Reading Code backwards

There is this nice book by Jason Turner called "Copy and Reference Puzzlers". It contains awesome brain teasers. You can do them alone, but they are also a nice warmup for meetings. We started our team meetings with a small competition who gets the right answer first.

Puzzler example

The book contains puzzlers, which read e.g. like the following:

char char_1{'j'};
char_1 = 'c';
char char_3{'s'};
char_1 = 'i';
char char_5{'d'};
char_5 = 'h';
char char_7{'t'};
char_1 = 'x';
char char_9{'p'};
char_1 = 'o';
char char_11{'u'};
std::cout << char_9 << char_11 << char_7 << char_3;

The task is to find the output that is printed by the given code.

The initial approach

My initial approach was to step through the code one line by line. This means reading the puzzlers top-to-bottom, front-to-back. After all, that is what the compiler does or how a debugger works.

However this approach places some high cognitive load on the reader. You need to keep track of all variables. A debugger does this automatically, and of course pen and paper will help. But it requires some effort to stay focussed. Naturally, some of the later puzzlers are way longer - even up to multiple pages. Finally, it is quite annoying to learn that some (in some puzzlers even most) of the variables are not even used for the answer. What a waste of brain power to keep track of all of them throughout the complete code.

The backwards approach

There surely is a better way to solve those types of puzzlers with a low cognitive load and without tracking variables in your head or on paper. This is by reading it back-to-front, or bottom-to-top.

In the given example: First start with the cout statement and pick the char_9 variable. Now follow the code upwards when this variable was changed last. Note that down as the first char of the answer and continue with the next variable, in this case char_11. Not only does this ignore all the variables that are not even needed, it will also allow you to stop for the current char, once you have found the lat time it was set.


While reading bottom-to-top is not a general approach for all source code, I found that it is quite useful also for non-pathologic code. For example I personally found it quite useful for getting a brief understanding of unit tests. Those often follow the "three As" (Arrange, Act, Assert). Mostly you are not interested in all of the details of the "Arrange" part and can simply skip to the assert statement directly. Reading code backwards also helps you to focus on the outcome.

For me it was very nice to experiment with a different approach to reading code.