Accessing parent functions from a nested class

[bsaunder]

Hi,

This is my first post, so I hope it is ok. I am writting a C++ application and I have been asked to incorporate a log file mechanisim. The structure of the code I have so far is something like:

class A_Class {
   void someFunctionsInA();
};
 
class B_Class {
   void someFunctionsInB();
};
 
class App_Class {
public:
   int x;
   A_Class a;
   B_Class b;
 
   int writeToLogFile(string s);
};
 

I want the functions in the classes a and b to be able to use the function writeToLogFile(string) but I am not sure how this should be done. Can you advise me please?

Do I have to create a separate writeToLogFile(string) function in each of the classes a and b? I would prefer to keep all the code that deals with the log file in one place.

[bsaunder]

ignoring the "int x;" - that was not meant to be there :-P

[bsaunder]

I think I have found one answer: use a static function.

class A_Class {
public:
   void someFunctionsInA();
};
 
class B_Class {
public:
   void someFunctionsInB();
};
 
class App_Class {
public:
   A_Class a;
   B_Class b;
   static int writeToLogFile(string s);
};
 

Then in the A_Class and B_Class functions I can use something like the following to use writeToLogFile(string s) function:

void A_Class::someFunctionsInA() {
App_Class::writeToLogFile("This log message was generated in the A class function.");
}


I guess this solution is ok if writeToLogFile can be made into a static function but if I wanted to have several applications running at the same time each with it's own log file then I guess that using a static function isn't going to work?

Ben.

[weaknessforcats]

Where's the logfile object????

I would expect you wouild have a logifle class to handle the log file.

class App_Class
{
private:
   A_Class a;
   B_Class b;
 
   LogFile  theLog
 
public:
 
    void AMethod();
};
 

The you could:

void App_Class::AMethod()
{
     //get a data string from a
     theLog->Write(data);
}
 

By using a LogFile pointer you assocuate your App_Class to a particular LogFile. Different App_Class objects could use different log files.

Now you can design a object-oriented hierarchy of log files so that you can record different error information as needed.

I don't think you need a static function because the App_Class has no idea how to write to a a LogFile since it is an App. Only a LogFile should know how to write to a log file.

[bsaunder]

Thank you for your reply ...

Where's the logfile object????

I don't think there is any need for so many "?" please try not to sound so rude!

I would expect you wouild have a logifle class to handle the log file.

As you have guessed I did actually already create a log file class however I found that I wanted to do some pre-processing of the string before I sent it out to the log file object. I don't just want to send the string to the log file (for example I want to display every log message on the HMI screen, among other things which means interacting with other object). Thus I wrote a single function that would do all the things I wanted to do with the log string, including sending it to the log file object.

I didn't want to confuse the issue by listing all the classes and all the objects in my code. What I was asking was how to use this function in the inner classes A_Class and B_Class. The fact that it is a log file function is (I thought) irrelevant because there maybe cases when I might want to use any generic application function in the inner classes.

I don't think you need a static function because the App_Class has no idea how to write to a a LogFile since it is an App. Only a LogFile should know how to write to a log file.

Indeed, the App_Class has no idea how to write to a log file since this is the log file objects job but this is what I have done already. However it is not the App_Class that I am asking about, it is the inner classes A_Class and B_Class. If all I was worried about was the log file object then I would pass this as a pointer to the inner classes but I want to pass all of the functionality of writeToLogFile(string s). In my original postI asked if I would have to write this functionality out in each of the inner classes, I would then have to pass to them the pointers for all the objects involved (including the log file object).

I don't think that a static function is the right way forward.

With this new information are you able to provide some more feedback?

Thanks

[weaknessforcats]

You might be needing a Visitor. There's an article on the Visitor design pattern in the C/C++ HowTos forum.

[weaknessforcats]

I don't think there is any need for so many "?" please try not to sound so rude!

I did not intend to be rude. Repeating characters must be some sort of arcane text-messaging rule. I my case I have a stuttering keyboard. Sorry.

[bsaunder]

Thanks weaknessforcats,

I had a look at the Visitor HowTo post and to be honest I got myself a little confused about how to apply it to my own problem. However, I did not know about "forward references" and so after reading this post I realised that I could just make a pointer to the App_Class object in the A_Class and B_Class objects. Below is the current version of my test code that compiles and does what I want:

#include "stdafx.h"
#include <iostream>
#include <string>
#include "LogFile.h"
 
using namespace std;
 
const string LOGPATH = "D:\\Documents and Settings\\ben\\My Documents\\temp\\";
const string LOGNAME = "benTestLogFile";
const string LOGEXTN = "txt";
 
class App_Class; // forward reference
 
class A_Class {
private:
    App_Class* appyInAPtr;
public:
    A_Class(App_Class* appyCaller);
    void someFunctionsInA();
};
 
class B_Class {
private:
    App_Class* appyInBPtr;
public:
    B_Class(App_Class* appyCaller);
    void someFunctionsInB();
};
 
class App_Class {
public:
    App_Class();
    A_Class a;
    B_Class b;
 
    LogFile log;
 
    bool writeToLogFile(string s);
};
 
App_Class::App_Class() : a(this), b(this)
{
}
 
A_Class::A_Class(App_Class *appyCaller) {
    appyInAPtr = appyCaller;
}
 
void A_Class::someFunctionsInA() {
    appyInAPtr->writeToLogFile("Running the A class function.");
}
 
B_Class::B_Class(App_Class *appyCaller) {
    appyInBPtr = appyCaller;
}
 
void B_Class::someFunctionsInB() {
    appyInBPtr->writeToLogFile("Running the B class function.");
}
 
bool App_Class::writeToLogFile(string s) {
    log.write(s);
    cout << "[IN LOG FILE]: " << s << endl;
    return true;
}
 
 
//int _tmain(int argc, _TCHAR* argv[])
int main()
{
    // Local Variables
    char endChar;
    App_Class appy;
 
    // set up the log file
    appy.log.setLogFile(LOGPATH, LOGNAME, LOGEXTN);
 
    cout << "Output is: " << endl << endl;
    appy.writeToLogFile("Testing log file:");
    appy.a.someFunctionsInA();
    appy.b.someFunctionsInB();
    cout << endl << "done!" << endl;
    cout << endl;
 
    cout << "Press any key (and enter) to end." << endl;
    cin >> endChar;
 
    return 0;
}
 

The output of the application is:

> Output is:
>
> [IN LOG FILE]: Testing log file:
> [IN LOG FILE]: Running the A class function.
> [IN LOG FILE]: Running the B class function.
>
> done!
>
> Press any key (and enter) to end.

And I get in my log file benTestLogFile_2008_02_27.txt (the date is automatically appended by the log file name):

> [2008/02/27 12:42:21]: Testing log file:
> [2008/02/27 12:42:21]: Running the A class function.
> [2008/02/27 12:42:21]: Running the B class function.

This compiles ok but I do get a couple of the following compiler warning:

warning C4355: 'this' : used in base member initializer list

I guess these are nothing much to worry about?

[weaknessforcats]

warning C4355: 'this' : used in base member initializer list

The this pointer is not valid in a constructor initilizer list since the object is not yet built. Potentially, you could call a derived method that uises variables that aren't initialized.

I would tuse the this pointer in the constructor initializer list.

What you are trying to do is get the derived address into the base object.

The Visitor uses a base method that has a Visitor* argument. That base method is overriden by a derived method that also takes a Visitor* argument.

The derived method calls a Visitor->VisitDerived(this).

VisitDerived is a derived class of Visitor.

So you end up in a VisitDerived object with the address of the derived object in the other hierarchy.

Visitor is great.

Say you have a Car derived form Vehicle and you suddenly need to do a cost analysis on various Cars but there is no cost in formation in the Car class.

You write a CarCostAnalyzer class (derived from Visitor) and you call the Car object using Vehicle->Visit(CarCostAnalyzer*).

Boom! You end up at CarCostAnalyzer::VisitDerived with the this pointer of the Car object.

Now you can access all Car methods regardless of whether they are in the Vehicle base class or not.

[bsaunder]

Thanks for your patients Weaknessforcats,

I wouldn't use the this pointer in the constructor initializer list.

I guess anything that generates a warning has to be questionable and so I guess I shouldn't use the this pointer in the constructor initializer list either.

What you are trying to do is get the derived address into the base object.

Does this mean getting the derived address of the App_Class object into the base objects of A_Class and B_Class? (Although App_Class is not a derived class, neither is A_Class or B_Class).

The Visitor uses a base method that has a Visitor* argument. That base method is overridden by a derived method that also takes a Visitor* argument.

The derived method calls a Visitor->VisitDerived(this).

VisitDerived is a derived class of Visitor.

So I make a Visitor class which is derived from some base class that has a method with a Visitor* argument. Then the Visitor class should overwrite this method?

Should my Visitor class be derived from App_Class or from A_Class/B_Class?
Should the VisitDerived class be the App_Class or the A_Class/B_Class or is this some completely different class?

So you end up in a VisitDerived object with the address of the derived object in the other hierarchy.

So VisitDerived should be the A_Class/B_Class thus ending up with the address of the derived object in the other hierarchy (APP_Class)? That's what I want right?

So in my case I would want to create a Visitor class and derive it from App_Class. I would need to put into the App_Class an empty method that takes an argument Visitor* and then in the Visitor class I should overwrite this method.

So far I have this:

class Visitor;        // forward reference
 
class A_Class {
public:
    void someFunctionsInA();
};
 
class B_Class {
public:
    void someFunctionsInB();
};
 
class App_Class {
public:
    A_Class a;
    B_Class b;
 
    LogFile log;
 
    bool writeToLogFile(string s);
 
    void runVisitor(Visitor *v) {};
};
 
class Visitor : public App_Class {
    void runVisitor(Visitor *v) {
        // Not sure what to put in here.
        // v->writeToLogFile(s)
    };
};

This compiles ok. The next thing I thought I was to do was to make the A_Class and B_Class derived from Visitor but I can not do this. It seems that I am creating a circle of declarations. I must have miss-understood something.

lets have a look at the example ...

Say you have a Car derived form Vehicle and you suddenly need to do a cost analysis on various Cars but there is no cost in formation in the Car class.

Or in my case: I have a class A_Class that is not derived from anything and I suddenly need to record its progress in a log file but there is no log file function in the A_Class class.

You write a CarCostAnalyzer class (derived from Visitor) and you call the Car object using Vehicle->Visit(CarCostAnalyzer*).

Ah! So I should write a completely new class? called logFileInterface and derive this from Visitor (is this the VisitDerived class mentioned above?) which in turn is derived from App_Class? But then I can't call A_Class->runVisitor(logFileInterface*) because the A_Class isn't a Visitor object.

...

I hope you can see why I am getting confused. I am sure that once I figure this out it will all suddenly become clear. I will keep on trying to figure this out by reading your comments above and the HowTo post on C++ Visitors.