Question

map<pair<double,double>,double> vectorDoubleMap;

vectorDoubleMap[ pair<double, double>(10, 10) ] = 1;          //1.
vectorDoubleMap.insert( pair<double, double>(10, 10), 1);     //2.

‘1.’ statement compile done, but ‘2.’ statement not compiled.

What’s the difference between these two statement?

Answer 1

map::insert takes a single pair argument holding key and value, not key and value as separate arguments. If you pass a second argument, the first is an iterator used as a hint for alikely insertion position, whereas the second is still a pair.

So you’d have to write one of these:

vectorDoubleMap.insert(pair<pair<double, double>, double>(pairs<double, double>(10, 10), 1));
vectorDoubleMap.insert(make_pair(make_pair(10, 10), 1));

You can see how make_pair makes the syntax much more concise in this case. (As the numbers are entered as integer literals, integer pairs will be constructed at first, but there is a conversion constructor for pairs of one kind to another, so this is all right and will be converted to double as needed. The compiler might even optimize the integers away and directly use doubles in the code it generates.)

If the key is not present in the map, the two statements will be equal in effect and similar in performance. If the key is already present, the […]=… form will overwrite it, whereas the insert(…) form will keep the old value.

Note that C++17 will likely offer two more ways to accomplish the same thing. insert_or_assign is roughly equivalent to your first line, since it will overwrite existing values. try_emplace is more like the second, since it won’t overwrite if the key is already present. The key benefit for both of these is that they accept arguments for a constructor, and construct the object in place. Which isn’t of much use for double, but might be useful for other objects which are hard to construct, copy or move. But still, the fact that try_emplace takes separate arguments and not a single pair makes that one a nice replacement for the insert discussed above.

vectorDoubleMap.insert_or_assign(make_pair(10, 10), 1);
vectorDoubleMap.try_emplace(make_pair(10, 10), 1);