Infix attr accessor

[hadley]

Maybe purrr would be a good place for:

`%@%` <- function(x, name) attr(x, name)

Need to think about if this should be S3 generic or not, and need to check if this works as is for S4.

[lionel-]

I've always been frustrated that @ does not work as an attribute accessor... Especially since Kevin Ushey mentioned on twitter that S4 slots are actually attributes.

I suppose it would be highly inappropriate to force the matter and export a redefinition?

`@` <- function(x, slot) {
  slot <- as.character(substitute(slot))

  if (isS4(x)) {
    env <- new.env(parent = parent.frame())
    env$`@` <- base::`@`
    eval(call("@", x, slot), envir = env)
  } else {
    attr(x, slot)
  }
}

mtcars@names

[hadley]

Yeah, we can't override existing functions, so I think %@% is the best we can do. It might be worth trying to push upstream to R core though.

[npjc]

would this follow then?

x %@% "foo" <- "bar"    # attr(x,"foo") <- "bar"     // x.attr("foo") = "bar";

OR use %<@% analogous to the %<>% relationship %>%

if so then what of

attributes(x) <- list(mycomment = "really special", foo = 3:2,bar = paste(1:6))

x %<@% list(mycomment = "really special", foo = 3:2,bar = paste(1:6))

then you need an .append = TRUE or such...

[lionel-]

The first one works:

`%@%` <- attr
`%@%<-` <- `attr<-`

mtcars %@% "names"
mtcars %@% "names" <- 1:11

The idea of overloading these operators to work with lists seems interesting.

`%@%` <- function(x, slots) {
  if (length(slots) > 1) {
    attributes(x)[slots]
  } else {
    attr(x, slots)
  }
}

`%@%<-` <- function(x, slots, value) {
  if (length(slots) > 1) {
    attributes(x)[slots] <- value
    x
  } else {
    attr(x, slots) <- value
  }
}

mat <- as.matrix(mtcars)
mat %@% c("dimnames", "dim")
mat %@% c("dimnames", "dim") <- list(NULL, c(2, 16, 11))

But it's not good to have the output type change as a function of the number of slots requested. So we'd have to choose one of those two behaviours. The most useful is the list of attributes accessor probably, and we can still hope that R core will enable single attribute lookup with @ in the future :)

[lionel-]

it's not good to have the output type change as a function of the number of slots requested.

One possibility is to make the behaviour depend on whether the slot is a vector or a list. But the syntax becomes kind of heavy then

mat %@% list("dimnames", "dim") <- list(NULL, c(2, 16, 11))

[hadley]

I'd rather keep it simple and stick with just getting of single values.

[hadley]

Closed in 97ddec7