arithmetic exponentiation operator proposal

[bindsdev]

I propose that syntax and functionality for exponentiation is implemented. Currently, if you want to raise some number x to a power of 2, you must do x * x. This is fine if you only want to raise x to itself once. But, if there is ever a case where you want to raise x to a power greater than 2, it could get ugly quickly. The operator would be like Python's exponentiation operator, which is **. Operator overloading for this could be implemented using Raise and RaiseWith(U) or Exponent and ExponentWith(U) interfaces.

With this operator, raising x to a power of 3 could look like x ** 3 instead of x * x * x.

[nigeltao]

Be careful, though, that x**y isn't ambiguous with the existing x * *y syntax where the first star is multiplication (a binary operator) and the second star is a pointer dereference (a unary operator).

Related: the deref p syntax idea from #1454 (which is admittedly different from C++ syntax).

[bindsdev]

Be careful, though, that x**y isn't ambiguous with the existing x * *y syntax where the first star is multiplication (a binary operator) and the second star is a pointer dereference (a unary operator).

Related: the deref p syntax idea from #1454 (which is admittedly different from C++ syntax).

Indeed. With this concern, the operator could be potentially modified to ^+. I do not see any ambiguity with this as + itself is not an existing unary operator from my knowledge. My first thought was ^* but x ^ *y (XOR and dereference) could become ambiguous with x^*y.

[OlaFosheimGrostad]

x^^n

[tkoeppe]

Does this really need to be an operator, rather than, say, a standard library function? Is this operation common enough to warrant a dedicated syntax (and requiring everyone to learn that syntax)?

[bindsdev]

Does this really need to be an operator, rather than, say, a standard library function? Is this operation common enough to warrant a dedicated syntax (and requiring everyone to learn that syntax)?

This is a fair question. But personally, I would much rather have an operator than a function call, mainly to not have to append .pow or .raise to every number I want to exponentiate. If unnecessary syntax is a concern though, I suppose a function would be fine

[chandlerc]

FYI, moving from an issue to a discussion.

[YagaoDirac]

This should be done with a math std library. It shouldn't be done as language feature. Carbon as a container language shares the same philosophy with c++.
But also, it can be done like:
{scope.enable_rich_arith_operator(true);
var x:int = 42;
x**2;//OK
}