How to perform a `two-step` method dispatch?

[niheaven]

Hey there!

I'm trying to set up a "two-step" method dispatch in S7, I'd like my generic function to automatically pick different versions depending on the classes of the arguments, and each version should accept slightly different parameter names.

For example:

method(fun, list(class_character, class_integer, class_numeric)) <- function(x, int_1, num_1) { ... }
method(fun, list(class_numeric, class_integer, class_character)) <- function(x, int_2, chr_1) { ... }

Then, when I call it, the right version kicks in based on the order of classes—and I'd use the matching parameter names:

fun(x, int_1 = xxx, num_1 = xxx)  # x is character, and uses the first version
fun(x, int_2 = xxx, chr_1 = xxx)  # x is numeric, and uses the second version

Any idea if this is possible? And if so, what's the cleanest way to pull it off?

Thanks!

[hadley]

Can you provide a bit more reasoning about why you want to do this? Without the context, it sounds very complicated.

[niheaven]

I'm currently porting an existing C++ library to R, and I've decided to use S7 classes to mimic C++ classes. For generics and methods, I'd like to keep the same names as in C++, like add() and remove(). But here's the catch: some method names, like add() or remove(), appear across multiple classes, and their implementations differ not just in behavior, but also in parameter names (and, of course, types).

So I'm weighing two options:

1. A two-step dispatch: first dispatch by class, then further dispatch based on the numbers (or names) of arguments.
2. Single-class dispatch, then manually inspect the arguments inside the method body to pick the right logic path.

That's where I'm stuck. Any thoughts?

[lawremi]

It sounds like you would be better off using a message passing, reference-based approach such as R6 or S4 reference classes.

…

On Tue, Jan 27, 2026, 7:02 PM Hsiao-nan Cheung ***@***.***> wrote: *niheaven* left a comment (RConsortium/S7#589) <#589 (comment)> I'm currently porting an existing C++ library to R, and I've decided to use S7 classes to mimic C++ classes. For generics and methods, I'd like to keep the same names as in C++, like add() and remove(). But here's the catch: some method names, like add() or remove(), appear across multiple classes, and their implementations differ not just in behavior, but also in parameter names (and, of course, types). So I'm weighing two options: 1. A two-step dispatch: first dispatch by class, then further dispatch based on the numbers (or names) of arguments. 2. Single-class dispatch, then manually inspect the arguments inside the method body to pick the right logic path. That's where I'm stuck. Any thoughts? — Reply to this email directly, view it on GitHub <#589 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AABGT3XS4SP6NRI4IGSDDG34JARCXAVCNFSM6AAAAACS76CLS2VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZTQMBYGY4DGMBTGQ> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[hadley]

Yeah, sounds like S7 probably isn't the right approach for your problem.