std::map Memory Issue

[Josh]

driver->loadedassets is an std::map with the following format:

std::map<std::string,AssetBase*> loadedassets

 

This leaks memory:

   	 if (driver->loadedassets[rpath]!=NULL)
       //if (driver->loadedassets.find(fontpath)!=driver->loadedassets.end())
       {
           reference = (FontReference*)driver->loadedassets[string::Lower(rpath)];
           if (reference!=NULL)
           {
               return (Font*)(reference->Instance());
           }
       }

 

This doesn't leak memory:

   	 //if (driver->loadedassets[rpath]!=NULL)
  	 if (driver->loadedassets.find(fontpath)!=driver->loadedassets.end())
       {
           reference = (FontReference*)driver->loadedassets[string::Lower(rpath)];
           if (reference!=NULL)
           {
               return (Font*)(reference->Instance());
           }
       }

[Roland]

driver->loadedassets[rpath]

If rpath does not match the key of any element in the container, the function inserts a new element with that key and returns a reference to its mapped value. Notice that this always increases the map size by one, even if no mapped value is assigned to the element (the element is constructed using its default constructor).

 

driver->loadedassets.find( ....... )

Searches the container for an element with rpath as key and returns an iterator to it if found, otherwise it returns an iterator to map::end (the element past the end of the container).

 

 

Use find if you don't want to add something.

 

Seriosly

[Josh]

Yeah, but if it's a pointer, it shouldn't add anything, because the default value is NULL. The problem was solved with a Clean & Rebuild. Thanks anyways for the tip.

[Roland]

Hmm... is really a pointer's default value NULL. Not to sure about that.

However... Great that its solved

[Canardia]

VS is caching a lot of things, because it's slow, so it tries to overcome the slowness with caching, which causes even more problems.

Use CodeBlocks and MinGW, and problem is solved, and you also create more cross-platform code, so less time lost on porting to iPhone and Android and Linux.

VS is the worst C++ compiler in the world, MinGW is faster by 25%, and Intel C++ is faster by 47%. Just stop using VS and you have a better life, and your customers too.

 

Same with SharePoint btw, totally horrible how slow it is because of C# garbage collection, and how you need to spam a SharePoint server every minute with a wakeup webservice. Apache does not have any of those problems, it just works.

[Roland]

VS is caching a lot of things, because it's slow, so it tries to overcome the slowness with caching, which causes even more problems.

Use CodeBlocks and MinGW, and problem is solved, and you also create more cross-platform code, so less time lost on porting to iPhone and Android and Linux.

VS is the worst C++ compiler in the world, MinGW is faster by 25%, and Intel C++ is faster by 47%. Just stop using VS and you have a better life, and your customers too.

 

Same with SharePoint btw, totally horrible how slow it is because of C# garbage collection, and how you need to spam a SharePoint server every minute with a wakeup webservice. Apache does not have any of those problems, it just works.

And what has this to do with maps ???

[Josh]

Nothing.

[Benton]

You have memory problems remembering your STD's?

 

Joking joking...

[Canardia]

And what has this to do with maps ???

It is directly related to Josh's problem with maps, where the compiler cached things in the VS cache file and he needed to clean rebuild to make it work. For example codeblocks and mingw doesn't have such cache file, and it still compiles much faster, and creates faster exes, and the syntax highlighting works also faster and more accurate.

[wh1sp3r]

for intelisence and highlighting, i am using Visual Assist X perfect tool, i can't live without it.

 

http://www.wholetomato.com/

 

Also, it has refractoring functions, it makes your life easier

[DaDonik]

for intelisence and highlighting, i am using Visual Assist X perfect tool, i can't live without it.

 

Thats so true, once you have it you won't go back.

[Furbolg]

Jea Visual Assist (X) rocks, just don't do the same error as me, buy the standard license not the personal license.

[Roland]

Thats so true, once you have it you won't go back.

Yeah that one is great, but 249$... ouch

 

 

@Metadron

Aha. I see. It was a comment on the solution

[wh1sp3r]

i am happy as i can buy it as academic license

[Furbolg]

Yeah that one is great, but 249$... ouch

 

But its worth every dollar, just the better syntax highlighting and there are soo much awesome featues.

 

 

i am happy as i can buy it as academic license

 

You are a lucky one

[Josh]

***************************************

IMPORTANT

***************************************

 

It seems Roland is right. Well, sometimes. I am getting unreliable behavior when reading with the [] operator, if the key does not exist in the map. I recommend always using find() instead of the [] operator!

[Mumbles]

It seems Roland is right.

 

Find a time when he isn't. It will make headline news

 

 

Well, sometimes.

 

Treason! Orrf with is head!

[Furbolg]

***************************************

IMPORTANT

***************************************

 

It seems Roland is right. Well, sometimes. I am getting unreliable behavior when reading with the [] operator, if the key does not exist in the map. I recommend always using find() instead of the [] operator!

 

I can only suggest to use the MSDN (msdn.microsoft.com)

 

The MSDN says under "Remarks" the following:

http://msdn.microsof...ibrary/fe72hft9

 

If the argument key value is not found, then it is inserted along with the default value of the data type.

 

operator[] may be used to insert elements into a map m using m[_Key] = DataValue; where DataValue is the value of the mapped_type of the element with a key value of _Key.

 

When using operator[] to insert elements, the returned reference does not indicate whether an insertion is changing a pre-existing element or creating a new one. The member functions find and insert can be used to determine whether an element with a specified key is already present before an insertion.

[Josh]

Right but the value is a pointer, so the default value is NULL. Or it would be if STL was designed by any sane person.

mymap["whatever"]=NULL;

 

Ah well, thanks for the tips.

[Furbolg]

The default value of an pointer is undefined (in release mode).

 

prove

 

#include <windows.h>
#include <iostream>
#include <stdio.h>
int main(int argc, char** argv)
{
int* test;
int* test2 = NULL;
std::cout << test << std::endl;
std::cout << test2 << std::endl;
system("pause");
return 0;
}

[Roland]

The default value of an pointer is undefined (in release mode).

 

prove

 

#include <windows.h>
#include <iostream>
#include <stdio.h>
int main(int argc, char** argv)
{
int* test;
int* test2 = NULL;
std::cout << test << std::endl;
std::cout << test2 << std::endl;
system("pause");
return 0;
}

Exactly. In fact nothing except classes/struct's have default values in C++, and if you have no constructor which initializes the class/struct state, even they are in the dark undefined area.

[Josh]

We can see from this program, that the [] operator will insert NULL into the map when the key is not present, at least on the MS compiler:

#include "stdafx.h"
#include <stdlib.h>
#include <list>
#include <sstream>
#include <map>
#include <iostream>

class Foo
{
public:
int i;
Foo();
};

Foo::Foo() : i(37)
{
}

int _tmain(int argc, _TCHAR* argv[])
{
std::map<std::string,Foo*> map;

Foo* foo = map["Hello"];

// (add breakpoint here)

std::cout << (foo->i) << std::endl;

return 0;
}

[Roland]

Apparently this happens by some internal std::map magic . Not by a pointer automatically having a default of NULL. Anyway that's good to know.

[Canardia]

Not sure what your problem is Josh, but any map query will always add the element you are asking for, unless you use find. Find will do the lookup without adding any new elements to a map. And no, STL is not stupid, it's made for maximum speed, and you have to take care of all things which are not default. That keeps the speed to it's maximum.

[Josh]

It is stupid, it shouldn't add a value in when you're reading from it. They only did this because it has to return a pointer that can be gotten or set:

a = map[2]
map[2] = a

So it's a result of a limitation of the C++ language, because none of this stuff was implemented when they originally designed the thing. Which is why we call C++ an octopus made by nailing extra legs to a dog.