Find a solution to the problem of serialization of trackcpp objects

[xresende]

one idea is to add getstate and setstate methods to every c++ class at trackcpp that is exported to python through SWIG. These methods could serialize its objects into strings accessible from python.

[fernandohds564]

I was looking at the pickle documentation and it seems the best way of doing this is defining the methods setstate and getstate in the python level. These methods could call other cpp methods which do the serialization de facto. For example, in the pyaccel.accelerator.Accelerator class, we could define the getstate with a call to read_flat_file and the setstate with a call to write_flat_file with the appropriate arguments for them to return a trackcpp.String object and a trackcpp.Accelerator object, respectively.

[fernandohds564]

I have implemented both methods for pyaccel.accelerator.Accelerator class and it is working. 😃
It is very fast! a list with 20 machines takes only 1.5s to pickle and 1.8s to unpickle.
It also consumes only 630KB of space in disk, because of the compression. 😄

If we also want this for the trackcpp.Accelerator class, we need to change the interface.

[xresende]

sorry, it took me some time to respond to your comments, @fernandohds564 ... but for the next time, i promise to either respond faster or never.

I think for now we do not need to implement serialization in trackccp.Accelerator, we can close this issue, right?