Stream: brlcad

Topic: C++

view this post on Zulip Himanshu (Jul 04 2021 at 14:52):

Just want to ask a silly question. During push_back operation in vectors, why previous elements that are already in the vectors are rearranged when I am using a move constructor? Sorry I asked a outside topic question.

view this post on Zulip Vikram Atreya (Jul 04 2021 at 14:53):

Can you put a code snippet maybe

view this post on Zulip Himanshu (Jul 04 2021 at 14:58):

https://paste.ofcode.org/?edit=gHRfmaMgbRbYasTEfaykAV

view this post on Zulip Himanshu (Jul 04 2021 at 14:59):

Move Constructor for: 10
Destructor freeing data for nullptr
Constructor for: 20
Move Constructor for: 20
Move Constructor for: 10
Destructor freeing data for nullptr
Destructor freeing data for nullptr
Constructor for: 30
Move Constructor for: 30
Move Constructor for: 10
Move Constructor for: 20
Destructor freeing data for nullptr
Destructor freeing data for nullptr
Destructor freeing data for nullptr
Constructor for: 40
Move Constructor for: 40
Destructor freeing data for nullptr
Constructor for: 50
Move Constructor for: 50
Move Constructor for: 10
Move Constructor for: 20
Move Constructor for: 30
Move Constructor for: 40
Destructor freeing data for nullptr
Destructor freeing data for nullptr
Destructor freeing data for nullptr
Destructor freeing data for nullptr
Destructor freeing data for nullptr
Destructor freeing data for:10
Destructor freeing data for:20
Destructor freeing data for:30
Destructor freeing data for:40
Destructor freeing data for:50

view this post on Zulip Himanshu (Jul 04 2021 at 15:01):

I want to know why after 30 is moved to the vector then previous elements values like 10 and 20 are moved too

view this post on Zulip Himanshu (Jul 04 2021 at 15:04):

also after 40 is moved to the vector then why previous elements like 10, 20, 30 are not rearranged. Am I missing something?

view this post on Zulip Daniel Rossberg (Jul 04 2021 at 17:59):

In every push_back() it has to make sure that there is enough memory for the additional element. If the additional memory can be provided in place (i.e., there is enough room after the vector), only the new element needs to be copied. But, this seems to be not the case for the majority of your push_back() calls. Probably because of the new operator in the Move constructor, which clutters the heap. Then, the Move objects have to be copied from the old memory of the vector to the new one.

BTW, you can reserve memory for a vector to make sure that it has enough memory for a distinct amount of elements.

view this post on Zulip Himanshu (Jul 05 2021 at 07:39):

Daniel Rossberg said:

In every push_back() it has to make sure that there is enough memory for the additional element. If the additional memory can be provided in place (i.e., there is enough room after the vector), only the new element needs to be copied. But, this seems to be not the case for the majority of your push_back() calls. Probably because of the new operator in the Move constructor, which clutters the heap. Then, the Move objects have to be copied from the old memory of the vector to the new one.

BTW, you can reserve memory for a vector to make sure that it has enough memory for a distinct amount of elements.

One more doubt that why destructor is not being called in reversed order but it called destructor from 10 not from 50?

view this post on Zulip Daniel Rossberg (Jul 05 2021 at 17:22):

This depends on the kind of iterator used to destruct the objects in the vector. I.e., this looks like a usual forward iterator. You can set a break point in Move::~Move() to investigate the implementation of the std::vector class. The standard template library (STL) is a library written in C++.

view this post on Zulip Himanshu (Jul 05 2021 at 17:25):

Thanks @Daniel Rossberg :grinning:

view this post on Zulip Himanshu (Jul 05 2021 at 17:26):

yeah I debugged it whenever it comes to destructor or move constructor it redirects to some other cpp files where it is implemented in some sort of library.

view this post on Zulip Daniel Rossberg (Jul 05 2021 at 17:27):

.cpp or .h?

view this post on Zulip Himanshu (Jul 05 2021 at 17:29):

sorry it's .h

view this post on Zulip Himanshu (Jul 05 2021 at 17:30):

it redirects to stl_vector.h

view this post on Zulip Himanshu (Jul 05 2021 at 17:30):

then something to new_allocator.h

view this post on Zulip Himanshu (Jul 05 2021 at 17:31):

With all the power and flexibility of C++ comes complexity.

view this post on Zulip Daniel Rossberg (Jul 05 2021 at 17:32):

Right. I'ts implemented in header files only. I.e., it's enough to write #include <vector>. You don't need to bind to a vector library.

view this post on Zulip Himanshu (Jul 05 2021 at 17:34):

Daniel Rossberg said:

Right. I'ts implemented in header files only. I.e., it's enough to write #include <vector>. You don't need to bind to a vector library.

but still idk why vscode redirects me to some sort of .h files and I included vector header file.

view this post on Zulip Daniel Rossberg (Jul 05 2021 at 17:39):

What do you mean by "redirect"? You included vector and this includes what's needed to implement std::vector, e.g. stl_vector.h, new_allocator.h, ....

There was once a vector.h header, which was nothing else than the vector without std::. The vector header was something like namespace std; #include <vector.h>.

view this post on Zulip Himanshu (Jul 05 2021 at 17:44):

Daniel Rossberg said:

What do you mean by "redirect"? You included vector and this includes what's needed to implement std::vector, e.g. stl_vector.h, new_allocator.h, ....

There was once a vector.h header, which was nothing else than the vector without std::. The vector header was something like namespace std; #include <vector.h>.

I created a breakpoint on move constructor and whenever debugging starts it jumps to another file like move.h

view this post on Zulip Daniel Rossberg (Jul 05 2021 at 17:46):

What is it for a move.h file? Where does it belong to? Is there already another Move class in the STL?

view this post on Zulip Daniel Rossberg (Jul 05 2021 at 17:52):

BTW, there is an implicit constructor Move::Move(void), which will be called to initialize the vector elements. When you break in this function, the place may look odd, because you haven't wrote code for this (but the compiler did). The break point's location is often in the declaration then.

view this post on Zulip Daniel Rossberg (Jul 05 2021 at 17:53):

However, I'm nor sure how it looks in Visual Studio.

view this post on Zulip Himanshu (Jul 06 2021 at 14:41):

Daniel Rossberg said:

However, I'm nor sure how it looks in Visual Studio.

redirects.gif

view this post on Zulip Himanshu (Jul 06 2021 at 14:42):

omg the quality looks cheap after converting to gif lol

view this post on Zulip Daniel Rossberg (Jul 07 2021 at 15:31):

For me, the quality looks good. The only issue I saw were the many frames which prevents converting (at least with ImageMagic). But, MPlayer can handle it.

view this post on Zulip Daniel Rossberg (Jul 07 2021 at 15:36):

The header files you see there belong all to the STL. You can see the path below the tab (usr > include > c++ >...) E.g., move.h contains the implementation of moving values in STL containers.

view this post on Zulip Himanshu (Jul 07 2021 at 19:29):

Daniel Rossberg said:

The header files you see there belong all to the STL. You can see the path below the tab (usr > include > c++ >...) E.g., move.h contains the implementation of moving values in STL containers.

I don't want to see those header files while debugging but I manage it.

view this post on Zulip Himanshu (Jul 07 2021 at 19:30):

Thanks @Daniel Rossberg :innocent:

view this post on Zulip Daniel Rossberg (Jul 08 2021 at 06:13):

:laughing: You want to use the vector STL container and you got it. It's part of your program now.
If you break inside one of your methods and this method was just called from a STL method, the debugger has to show it to you. It would be a bad debugger, if it wouldn't do this :grinning_face_with_smiling_eyes:

However, there is a shortcut through the STL stuff while debugging: Navigate with the Call Stack to the next function which doesn't belong to the STL.

view this post on Zulip Himanshu (Jul 08 2021 at 06:40):

Daniel Rossberg said:

:laughing: You want to use the vector STL container and you got it. It's part of your program now.
If you break inside one of your methods and this method was just called from a STL method, the debugger has to show it to you. It would be a bad debugger, if it wouldn't do this :grinning_face_with_smiling_eyes:

However, there is a shortcut through the STL stuff while debugging: Navigate with the Call Stack to the next function which doesn't belong to the STL.

:neutral: :grimacing: Now I understand why it is switching to other files while debugging.

view this post on Zulip Himanshu (Jul 23 2021 at 15:09):

Daniel Rossberg said:

In every push_back() it has to make sure that there is enough memory for the additional element. If the additional memory can be provided in place (i.e., there is enough room after the vector), only the new element needs to be copied. But, this seems to be not the case for the majority of your push_back() calls. Probably because of the new operator in the Move constructor, which clutters the heap. Then, the Move objects have to be copied from the old memory of the vector to the new one.

BTW, you can reserve memory for a vector to make sure that it has enough memory for a distinct amount of elements.

Came to know about .reserve() so it will not expand untill it hits the given range. Now move constructor and assignment will not be copied from old memory of the vector to new one everytime it is being called.

view this post on Zulip Sean (Mar 10 2023 at 18:21):

Here's a great talk that just posted from CPPCON 2022 about design patterns. Excellent speaker.. https://www.youtube.com/watch?v=PEcy1vYHb8A

view this post on Zulip Sean (Mar 10 2023 at 18:28):

Most of the CPPCON talks are available as of today: https://www.youtube.com/@CppCon

Last updated: Oct 09 2024 at 00:44 UTC