Python: Solving slow eval

Welcome to another one of my interesting findings, where I explore some stuff for fun and profit just fun

A while back I was solving AOC and for a problem (day 13), I had this sort of clever idea of using eval to skip writing a if else ladder. The input file had the following structure

  Starting items: 91, 65
  Operation: new = old * 13
  Test: divisible by 5
    If true: throw to monkey 7
    If false: throw to monkey 4

I had to parse the file to get all values, but when parsing the second line I thought why not use the python interpretor itself. And so, I created a variable called old and whenever it needed updating I would use the following line

old = eval( "line 2 after the equal part" )

The clever part was that I was skipping having to translate whatever symbol was present in the input (one of +, - , / or *) This was fine for the test input, but it became slow when I ran it against the larger input.
I ran python -m cProfile main.py to find out that it was eval that took a bunch of time. It was being invoked a lot of time because it was called in the update loop.
So, I thought,
why not store the function into a lambda before hand and then just call the lambda instead of calling eval? And that’s exactly what I did, and it worked. The unoptimized program took about 6 seconds when running on the complete input. But after I stored the operation in a lambda, it was basically instant. So, here is the main “trick”

operation = eval("lambda old: " + lines[2].split("=")[1].strip())

Notice the parameter name of the lambda “old” it is necessary because the string is “old” in the input.

So, why is running eval repeatedly slow?
Eval is used for evaluating dynamic expressions, but we already have a fixed expression, so eval is doing extra work trying to evaluate the same expression many times, if we compile the expression to a lambda, we can just execute a lambda and it is a lot faster.