调用类的函数，但作为参数传递

Question

I am quite new in C++ programming (but much more used to IDL and Python). I am actually trying to re-write a code that I have already written in IDL into C++, to improve its efficiency. In my project, I wish to call a function within a class, from another class. Below, I copied what I think is the essential part of the code required to understand the problem. However, I simplified it in order to improve its clarity.

When compiling with g++, I obtain the following error:

MALA.cpp: In member function ‘void MALA::update_position_MH(Model_def*, Data*, Config*, int)’:
MALA.cpp:277:52: error: no matching function for call to ‘Model_def::call_model(Data&)’
propos_model=(*model_class).call_model(*data_struc); // PROBLEM HERE! TO DEBUG!
                                                ^
In file included from MALA.h:13:0, 
                 from MALA.cpp:29:
 model_def.h:34:12: note: candidate: Eigen::VectorXd Model_def::call_model(Data*)
      VectorXd call_model(Data *data_struc); // call a model using its name
        ^
 model_def.h:34:12: note:   no known conversion for argument 1 from ‘Data’ to ‘Data*’

What I do not understand is that the code compiles without errors if I comment the line (see at the very end of the given code):

 propos_model=(*model_class).call_model(*data_struc); // PROBLEM HERE!

But then, why I do not get an error as well for the line:

vars_new=new_prop_values((*model_class).vars, m); // This works fine

What am I doing wrong when passing the structure 'Data' as an argument of the function call_model?

Here is the simplified code, spread in different files:

This is the model_def.h (simplified)

#include <Eigen/Dense>
#include <string>
#include "config.h"
#include "data.h"

class Model_def{
       string model_name;
       bool relax[];
       int plength[];
       VectorXd cons;
       long Ncons, Nvars, Nparams;
       string vars_names[];
       string cons_names[];
       string params_names[];
    public:
       Model_def(string m_name, bool rlx[], int plgth, VectorXd params_in, string params_in_names); // The constructor
       Eigen::VectorXd params;
       Eigen::VectorXd vars;
       Eigen::VectorXd call_model(Data *data_struc);
};

This is the model_def.cpp (simplified)

#include <Eigen/Dense>
#include <iostream>
#include <iomanip>
#include "model_def.h"

Eigen::VectorXd Model_def::call_model(Data *data_struc){

    bool passed=0;

    if(model_name == "Model_MS_Global"){
       passed=1;
       return Model_MS_Global(params, plength, (*data_struc).x); // This function is in a dedicated cpp file (not shown here)
     }

// Other things happening...
}

This is the data.h

#include <Eigen/Dense>
#include <string>
struct Data{
    Eigen::VectorXd x;
    Eigen::VectorXd y;
    long Nx; // Ny is not checked but should be as long as x
    string xlabel[];
    string ylabel[];
};

This is the MALA.h (simplified)

#include <Eigen/Dense>
#include <string>
#include "model_def.h"

class MALA{
    private:
        int seed; // to generate random numbers
        double epsilon1;
        Eigen::MatrixXd epsilon2;
        double A1;
        double delta;
        double delta_x;
    public:
        void update_position_MH(Model_def *model_class, Data *data_struc, Config *cfg_class, int m);
 };

This is the MALA.cpp (simplified)

void MALA::update_position_MH(Model_def *model_class, Data *data_struc, Config *cfg_class, int m){

    VectorXd vars_new;
    double* u=uniform_01(1, &seed ); // Generate uniform random number between 0 and 1
    VectorXd propos_model;

    if((*cfg_class).proposal_type == "Random"){
        vars_new=new_prop_values((*model_class).vars, m); // This works fine
    }

    propos_model=(*model_class).call_model(*data_struc); // PROBLEM HERE!

 }

Answer 1

data_struc is a pointer, and call_model wants a pointer, so you just pass it directly. Putting a * in front "dereferences" it and makes it not a pointer anymore.

 propos_model=(*model_class).call_model(data_struc);

Or, a bit better

 propos_model = model_class->call_model(data_struc);