Originally posted by: beyondzhao

From the website of CP optimizer, https://www.ibm.com/support/knowledgecenter/SSSA5P_12.5.1/ilog.odms.cpo.help/CP_Optimizer/Advanced_user_manual/topics/advcpoptimizer.html, I learned something useful about the cp optimizer extensions. But the version that I use is Python, namely docplex.cp. 

I just want to know whether it's possible to use CP Optimizer extensions in Python. And how to implement same functions such as defining a new propagator, writing goals, and so on.

Originally posted by: Olivier OUDOT

Hello Xiqiang,

Python interface to CPO is limited to the modeling functions, and more generally to all what can be exported in a CPO file format.

All functions that you are talking about have to be implemented as "callbacks" that are called from the solver. These are available from the native C++ interface, but not from Python.

Hope this helps.

Originally posted by: beyondzhao

Hi Olivier, 

Thanks for your reply.

Does you mean that it's impossible to custom steps of search algorithms in Python?

Is there any other way defineing a new propagator in Python? Or I have to switch to C++ interface?

Originally posted by: Olivier OUDOT

Hello Xiqiang,

Yes, it is what I mean. You can do that only in C++ and Java, C++ being the more complete and powerful API.

Anyway, it is usually not a good idea to do that because it avoids solver to use existing or future specific enhancements on your model. 

Did you have a look to Search Phases that may help you to solve your problem ? They are available in Python using modeling function search_phase().