[Open Babel] Antwort: Re: Serializing OBMol to memory

[<ern...@ba...>]

>On the other hand, if you're scanning through an entire database of 
molecules, then this strategy won't help.  In that case, the serialization 
methods you're talking >about would be a benefit ... but there is no such 
method.

If I can identify the main brakesman in the instantiation of the OBMol and 
some prcomputed data to speed this up can be serialized as ancillary data 
along with the Molfile, no full serialization may be needed.

>In my experience, though, parsing a MOL file is not the bottleneck, it's 
the SMARTS matching that is slow.  Serializing would help somewhat, but 
the OB SMARTS pattern >matcher is overdue for some serious optimizing.

Unfortunately I'm out of my league with the intrinsics of the matcher, so 
i'll stick to fast instantiation for the moment. :-)

For my sample search.

Full search: 1844 ms

Full search without real match: 625 ms

Full search without real read and real match, only returning true: 94 ms

So the match eats about 2/3 of the time BUT the reading of the molecule 
takes the majority of the rest. If i can seriously speed this up, another 
~25% have been shaved off.

I have taken a brief look at mdlformat.cpp. Aside from plain text parsing 
I found two calls that might be of interest:

mol.AssignSpinMultiplicity();

and the _mapcd stuff.

I think I now have to start commenting out code and test to find 
interesting candidates. :-)

Mit freundlichen Grüßen

Ernst-Georg Schmid
_________________________________________

BBS-S&T-APS-AD

Bayer Business Services GmbH
Gebäude 4810
51368 Leverkusen, Deutschland

Tel: +49 214 30 50250 
Fax: +49 214 30 22999
E-Mail: ern...@ba...
Web: http://www.BayerBBS.com

Geschäftsführung: Vorsitzender Daniel Hartert   |   Arbeitsdirektor 
Norbert Fieseler
Vorsitzender des Aufsichtsrats: Klaus Kühn
Sitz der Gesellschaft: Leverkusen   |   Amtsgericht Köln, HRB 49895