Modern Python Standard Library Cookbook

How it works...

Whenever the function is invoked, functools.lru_cache saves the returned value together with the provided arguments.

The next time the function will be called, the arguments are searched in the saved arguments and, if they are found, the previously returned value is provided instead of calling the function.

This, in fact, changes the cost of calling our function to being just the cost of a lookup in a dictionary.

So the first time we call fibonacci(5), it gets computed, then next time it will be called, it will do nothing and the value previously stored for 5 will be returned. As fibonacci(6) has to call fibonacci(5) to be able to compute, it's easy to see how we provided a major performance benefit for any fibonacci(n) where ...