Abstraction

Abstraction is separating what something does from how it does it. The interface is the what: the operations you can use. The implementation is the how: the details underneath, hidden from you. The interface is a promise, and the implementation is free to change as long as it keeps that promise.

It hides the memory addressing. Underneath, the machine takes the array’s starting address and adds $2\times (\text{bytes per element})$ to compute the element’s address, then reads from that address. You write numbers[2] and never think about addresses or bytes.

It hides the fixed-size backing array: its capacity, how full it is, and the resizing. Without it, the client would have to commit to a capacity up front, track how many slots are in use, and, when the array fills up, allocate a larger array, copy every element across, and reassign the variable. The dynamic array does all of that on its own.

The interface says add appends a value; it says nothing about cost. Inside, most adds are a single write, but when the backing array is full, add allocates a larger array and copies every element. If you only know the interface, you cannot see that variation and cannot explain why a program that adds millions of elements spends time copying. If you know the internals, you can predict that behavior instead of being surprised by it.