• 0

C++ (STL): remove specific object from list


Question

Hello,

I have learned that I can create a linked list of objects, like this:

list< MyClass* > = myList;

I also know that if it were a list of primitive types, such as a list<int>, I could remove a specific element by calling

remove(myList.begin(), myList.end(), 8);

which for instance would remove any element that is 8.

However let's say MyClass has an attribute name_. I have a list of objects of type MyClass, and I'd like to create a function that finds and removes the object that has that name.

Stub:

void remove(string &name) const

{

remove(myList.begin(), myList.end(), ???);

}

The thing is, if I supply the string &name as an argument to the remove algorithm, it will attempt to compare MyClass objects with strings, which doesn't make sense. What I'd like it to do is compare the name_ attribute of its MyClass objects with the string &name.

I thought of supplying a boolean function that would return element->name_ == name, but the function would need two arguments (list<MyClass*> element, string &name) and I don't think that would work.

As you see I'm stuck.

10 answers to this question

Recommended Posts

  • 0

The way I have done this (which is probably wrong as I taught it to myself) is to use a std::map.

This way you would do something like:

 std::map&lt; std::string name, MyClass* &gt; MyList;
MyList["ClassName"] = PtrToClass;
MyList.erase( the name of the class (the std::name), in this case "ClassName" );

Even if this is wrong hopefully someone can show me how to do it right.

  • 0

for a list:

std::list&lt;MyClass*&gt; MyList

a remove function could be like:

void remove(string &amp;name) const
{
  std::list&lt;MyClass*&gt;::iterator iter;
  for (iter = MyList.begin(); iter != MyList.end(); iter++)
  {
	if ((*iter)-&gt;name_.equalis(name))
	{
	  MyList.erase(iter);
	  break;
	}
  }
}

  • 0

Lant's suggestion is probably the correct one if changing the container is practical which would depend on the particular circumstances.

dev's suggestion is probably the easier (and by no means wrong) if you have to stick with list<>.

The more STL solution would be to use one of the supplied algorithms, in this case remove_if would seem appropriate. The link shows some sample code which uses 'compose1' and 'bind2nd' to construct the 'predicate' function. This is also the STL way of doing things, i.e. you are reusing generic components instead of writing lots of special case code.

If all this 'predicate' stuff is a bit confusing try a different approach. Give your class an overloaded equality operator with another class of the same type which only compares their name_ attribute. Then pass a dummy instance of such a class to remove with the name_ attribute set to the desired value to remove.

As you can gather there are really a load of approaches and this can make C++ seem overwhelming at first. Just remember no one way is necessarily the best. But usually the professionals prefer the solution which reuses existing generic code rather than the individual approach.

  • 0

Ok I'm starting to make some sense of out bind2nd and compose1. Really cool stuff.

Another important question I have about the STL: let's say I have a vector of pointers to dynamic objects, for example:

std::vector< MyClass* > myVector;

MyClass * object1 = new MyClass();

MyClass * object2 = new MyClass();

myVector.push_back(object1);

myVector.push_back(object2);

When I call myVector.clear(), the cplusplus reference says that it will call the destructors for each of the elements. The thing is, I'm not sure if I need to call delete for each of these pointers or not.

  • 0
  Eoin said:
dev's suggestion is probably the easier (and by no means wrong) if you have to stick with list<>.

i tend to go for more verbose code due to me favouring the C way of doing things over C++. Only times i use C++ parts is when i want a list of things and don't want to make my own linked list, or i need classes with subclasses. Most of the time i avoid the STL due it not liking threading very well (in my experience anyway)

  Dr_Asik said:
Ok I'm starting to make some sense of out bind2nd and compose1. Really cool stuff.

Another important question I have about the STL: let's say I have a vector of pointers to dynamic objects, for example:

std::vector< MyClass* > myVector;

MyClass * object1 = new MyClass();

MyClass * object2 = new MyClass();

myVector.push_back(object1);

myVector.push_back(object2);

When I call myVector.clear(), the cplusplus reference says that it will call the destructors for each of the elements. The thing is, I'm not sure if I need to call delete for each of these pointers or not.

afaik you will need to delete the objects, i wasn't aware that calling clear() would call the destructors of the objects but apparently it does (Y)

  • 0
  Dr_Asik said:
When I call myVector.clear(), the cplusplus reference says that it will call the destructors for each of the elements. The thing is, I'm not sure if I need to call delete for each of these pointers or not.

Yeah I think you still need to call delete. I'm pretty sure the destructors only get called for objects stored by value

std::vector< MyClass > myVector;

MyClass object1;

myVector.push_back(object1);

If you need to store them as pointers consider using a smart point like boost::shared_ptr<>.

  • 0

Hm I'm really puzzled at how I am supposed to deallocate memory when erasing specific elements from a list. I thought about using auto_ptr (I can't use boost now), but the cplusplus reference advises against this:

Warning: It is generally a bad idea to put auto_ptr objects inside C++ STL containers. C++ containers can do funny things with the data inside them, including frequent reallocation (when being copied, for instance). Since calling the destructor of an auto_ptr object will free up the memory associated with that object, any C++ container reallocation will cause any auto_ptr objects to become invalid.

http://www.cppreference.com/cppmisc/auto_ptr.html

So what should I do?

  • 0

Well you can still manually call delete whenever you remove a element from a container, but ideally you do want a smart pointer to do this automatically.

If Boost IS an option then use shared_ptr as I suggested or the pointer containers. Or if you have access to tr1 the tr1::shared_ptr is the boost::shared_ptr basicially.

Otherwise you could try finding another reference counted smart pointer. The problem is inside STL containers objects can get copied as the container grows so if you used auto_ptr then at times there could exist two auto_ptr's to the same object. You need to use a reference counted smart pointer in this case as it ensures the object is deleted only when there are no smart pointers left pointing to it.

I know Scott Meyers published a reference counted smart pointer. Haven't used it myself but I imagine it should work fine. Of course it's published in one of his books but maybe you can find a listing of the code if you look around his online articles. His site maintains a list.

Others surely exist out there too but ideally you should use the boost or tr1 shared_ptr if you can.

  • 0

Well, as I really have to stick to absolute basic stuff, the STL, and not use any sort of loops, I have developped a very heavy and unelegant solution. Basically I selectively copy the elements that must not be removed into a new container. But as the containers contain pointers to these elements, I have to really copy objects, a costly process. Then I call delete on every pointer of the old container, erase the container, and assign the new container to the old container. Then the new container is discarded.

This topic is now closed to further replies.
  • Recently Browsing   0 members

    • No registered users viewing this page.
  • Posts

    • Do they tell Google not to scrape their content via something like robots.txt? Do they specify anywhere that certain agents aren't to scrape? If not, tough. Plus there's no obligation on anyone's part to adhere to any directives that might be in this file anyway...
    • DMDE 4.3.5.823 Beta by Razvan Serea DMDE is a software designed to effectively recover lost data. It retrieves files and folders swiftly and stores them in the user-defined location. It is an easy to use yet powerful tool that will assist both novice and experienced users in getting back lost files in just a few simple steps. Free Edition includes all basic features but a single recovery operation recovers up to 4000 files in the current panel only (you should first open a subdirectory in the current panel and then recover files in the panel). In paid licenses there is no this restriction, and recovery of nested directories is allowed. Can paid versions recover more files than the free version of DMDE? If a file cannot be recovered in the DMDE Free Edition (or it is damaged after recovery) the same will occur in the paid versions. DMDE paid versions are capable of recovering the same files. The only difference is that paid versions can recover all found files in one go, as well as restore the directory structure presented in the free version. Professional Edition provides additional features: rights to provide data recovery services portable use on different computers one-time activation on client computers (including remote use) data recovery reports (include logs and file checksums) read support for E01 disk image files using logs when copying a disk (resume copying, multiple passes) customizable I/O handler script recovery of NTFS alternate data streams DMA access in DOS (for ATA interface) DMDE key features: Portable run without installation Support for NTFS, FAT12/16, FAT32, exFAT, ReFS, Ext2/Ext3/Ext4, btrfs, HFS+/HFSX, APFS Thorough FS and Raw scan, FS reconstruction for data recovery in complex cases Simple partition manager for express search, diagnostics, and restoration of partitions Disk cloning and disk image creating, including I/O error handling, reverse copying, and other features RAID constructor for virtual RAID reconstruction supporting levels RAID-0, RAID-1, RAID-4, RAID-5, RAID-6, delayed parity, custom striping, JBOD/spanned disks; automatic calculation of RAID configurations Cluster map to investigate file allocation Disk editor compatible with the most recent Windows versions which allows viewing, editing, and navigating through different disk structures using built-in and custom templates NTFS tools to work bypassing NTFS driver (copy, delete file, create, repair directory) Support for various device I/O interfaces and settings to work with damaged devices, disk images, NTFS compression and encryption, national names, large disks, large files, large sectors, and other features DMDE 4.3.5.823 Beta changelog: Expanded built-in signatures for RAW search functionality Added file list export to HTML format (DMDE Professional Edition only, view sample) Improved handling of I/O errors with selective skipping by error code Enabled preview support for additional image (graphic) file types (Windows only) Improved extfs reconstruction when copies of superblocks with group descriptors are found Fixed potential hang during Btrfs volume reconstruction Resolved issue with cluster list creation when subfolders are present Other improvements and fixes Download: DMDE 64-bit | 2.4 MB (Free, paid upgrade available) Download: DMDE 32-bit | 2.0 MB Link: DMDE Home Page | DMDE Manual | Screenshot Get alerted to all of our Software updates on Twitter at @NeowinSoftware
    • The BBC might have gone about this the wrong way, but if there is a revenue sharing program then they and all other "providers" of data should be included in the plan.
    • PicView 3.1.4 by Razvan Serea PicView is a fast, free and fully customizable image viewer for Windows 10 and 11. It supports a vast range of image file types, including WEBP, GIF, SVG, PNG, JXL, HEIC, PSD and many others. Additional features includes viewing EXIF metadata, image compression, batch resizing, viewing images within archives and comic books, image effects, image galleries, and more. Available in portable and installable versions. PicView 3.1.4 changelog: What's new Mouse Side Buttons Customization: You can now change how the mouse side buttons work. Choose to navigate file history or switch between directories. Find this setting under the Mouse tab in the settings window (#199). Improvements Directory Navigation: Improved navigation between directories. If Search subdirectories is enabled, PicView moves to the next directory in the list; if off (or if there are no directories in the list), it navigates as before. PicView now remembers your startup directory (including subdirectories) and restores it on the next launch. Performance: Preloader has been fine-tuned for better performance and lower memory usage. Bug Fixes Fixed incorrect saved settings path when saving in portable mode (#213). Corrected cases where keybindings and file history were not saved when there was no write permission. Restoring the window from a maximized state now correctly keeps its position when auto-fit is enabled. Fixed start-up menu text alignment and missing text issues. Translations Hebrew translation by @Y-PLONI (#212) Hungarian translation by @JohnFowler58 (#209) Japanese update by @coolvitto (#214) Miscellaneous Updated to Avalonia 11.3.1. Now using ZLinq and ZLinq.FileSystem to reduce memory allocation and possibly improve performance. Future versions managed via Scoop will now preserve configuration files (#15555). Experimental settings (not yet available in the UI—edit UserSettings.json directly): Disable the file watcher (not recommended, untested). Choose preload amount and direction (forward/backward). Download: PicView 3.1.4 | Portable ~50.0 MB (Open Source) Download: PicView ARM64 | Portable ARM64 Links: PicView Home Page | Github Project Page | Screenshot Get alerted to all of our Software updates on Twitter at @NeowinSoftware
    • The design looks good on their phones, but I can't say that I like the look of these.
  • Recent Achievements

    • Week One Done
      Crunchy6 earned a badge
      Week One Done
    • One Month Later
      KynanSEIT earned a badge
      One Month Later
    • One Month Later
      gowtham07 earned a badge
      One Month Later
    • Collaborator
      lethalman went up a rank
      Collaborator
    • Week One Done
      Wayne Robinson earned a badge
      Week One Done
  • Popular Contributors

    1. 1
      +primortal
      681
    2. 2
      ATLien_0
      276
    3. 3
      Michael Scrip
      221
    4. 4
      +FloatingFatMan
      170
    5. 5
      Steven P.
      164
  • Tell a friend

    Love Neowin? Tell a friend!