The art of Engineering (2): the transformation from general type to specific type needs to be tested for legitimacy

void foo() {
	QVariant a = 10;
	QVariant b = 20;
	int c = add(a, b);
	return c;
}
int add(const QVariant &a, const QVariant &b) {
	int left = a.value<int>();
	int right = b.value<int>();
	return a + b;
}

Here's add function , Although it's ours foo It will be called internally , But if it's not a performance bottleneck , Better be right add Check the parameters of .
The reason is that the subsequent maintainers are not as clear about the internal implementation of the function as we are , Some wrong types may be passed in .
This involves a defensive programming problem , We are the function writers , Try to prevent others from introducing a wrong thing , And try to solve this wrong thing .

This is similar to the question whether the smart pointer can take out the raw pointer and pass it into the function as an argument . In some cases , If we have enough knowledge about the internal implementation of functions , Ensure that it does not release the resource indicated by the pointer , We can pass the bare pointer .
One is the writer's perspective , One is the perspective of the caller .