[antlr-interest] Re: symbol table in C++

[lgcraymer]

If you follow the canonical C++ paradigm, non-inlined method code 
should be defined in .cpp files, with class definitions, including 
method prototypes, in a .h file.  For your example, A::hello() might 
be defined in an A.cpp file that looks like

#include "A.h"     // defines class A
#include "B.h"     // defines calss B

void A::hello() {
    pb->hello();
}

with A.h something like

class B;
class A {
private:
     B *pb;
public:
     A(B *);

    void hello();
};

--Loring


--- In antlr-interest at yahoogroups.com, "Fan Yang" <yhhf_dy at y...> 
wrote:
> Thank you for your help.
> 
> > The trick is that Java uses references--each of the cross-
> referenced 
> > items is actually a pointer, but the compiler hides this--and 
all 
> > pointers are of fixed size which makes memory allocation easy.  
To 
> > do the same thing in C++, you will have to make the references 
> > explicit--"HashTable& foo" for example instead of "Hashtable 
foo".  
> 
> That's exactly what I did. I have used pointer/reference 
throughout 
> the new implementation. 
> 
> > You may also need to declare a forward reference to the class--
that 
> > should just be a line of the form class Hashtable;
> 
> I have used forward references in the new implementation. But the 
> problem is that you can't reference members of the type which is 
> defined by forward reference. For example, there are compile 
errors 
> if you use forward referennce for hashtable or symboltable. The 
> reason is that JavaHashtable calls SymbolTable.LookupDummy(), and 
> SymbolTable calls JavaHashtable.get(). 
> 
> The following code snip simplifies the former problem. 
> class B;
> class A{
> private:
>  B*pb;
> public:
>  A(B*b):pb(b){}
>  void hello(){
>    pb->hello(); // complie error 
>  }
> };
> 
> class B{
> private:
>  A*pa;
> public:
>  ...
>  void hello(){
>  }  
> }
> 
> I know Template can overcome this simple problem. but the symbol 
> table program is much more complex than above example. I'm not 
sure 
> it's a right way to do it. Does there exist some kind of patterns 
to 
> deal with conversion from java to c++ code. thanks.
> 
> Fan


 

Your use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/