Can I make a custom operator with function as one of inputs?

[DenisSvirin]

Is it possible to create function like this: function_1(function_2, variable_1), so I can apply function_2 to variable_1 inside function_1?
Preferebly without drastic backend changes

[MilesCranmer]

Not sure I understand, sorry - why can’t you just make a new function equal to the composition?

[MilesCranmer]

In this example, what would F do to cos?

[DenisSvirin]

f(cos, variable_1):
return cos(10 + 4 * variable_1) + cos(30 + 2 * variable_1)

So in general f is like: function_2(10 + 4 * variable_1) + function_2(30 + 2 * variable_1)

[MilesCranmer]

Ah I see what you mean. So f is a function on a function.

Unfortunately this isn’t possible right now; all functions must be defined as f: ℝ × ℝ → ℝ. In the future this will be extended to tensor operations. However, I think functions on functions is really difficult to set up, unfortunately.

[MilesCranmer]

Maybe in the future I’ll make it so arbitrary datatypes are supported, and you could pass cos as a column of your dataset. But if cos is an operator, it will never show up in a leaf of an expression tree by itself without an argument.

[DenisSvirin]

Oh, ok.
Thanks