I was trying to make a circular buffer in C++. I was also trying to expose iterators for using the buffer with STL algorithms. I kept trying to think about how to add the functions needed to manipulate the existing internal iterators to not exceed the bounds of the buffer. Then I realized I was "too close" to the problem. There was no way I could properly control the internal iterators of the storage vector I was using. Not without giving too much power to the user of my library. So I abstracted the iterators up one level. Hid all the details of the internal iterator and made a new iterator.

The solution of abstracting the iterator was not the epiphany. The epiphany was if you are struggling with how to solve a particular problem. You keep running into problems with how to represent something, there is too much power available at a particular representation, or the object you are trying to make work just don't fit. This is when you should consider abstracting a level up. Take a higher look at the problem and simplify the interface.

Abstraction could be a number of things. Divide and conquer, hiding details, specializing an object, etc. Whatever tool is needed to make the problem more consumable to your brain.

Add Comment