DEV Community

Magne for This is Learning

Posted on

Side-effects vs. side-causes

The concept of "side-causes" doesn't seem to be a very established term yet, but I think it should be, for conceptual clarity.

The current discussion in functional programming revolves very much around side-effects (and how to avoid them). I've noticed a tendency to label every undesired influence or reliance on externals as "side-effects". Which doesn't really make sense, if you think about it in terms related to our physical world (from where the concept of an 'effect' assumedly was derived, originally). In the physical world we have both causes and effects, which are clearly distinguished (albeit sadly often confused or reversed). So we'd do well to distinguish these concepts in programming as well.

The colloquial definition seems to be that side-causes are hidden inputs, and side-effects are hidden outputs.

But, with some help of our AI language model / search engine of choice (, to word it more precisely and list useful examples, we can say the following.

Side-effects and side-causes are related but distinct concepts in programming:

Side Effects

A side effect is a change in system state or observable behavior that occurs as a result of executing a function or expression, beyond just returning a value. Some examples of side effects include:

  • Modifying a variable or data structure outside the function's local scope
  • Writing to a file, network, or database
  • Triggering an exception or error
  • Calling other functions that have side effects
  • Outputting data to the console or UI

Side effects make a program's behavior dependent on the order of execution and previous state, which can make code harder to reason about, test, and maintain.[1]

Side Causes

A side cause refers to an implicit input or dependency that a function relies on, beyond its explicit arguments. Some examples include:

  • Reading from global variables or module-level state
  • Getting the current time or generating random numbers
  • Reading configuration from the environment or system properties
  • Calling impure functions that depend on external state

Side causes violate referential transparency, meaning you cannot simply replace a function call with its result because the result may be different each time due to implicit inputs. This makes code harder to understand, test in isolation, optimize, and run in parallel.

In summary, [explicit (my note)] side effects are about outputs/changes beyond the return value, while side causes are about implicit inputs beyond the arguments. Both can make code harder to reason about, especially in combination. Functional programming aims to minimize or isolate side effects and side causes for better modularity and determinism. [1][3]

In addition here, we can make a further distinction between two variants of side-effects. Importantly, side-causes often result in implicit side-effects. Because:

  • if a function has a side-cause,
  • and the side-cause changes from an invocation of the function to another,
  • and the function uses it in computing its result,

then, even if the function does not affect the outside world apart from returning its result, the result is affected. This can be seen from the outside of the function (by anything that relies on its result). So even though the function didn't directly affect the outside world, through a write operation or other kind of output occurring during the function execution, it still introduced some side-effect into it's own result. This is another way to view side-effects. Side-effects can be explicit (hidden extra outputs), but also implicit (silently modified results).

I think the reason that side-causes are often erroneously called side-effects is because side-causes generally introduce (implicit) side-effects, if the side-cause affects the function's result in any manner. (If not, why is the function even relying on that side-cause in the first place?)

I hope this helped give some conceptual clarity, and can be a post to link to for distinguishing these two terms specifically.

PS: Funnily enough, it seems the AI derived its understanding from the following citations, where the only citation it used that mentioned side-causes was my own brief comment on the Software Engineering Stack Exchange site:

"read from any input stream" is strictly speaking a side-cause, not a side-effect. Writing to an output stream would be a side-effect. See:

Citations used by the AI:

The top Google results I've found previously mentioning "side-causes" are:

Do you know of any earlier reference to "side-causes"?

Top comments (0)