this returned destructor called...

"*this" returned destructor called...

Hi!

While testing a project I noticed a weird behaviour.
Here is the draft class:

Code:

Class MyClass { public: MyClass(); ~MyClass(); MyClass self() { return *this; } }

And here comes the problem:

Code:

MyClass anObj; anObj.self(); // PROBLEM HERE - after returning from self() function, de destructor of MyClass gets called // do something

If I change the self() to return a reference:

Code:

MyClass& self() { return *this; }

then the destructor isn't called.

Do you have a good explanation for this one?

Excuse my rudeness but I don't think it is weird or something. Maybe don't I just understand anything. ^^

So, I was saying that it is perfectly natural that the destructor is called, I don't know if you have realized that yet but in fact no function really "returns" an object, it's obvious that we could never make the EAX register huge enough to hold so much data.
Therefore, when we call a function that returns an object, the program actually calls the function with an extra argument, a pointer to a variable that will be initialized by the function. (sorry if I'm wasting your time and if you already knew that)
So, with that said, it appears clearly that once the object was created by the called-function, it must be deleted by the caller even if it was not used or even assigned to a "visible" variable.

when you return a value like that, you create a "temporary variable" So this line:

Code:

MyClass self() { return *this; }

actually attempts to create a copy of *this to return. The copy is temporary and is destroyed right after it is returned. This is why the destructor was called.

on the other hand:

Code:

MyClass& self() { return *this; }

This does not create a temporary. It is a reference to *this. Of course there's really no reason to do this one because an assignment would be no different in either of these cases:

MyClass instance;
MyClass & firstref = instance.self();
MyClass & secondref = instance;

Thank for your answers lyx and FillYourBrain.

Quote:

Originally posted by FillYourBrain
Of course there's really no reason to do this one because an assignment would be no different in either of these cases:

MyClass instance;
MyClass & firstref = instance.self();
MyClass & secondref = instance;

The purpose was to overload the operator [] and =, to create an array class which is able to handle e.g. common errors like overindexing.

Code:

MyArray::MyArray( unsigned int arraySize_in ) { if( arraySize_in == 0 || arraySize_in >= MAX_ARRAY_SIZE_VALUE ) { arraySize_in = MAX_ARRAY_SIZE_VALUE; } myArray = new ( int[arraySize_in] ); for( unsigned int i = 0; i < arraySize_in; i++ ) { myArray[i] = 0; } myIndex = 0; myLastIndex = arraySize_in - 1; } // copy constructor MyArray::MyArray( MyArray& array_in ) { myIndex = array_in.myIndex; myLastIndex = array_in.myLastIndex; myArray = new ( int[myLastIndex + 1] ); for( unsigned int i = 0; i <= myLastIndex; i++ ) { myArray[i] = array_in.myArray[i]; } //*this = array_in; } MyArray::~MyArray() { delete[] myArray; myArray = 0; myIndex = 0; } MyArray& MyArray::operator [] ( int index_in ) { // range check if ( index_in > myLastIndex ) { myIndex = myLastIndex; } else { myIndex = index_in; } return *this; } void MyArray::operator = ( const int val_in ) { myArray[myIndex] = val_in; }

This way, the following code part will never cause problems:

Code:

MyArray anArray(20); anArray[120] = 12;

I think that standard containers implement methods that allow you to do safe indexing. I don't have much knowledge in STL but the BS says that at() method is used to do range controled indexing.

just store the size of the array. This will allow you to do a check on your operator=()

Code:

void MyArray::operator = ( const int val_in ) { if(val_in < m_arrsize) myArray[val_in] = val_in; }

Quote:

Originally posted by FillYourBrain
just store the size of the array. This will allow you to do a check on your operator=()

Code:

void MyArray::operator = ( const int val_in ) { if(myIndex < m_arrsize) myArray[myIndex] = val_in; }

And where to get myIndex from?? Right that's the purpose of [] operator redefinition. It returns *this just because the object type is needed as = operator is also redefined.

Quote:

Originally posted by lyx
I think that standard containers implement methods that allow you to do safe indexing. I don't have much knowledge in STL but the BS says that at() method is used to do range controled indexing.

Of course STL would do it, however I intend to create a good example on operator overloading. On the other hand, I think sometimes is easier to implement your own class for such purposes rather then using STL.

I changed that in my post. Just used the wrong variable name. No big deal there.

by the way, if you index outside of the boundaries of vector it will crash.

I really should say, unexpected results like returning a *this for indexing into an array of ints doesn't make much sense. What's the person going to do with that information?

Quote:

Originally posted by FillYourBrain
I really should say, unexpected results like returning a *this for indexing into an array of ints doesn't make much sense. What's the person going to do with that information?

*this is only returned in order to correctly evaluate the = redefined operator.

Code:

MyArray anArray(20); anArray[120] = 12;

When going to this line,
anArray[120] = 12;

first operator [] is evaluated, whereas myIndex is set to max allowed value (as it exceeds the allowed range, which is 0-19).
myIndex gets the value myLastIndex, which is 19.
*this is returned, so the redefined operator = from class MyArray is evaluated:

myArray[19] = 12;

But I got your problem - what happens in this case:

int y = myArray[15];

Well, nothing, it won't compile.

but it's not correct.

myArray[num]

should return a reference to the type stored in the array. NOT the array class itself.

I see, it is an example.
Um... for the use of the STL, I don't use it.

Anyway, I was thinking the same as FillYourBrain but he posted before I do. ^^

Quote:

Originally posted by FillYourBrain
but it's not correct.

myArray[num]

should return a reference to the type stored in the array. NOT the array class itself.

As I told you, this was meant to be an example on operator overloading.

I overloaded () to return the given element of the array, but a getAt( int index_in ) or setAt( int index_in ) would also do it.
This way there's no possibility to exceed the array range, which might be more important.

Instead of:

int y = anArray[15];

use

int y = anArray(15);

or

int y = anArray.GetAt(15);

Or you have better ideas on preventing range excess in:

anArray[ index ] = value;

?

the return value isn't even used in this example:

anArray[ index ] = value;

so I have difficulty following you. returning *this accomplished nothing here.

Maybe do you absolutely have to have an example of the equal operator overloading. If so, then create a reference class and overload his = operator instead of doing so in the array class.

Quote:

Originally posted by FillYourBrain
the return value isn't even used in this example:

anArray[ index ] = value;

so I have difficulty following you. returning *this accomplished nothing here.

Not quite, maybe I could not make myself clear enough.

The object MUST be of MyArray type, as operator = is redefined in this class. If operator [] would return an int, then operator = defined for int type would be called, not the one I WANTED to.

Just try out both cases, you'll see what happens.

What about my suggestion, why not write your *own* int reference class then?

Quote:

Originally posted by lyx
Maybe do you absolutely have to have an example of the equal operator overloading. If so, then create a reference class and overload his = operator instead of doing so in the array class.

The example presents also an alternative way of handling some situations which can cause errors like over indexing.

Yes, but this is the task of your array class which returns a int reference class which task is to overload the = operator, and nothing more. Is that right with you?

Quote:

Originally posted by lyx
What about my suggestion, why not write your *own* int reference class then?

I'm planning to extend MyArray class to template class.

Or should I write another template class for operator = for all the primary types (char, int, etc.) ?
I rather live with this small inconvenience - that assignements like int x = anArray[ index ] won't be supported / compiled.

I might change my mind :)

Then write a template for the element class as well.

[EDIT]Don't answer to my older posts, just to the newest ones.

wait, wait, wait. I see where the confusion lies. First of all..

anArray[ index ] = value;

is to set an "int" inside the array. you are supposed to return a reference to an int in operator[]. This would set the thing with value.

I misunderstood but the example is still wrong. It makes no sense to return a *this because that's not what the array contains. You can simply return a reference to the corrected int instead. operator= on the array class doesn't make sense as a solution.

in addition, setting the last element as an alternative to the incorrect one they asked for is not safe as it would overwrite their valid data.

But FYB (could I call you that? ^^), he *wants* to overload a = operator...

but...

myArray = 4;

what does that mean? When you set an array equal to 4. what is that? Overloading operators is there to make things more readable, not more difficult to understand. really, it'll work that way (starts to become a state engine for no reason though) but he's writing a chapter in a book on operator overloading. It's incorrect to make an equals operator that makes no logical sense like that. that's all I'm saying. Also, it's more correct to return a reference to the thing you're actually expecting.

what would this give you?

int i = myArray[3];

it would return a myArray reference? not terribly helpful if you ask me!

Read all what have been posted, I said that I agree with you, however, if he really want that much his overloaded = operator, he should make a reference class. And that if he wants a template class, he should make both templates, that's all.

Quote:

Originally posted by FillYourBrain
wait, wait, wait. I see where the confusion lies. First of all..

anArray[ index ] = value;

is to set an "int" inside the array. you are supposed to return a reference to an int in operator[]. This would set the thing with value.

I misunderstood but the example is still wrong. It makes no sense to return a *this because that's not what the array contains. You can simply return a reference to the corrected int instead. operator= on the array class doesn't make sense as a solution.

Ok, FillYourBrain, I got it. There's no need to overload operator =...

Code:

int& MyArray::operator [] ( int index_in ) { // range check if ( index_in > myLastIndex ) { myIndex = myLastIndex; } else { myIndex = index_in; } return myArray[myIndex]; }

Quote:

in addition, setting the last element as an alternative to the incorrect one they asked for is not safe as it would overwrite their valid data.

Of course, it's just a temporary workaround, I planned to do some error handling and logging here.

yeah, carlos. Sorry I misunderstood you for about a whole page. The proposal that you were making was strange and unexpected so I went off in another direction.

All those words for finally you accepting that the operator isn't needed...

Quote:

The object MUST be of MyArray type, as operator = is redefined in this class. If operator [] would return an int, then operator = defined for int type would be called, not the one I WANTED to.

I believed you WANTED to write your own operator, didn't you?

he could certainly justify an operator=(MyArray & arr). It makes sense to set an array equal to an array. That would be a good example. Not that Carlos is looking for examples from me :p

Quote:

Originally posted by lyx
All those words for finally you accepting that the operator isn't needed...
I believed you WANTED to write your own operator, didn't you?

To correct myself: the operator = is not needed for that kind of assignements. However, I will surely redefine it, for cases like:

1.) anArray = value;

Quote:

Originally posted by FillYourBrain
he could certainly justify an operator=(MyArray & arr). It makes sense to set an array equal to an array. That would be a good example. Not that Carlos is looking for examples from me
[/B]

Thanks FYB! :)

2.) anArray = anotherArray;

etc...

Not that but, we were saying since the beginning that there were no need for such a redefinition. Remember, you said:

Quote:

myArray[num]

should return a reference to the type stored in the array. NOT the array class itself.

And I agreed.

I feel like I wasted my time for something that wasn't worth it, that's all... A problem that did even never exist for someone who claims to need what he doesn't need. At least, one must know what he wants to ask when submitting a question.

lyx, take it easy man. no big deal here. don't get stressed. not good for the heart :p

Quote:

Originally posted by lyx
I feel like I wasted my time for something that wasn't worth it, that's all... A problem that did even never exist for someone who claims to need what he doesn't need. At least, one must know what he wants to ask when submitting a question.

Look, the way I first made my class was redundant, and FillYourBrain lead me to the correct solution.

P.s.
Lyx, sorry if I caused you such a great problem, headache, maybe nearly brought you to suicide :)
You just can ignore the topics and don't bother yourself if you are such a sensitive person.

The problem is that things get clear while discussing it. To discuss something you need people who are able to DISCUSS, real team-players. If you don't feel like this, you will probably cause conflicts here in the (near) future.

Sometomes people ask silly things, not only the newbies. Nobody is perfect.

I extended my Class to dinamically allocate memory in case of over-indexing.

Code:

int& MyArray::operator [] ( unsigned int index_in ) { // range check if ( index_in > myLastIndex ) { unsigned int anArraySize; // when the range is exceeded, elements of myArray are copied to a temporary array // myArray is deleted, then a new myArray of size index_in is created, and the previously saved // data are restored // MAX_ARRAY_SIZE_VALUE can be set to infinite, error handling WILL be introduced if call to new(MAX_ARRAY_SIZE_VALUE) fails if( index_in == 0 || index_in >= MAX_ARRAY_SIZE_VALUE ) { anArraySize = MAX_ARRAY_SIZE_VALUE; index_in = MAX_ARRAY_SIZE_VALUE - 1; } else { anArraySize = index_in + 1; } // create temp array int* aTempArray = new ( int[myLastIndex + 1] ); // store myArray data for( unsigned int i = 0; i <= myLastIndex; i++ ) { aTempArray[i] = myArray[i]; } // delete myArray delete[] myArray; // create the new myArray (size increased), error handling will be introduced! myArray = new ( int[anArraySize] ); for( i = 0; i <= myLastIndex; i++ ) { myArray[i] = aTempArray[i]; } // initialize the rest for( i = myLastIndex + 1; i <= index_in; i++ ) { myArray[i] = 0; } myLastIndex = index_in; myIndex = index_in; delete[] aTempArray; } else { myIndex = index_in; } return myArray[myIndex]; }

how do you feel about memcpy instead of that copying for-loop?

Quote:

Originally posted by FillYourBrain
how do you feel about memcpy instead of that copying for-loop?

I thought of that, but I tried to build first an easy-to-understand example. Later it will be optimized and error handling will be also introduced.
Anyway thanks for the idea!

that's kinda what I thought.

This is an interesting type of class. If you haven't already, you should take a look at the code for the STL vector. Its actually really fun how they use buffer copies, copy constructors and placement new to give it some speed. Such an exploration could be beneficial in your book.

Thank you FillYourBrain!
I really appreciate your help.

By the way, STL will be also discussed in my book, originally I planned just a basic introduction, but you made me realize that I could eventually compare STL containers with some self-made classes.

First, I will talk about the code, then I will take out your comments.

You're wasting execution time by allocating an extra array. aTempArray is completely useless the way you use it.

Instead of:

Code:

allocate temporary array (old size) copy array into temporary array free array allocate array (new size) copy temporary array into array free temporary array

You should be doing:

Code:

allocate temporary array (new size) copy array into temporary array free array assign temporary array to array

Another thing is that you should put

Code:

myIndex = index_in;

out of the conditionnal branches because it is done either way.

And, you shouldn't be using myIndex as the index right now because myIndex has the value of index_in, so instead of using a member variable, use a local one.

Lastly, the

Code:

if (index_in == 0)

makes no sense as you use unsigned indexes and we have index_in > myLastIndex; thus, index_in cannot be equal to zero, even if myLastIndex's value were zero, index_in would be greater than zero.

Quote:

I thought of that, but I tried to build first an easy-to-understand example. Later it will be optimized and error handling will be also introduced.

It sounds quite funny, you are teaching one of the most difficult point in the C++ language, though you bother with your students not understanding memcpy()?

Quote:

Lyx, sorry if I caused you such a great problem, headache, maybe nearly brought you to suicide
You just can ignore the topics and don't bother yourself if you are such a sensitive person.

Just don't make fun of me with your suicide thing, I knew someone who killed himself while ago and there's nothing to laugh about, even if I am only 15 and you are older and know much more than I do, it does not give you the right to play with me.

Quote:

The problem is that things get clear while discussing it. To discuss something you need people who are able to DISCUSS, real team-players. If you don't feel like this, you will probably cause conflicts here in the (near) future.

Sorry but I discuss of the solutions and not of the problem, it is your task to know what you want us to solve. Besides, if you want me to leave, I will do, once I make sure that I cannot give to this place anything but ruination.

Quote:

lyx, take it easy man. no big deal here. don't get stressed. not good for the heart

Thanks for your regards but it has been while since I forgot the true meaning of that. ^^

[edit]Corrected some grammar mistakes.

lyx, thank for your help, I made the code in a hurry, and it needed some optimization indeed.

On the other side, I didn't try to make fun of you, just felt *really* offended by you. Maybe you had a bad day, I don't know, but cannot accept the "I_helped_you_therefore_I_am_authorized_to_kick_yo ur_head_off" style. Sorry!

Quote:

On the other side, I didn't try to make fun of you, just felt *really* offended by you. Maybe you had a bad day, I don't know, but cannot accept the " I_helped_you_therefore_I_am_authorized_to_kick_you
r_head_off" style. Sorry!

In fact, I felt really irritated by you changing your mind that way after going that far into the problem, I just thought you would have done it sooner if you wanted. Besides, the "not_a_thank_to_those_who_helped_me_that_far" style isn't better, is it?
Anyway, I could say that I'm sorry.
Consider we are quits.

Quote:

Originally posted by lyx
In fact, I felt really irritated by you changing your mind that way after going that far into the problem, I just thought you would have done it sooner if you wanted.

In fact, I didn't realize that I overcomplicated my class for no reason. So I didn't changed my mind, just corrected the code.
Anyway, this is how you react to people who dare to change their minds? Wow!

Besides, the "not_a_thank_to_those_who_helped_me_that_far" style isn't better, is it?
Have you just re-checked my answers? I just can't thank every answer, we were discussing things. As an alternative I could change my signature to "Thank you very much!", but I won't :)

Quote:

Originally posted by lyx

Anyway, I could say that I'm sorry.
Consider we are quits.

I could say, too...

Quote:

Have you just re-checked my answers? I just can't thank every answer, we were discussing things. As an alternative I could change my signature to "Thank you very much!", but I won't

Yes, I checked them all again. ^^
Yet, I don't ask for a thank for each post, only once everything is over. Besides, it was just a "plus" to make me irritated.

Quote:

Anyway, this is how you react to people who dare to change their minds? Wow!

After an hour trying to solve the old problem, yes. ^^ It wouldn't had been so if you did sooner. :)
Actually I'm not a very patient person and making me look for solutions to finaly take the first one doesn't do good. ^^

wow, lyx. You're nuts. I've never seen so much made over nothing.

Carlos, imagine this guy as a moderator. :D

[quote]wow, lyx. You're nuts. I've never seen so much made over nothing.

Carlos, imagine this guy as a moderator.[quote]
To make things of nothing is the way to get good marks when it comes to French (English for you guys) tests. :D

However, this time, I'm just saying what I think. But, not all what I have said so far is true the way most people will see it. Want to know what I'm really saying? ;)

Then take a look at the "not_a_thank_to_those_who_helped_me_that_far" stuff I was talking about. What I mentionned is really nothing, because I don't work for a thank, if so, I wouldn't help at all and I think you realized that; however, I said that it *wasn't better* than the "I_helped_you_therefore_I_am_authorized_to_kick_yo ur_head_off" style. Here, the "not better than" is to be understand as "is the same as". Got it? ;)

Don't worry, I have a very twisted mind so if you haven't noticed what I meant, it is not serious. ^^

Though, the rest of my speech was but reality, I really got irritated, you know. ^^

In a word, I'm not mad, just a very strange person.

Quote:

In a word, I'm not mad, just a very strange person.

That's more than one word.

Lol, here, we say that quite commonly to make a summary. Don't know if you do the same, maybe it is just a mistake.

no, people do that here too. And it's kinda annoying.

It's like "I could care less" which is flat out wrong but a lot of people say it all the time. The saying is "I COULDN'T care less". Reason? Because the saying is supposed to suggest that you care so little that there's no room to care any less, thus I couldn't care any less than I do. While "I COULD care less" makes no sense as an insult.

Quote:

Originally posted by FillYourBrain
wow, lyx. You're nuts. I've never seen so much made over nothing.

Carlos, imagine this guy as a moderator. :D

I think he's driving he's colleagues and teachers mad.

Hey Lyx, just don't get too far with this way of thinking, or you'll find yourself soon in a psychiatry :D
You are really clever in spite of your young age (are you really only 15?), or at least you have good C/C++ knowledge.

However, there are lot of things one must learn when working together or discussing with other people.

E.g. in case you will be working in a team, your colleagues probably won't tolerate this style, and will probably avoid you. I know this, because we have such guys here at the company. They are probably good experts, but not very social.

Sorry this thread went in such a direction. I've never met such situation in the past three years.

Again, sorry if I offended you with my "inconsequency".

If you really feel you cannot afford my style, just please ignore me. Else your constructive ideas are always welcome!

Yes, I'm 15 but I've programmed for almost 3 years... I have always been interested in programming, so I've learn a lot in the domain.
I started with Pascal, then learnt Delphi, C (for a long time I've been using much more C than C++) then I took a look at win32 API. I took a break with HTML and JavaScript, later CSS. Meanwhile I started C++. At the end, I stopped learning C++ as I couldn't figure out what was the meaning of the word "template" (a year ago, I didn't know as much English as today) so I just did without. Well, at the time, I also did some DirectX 7 and OpenGL. Then, I read some assembly x86 documents, DOS assembly, then Windows assembly, PMODE assembly and finally return to Win32/COM/OLE/ActiveX assembly. (not very useful you'd say, well, really useless it is but it was amusing, it took me about 4 or 5 days to make a very simple ActiveX component which displayed a text with a color from an entry in the HTML page) At that time, the first .NET distributions I heard of were released, it made me realize that there were nothing more to do with COM things. (or so I believed) So I gave up, tried Java but gave up soon too as it was like .NET something "managed". (besides they have their "simple development" strategy and I hate to see fools who do not know anything saying they do know programming, and I saw some...)
Well, the story does not end here, unfortunately for you. ;) (I haven't learnt templates yet) Next, I tried to tackle the network technologies, sockets programming (HTTP/SMTP/POP3 only) and also PHP/MySQL, XML/XSLT with XHTML. Eventually, I returned to C++ after a long time not using it. I used my C++ skills again by learning DirectX 9. I completed my knowledge by reading the BS.
More recently, I've studied Design Patterns and Data structures as well as some algorithm things.

That's all! Well, you must wonder, how I can prove that. The answer is simple: I can't, I spent all my time studying so I didn't do any consequent program but only exercises.

Thanks for the compliment, and, yes, some people just avoid me when others think of me as mad man; but in reality, I often look like a stupid person who does stupid things.

For the team works, well, I actually don't feel like helping in a team or something... And it is something we learn at school. The fools, they think by giving work that cannot be done alone, we'll get mixed together, but they have underestimated us. ;)

Never met such a situation in the past three years, eh? I did, but, well, it did almost never end that peacefully, everyone got angry and finally asked me to leave most of the time lol. But not because I got irritated for the same reason but because some were saying silly things and I corrected it, then they didn't appreciate and... well you see, I am 15 after all. ^^

[edit]grammar...

Ok lyx, so you're the mad antisocial genius type? :D
Then you're welcome, this site really needed you!

Well, actually, I was looking for a student to whom teach what I learnt these years during or I could write some tutorials but I almost completely lack of pedagogy so a "disciple" would be better, but can't find anyone valuable. I mean, someone who could learn somehow like I did but with some help from me. But I have a learning strategy which is really awful, I try to handle all the theory in one go, then I just do exercises to practicize a bit.
I had a "teacher" once, but it has been while since he cannot teach me anything. Mainly because we took two different way, he continued to Java and I to assembly at that time. And also because he don't learn as fast as I do, and that because he is lazy...

And I found it strange, here, everyone is student somewhere (or are professionnals) whereas in France, everyone is doing this on his own. (I mean there are almost no useful classes in the domain)

I like programming mainly for it's purpose: creating applications.

It seems to me that you learn in a "l'art pour l'art" fashion. Why don't you start some projects on your own. With the knowledge you gathered, you probably could write some nice programs.

My intention was always to create something functional, never understood those who knew the theory, but never made a single program on their own, only if it was needed.

I'd suggest you to start an exciting project, e.g. some DirectX game. Writing tutorials can be also fun. I wrote some articles for PC-World, then I realized I could write a book. Now I'm working on it. There is always a way to use your creativity.

And I'd advice you to forget this disciple thingie for a while. To teach somebody you need to be *really* patient with your students.

Master wasn't that patient with me, but I understood quite fast...
But well, I forgot the idea long already, but I was just thinking it can be amusing. Because yes, I do it because I find it fun. ^^
Oh, and one more thing, you don't choose your students, you do choose your disciples, or so is my opinion.

Well, the way I feel about learning is, indeed, just for the knowledge, in the "l'art pour l'art" way like you said.
It is just great to know, nothing more. Socrate said that the only good was knowledge.
I, as an human, live for two things: to know and to feel. To make both my spirit and heart happy.

To start a project? Well, why not, but DirectX, well, I don't feel like doing graphics right now. ^^
But, at present, I'm doing some works on sockets, not something very impressive, just for some friends of mine.

And for tutorials, I think it can be very fun, but, I don't know where to begin and what to do in what order...