S7 and a new := named-bind operator

@lawremi

The dev version of the S7 package introduces a new := operator:

`:=` <- function(lhs, rhs) {
  cl <- sys.call()
  if (!is.symbol(cl[[2L]])) {
    stop2("Left-hand side of `:=` must be a symbol.")
  }
  if (!is.call(cl[[3L]])) {
    stop2("Right-hand side of `:=` must be a function call.")
  }
  if ("name" %in% names(cl[[3L]])) {
    stop2("Right-hand side of `:=` must not already supply a `name` argument.")
  }

  cl[[1L]] <- quote(`<-`)
  cl[[3L]]$name <- as.character(cl[[2L]])

  # Can't use eval because it introduces a new frame
  invisible(.Call(S7_eval_bare_, cl, parent.frame()))
}

This is designed to support creating new classes and generics (and various more exotic objects) where you want to give an object a name and bind it to that name.

# with <-
foo <- new_class("foo")
bar <- new_generic("bar")

# with :=
foo := new_class()
bar := new_generic()

This feels like a useful and general tool and @lawremi thinks there's some chance it will make it to base R.

The downside is that it conflicts with the existing definitions of := in data.table and rlang which both error to suggest to the user they've done something wrong. (It doesn't actually affect the operation of these packages since both data.table and rlang override the meaning of := via NSE.)

We don't have any solutions to this problem yet (although we're experimenting with a .onLoad hook approach for rlang that we can share when it's done), but I wanted to give y'all a heads up that we're seriously exploring it and give you an opportunity to give feedback to the S7 team.

(cc @t-kalinowski)

there's some chance it will make it to base R

Do you mean adding `:=` as defined above to the base namespace, or to 'methods', or to a new base package? Right now the operator seems limited to functions that accept a 'name' argument. It seems it could easily be a custom %...% operator, similar to what 'magrittr' exports. In theory, data.table could drop its `:=` export, since no reasonable package needs to import and use it, although the error message is certainly more convenient than "function not found" or (worse) performing an unintended assignment in some environment. Related: #7453 (our %notin% should match the behaviour of the new operator in R >= 4.6, but some packages do import it for re-exporting)

It will be added to S7 in the near release, with the hope (pending approval by R core) of being added to base at some point in the future.

Or maybe a new package that just contains this operator.

Navigation Menu

Repository navigation

S7 and a new `:=` named-bind operator #7801

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

S7 and a new := named-bind operator #7801

Description

Activity

aitap commented on Jun 24, 2026

hadley commented on Jun 24, 2026

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions

S7 and a new `:=` named-bind operator #7801