Let's say `(new function(input) { return input })` is a valid expression. If it then doesn't matter (in any circumstance) whether you do
let myFunction = (new function(input) { return input })
myFunction(1)
myFunction(2)
(new function(input) { return input })(1)
(new function(input) { return input })(2)
I think that's not a bad example actually, because I assume it feels more natural to people to be able to treat an expression that creates a function in the way above and expecting that it won't change how the program behaves. And this is really what functional programming is about - it is exactly this feeling and guarantee that FP enforces throughout the whole application, not just in some places.
If you are talking about creating a named function (i.e. a class member) that is not an expression, then there isn't really much point in talking about it since we are now talking about definitions, not expressions anymore. Unless you can change/create definitions programmatically - in that case the code that does it programmatically is the one that needs to be considered, not the definition itself.
let myFunction = ...
A language could certainly treat assignments as expressions (some do) and have them return something, such as the value that is assigned to the variable or always null/unit. But we are now talking about a totally different thing.
Turns out there are non-expressions present; yet somehow if just the expressions are referentially transparent, we have FP?
We have to say something about the kinds of non-expression thingys that are required/allowed and what their properties must be.
Assignments can be expressions, but they are not referentially transparent. If an assignment is replaced by its value, it's no longer performing the assignment. Hence, assignments are easily found to be incompatible with FP under the RT definition, which underscores that it is useful.
Why "just"?
> We have to say something about the kinds of non-expression thingys that are required/allowed and what their properties must be.
Well, if there are syntactical constructs that will execute code at runtime, then we can essentially consider them expressions, even though sometimes they are restricted in how they can be used. For example, in Javascript and Java there is an if-syntax (call it if-statement or whatever) that will execute code but it doesn't return anything and you can "assign" the result to a variable. Obviously this breaks referential transparency (or can't do anything meaningful). We can still see it as an expression for the purpose of the definiton.
Other than that, as I said in my parallel post, if you have something like templates (which are not executing code at runtime), then they can do whatever you want, the only thing that matters is what expressions they will produce and if those are RT or not to fullfill the definition of FP.
> If an assignment is replaced by its value, it's no longer performing the assignment
And the funny thing is, it doesn't matter. Your code might stop compiling because you refer to a variable that isn't declared, yeah. And that's about it. But otherwise in a fully RT program it won't change the bebaviour of the program. EXCEPT if it is a reassignment. Which is why reassignments are violating RT and are therefore not a thing in FP - there can be specific exceptions (i.e. shadowing something in a different scope) but those aren't really reassignments, they just sometimes use the same syntax.
It is a syntax level term, muddying execution model is just not correct usage, see https://stackoverflow.com/questions/210835/what-is-referenti...
What you are looking for is simply and plainly “pureness”.
Referential transparency is not a property of programming languages but expressions (or more general, parts of computer programs).
> It means that one can replace a reference to a thing with the thing itself.
That's not the common definition at all. Read the link I posted. In any case, I'm talking about evaluatable expressions, you talk about "references" - however you define that.
Even in your own stackoverflow link, the accepted answer says this:
> the thing that an expression refers to
What’s missing from side-effect freedom or pureness? That is the property that allows replacing an expression with its value.
Not sure what you are talking about now.
> What’s missing from side-effect freedom or pureness? That is the property that allows replacing an expression with its value.
I think you are just unfamiliar with the terminology, that's all.
See Wikpedia again:
> If a pure function is called with arguments that cause no side-effects, the result is constant with respect to that argument list (sometimes called referential transparency or idempotence), i.e., calling the pure function again with the same arguments returns the same result. (This can enable caching optimizations such as memoization.)
(https://en.wikipedia.org/wiki/Functional_programming#Pure_fu...)
There you go. All 3 terms in once sentence. Maybe, again, you disagree with Wikipedia and the definition and that's your good right, but I don't see any point in prolonging the discussion over that.
In other words, this is not even a property of a programming language. It is a property of the code itself. It's just that some languages make it easy or impossible to build a program in that style - or they even enforce it. Haskell however has escape hatches. And Scala is considered an FP language because, unlike Javascript/Lisp/Clojure/ML/Rust/... it really allows and supports(!) to make the whole program referential transparent.