Guess what I found in MFC's source code?
-
I was just debugging an MFC program and surprise, surprise there is a Goto statement insinde wincore.cpp at line 2438: goto LReturnTrue; Hmmm...makes you wonder...;P Rob
gotohas been the subject of a lot of debute, but the only reasonable point is that some say usinggotois just the signs of a bad implementation. The jump agotostatement uses is really no different that the jumps used for conditionals and switches. I really don't see any valid argument about not using them.Microsoft MVP, Visual C# My Articles
-
gotohas been the subject of a lot of debute, but the only reasonable point is that some say usinggotois just the signs of a bad implementation. The jump agotostatement uses is really no different that the jumps used for conditionals and switches. I really don't see any valid argument about not using them.Microsoft MVP, Visual C# My Articles
-
1st, the code you refer to is not part of MFC, this is the implementation of standard C++ library that comes with VC. 2nd, these are templates. Declaration and definition of template classes should completely be in header files to be visible to compiler in producing concrete types (instantiation). 3rd, this is not even wrote by Microsoft - it was written by P.J. Plauger (and sold to Microsoft). I agree it could easily won some prizes on "obfuscated C++" contest :-D
i found this in "The C++ Language" from Bjarne Stroustrup, Chapter 13.7 : Organisation of the source code. (please appology mu english and my poor abilities in the translation (i've got the book, but in french). [...] In fact, the definition of out(), and all the declarations it depends on, are included into several compilation units. The compiler must generate the code (uniquely) when necessary, and optimize the reading multiple definitions process. In this strategy, the template functions are treatd exactly as inline functions. In this case, an obvious problem appears. Every element depending on the definition of out() is added into each file that use this function. Quantity of information that has to be treatd by the compiler is so considerably grown up. Another problem reside in the fact that users can become dependant on some declarations included uniquely for the out definition. This danger can be minimized while using namespaces, avoiding coding macros and reducing quantity of informations so included. The strategy of separate the compilation is the logical solution to this reasoning : if the template definition don't appear in the user code, none of its dependances can affect this code. This is why we split here the original out.c file into the following two :
// out.h
template<class T> void out (const T&);// out.c
#include <iostream>
#include "out.h"
export template<class T> void out (const T& t) { std::cerr << t; }the out.c file now contain all the requiered information to define out(), and out.h only contain useful information to call it. A user will only have to include the declaration (interface) :
#include "out.h"[...]. i let you notice that the definition and the implementation are well splited !!!
TOXCCT >>> GEII power
-
Yes, and besides that, my version has only one exit point AND it closes the handle (if the handle was created). So you see, the short way, and more importantly the *exotic* way, are almost always less desireable than the *right* way. And before anyone becomes pedant, I am aware there are several "right" ways, but my point is that some ways are more right than others, and I will always stress maintainability over short/exotic. I hope I have been sufficiently pellucid, even for those who's most prominent physical feature is their single eyebrow which is more-or-less centered on their sloping foreheads. Thanks for playing our game. You may now grunt amongst yourselves. ------- sig starts "I've heard some drivers saying, 'We're going too fast here...'. If you're not here to race, go the hell home - don't come here and grumble about going too fast. Why don't you tie a kerosene rag around your ankles so the ants won't climb up and eat your candy ass..." - Dale Earnhardt "...the staggering layers of obscenity in your statement make it a work of art on so many levels." - Jason Jystad, 10/26/2001
-
I was just debugging an MFC program and surprise, surprise there is a Goto statement insinde wincore.cpp at line 2438: goto LReturnTrue; Hmmm...makes you wonder...;P Rob
I personally seldom use a Goto, but for those times when it makes sense, I am glad it is there. This one one of the reasons I did not like Java when it first became popular, since they were telling me I "could not" use a goto if I wish. Of course, I also did not like that they made no provisions for enums. It surely could not have been that hard to implement, but they figured they knew best :confused: It was funny how people would attack the poor little Goto, saying it made code hard to follow and led to spaghetti code while at the same time, these great thinkers would abuse classes is such a way that you would welcome pasta ;) As you can tell, you have touched a religous issue :) Rocky <>< www.HintsAndTips.com www.GotTheAnswerToSpam.com
-
gotohas been the subject of a lot of debute, but the only reasonable point is that some say usinggotois just the signs of a bad implementation. The jump agotostatement uses is really no different that the jumps used for conditionals and switches. I really don't see any valid argument about not using them.Microsoft MVP, Visual C# My Articles
-
That's misleading code. When someone sees a "do" they're expecting an honest-to-god loop. I would do it this way:
error myroutine()
{
HFILE hFile = NULL;
error = success;
if ((hFile = CreateFile(...)))
{
ReadFile(...);
if (dwRead)
{
WriteFile(...);
if (!dwWritten)
{
error = write;
}
}
else
{
error = read;
}
CloseHandle(hFile);
}
else
{
error = create;
}
return error;
}No goto's, no fake loops, the calling function can always programatically respond to the "error" reported by the function, and it's more maintainable by new employees. I've never seen a valid reason to ghave goto's in a c++ function, and I've been doing C++ for 14 years. ------- sig starts "I've heard some drivers saying, 'We're going too fast here...'. If you're not here to race, go the hell home - don't come here and grumble about going too fast. Why don't you tie a kerosene rag around your ankles so the ants won't climb up and eat your candy ass..." - Dale Earnhardt "...the staggering layers of obscenity in your statement make it a work of art on so many levels." - Jason Jystad, 10/26/2001
Personally, I use the fake loops often BUT I always include a warning that it IS a fake loop. I find it to be helpful for scoping objects and minimizing indentation. I view it as a personal preference and I prefer that to using gotos which I have also managed to avoid for a long, long time. __________________________________________ a two cent stamp short of going postal.
-
I was just debugging an MFC program and surprise, surprise there is a Goto statement insinde wincore.cpp at line 2438: goto LReturnTrue; Hmmm...makes you wonder...;P Rob
Every if is a goto, every for, while and do. The ?: operator is, function calls are, and returns are gotos to an adress that where picked up at a crowded plaza. try going without goto, and you will arrive nowhere. ;P
Flirt harder, I'm a Coder
mlog || Agile Programming | doxygen -
I personally seldom use a Goto, but for those times when it makes sense, I am glad it is there. This one one of the reasons I did not like Java when it first became popular, since they were telling me I "could not" use a goto if I wish. Of course, I also did not like that they made no provisions for enums. It surely could not have been that hard to implement, but they figured they knew best :confused: It was funny how people would attack the poor little Goto, saying it made code hard to follow and led to spaghetti code while at the same time, these great thinkers would abuse classes is such a way that you would welcome pasta ;) As you can tell, you have touched a religous issue :) Rocky <>< www.HintsAndTips.com www.GotTheAnswerToSpam.com
Rocky Moore wrote: I personally seldom use a Goto, but for those times when it makes sense, I am glad it is there. This one one of the reasons I did not like Java when it first became popular, since they were telling me I "could not" use a goto if I wish. Of course, I also did not like that they made no provisions for enums. It surely could not have been that hard to implement, but they figured they knew best I think the Java creators were striving to achieve the complete opposite of C++: simplicity. Everything is either a class type or a primitive type (char, int, float, double, byte). They also figured
gotos are so seldomly the superior route that not having them would force people to write proper code (without them). In other words, get rid of the gun so that people who don't know how and when to use it properly don't get any funny ideas. :-) One thing I think Java should have had was Properties. They don't add much to the complexity and improve readability a lot:int s = obj.Something;vs.int s = obj.getSomething();Regards, Alvaro
Give a man a fish, he owes you one fish. Teach a man to fish, you give up your monopoly on fisheries.
-
Rocky Moore wrote: I personally seldom use a Goto, but for those times when it makes sense, I am glad it is there. This one one of the reasons I did not like Java when it first became popular, since they were telling me I "could not" use a goto if I wish. Of course, I also did not like that they made no provisions for enums. It surely could not have been that hard to implement, but they figured they knew best I think the Java creators were striving to achieve the complete opposite of C++: simplicity. Everything is either a class type or a primitive type (char, int, float, double, byte). They also figured
gotos are so seldomly the superior route that not having them would force people to write proper code (without them). In other words, get rid of the gun so that people who don't know how and when to use it properly don't get any funny ideas. :-) One thing I think Java should have had was Properties. They don't add much to the complexity and improve readability a lot:int s = obj.Something;vs.int s = obj.getSomething();Regards, Alvaro
Give a man a fish, he owes you one fish. Teach a man to fish, you give up your monopoly on fisheries.
Alvaro Mendez wrote: simplicity. Code with gotos is no less simplistic. In some code it can even add to the readabilty. Alvaro Mendez wrote: force people to write proper code (without them). In other words, get rid of the gun so that people who don't know how and when to use it properly don't get any funny ideas. That is just the point, who is it that told them they possessed all knowledge and only they new what "proper code" is? Just like with guns, we should still be allowed to possess and use them regardless that some people abuse them. Alvaro Mendez wrote: One thing I think Java should have had was Properties. They don't add much to the complexity and improve readability a lot: Yep, properties can be handy. So can enums and gotos :) Rocky <>< www.HintsAndTips.com www.GotTheAnswerToSpam.com
-
You could also write fo in such a way that it returns to a calling function for the more code part instead of having "more code" itself. Voila - no goto needed. ------- sig starts "I've heard some drivers saying, 'We're going too fast here...'. If you're not here to race, go the hell home - don't come here and grumble about going too fast. Why don't you tie a kerosene rag around your ankles so the ants won't climb up and eat your candy ass..." - Dale Earnhardt "...the staggering layers of obscenity in your statement make it a work of art on so many levels." - Jason Jystad, 10/26/2001
Heh, you're right :-O "After all it's just text at the end of the day. - Colin Davies "For example, when a VB programmer comes to my house, they may say 'does your pool need cleaning, sir ?' " - Christian Graus
-
You are not really trying to convince me that
error myroutine(){ HFILE hFile = NULL; error = success; if ((hFile = CreateFile(...))) { ReadFile(...); if (dwRead) { WriteFile(...); if (!dwWritten) { error = write; } } else { error = read; } CloseHandle(hFile); } else { error = create; } return error; }
is more readable and maintainable than:error myroutine(){ File f; if(!f.Create(...)) return create; if(!f.Read(...)) return read; if(!f.Write(...)) return write; return success; }Nenad Caklovic wrote:
error myroutine()
{
File f;
if(!f.Create(...))
return create;
if(!f.Read(...))
return read;
if(!f.Write(...))
return write;
return success;
}This kind is more elegant and neat to another style (a deep-nested conditioning), at least what I think. I always use this style in my code. Maxwell Chen
-
toxcct wrote: have you ever heard or break keyword ? A break only breaks the loop belonged to that break. As:
int main()
{
int i, j, k;
printf("Begin:\n");
for(i = 0; i < 3; i++) {
for(j = 0; j < 3; j++) {
for(k = 0; k < 3; k++) {
if(k == 1)
break;
printf(" ~ k = %d.\n", k);
}
printf(" ~ j = %d,\n", j);
}
printf(" ~ i = %d,\n", i);
}
printf("End.\n");
return 0;
}Result:
Begin:
~ k = 0.
~ j = 0,
~ k = 0.
~ j = 1,
~ k = 0.
~ j = 2,
~ i = 0,
~ k = 0.
~ j = 0,
~ k = 0.
~ j = 1,
~ k = 0.
~ j = 2,
~ i = 1,
~ k = 0.
~ j = 0,
~ k = 0.
~ j = 1,
~ k = 0.
~ j = 2,
~ i = 2,
End.Maxwell Chen
-
gotos are inherited from assembly code and that is the main reason it exist in C/C++. in all ways, i find MFC and other microsoft library codes awful to read. for example, they implement some of their functions into the header files instead of writing it into CPP files. moreover, they abuse of the imbricated operations...
TOXCCT >>> GEII power
toxcct wrote: MFC and other microsoft library codes awful to read. for example, they implement some of their functions into the header files instead of writing it into CPP files. The whole STL implementation is all written in header files. For ex, <vector> ... Maxwell Chen
-
toxcct wrote: MFC and other microsoft library codes awful to read. for example, they implement some of their functions into the header files instead of writing it into CPP files. The whole STL implementation is all written in header files. For ex, <vector> ... Maxwell Chen
-
we've powerful loop that perform clean jumps. we are not writing assembly here, so,
gotos must be avoided as much as possible. Don't forget a thing : your program must stay "DEBUGGABLE"
TOXCCT >>> GEII power
toxcct wrote: your program must stay "DEBUGGABLE" When talking about what non-debuggableness, it goes to exception-handling instead of goto jumps. In the old days when I was coding ADO stuff, for example:
bool Some_ADO_Type::FooBar(...)
{
// ...
// Everything is OK in the block.
return true;
} // Crashed here right after the (debug) yellow arrow passed this closing bracket.I traced deep into the hell bases... no idea why it threw such an exception, and no way to catch with VC++ 6. Finally, VC++ 7.x supports ISO C++, and there is such way:
bool FooBie()
try
{
return true;
}
catch(...)
{
return false;
}to catch such kind of god-d__n evil exceptions! Maxwell Chen
-
toxcct wrote: your program must stay "DEBUGGABLE" When talking about what non-debuggableness, it goes to exception-handling instead of goto jumps. In the old days when I was coding ADO stuff, for example:
bool Some_ADO_Type::FooBar(...)
{
// ...
// Everything is OK in the block.
return true;
} // Crashed here right after the (debug) yellow arrow passed this closing bracket.I traced deep into the hell bases... no idea why it threw such an exception, and no way to catch with VC++ 6. Finally, VC++ 7.x supports ISO C++, and there is such way:
bool FooBie()
try
{
return true;
}
catch(...)
{
return false;
}to catch such kind of god-d__n evil exceptions! Maxwell Chen
-
Maxwell Chen wrote: The whole STL implementation is all written in header files. yes, and it is what i reproach them. read the post i wrote just before yours here...
TOXCCT >>> GEII power
So you mean that there would be the header <vector> and the implementation, vector.cxx, in gcc or others!? Maxwell Chen
-
i found this in "The C++ Language" from Bjarne Stroustrup, Chapter 13.7 : Organisation of the source code. (please appology mu english and my poor abilities in the translation (i've got the book, but in french). [...] In fact, the definition of out(), and all the declarations it depends on, are included into several compilation units. The compiler must generate the code (uniquely) when necessary, and optimize the reading multiple definitions process. In this strategy, the template functions are treatd exactly as inline functions. In this case, an obvious problem appears. Every element depending on the definition of out() is added into each file that use this function. Quantity of information that has to be treatd by the compiler is so considerably grown up. Another problem reside in the fact that users can become dependant on some declarations included uniquely for the out definition. This danger can be minimized while using namespaces, avoiding coding macros and reducing quantity of informations so included. The strategy of separate the compilation is the logical solution to this reasoning : if the template definition don't appear in the user code, none of its dependances can affect this code. This is why we split here the original out.c file into the following two :
// out.h
template<class T> void out (const T&);// out.c
#include <iostream>
#include "out.h"
export template<class T> void out (const T& t) { std::cerr << t; }the out.c file now contain all the requiered information to define out(), and out.h only contain useful information to call it. A user will only have to include the declaration (interface) :
#include "out.h"[...]. i let you notice that the definition and the implementation are well splited !!!
TOXCCT >>> GEII power
Here is what I found simply by Googling: http://www.comeaucomputing.com/techtalk/templates/#whylinkerror[^] A little snippet:
Why do I get a link error when compiling my templates?
Some compilers require that all function templates need to be made available in every translation
unit that it is used in. In effect, for those compilers, the bodies of template functions must be
made available in a header file. To repeat: that means those compilers won't allow them to be
defined in non-header files such as .cpp files. Furthermore, some compilers also require that
some functions be defined inline inside a class, and not outside of one.
Note that not all compilers require this. For instance, Comeau C++ does not in all cases (check
out http://www.comeaucomputing.com/4.0/docs/userman/ati.html for details on our current setup).
In short, Comeau C++ supports many models, including one which comes close to what the export
keyword's intentions are (as an extension).And on this point, note that the C++ export keyword is intended to alleviate the original
question. However, currently Comeau C++ is the only compiler known to support export. Therefore,
some compilers use such extensions referred to in the previous paragraph, or you need to put the
bodies of your functions in headers, if you are using a compiler which does not support export.So, the ability to put template implementations in the cpp file may be standard, I don't know, and you can't deduce it from reading "The C++ language", because Stroustrup is NOT responsible for that C++ standard, he is not Linus you know ;) But since very few compilers, and apparently none of the most used, are able to do it, you should just avoid trying. "After all it's just text at the end of the day. - Colin Davies "For example, when a VB programmer comes to my house, they may say 'does your pool need cleaning, sir ?' " - Christian Graus
-
Every if is a goto, every for, while and do. The ?: operator is, function calls are, and returns are gotos to an adress that where picked up at a crowded plaza. try going without goto, and you will arrive nowhere. ;P
Flirt harder, I'm a Coder
mlog || Agile Programming | doxygen