Python decorator for auto-memoizing pure functions
Memoization is an old computer science trick, that stores previous computation results to prevent the need to recalculate them. Often a memoizing program will lazily memoize results as they are used, so that next time they don't need to be calculated.
I was reminded of the wide utility of this technique by watching this MIT lecture recently, and reminded that memoization is a core concept in Dynamic Programming, which is a great technique for optimisation of algorithms.
MIT 6.006, Fall 2011, Introduction to Algorithms Lecture 19
Python decorators are well described in the 'Professional Python' book and I have been using them for years to wrap loggers and other things around functions.
But having watched this MIT video recently, it made me realise that it was possible to write a Python decorator that did the work automatically.
For the Fibonacci sequence generator, this turns an exponential time calculation into a linear time calculation. So, fib(300) will take a long time to calculate without memoization, as many of the sub-problems are re-computed a large number of times. Because fib(n)->m always yields the same m for n, in all cases (i.e. it is a pure function), this is a simple but very effective optimisation to make.
I left the print statements in this program so that you can see a trace on the console of what it does, but you would take those out if deploying this in production code.
@whaleygeek
13th Nov 2020