ObjectLoader & batch processing

[Sven Thommesen]

I've started looking at how to use the ObjectLoader/ObjectSaver
functionality to do some serious batch-mode processing. Several questions
have come up, which I hope one of the gurus may be able to answer:

a) I see the ObjectLoader code parsing the relevant files for "@begin" then
the data then "@end", ONCE, then closing the file (?). If one wanted to run
a thousand runs varying the input parameters, how is it envisioned that the
data for each run get changed? Externally to the Swarm app, or is there some
relatively simple way to do it in Swarm?

b) If done within Swarm, I need some way to completely reset the sim between
runs. A couple of sub-questions: 

(i) To get rid of generated objects from the previous run, is it enough to
nuke the modelSwarm (and what's the proper way to do that), or do I have to
do a recursive descent to nuke individual objects? (In which case a
'cleanup' method needs to be defined for all swarms ...)

(ii) What scheduling objects etc. need to be cleared away?

  [ This ties in with one of the items on our wish list: a "start over"
button on the control panel... ("reload" in mousetraps does not seem to work.) ]

c) Regarding the objects being loaded with ObjectLoader (saved with
ObjectSaver), are we limited to just a plain linear list of simple data
types? I tried to put a struct as a data item in a  batchSwarm, and this
compiles fine and does as intended. However:

(i) the probe machinery, when asking for a complete probe, displays the name
of the struct but none of the data. OK, the documentation explains why. Now,
when I define a custom probe map, should I not be able to specify parts of
the struct that are simple data types? I've been unsuccessful specifying
"structName.varName" and "structName->varName".

(ii) the ObjectSaver does not save structs! Looks like it declines to save
data members marked as non-interactive, which happens automatically to
structs. Any way around this? 

Sven