TL;DR

Encountered an out of bounds error when calling model.orig_vars().

When calling this method, this underlying piece of code is called:

This loops over all variables (n_vars) and retrieves them from SCIP. The problem here is that the number of original variables can be different from the number of variables currently in the problem.

When querying the original variables, n_vars should (I believe) be retrieved via ffi::SCIPgetNOrigVars(self.raw). Also see; https://www.scipopt.org/doc/html/group__GlobalProblemMethods.php#ga6072e2a6e7e1d54410e19ddcaae32cd6

--- Context:
I was working on the Branch-and-Price example found here. One of the bonus exercises asks us to initialize the problem with a smart selection of variables. When SCIP starts solving the problem, all variables are transformed. As a result, new variables (with new indexes) are created by SCIP. This is an issue when you are using the Ryan-Foster branching rule, as you need to know what the pattern looks like when branching. If you have a variable that is transformed from an original variable, then you need to figure out during solving which original variable it relates to because then you can retrieve the object representation. For this I wanted to use the orig_vars() method, to check which transformed variables represent them.