If you're that concerned about disk reads, I'd just copy the file into shared memory space (eg: /dev/shm) on system or web server startup, then read the script from there instead.

Or, use a system that only has to read the file once upon web server instantiation.

This seems like premature optimization.

Update: I thought some more about this. If you're unit testing your code (which you should be for sure!), you'd have to run the tests again on this automatically re-written code in case something is lost in translation.

I'm all for doing things for education and learning purposes, but I don't think the risk is worth it if the sole objective is to use the code to try to make something a fraction of a nanosecond (obviously estimated) more efficient.