Currently, loading a model with a different order than the current instance has some unintended behavior. If the on disk model has an order lower than the one we are using, we need to expand the nodes to fit the current order by appending empty tokens to the memory chain; if the order is higher we need to drop the additional tokens.
This will not work without fault either but should be good for now. In long terms, think of a way of automatically rebuilding the edge map if a model of differing order is encountered. This would require recalculating each path in the graph and then changing the nodes and edges accordingly, which for larger models could be very time consuming.
Currently, loading a model with a different order than the current instance has some unintended behavior. If the on disk model has an order lower than the one we are using, we need to expand the nodes to fit the current order by appending empty tokens to the memory chain; if the order is higher we need to drop the additional tokens.
This will not work without fault either but should be good for now. In long terms, think of a way of automatically rebuilding the edge map if a model of differing order is encountered. This would require recalculating each path in the graph and then changing the nodes and edges accordingly, which for larger models could be very time consuming.