std Pro Tips for C++

Published Sep 04, 2017Last updated Nov 25, 2017
std Pro Tips for C++


Another mentoring session, another time to sneak in some pro-tips. This time, it was all about the C++ Standard Library being really helpful.

Swapping things

If you have a need to swap things around in your code, please don't do this:

double temp = a;
a = b;
b = temp;

There's a super handy tool for the job:


It'll work for a lot of things, including containers and your own types (given some conditions), and it's actually really, really smart (for a nerdy example, it will utilize built-in CPU intrinsics to swap out primitive values in just a few micro-ops).

Updating minimums

In dynamic programming, there's often a need to implement a "running minimum/maximum" algorithm. In simple terms, that means taking values one after another and ending up with the smallest and biggest ones. That's one way to do that:

void update(int element) {
    if (element < minimum) {
        minimum = element;

But three lines is three times more than it needs!

minimum = std::min(element, minimum);

This also works wonders when you have to create a third value out of two singles, e.g. some composition of two objects.

While we're on minimums...

... you've probably seen code that did this:

int minimum = 9999999;
int maximum = -9999999;

That's lousy! Setting aside weird magic numbers, there are bigger numbers than 9999999 that'll fit in an int. Standard library comes to the rescue again:

#include <limits>
int minimum = std::numeric_limits<int>::max();
int maximum = std::numeric_limits<int>::min();

There's also a catch there — this won't work properly for double. To get the real smallest negative double, you'd need to use std::numeric_limits<double>::lowest() instead. However, IEEE754 floating point numbers have values even smaller (bigger) than that! Specifically, a float/double that will compare smaller/greater to every other value can be obtained by std::numeric_limits<double>::infinity() (add a minus sign in front for the minus-infinity), which is probably the best way to initialize values for such an algorithm.

Hope that was helpful!

Discover and read more posts from Bartek Banachewicz
get started
Enjoy this post?

Leave a like and comment for Bartek