map erase issue

Yu Zhiquan

When I use map erase in "for", it will be error, I do not know the reason, who can tell me details, and how to correct it? thanks a lot! following is code: #pragma warning(disable:4786) #include #include #include using namespace std ; typedef map STRING2INT; void main() { STRING2INT MyMap; STRING2INT::iterator MyIterator; MyMap["January"] = 1; MyMap["February"] = 2; MyMap["March"] = 3; MyMap["April"] = 4; MyMap["May"] = 5; MyMap["June"] = 6; MyMap["July"] = 7; MyMap["August"] = 8; MyMap["September"] = 9; MyMap["October"] = 10; MyMap["November"] = 11; MyMap["December"] = 12; for (MyIterator = MyMap.begin(); MyIterator!=MyMap.end(); MyIterator++) { MyMap.erase(MyIterator); } cout << MyMap.size() << endl; }

Abebe

don't delete elements from a map while iterating through it, you'll probably end up accessing pairs that are deleted. Er zit een korstje op mijn aars.

Michael Dunn

When you delete an element, any iterators pointing at that element become invalid. So after calling MyMap.erase(MyIterator) you can't use MyIterator again (like in the ++ expression) unless you reset it first. --Mike-- Ericahist | CP SearchBar v2.0.2 | Homepage | RightClick-Encrypt | 1ClickPicGrabber There is a saying in statistics that a million monkeys pounding on typewriters would eventually create a work of Shakespeare. Thanks to the Internet, we now know that this is not true.

Joaquin M Lopez Munoz

Just to add to the previous comments, the following

for (MyIterator = MyMap.begin(); MyIterator!=MyMap.end(); )
{
MyMap.erase(MyIterator++);
}

is correct. Joaquín M López Muñoz Telefónica, Investigación y Desarrollo

miropl

... and just one line more. typedef map< char*, int > STRING2INT; instead typedef map STRING2INT; // it's never going to work Miro

Joaquin M Lopez Munoz

I didn't see the declaration of STRING2INT, as the original poster forgot to tick the "Do not treat <'s as HTML tags" checkbox :) Now that you have revealed the definition of STRING2INT to the world, another subtler flaw comes to light. The type

map< char*, int >

does not guarantee the behavior one would assume at first sight. Even worse, it can seem to work until less expected. The problem is that what's being used as a key are pointers to chars, and pointing to equivalent contents does not imply pointer equality. Consider:

typedef map< char*, int > STRING2INT;
STRING2INT m;
m["January"]=1; // #1
cout<<m["January"]<<endl; // #2

No guarantee is made that the first pointer in #1 is the same as that in #2. Usually, this equality happens to work because the compiler is smart enough not to duplicate literals, but the following will fail for sure

m["January"]=1;
string str="January";
cout<<m[str.c_str()]<<endl;

Of course, the simple solution is to define STRING2INT as a map< string, int >. Regards, Joaquín M López Muñoz Telefónica, Investigación y Desarrollo

miropl

Yes, char* is difficult to manipulate, but char* also is a pointer to null terminated string - the basic C/C++ strings primitive type (well know "C" string). A little modification of "The problem is that what's being used as a key are pointers to chars". No the key are pointers to null terminated strings. Now let's see the code. typedef map STRING2INT; // Thank's Mike m["January"]=1; string str="January"; // cout< 012345678901 the next (12) byte is '\0' // |||||||||||| char somearr[] ="MonthJanuary"; char* psomearr = &somearr[5]; // lets take the address of the fifth byte char** ppkey = &psomearr; // let's try indirect cout<. Best regards, Miro

ZoogieZork

miropl wrote: _const char* key = str.c_str(); // init that's ok cout< That shouldn't even compile if the type of m is map<char*,int>, since key is a const char* and not a char* (as referenced in the template). In fact, on VS.NET 2k3 and GCC 3.2.3, the following code:

#include <map>
#include <iostream>

using namespace std;

typedef map<const char*,int> mymap_t;

int main(int argc, char* argv[]) {

mymap\_t m;

m\["January"\]=1;
string str="January";

const char\* key = str.c\_str();
cout << m\[key\] << endl;

return 0;

}

Prints out 0 instead of 1. The reason why is the crux of what Joaquín was saying -- it doesn't matter what the char* is pointing to (whether it's a zero-terminated C string or a big buffer of random char crud), the key is treated as a simple char pointer, and thus takes on pointer comparison semantics when used in the map. Here's a simpler example of how things can go wrong with char*:

#include <conio.h>
#include <map>
#include <iostream>

using namespace std;

typedef map<char*,int> mymap_t;

int _tmain(int argc, _TCHAR* argv[]) {

mymap\_t m;

m\["January"\]=1;
char \*str = strdup( "January" );

cout << m\[str\] << endl;  // Prints out 0, not 1.

free(str);

return 0;

}

- Mike_

souldog

By the way, if you want to make the map empty then just call clear()

f2

Yu Zhiquan wrote:

for (MyIterator = MyMap.begin(); MyIterator!=MyMap.end(); MyIterator++) { MyMap.erase(MyIterator); }

the moment u call MyMap.erase(MyIterator), the size is reduced and begin() has changed. so in your for loop, MyIterator++ is directing u to the "3rd" element. u should use while loop to delete by check the size, not begin till end. from, -= aLbert =-