Stuff We Didn’t Cover

`static` member data

static members of a class are not associated with objects that are instances of the class but instead there is one value for all instances of the class. In this sense, they act as global data in the class’s namespace.

Memory management

We’ve been relying mainly on objects that manage their own memory internally (like std::vector). We saw one instance of allocating memory on our own using new or by make_unique. The latter is the safest way to deal with allocating memory.

Regular expressions

Regular expressions are a powerful way to search for patterns and extract substrings out of strings. The C++ regular expressions library provides all the capability needed to work with regular expressions.

`std::optional`

std::optional allows a function to return something that may or may not contain a value. Here’s a simple example (inspired by reddit ;)

Listing 156 optional.cpp

#include <iostream>
#include <optional>

std::optional<std::string> valid(const std::string& a_in) {

    if (a_in.size() < 2) {
        return std::nullopt;
    } else {
        return a_in;
    }
}

int main() {

    auto a = valid("test");
    std::cout << a.value_or("") << std::endl;

    auto b = valid("a");
    std::cout << b.value_or("") << std::endl;

}

Profiling

A profiler will allow you to time independent functions or even lines of code and generate a report telling you where your code spends most of its time.

GNU gprof is the GNU profiler than can be used with g++.

We can compile our code to be instrumented for profiling by simply adding the -pg flag to the compiler arguments. Then we run as normal and use gprof to examine the timing.

Here’s an example, using our Poisson Equation and Relaxation application. The GNUmakefile there will compile with profiling if we set PROFILE=TRUE:

make PROFILE=TRUE
./poisson
gprof poisson > gprof.out

We can then look at the gprof.out file to see where we are spending the most time.

Tip

We need to make sure that the program runtime is long enough that we sample functions many times. For the Poisson problem, this means using a bigger grid (like \(256^2\)).

Tip

We should profile the optimized build, e.g., using -O2 or higher and disabling asserts with -DNDEBUG.