nLab imperative programming

Idea

In computer science, under imperative programming one understands a programming language-style in which a program largely consists of

1. a succession of commands (whence “imperative”)

2. for (in particular) accessing and modifying the global state of the computing environment

   (such as the RAM and I/O-devices).

A typical example of imperative pseudocode would be of the form

n <- Read

n++

Write n

being the sequence of commands to, in this case:

1. read a (numerical) datum (from some input device) and store it in a global variable n

2. increment n

3. write the value of n (to some output device).

Low-level languages are often imperative, reflecting the actual operation of the underlying computing machine; and medium-level imperative languages may be felt to have a transparent flow logic.

In its naive form the access and modification of global state variables means that imperative programs are hard to verify (since there is no guarantee that the global state is in the assumed form). This is generally in contrast to pure functional programming languages, which enforce that every procedure be a pure function in that its output and side-effect depends deterministically on its input (instead of also on the value of some global variables).

On the other hand, much of the features of imperative programs with access to a global state can be emulated/encapsulated in pure functional programming languages via the scheme of monadic data types (see there for more). With the syntactic sugar of do-notation [Launchbury 1993 §3.3] this allows purely functional languages to look essentially like imperative code while remaining verifiable.

Related concepts

• functional programming

• object-oriented programming

• monad (in computer science)

References

See also:

• Wikipedia, Imperative programming

The origin of do-notation for emulating imperative programming in functional programming languages via monadic data types:

• John Launchbury, §3.3 in: Lazy imperative programming, Proceedings of ACM Sigplan Workshop on State in Programming Languages, Copenhagen (1993) [pdf, pdf]

Conversely, building features of functional programming languages into imperative languages:

• David R. Cok, Reasoning about functional programming in Java and C++, ISSTA ‘18: Companion Proceedings for the ISSTA/ECOOP 2018 WorkshopsJuly 2018 Pages 37–39 (doi:10.1145/3236454.3236483)