[Moved] A question about pointers.

cazador7907

What is the proper way to initialize a pointer? Can it be initialized to NULL?

Taamalus

Int[or whatever] * ptrName = NULL (or 0);

That's what I do - It does not mean it's an absolute way. Usually I know what I want the pointer to do.

cazador7907

The the way to test for NULL is simply:

If (pointer == null)

or

if (pointer == 0)

Yes?

Taamalus

^ This may not work.
http://www.codeguru.com/forum/archive/index.php/t-358371.html

Simply go
@
if (ptrName)
cout << “Pointer points to something”;
else
cout << “Pointer is NULL”;
@

Yet, make sure, the pointer is initialized, as described in more detail in the link. ;)

tobias.hunger

You can of course also initialize member pointers in the initializer list of the constructor.

andre

Initialization in the initializer list is a good idea. Did you know you can even new objects there? Things like this are legal:

@
MyQObject::MyQObject(parent):
m_timer(new QTimer(this))
{
...
}
@

For pointers that you wish to initialize elsewhere: NULL is more of a C construct. For C++, using 0 is more common.

giesbert

If you don't ant to check your pointers for 0, you can also use the so called NullObject pattern.

This means you initialise a pointer to a static object with the same interface that just does nothing. The you can avoid this (if(0 != p) ...

"see here":http://en.wikipedia.org/wiki/Null_Object_pattern

andre

Nice one Gerolf, I was not familiar with this pattern.
Though of course you don't need to actually do if(0 != ptr) { }, as Taamalus points out.

goetz

[quote author="Taamalus" date="1300504867"]^ This may not work.
http://www.codeguru.com/forum/archive/index.php/t-358371.html
[/quote]

What's the rationale behind this? To my knowledge even in poor old C (without ++) "0" is the all-valid constant for a null pointer (regardless of the internal representation for the actual machine). So

@
if(ptr == 0)
doFancyThings();
@

is always valid code (See "comp.lang.c FAQ list, Question 5.2":http://c-faq.com/null/null2.html and "Question 5.5":http://c-faq.com/null/machnon0.html).

Though the short comparison you suggested, is always valid too ("C FAQ, 5.3":http://c-faq.com/null/ptrtest.html) and also more readable, so to prefer over the "noisy" one, IMHO.

tobias.hunger

Somewhat unrelated, but the assumption that accessing the 0 pointer will always results in a segmentation fault is not necessarily true. See https://lwn.net/Articles/342330/ for a very interesting description of what can happen when 0 suddenly becomes a valid pointer.

giesbert

I also saw something like this:

@
int CClass::foo()
{
if(0 == this)
return 0
do stuff
}
@

with this syntax, this is absolutly safe:

@
CClass* p = 0;
p->foo();
@

This was some method in I think MFC API

goetz

[quote author="Gerolf" date="1300560358"]I also saw something like this:

@
int CClass::foo()
{
if(0 == this)
return 0
do stuff
}
@

with this syntax, this is absolutly safe:

@
CClass* p = 0;
p->foo();
@

This was some method in I think MFC API[/quote]

Completely weird... I would not suppose this to work on other compilers than Microsoft's :-)

giesbert

don't know, I think ist CWnd::getSafeHWnd...
and as no members are accessed, it could work.
This is in fact a parameter (hidden behind C++).

Taamalus

[quote author="Volker" date="1300533129"][quote author="Taamalus" date="1300504867"]^ This may not work.
http://www.codeguru.com/forum/archive/index.php/t-358371.html
[/quote]

What's the rationale behind this? To my knowledge even in poor old C (without ++) "0" is the all-valid constant for a null pointer (regardless of the internal representation for the actual machine). So

@
if(ptr == 0)
doFancyThings();
@

is always valid code (See "comp.lang.c FAQ list, Question 5.2":http://c-faq.com/null/null2.html and "Question 5.5":http://c-faq.com/null/machnon0.html).
[/quote]
Thanks, I stand corrected. :) Still, there is no way I will use it, but from now on, just as a personal preference.
http://en.wikipedia.org/wiki/Pointer_(computing)#Null_pointer
this link is just to save face :D on my preferences

Also thanks Gerolf! Re: Null Object Patterns, not for C++ but for LISP! Cheers! :)