Hi,
Consider the results of the following Particle Sets job (intersecting):
I expect, based on this, that if I supply these three slots to a new job, I should get the union - ~28+143+147=318k particles.
However, instead I get this when supplying all three to an NU-refine job:
What am I missing, or is this a bug?
Cheers
Oli
@olibclarke Please can you share with the developers job reports and particles *.cs files for
I will send you a private message regarding file transfer arrangements.
Hey @olibclarke,
Thanks for reporting. I took a look at your cs files and it appears this discrepancy is a unintuitive behaviour that results from using Particle Sets Tool to intersect based on blob/path (and blob/idx, the index of the particle in that file), rather than based on uid. When connecting the particles to the downstream NU-refine job, the job takes the union of all inputs based on uid.
These two can result in differing sets particularly in the case where symmetry expansion was done at some point – then uids across all particles will be unique, but more than one particle can point to the same blob/path and blob/idx.
Could you try running particle sets tool with intersecting based on uid, and seeing if the total size of the outputs then adds up to the same number as when they are connected to a downstream job? As well, let us know what the workflow looked like up until now, and likely we can figure out the most sensible next step (in particular, do you recall doing symmetry expansion at some point in this workflow?)
Note: another confusing thing is that, when intersecting based on blob/path, the intersect and intersect_B outputs can actually have differing uids, which means they both need to be connected to any downstream job to retain all particles input
Best,
Michael
Hi @mmclean,
Thanks! This workflow did not include symmetry expansion, as it is an asymmetric complex. I just tested, and yep, when intersecting on UID, it behaves as expected.
Cheers
Oli
