Allowing all binary C++ operators in WebIDL

[jrouwe]

This expands the list of supported operators by the 'Operator' WebIDL attribute to all binary operators that are supported by C++.

Fixes #9415

[sbc100]

Seems reasonable to me. @kripken WDYT?

[kripken]

Hmm, I think the reason we focused on the += form is because the C++ x += y turns into JS x.add(y). Both of these operations are asymmetrical, with x playing a more central role. OTOH, with this change, we'd have C++ x * y turn into JS x.mul(y), so something symmetrical turns into something asymmetrical. So there might be some risk of confusion there.

But maybe that risk is very small? Anyhow, if there is real-world use case I'm not opposed to it. Out of curiosity @jrouwe can you explain a bit about the use case you have in mind for this? I guess the library has only * and not *=?

[jrouwe]

The use case is for a quaternion or matrix class.

These classes look like:

class Vector { ... };

class Matrix {
public:
  Vector operator *(const Vector &) { ... };
};

In this example the operation Matrix * Vector => Vector cannot be expressed as an operator *= because the return type is not a Matrix but a Vector.

Obviously I could define a function like MultiplyVector in C++ instead of using operator *, but I want my library (JoltPhysics) to have an intuitive interface in C++:

Matrix matrix = ...;
Vector vector = ...;
Vector result = matrix * vector;

B.t.w. the real class lives here:
https://github.com/jrouwe/JoltPhysics/blob/9fb538ef9b945fe4d1ccab18bf979715fd90a312/Jolt/Math/Quat.h#L160

The IDL lives here:
https://github.com/jrouwe/JoltPhysics.js/blob/fa0df81fb3cffb4946a9e5740ecd68ced00f1fb7/JoltJS.idl#L382-L384

[kripken]

Thanks for the info @jrouwe , yeah, that makes sense. I agree we should support this.

[kripken]

lgtm with an update to the docs:

https://emscripten.org/docs/porting/connecting_cpp_and_javascript/WebIDL-Binder.html#operators

File is at site/source/docs/porting/connecting_cpp_and_javascript/WebIDL-Binder.rst.

[jrouwe]

Documentation updated.

[kripken]

Can we error somehow if this is not one of the allowed operators? For example if it is a unary operator. Before this PR we'd error on unsupported things, and it would be good to keep doing so.

[jrouwe]

I've made the list explicit (based on https://en.cppreference.com/w/cpp/language/operators). Note that previously operators would be allowed as long as they contained a '=', so 'x=' would have been a valid operator (but doesn't compile obviously). Those will be caught now as well.

[kripken]

Suggested change

	if any(x == operator for x in ["+", "-", "*", "/", "%", "^", "&", "|", "=",
	"<", ">", "+=", "-=", "*=", "/=", "%=", "^=", "&=", "|=", "<<", ">>", ">>=",
	"<<=", "==", "!=", "<=", ">=", "<=>", "&&", "||"]):
	if operator in ["+", "-", "*", "/", "%", "^", "&", "|", "=",
	"<", ">", "+=", "-=", "*=", "/=", "%=", "^=", "&=", "|=", "<<", ">>", ">>=",
	"<<=", "==", "!=", "<=", ">=", "<=>", "&&", "||"]:

[kripken]

Unless I'm missing something?

[jrouwe]

Sorry, I don't use python very often... I've merged your suggestion.

[kripken]

No worries, looks good!