Conditional branches of expressions

[MilesCranmer]

Here's how you can search for conditional branches of expressions (including if statements) using PySR:

PySRRegressor(
    binary_operators=[..., "cond(x::T, y::T) where {T} = x >= T(0) ? y : T(0)"],
    ...
    extra_sympy_mappings={"cond": lambda x, y: sympy.Piecewise((0, x < 0), (y, True))},
)

The ::T where {T} is just Julia notation for templating a function for arbitrary type. It's not normally required, but I want to declare the type of the 0 there as the same type, so I used it.

Using the Piecewise function, SymPy should correctly convert the output of this to an executable function when calling lambdify.

---

A simpler version which is also valid: "cond(x, y) = x >= 0 ? y : 0*y". Though this might be slower due to the type conversions.