My friend and I are arguing about the concept of anonymous (lambda) functions in C++. The utility of those is undeniable when considering rapid development and flexibility.
I'm arguing that they can't be worth it. Either the compiler would have to check for the format of the function, requiring its return type and arguments to be strict (identical to using a function pointer), or it would ignore everything that rules C++ versus scripting languages, and then debugging would be a total nightmare.
There is one key thing that distinguishes function pointers from function objects/anonymous functions: partial application. Given a function
int add(int a, int b) { return a+b; }
it is not possible with function pointers to create a new function f with one argument that adds a variable n to its argument, i.e., f(x) = add(x,n). With function objects it is, e.g., to add n to all the elements in a vector:
int n = ...;
vector v;
transform(v.begin(), v.end(), v.begin(), bind2nd(ptr_fun(add), n));
With anonymous functions in C++11, the syntax becomes more convenient (no ptr_fun and bind2nd):
transform(v.begin(), v.end(), v.begin(), [=](int x) { return add(x,n); } );
There is one key thing that distinguishes function pointers from function objects/anonymous functions: partial application. Given a function
int add(int a, int b) { return a+b; }
it is not possible with function pointers to create a new function f with one argument that adds a variable n to its argument, i.e., f(x) = add(x,n). With function objects it is, e.g., to add n to all the elements in a vector:
int n = ...;
vector v;
transform(v.begin(), v.end(), v.begin(), bind2nd(ptr_fun(add), n));
With anonymous functions in C++11, the syntax becomes more convenient (no ptr_fun and bind2nd):
transform(v.begin(), v.end(), v.begin(), [=](int x) { return add(x,n); } );
http://scholar.google.com.au/scholar?q=Anonymous+functions+C%2B%2B&btnG=&hl=en&as_sdt=0%2C5
Scholar About 1,650 results (0.17 sec)
- Badges
- Science topic
- Similar topics
- Computer Science