Initializing data structure with meset

Question

0.00/5 (No votes)

See more:

Hi, I've seen some code where data structure was initialized with memcpy.
Was curious if it is a good idea?
e.g.

C++

auto_ptr<MyDataStructure> tmp(new MyDataStructure());
memset(tmp.get(),0x0,sizeof(MyDataStructure));

Posted 20-May-13 4:51am

Dan page

Updated 20-May-13 12:05pm

CPallini

v3

Add a Solution

4 solutions

Add a Solution

Add your solution here

Treat my content as plain text, not as HTML

Preview 0

…

Existing Members

Sign in to your account

...or Join us

Download, Vote, Comment, Publish.

Your Email
Password
Forgot your password?

Your Email
This email is in use. Do you need your password?
Optional Password

I have read and agree to the Terms of Service and Privacy Policy
Please subscribe me to the CodeProject newsletters

When answering a question please:

Read the question carefully.
Understand that English isn't everyone's first language so be lenient of bad spelling and grammar.
If a question is poorly phrased then either ask for clarification, ignore it, or edit the question and fix the problem. Insults are not welcome.
Don't tell someone to read the manual. Chances are they have and don't get it. Provide an answer or move on to the next question.

Let's work to help developers, not make them feel stupid.

This content, along with any associated source code and files, is licensed under The Code Project Open License (CPOL)

Andreas Gieriet · Answer 1 · 2013-05-20T11:49:00

I cosider this a broken way of initialization in C++.
It breaks the OOP encapsulation and information hiding - you override with brute foce whatever the constructor did/was supposed to do.

Why does the client of a class know better than the class itself how an instance is initialized.
This approach may produce leaks (a constructed memory address gets overridden by 0), initialize enum members to 0 (while that enum may not have a 0-value), etc.

This initialization idiom is most likely a legacy approach of some C-veteran. C does not know constructors, so it was the only way and safe those days to initialize the allocated memory after a alloc/calloc/malloc.

The new operator calls also malloc of the appropriate size and then calls the constructor of that class by providing that memory as the this pointer to the constructor.

My advise:
- in C++: never ever do this - you have constructors for this.
- in C: definitively do it (or use calloc).

Cheers
Andi

BCN-163 · Answer 2 · 2013-05-20T15:12:00

Solution 3

if your structor has virtal functions, disaster comes. e.g code as follow, run it and see what will happen.

C++

typedef struct tagTestStructor
{
	virtual void testOut(){ std::cout << "this is a virtual function" << std::endl;}
	int m_nVal;
	char m_cVal;
	// ...
}TestStructor;

int _tmain(int argc, _TCHAR* argv[])
{
	TestStructor objTest;
	memset(&objTest, 0, sizeof(objTest) );

	// use a pointer to test
	TestStructor* pObjTest = &objTest;
	pObjTest->testOut();
}

Posted 20-May-13 15:12pm

BCN-163

Comments

«_Superman_» 20-May-13 23:38pm

Your point in absolutely valid. 10 points for this.
However, it may not be applicable to this case because the destination in memset is tmp.get() and not &tmp.
Although I don't know how auto_ptr is implemented, the get method could return an address after the virtual pointer so that the vptr is not initialized to 0.

Also, I agree with others that this sort of initialization is not the right way to do it.
Another thing I would like to add here is that auto_ptr has been deprecated and should not be used. unique_ptr must be used instead.

BCN-163 22-May-13 22:37pm

you could use follow code to test get method of auto_ptr.
typedef struct tagTestStructor
{
virtual void testOut(){ std::cout << "this is a virtual function" << std::endl;}
int m_nVal;
char m_cVal;
// ...
}TestStructor;

int _tmain(int argc, _TCHAR* argv[])
{
TestStructor* pTmpNew = new TestStructor();
auto_ptr<teststructor> apTmp(pTmpNew);

assert(apTmp.get() == pTmpNew);

memset(apTmp.get(),0x0,sizeof(TestStructor));

apTmp->testOut();
}

Andreas Gieriet 21-May-13 11:06am

My 5! It may/may not fail, depending on the C++ compiler. Things get worse with multi inheritance (diamond pattern) and casting...
Cheers
Andi

BCN-163 22-May-13 22:41pm

as I know, it will always fail. I have heard a lot about the different implement of C++ compiler. but I never met one. would you give me some tips? really appreciate.

Andreas Gieriet 23-May-13 4:17am

See Inside the C++ Object Model for some insight on how it might be implemented.
Cheers
Andi

BCN-163 23-May-13 4:44am

ok, 3q, I will read this book again.

JackDingler 23-May-13 11:14am

These days you may never need to worry about other compiler implementations. Hardware and software is becoming increasingly standardized and converging into common paths.

The two dominant compilers adhere well to the standards, and work with a wide variety of platforms. There's little need now to roll out odd flavors of compilers.

JackDingler 21-May-13 13:42pm

Though some flavor of C++ may be an exception....
Virtual function tables aren't cloned with every instantiation of the class. they exist outside of the class, so modifying the class contents will not destroy these tables.

That said, don't use memset to initialize objects...

BCN-163 22-May-13 22:44pm

what was set to 0 is the vptr which point to the virtual table. not the virtual table itself. I think

JackDingler 23-May-13 11:09am

It doesn't work that way though. There are no secret or hidden members in a class that you have to worry about. Links per instantiation of the class are unnecessary.

Any virtual tables will likely get a decorated name, and be accessible behind the scenes as globals.

«_Superman_» · Answer 3 · 2013-05-20T17:41:00

Solution 4

I agree with the others that this is not the right way to initialize it.

Another thing I would like to add is that auto_ptr should not be used at all because it has been deprecated.
Use unique_ptr instead.

See here - http://www.cplusplus.com/reference/memory/auto_ptr/[^]

Posted 20-May-13 17:41pm

«_Superman_»

Comments

Andreas Gieriet 21-May-13 11:05am

My 5!
Cheers
Andi

Jonathan Davies · Answer 4 · 2013-05-20T04:58:00

Solution 1

Always a good idea (essential) to initialize your data by whatever means. Otherwise one day, somewhere, your uninitialized structure may be read and will contain something random that looks like it might be real data but it isn't.

Posted 20-May-13 4:58am

Jonathan Davies

Comments

Andreas Gieriet 20-May-13 17:51pm

While proper initialization is crucial to all robust software, I consider this particular approach in C++ as broken.
See my reasoning in my solution #2.
Cheers
Andi

Initializing data structure with meset

4 solutions

Solution 2

Solution 3

Solution 4

Solution 1

Add your solution here

Preview 0

Existing Members

...or Join us