Abstraction is an essential element of both software and hardware design. Computers with billions of transistors and software with millions of code statements are much too large for any one person to hold in their head, so it is navigated by the use of wide-spread, hierarchical way.
Abstraction can be achieved by detail removal; in this sense, a silhouette is more abstract than a portrait and a stick figure more abstract than a silhouette.
In programming, detail removal is primarily achieved by naming things: instead of listing the steps needed to compute a the square root of we just say sqrt(x), defining elsewhere the detailed steps involved in the sqrt set of actions.
Abstraction can also be achieved by generalization; in this sense, a human
is more abstract than Ada Lovelace
and a legal entity
is more abstract than a human.
In programming, generalization is primarily achieved by parameterization: instead of naming square root sqrt and cube root cbrt and so on we make one function pow(x,y) which computes for any and , including square root () and cube root ().
A common way to use abstraction in computing is to tame complexity, keeping people from needing to think about most details most of the time. An abstraction where you can operate effectively only thinking about the abstraction and not thinking about the details behind it is called an abstraction barrier.
Cars provide the driver with an abstraction barrier: drivers can think about a pedal that speeds up, a pedal that slows down, and a wheel that steers without thinking about engines and fuel mixtures and differentials and so on.
Processors provide an abstraction barrier, allowing programmers to think about numbers instead of bits.
Well-designed functions provide an abstraction barrier in code, allowing programmers to think about the function as an entity instead of the individual steps used inside it to compute a value.
Most abstraction barriers are imperfect: there are situations where the abstraction fails and thinking about what’s behind them becomes important.
A car’s abstraction barrier fails when (for example) driving through pools of water, where the air-breathing nature of the engine fails.
A processor’s abstraction barrier files when (for example) integers overflow their available space or floating-point numbers fail to represent 0.1 exactly.
Computing has become the versatile and powerful tool it is by allowing new abstractions to be readily created using existing abstraction as their building blocks.
When one system uses abstractions to hide that it was implemented with another system, we call the one more-abstracted system higher-level and the less-abstracted system lower-level. This type of level is a fundamental property of many parts of computing and is often mention by computing professionals as a way of helping guide conversations and collaboration on a task.
The following are listed with the highest-level first and lowest-level last. Each item in the list uses abstraction to conceal the details of the level below it.