From: Bernd W. <Ber...@un...> - 2006-11-25 10:33:12
|
I've been noticing that the Molecule data structure - as available on the trunk - leaves a larger memory footprint than it did in the July snapshot (last stable release 14. July 2006). I have a couple of smiles strings (337) available, which I parsed to Molecule objects and serialized them (using an ObjectOutputStream) to a byte array. I did that using the (1) 20060714 release and (2) the latest trunk sources. Most of the molecules (about 2/3) increased in size, some of them are now 3 times as big. My personal favorite is c12-c3:c(-C(-c4:c(:c(-[H]):c5:c(:c:2:4):c(-C(=O)-c2:c(:c(:c(:c(-[H]):c-5:2)-[H])-[H])-[H]):c(-[H]):c:1-[H])-[H])=O):c(:c(:c(:c:3-[H])-[H])-[H])-[H] whose byte array size increased from 88kB using the july code to 426kB (using cdk-svn-20061125.jar). Does anyone have a clue why that has changed so dramatically? In KNIME (http://www.knime.org) we use serialization to not hold objects in memory, i.e. for us "size does matter". Thanks Bernd PS: If requested, I can send the little test class and the data around. |