Say I have a simple object created by from user input:
var input = { url: 'http://example.com/', path: 'abc', user: 'adam' };
And I am to write function(s) that update url and path properties according to some business logic, there are many approaches:
A. functional
input.url = validate(input);
input.path = convert(input);
B. side-effect
update(input);
C. create a new object internally
input = process(input);
For javascript, is it always better to use the functional approach, for easier debug and speed? Or a general thought process on deciding which is more appropriate.
I wouldn’t call the first approach functional. One of the key aspects of the functional paradigm is to avoid mutable state. So any kind of reassignment is not in the functional style. Once you’ve defined input is, you wouldn’t change it. This might be more functional:
const validatedInput = {
url: validate(input.url),
path: convert(input.path),
user: input.user
};
The advantages would be code that’s easier to reason about. You know what input and validatedInput are and that they won’t change after being created. However, this will be slower and use more memory than the other approaches, using plain JavaScript objects at least, as a whole new object is created. It’s possible that the change in performance is negligible in this instance but there are likely to be cases where it’s not.
Personally, I would go with a functional approach wherever possible to ensure clarity and correctness. But in cases where that is likely or proven to be too slow, use some mutation in a very limited scope. For example, change a variable only within a certain function. This will limit the likelihood of subtle bugs caused by interactions between lines of code that are far apart or in different files.