Generalized application-defined scalar functions

[trueqbit]

Another use of the recently established method for creating aliases and column pointers, this time for creating scalar functions from existing free-standing functions, lambdas and classic function objects.

So instead of going through the process of constantly creating your own user-defined function object ("boiler plate" coding), it's easy to turn existing functions (e.g. from libraries) into scalar functions like this:

// example for a freestanding function from a library
constexpr auto clamp_int_f = "clamp_int"_scalar.quote(std::clamp<int>);
// example for a stateless lambda
constexpr auto is_fatal_error_f = "IS_FATAL_ERROR"_scalar.quote([](unsigned long errcode) {
    return errcode != 0;
});
// example for a function object instance
constexpr auto equal_to_int_f = "equal_to"_scalar.quote(std::equal_to<int>{});
// example for a function object
constexpr auto equal_to_int_2_f = "equal_to"_scalar.quote<std::equal_to<int>>();

storage.create_scalar_function<clamp_int_f>();
storage.create_scalar_function<is_fatal_error_f>();
storage.create_scalar_function<equal_to_int_f>();
storage.create_scalar_function<equal_to_int_2_f>();

auto rows = storage.select(clamp_int_f(0, 1, 1));
auto rows = storage.select(is_fatal_error_f(1));
auto rows = storage.select(equal_to_int_f(1, 1));
auto rows = storage.select(equal_to_int_2_f(1, 1));

A nice advantage is that a stateless function object does not have to be constantly recreated, saving a few levels in the call stack.