Welcome to Software Development on Codidact!
Will you help us build our independent community of developers helping developers? We're small and trying to grow. We welcome questions about all aspects of software development, from design to code to QA and more. Got questions? Got answers? Got code you'd like someone to review? Please join us.
Terms for types of functions with respect to side effects
Mathematically, the purpose of a function is to return an output. However, in a programming context functions often have side effects. It is even common to call functions for the side effects alone. The classic example is print(x), which has a useful side effect and no output.
It seems to me that we can taxonomize functions into five classes based on human expectations about their outputs and side effects:
- Functions with no side effects, where the output is most interesting
- Functions with side effects, but the output is most interesting
- Technically, all functions are like this, because running things on the computer always has side effects like using up CPU and RAM.
- Functions with an output, but the side effect is more interesting than the output
- Example:
request(url, type="POST")- the output is only a detail if we care to check whether the request succeeded, which may even be irrelevant in some cases
- Example:
- Functions with no output, executed for their side effects
- Functions with no output or side effect to speak of (usually created for testing or as placeholders)
Is there a terminology for these classes?
2 answers
You are accessing this answer with a direct link, so it's being shown above all other answers regardless of its score. You can return to the normal view.
I don't think there are any formal names for the various versions you list.
First of all, please note that output in a programming context most often refers to printing something on a screen or to a file, or updating graphics. "Function output" is not a common programming term - almost every language instead terms such function return value or perhaps "result".
That being said, some semi-formal, widely accepted "language-agnostic" terms do exist:
-
Procedure or subroutine often refers to functions which do not return anything. I think this might originate from Pascal and/or Ada. C family languages tend not to use either of these terms but call everything functions.
-
Re-entrant refers to a function which can be safely used in a multi-thread program, because it has no side effects.
-
Thread-safe refers to a function which can be safely used in a mult-thread program. It has side effects, but those are guarded with semaphore/mutex or similar.
-
No-op is often used informally to describe functions which do not perform anything and have no side effects. Likely originating from the common assembler
nopinstruction (no operation, do nothing). -
Member function or method is something that interacts with a given object. It may have side effects, including updating the object data.
Many OO languages also support "read-only" member functions, which are guaranteed not to modify the current object (but may contain other side effects). Immutable objects is another common term used for objects that are never modified after creation - and so a function taking an immutable object as parameter cannot/will not modify that object (but may contain other side effects).
I know that 1 is sometimes called a pure function - although apparently a pure function must also not vary when the input is constant.
By negation, the other kind are called impure functions, although this seems to me biased towards the idea that side effects are bad.
I've seen procedure used for 4, but it seems to be used inconsistently and sometimes gets conflated with 3 and even 2 depending on the language.
5 is sometimes called a trivial or dummy function. Null function or no-op are also used.
0 comment threads
1 comment thread