT was the point where I found that this problem can > arise. Also, pubchem > (http://pubchem.ncbi.nlm.nih.gov/summary/summary.cgi?cid=518740&loc=ec_rcs) > lists the molecule, so I expect it to be a valid input. > > Krab > > > gilleain torrance wrote: >> >> Hi Krab, >> >> I'm not sure if descriptors are interruptible in the way you describe. >> It might be good for there to be a general ITimeout interface for >> classes, the same way that the ring finder uses. >> >> However, a small point : are you sure about that smiles of arsenious >> acid. It seems to be a ring system: >> >> http://www.daylight.com/daycgi/depict?4f315b41735d324f5b41735d314f32 >> >> gilleain >> >> On 11/14/11, Krab <sed...@flih.cz> wrote: >>> >>> Hi, >>> is there any way to interrupt computation of >>> org.openscience.cdk.qsar.IMolecularDescriptor? >>> >>> For example, the computation of BCUT of arsenious acid (O1[As]2O[As]1O2) >>> hangs for long time, I left it to run during weekend and it still did not >>> finish. >>> >>> I need to run the calculations in batch on server where I can't predict >>> which molecules a user would upload. So I need to stop each descriptor's >>> calculation after some timeout even if it's not finished. >>> >>> >From what I've seen in the source, I think the descriptors don't check >>> their >>> interrupted state, so interrupting a thread would be pointless. Now I can >>> imagine only to start new thread for each calculation and eventually use >>> Thread.stop(). >>> >>> Is there a better way? Also will this be safe? I presume it should be OK, >>> but I can't inspect every descriptor to tell whether an exception at >>> arbitrary location could leave any global object at inconsistent state or >>> any resource open. >>> >>> Thank you for your time. >>> >>> Krab >>> -- >>> View this message in context: >>> http://old.nabble.com/Interrupting-descriptor-computation-tp32839007p32839007.html >>> Sent from the cdk-user mailing list archive at Nabble.com. >>> >>> >>> ------------------------------------------------------------------------------ >>> RSA(R) Conference 2012 >>> Save $700 by Nov 18 >>> Register now >>> http://p.sf.net/sfu/rsa-sfdev2dev1 >>> _______________________________________________ >>> Cdk-user mailing list >>> Cdk-user@lists.sourceforge.net >>> https://lists.sourceforge.net/lists/listinfo/cdk-user >>> >> >> ------------------------------------------------------------------------------ >> RSA(R) Conference 2012 >> Save $700 by Nov 18 >> Register now >> http://p.sf.net/sfu/rsa-sfdev2dev1 >> _______________________________________________ >> Cdk-user mailing list >> Cdk-user@lists.sourceforge.net >> https://lists.sourceforge.net/lists/listinfo/cdk-user >> >> > > -- > View this message in context: > http://old.nabble.com/Interrupting-descriptor-computation-tp32839007p32839548.html > Sent from the cdk-user mailing list archive at Nabble.com. > > > ------------------------------------------------------------------------------ > RSA(R) Conference 2012 > Save $700 by Nov 18 > Register now > http://p.sf.net/sfu/rsa-sfdev2dev1 > _______________________________________________ > Cdk-user mailing list > Cdk-user@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/cdk-user > ------------------------------------------------------------------------------ RSA(R) Conference 2012 Save $700 by Nov 18 Register now http://p.sf.net/sfu/rsa-sfdev2dev1 _______________________________________________ Cdk-user mailing list Cdk-user@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/cdk-user
