C++Talk.NET

tojohere@gmail.com · Guest

Hello,
I want to make a member variable of my class as it can be modified only
in a specific member function(s).
Is that possible in any way....?
I tried many ways but finally I arrived in a dirty idea...( Also not
that perfect )... but it works...
The code is shown below....
-----------------------------------------------------------------------------------------------------------
class Boy
{
private:
const int m_nCount;
public:
Boy()
:m_nCount ( 0 )
{
}
void SetCount( int argCount)
{
*( const_cast <int *> ( &m_nCount ) ) = argCount;
// This can be possible from anywhere
// in the class if the programmer is
// that much crooked to do like this... Smile

}
int GetCount()
{
return m_nCount;
}
void SomeFunc()
{

m_nCount++; // Error : compiler will prevent this just
becoz its a const.

SetCount( GetCount() + 1); // This will work without any
problem.
// I want the programmer to
force
// to do like this when he
modify the m_nCount
}
};

int main(int argc, char* argv[])
{
Boy b;
b.SetCount( 10 );
int a = b.GetCount();
return 0;
}
-----------------------------------------------------------------------------------------------------------
I feels this as dirty idea because I heard that casting away cost is
not a good practice.
Does any body know any other better idea to do this?
Thanks in advance for you valued advice.

[ See http://www.gotw.ca/resources/clcm.htm for info about ]
[ comp.lang.c++.moderated. First time posters: Do this! ]

Mark Van Peteghem · Guest

You could define a new class Count that has the member variable
m_nCount, that is private, and the methods SetCount and GetCount.
Then derive Boy from that class:

class Count
{
private:
int m_nCount;
public:
Count(int c): m_nCount(c) {}

void SetCount( int argCount)
{
m_nCount = argCount;
}
int GetCount()
{
return m_nCount;
}
};

class Boy: public Count
{
public:
Boy(): Count ( 0 )
{
}

void SomeFunc()
{
// Error : compiler will prevent this
// because it's private in superclass.
m_nCount++;

// This will work without any problem.
// I want the programmer to force
// to do like this when he modify the m_nCount
SetCount( GetCount() + 1);
}
};

--
Mark dot Van dot Peteghem at q-mentum dot com
http://www.q-mentum.com -- easier and more powerful unit testing

[ See http://www.gotw.ca/resources/clcm.htm for info about ]
[ comp.lang.c++.moderated. First time posters: Do this! ]

pinto · Guest

why do not u try to use a parent class and put the m_nCount in the
parent

class human
{
protected:
void setcount(int _iCount){i_Count=_iCount;};
int GetCount(){return i_Count;};
private:
int i_Count;
};

class Boy: public human
{
};

[ See http://www.gotw.ca/resources/clcm.htm for info about ]
[ comp.lang.c++.moderated. First time posters: Do this! ]

Hendrik Belitz · Guest

[email]tojohere (AT) gmail (DOT) com[/email] wrote some weird stuff.

Why do you want to do such a thing at all? Is there a good reason for not
allowing all member functions access to the member variable?

--
- Abort, Retry, Fthagn? -

[ See http://www.gotw.ca/resources/clcm.htm for info about ]
[ comp.lang.c++.moderated. First time posters: Do this! ]

Michael Tiomkin · Guest

[email]tojohere (AT) gmail (DOT) com[/email] wrote:

Tojo · Guest

Hendrik Belitz wrote:

Tojo · Guest

Yeah I thought about this idea first,
But just for a variable, creating one base class felt as an
overhead to me. Also I thought about maintaining an internal structure
or class like this

class Boy
{
private:
struct Count
{
private:
int m_nCount;
public:
Count() :m_nCount(0) { }

void SetCount(int argCount)
{
m_nCount = argCount;
}
int GetCount()
{
return m_nCount;
}

} st_Count;

public:
Boy(): st_Count ( )
{
}
void SetCount(int argCount)
{
st_Count.SetCount( argCount );
}
int GetCount()
{
return st_Count.GetCount();
}

void SomeFunc()
{
// Error : compiler will prevent this
// because there is no ++ operator overloaded for Count
structure.
st_Count++;

// This will work without any problem.
// I want the programmer to force
// to do like this when he modify the m_nCount
SetCount( GetCount() + 1);
}
};

int main(int argc, char* argv[])
{
Boy b;
b.SetCount( 10 );
int a = b.GetCount();
return 0;
}

But you can see how much extra code I have to write to achieve such a
simple thing. Thats why I finally decided to use 'const' approach.
The 'cost' approch doesn't need to introduce any new classes or
structure....thats the benefit I am seeing in that style. But that too
feels as wrong even if it works....

Mark Van Peteghem wrote:

tony_in_da_uk@yahoo.co.uk · Guest

You're getting caught up in the doomed pursuit of over-protecting your
class. Just as private can be circumvented by body practices such as
"#define private public", const can be circumvented by things like
"reinterpret_cast<int&>(b) = 3". There's no point fighting this - if a
client programmer is dumb enough to do this, they deserve whatever
happens to them.

What you've already done isn't too bad, though it's a choice between
evils: having a const int as you've done mistakenly implies the class
doesn't need to change it after construction, which could confuse a
programmer maintaining your code, while omitting const doesn't convey
that changing the value requires special caution. Either way, a
comment should appear beside the declaration explaining that it must
only be changed in SetCount(). As a guideline, you shouldn't make it
const unless you anticipate attempts to try to change it
inappropriately. Will other programmers write or maintain the
functions in this class? Will the class be so large they can't easily
take in the complexity and ramifications of changing the count? If so,
then perhaps you should refactor your code. If you're really serious
about preventing changes to the value, then you could use an operating
system call to lock the memory (on UNIX, mprotect), but then attempts
to access it will generate signals that, if unhandled, may make your
app fall over, and are problematic to recover from.

Cheers, Tony

[ See http://www.gotw.ca/resources/clcm.htm for info about ]
[ comp.lang.c++.moderated. First time posters: Do this! ]

kanze@gabi-soft.fr · Guest

[email]tojohere (AT) gmail (DOT) com[/email] wrote:

Mark Van Peteghem · Guest

Tojo wrote: