Re: Archiver

[Marcus G. Daniels]

LRA> It isn't clear to me with your new "Archiver" whether you can use the 
LRA> -putShallow or -putDeep methods repeatedly to save data from multiple
LRA> time steps as a model runs, all in one data base.

You can, provided that it is feasible to keep all the objects around
until which time `save' is run.  If it is, then it is just a matter of
attaching timestamps to all the keys, e.g.

sprintf (key, "fish%06lu, getCurrentTime ());   

Unfortunately, the Archiver class isn't really set up to be used in
this way.  In its current form, it's really only intended to be used
for loading and saving a fixed, related set of objects.  

All the Archiver does is provide an interface for naming the objects
that get serialized, and it provides some `stream packaging'.

Using the serialization methods directly (like in Glen's recent post)
would probably make more sense in this case.  For you purposes,
that would probably mean direct use of the HDF5Object class (in order
to read the objects back in to R).  That might be more trouble than
its worth, unless the state in a fish is very complex.

I'd like it better if the Archiver did lazy evaluation and had `dirty'
flags for modified objects, but I haven't had time to implement that yet.

The goal with the HDF5 archiving capability is to be able to read and write
objects in a random access fashion.  The goal with the Lisp archiving
capability is to be able send objects over a wire.  The former hasn't
been achieved because the Archiver just saves all the objects that
are registered, without regard to whether they've ever been changed.

                  ==================================
   Swarm-Support is for discussion of the technical details of the day
   to day usage of Swarm.  For list administration needs (esp.
   [un]subscribing), please send a message to <address@hidden>
   with "help" in the body of the message.