1/ With 10000 particles the xml files are about 40Mb, it is already a
little heavy. With 100000 particles it would be about 400Gb and I
imagine it is quite difficult to open such a file, to edit the first
twenty line to define the simulation.
Save simulations to .xml.bz2, it is much smaller (no extra work for you,
except the longer extension), yade handles that transparently. Some text
editors (at least gvim does) let you edit bz2-compressed files also
transparently, without having to decompress them by hand.
2/ In my mind, the definition of the particle assembly with its
interactions and the definition of the simulation are two different
stuffs. For instance you can perform different simulations from a same
particle assembly, thus you don't need for each of these different
simulations a different xml file defining the same initial particle
assembly with the same initial interactions.
Or reciprocaly you can perform an identical simulation from different
particles assemblies...
But, maybe you will tell me that it is not a problem thanks to python
;-) in every case it is only a question, what is your opinion?
You anticipated right :-). I personally have the simulation defined in
python, but I load spheres assemblies from text files
(utils.spheresFromFiles, there is some function in c++, but I don't
remember). That way I can run the same tension-compression test on
different assemblies. And use the same assembly for different simulations.
Those files describe really just the position and readius, not
interactions.
If you load a .xml in python, you can however consequently modify engines,
for example, and keep the same body and interaction parameters as you saved
them.
The extras for python are that it can, for example, run tension, then
compression, then save the plot to some file and exit. You can read
different parameters from text files, then running 40 jobs with different
parameters overnight and compare results etc (see
http://yade.wikia.com/wiki/ScriptParametricStudy for a very simple example)
Vaclav
