看了下源码， CCTMXXMLParser.cpp 如下：
…
else if (elementName == “object”)
{
TMXObjectGroup* objectGroup = tmxMapInfo->getObjectGroups().back();
// The value for "type" was blank or not a valid class name
// Create an instance of TMXObjectInfo to store the object and its properties
ValueMap dict;
// Parse everything automatically
const char* keys[] = {"name", "type", "width", "height", "gid", "id"};
for (const auto& key : keys)
{
Value value = attributeDict[key];
dict[key] = value;
}
// But X and Y since they need special treatment
// X
int x = attributeDict["x"].asInt();
// Y
int y = attributeDict["y"].asInt();
Vec2 p(x + objectGroup->getPositionOffset().x, _mapSize.height * _tileSize.height - y - objectGroup->getPositionOffset().y - attributeDict["height"].asInt());
p = CC_POINT_PIXELS_TO_POINTS(p);
dict["x"] = Value(p.x);
dict["y"] = Value(p.y);
int width = attributeDict["width"].asInt();
int height = attributeDict["height"].asInt();
Size s(width, height);
s = CC_SIZE_PIXELS_TO_POINTS(s);
dict["width"] = Value(s.width);
dict["height"] = Value(s.height);
// Add the object to the objectGroup
objectGroup->getObjects().push_back(Value(dict));
// The parent element is now "object"
tmxMapInfo->setParentElement(TMXPropertyObject);
}