Hi, I am recently looking into the extensibility of the functionality in C++ language, so far, i couldn't figure out the way to actually get the things worked nicely. I am actually trying to do some add-on to the C++ function's returning type control with Assembly language, however i am stucked as i couldn't figure out which way to dive in. I wish to know if it is the right way to use Assembly to taking control of the things? Of which C++ is lacking of some functionality, which is i hope to redesign a class module for my own purpose, so could Assembly handle these things for me? Thanks in advance, Rgrds, Daniel.

That would not be a good use of assembly. What, exactly, are you trying to do?

Just trying to achieve something that the older C++ (vc6++) couldn't achieve.
I am just wondering if it makes sense to use Assembly to expand the functionality & power of it.
Or is there a better way?

While you could actually use assembly and wrap it inside macros or forced inline functions, using assembly is generally evil. Either you will lock your users into x86/64-architecture, or you end up writing assembly for many cpu-architectures (e.g. ARM is very widespread). Additionally, if you want to support several compilers (gcc, icc, msvc++, codewarrior) you eventually have to support multiple [inline] assembler syntaxes. In short, assembler makes your changes unportable.

Are you sure the changes you want to make can't be achieved by clever and clean c++ code?

Have you already checked ...
* C++ Templates - The Complete Guide (Vandevoorde, Josuttis)
* GoF [Design Patterns - Elements of Reusable Object-Oriented Software] (Gamma, Helm, Johnson, Vlissides)
* the boost libraries
* blitz++
?

Do you grok the "Guru of the Week" (gotw) articles?

Do you breath C++-03 and are you up to date (in knowledge) about C++0x?

If checked all (and a bit more), consider studying c++ compiler sourcecode (that is probably g++), and maybe write patches to add your features. Or rewrite a compiler from scratch (the latter will take you the next 5 years, at the minimum least and without the export keyword; c++ is not an LALR language and so somewhat full of ambiguities from the viewpoint of a compiler writer (how many interpretation do you find for (including template instantations) ?)).


Another solution would be an add-on, a kind of macro processor, to be placed on top of c++. But I personally wouldn't use it for many reasons.


edit: Out of curiousity, what feature do you miss exactly?

[Edited by - phresnel on January 5, 2009 9:05:19 AM]

Quote:Original post by danong
Or is there a better way?

Upgrade to a newer compiler - VC6 is almost universally recognised as one of the worst compilers in existence.

Thanks all for the reply and helps (appreciated).

Okay, frankly i'd found out that C++ doesn't support multiple-type-return from the function, either it must be different-type-return value + different-type-parameter to actually ask the compiler to tell what is the difference.

And the engine i'm designing somehow needs this kind of functionality, i mean stay the same with the parameter, but only different type return value, however it's not achievable.

and for some reasons, i'm using somesort of code like this :

 int foo::operator [] (const int index);char foo::operator [] (const int index);      // compiler couldn't work on this.

And so this is the reason i'm trying to deal with the front-end of the C++ compiler.

But now i'd found out that <typeinfo.h> could give me sth that i want,
but couldn't find the <typeinfo.cpp> source,
perhaps i'm wrong ;(

Perhaps i will just figure out another way of fixing this.



Thanks once again in advance.

What about this instead?:

void foo::get(int index, int  & out);void foo::get(int index, char & out);

operator [] overloading doesn't support multiple parameters , i had tried that once. ;(

Quote:Original post by danong
operator [] overloading doesn't support multiple parameters , i had tried that once. ;(

Why do you need to overload the [] operator?