[c++] Recommended way to insert elements into map

Possible Duplicate:
In STL maps, is it better to use map::insert than []?

I was wondering, when I insert element into map, what is the recommended way. Should I

map[key] = value;

or

map.insert(std::pair<key_type, value_type>(key, value));

I did the following quick test:

#include <map>
#include <string>
#include <iostream>

class Food {
public:
    Food(const std::string& name) : name(name) { std::cout << "constructor with string parameter" << std::endl; }
    Food(const Food& f) : name(f.name) { std::cout << "copy" << std::endl; }
    Food& operator=(const Food& f) { name = f.name; std::cout << "=" << std::endl; return *this; } 
    Food() { std::cout << "default" << std::endl; }
    std::string name;
};

int main() {
    std::map<std::string, Food> m0;

/*
1) constructor with string parameter
2) copy
3) copy
4) copy
*/
    m0.insert(std::pair<std::string, Food>("Key", Food("Ice Cream")));

/*
1) constructor with string parameter
2) default
3) copy
4) copy
5) =
*/
    // If we do not provide default constructor.
    // C2512: 'Food::Food' : no appropriate default constructor available
    m0["Key"] = Food("Ice Cream");
}

1. I realize by using member function insert, less value's function call will be involved. So, is using insert a recommended way?
2. Why default constructor is needed, when map[key] = value way is being used?

I know that insert doesn't overwrite existence key value pair, but map[key] = value does. However, is this the only factor I take into consideration, when try to choose among the both?

How about

1. Performance
2. Availability of value's default constructor
3. ???

To quote:

Because map containers do not allow for duplicate key values, the insertion operation checks for each element inserted whether another element exists already in the container with the same key value, if so, the element is not inserted and its mapped value is not changed in any way.

So insert will not change the value if the key already exists, the [] operator will.

EDIT:

This reminds me of another recent question - why use at() instead of the [] operator to retrieve values from a vector. Apparently at() throws an exception if the index is out of bounds whereas [] operator doesn't. In these situations it's always best to look up the documentation of the functions as they will give you all the details. But in general, there aren't (or at least shouldn't be) two functions/operators that do the exact same thing.

My guess is that, internally, insert() will first check for the entry and afterwards itself use the [] operator.

map[key] = value is provided for easier syntax. It is easier to read and write.

The reason for which you need to have default constructor is that map[key] is evaluated before assignment. If key wasn't present in map, new one is created (with default constructor) and reference to it is returned from operator[].

Use insert if you want to insert a new element. insert will not overwrite an existing element, and you can verify that there was no previously exising element:

if ( !myMap.insert( std::make_pair( key, value ) ).second ) {
    //  Element already present...
}

Use [] if you want to overwrite a possibly existing element:

myMap[ key ] = value;
assert( myMap.find( key )->second == value ); // post-condition

This form will overwrite any existing entry.