Thread: goto, why not

The linux kernel uses them.
microsoft also does.

Here's some msdn code

int ProcessFiles(const char* pszFiles, FILECALLBACK pFunc, void* pArgs)
{
HRESULT hr = S_OK;
int i;

WCHAR* wszArg = WideString(pszFiles);
WIN32_FIND_DATA FindFileData;
HANDLE handle = ::FindFirstFile(wszArg, &FindFileData);
int failed = 0;

if (handle == INVALID_HANDLE_VALUE)
{
// Then maybe it's a URL.
hr = (*pFunc)(wszArg, pArgs);
if (hr != 0) failed++;
goto CleanUp;
}
for ( i = wcslen(wszArg)-1; i >= 0 && wszArg[i] != '\\'; i--)
{
wszArg[i] = 0;
}

while (handle != INVALID_HANDLE_VALUE)
{
ULONG len1 = wcslen(wszArg);
ULONG len2 = wcslen(FindFileData.cFileName);
WCHAR* buffer = new WCHAR[len1 + len2 + 1];
if (buffer == NULL)
{
hr = E_OUTOFMEMORY;
goto CleanUp;
}
memcpy(buffer, wszArg, sizeof(WCHAR) * len1);
memcpy(&buffer[len1], FindFileData.cFileName, sizeof(WCHAR) * len2);
buffer[len1+len2] = '\0';

hr = (*pFunc)(buffer, pArgs);

if (hr != 0) failed++;

delete[] buffer;

if (! ::FindNextFile(handle, &FindFileData))
break;
}
if (handle != INVALID_HANDLE_VALUE)
::FindClose(handle);

CleanUp:
delete wszArg;

return failed;
}

Why would you have nested loops 4 deep?

you could just do this

int done = 0;
for (int a = 0; a < 100, done < 1; a++)
{
for (int b = 0; b < 100, done < 1; b++)
{
for (int c = 0; c < 100, done < 1; c++)
{
for (int d = 0; d < 100, done < 1; d++)
{
if (condition)
done = 2;
}
}
}
}
}

omg... no goto.... no 4 conditional break....

-Vulcan

Just as a side note and to say something a bit more constructive, I don't use goto's at all. The last time I used a goto was coding in GW-BASIC back in the early 80's. Even QBasic has been designed specifically so you do not have to use goto's. VB, which is designed from QBasic and the BASICs, also does not make use of a goto (only in error handling do you 'have' to use a goto). It is clear to me that Microsoft went to modules and no line numbers in their new line of BASICs specifically to remove the ability to jump around (goto's and gosub's) all over the place in code since it is not good programming practice.

I was taught that there should be only one entry into and one exit from a function or procedure and that the procedure should only do one thing, but do it very well. So, I try to follow this example as much as possible and do not exit out of my functions until the end of the function. It was hard to break the habit at first, but now I cringe at code that jumps all over the place. Besides, when you jump around, are you not slowing the entire program down? According to Randall Hyde's AoA and a lot of books, websites, and teachers I've had, jumping around a lot is not a good practice.

With a goto, I could exit as many times as I please and the code would be very ugly, hard to debug, and probably would not be any faster than if I had not used a goto - probably much slower. I've never used a goto in C/C++. The closest thing that I've used to a goto is an on error statement in VB which jumps to a label on an error. But I don't even like using those cuz your subs and functions get all cluttered up with error handlers and you have to explicity exit the sub or function prior to reaching the label. Very ugly. You can have this big ugly global error handler, but again error handling in QBasic and VB is just ugly.

I even cringe when I have to use a jmp or jx instruction in assembly, even though I must at times. To avoid this I try to structure asm program logic to 'fall through' to labels on certain conditions instead of jumping to them, even though a near jump is not that costly.


I said this before and I'll say it again.

A high level program that looks as though it has no choice but to use a goto is poorly designed. When it comes to goto's and higher level languages it seems to me that there is always a better way.

Bah! You hippocrites. Just code comfortably. After all, that shoe commercial says "Just do it!"