My current system is matlab and python code, mainly pattern based analysis where a graph structure is constructed and updated by analyzing each incoming bar.
The quickest way I imagine to move to Zorro is to make a Lite-C driver to call my code, return trading signal(s) and use the Zorro for order managements, etc.
Since my algo needs quite a bit of historical data (in the range of 10K bars), it doesn't make sense to recompute the whole thing for each incoming bar.
So my question is: is it possible for Zorro to start an R(matlab to R is easy..)(or Python) session, and keep it alive (so that states are maintained), and Zorro will call R/Python to incrementally compute on the next incoming bar?

Thanks.