Confused about s3 dispatch in r_safe about callr

[gaborcsardi]

This is not a callr issue. It is an S3 method lookup issue. E.g. this fails the same way:

f <- function(n) {
  point <- function(x, y) {
    stopifnot(is.numeric(x), is.numeric(y))
    structure(list(x = x, y = y), class = "point")
  }

  abs.point <- function(x) {
    sqrt(x$x ^ 2 + x$y ^ 2)
  }

  points <- mapply(point, runif(n), runif(n), SIMPLIFY = FALSE)

  sapply(points, abs)
}
f(10)
#> Error in FUN(X[[i]], ...) : non-numeric argument to mathematical function

I think it is a bug in base R, but R-core devs might not share this opinion. One workaround is to put the S3 method definition into .GlobalEnv (i.e. outside of a function in the Rmd for you).

[nbenn]

Thanks for the quick feedback.

Sorry to insist. Do you also not consider this a bug in my original scenario? There I did not have my s3 method definition wrapped in a function. My knitr code chunk is something like

point <- function(x, y) {
  stopifnot(is.numeric(x), is.numeric(y))
  structure(list(x = x, y = y), class = "point")
}

abs.point <- function(x) {
  sqrt(x$x ^ 2 + x$y ^ 2)
}

points <- mapply(point, runif(5), runif(5), SIMPLIFY = FALSE)

sapply(points, abs)

When I create my .md using rmarkdown::render, all is fine. However pkgdown wraps render with r_safe, which causes the markdown generation to fail. If you agree that something seems off, I will open an issue at pkgdown.

[gaborcsardi]

rmarkdown::render evaluates code in an empty environment by default and this messes up the S3 lookup as well. You can set its envir argument to .GlobalEnv to work around this. This is done here: https://github.com/r-lib/pkgdown/blob/d4e9509f99265a551f03ef2696d2f0393f24103c/R/build-articles.R#L114-L123

But it is not done here:
https://github.com/r-lib/pkgdown/blob/c8726f9242b2b67f572c16422c83cdc8160c7628/R/build-home.R#L72-L79

So maybe the latter is biting you, and then yes, this needs a fix in pkgbuild.

[nbenn]

Thanks for getting back to me.