dput() without .internal.selfref in data tables

Question

Someone recently shared their data table with me via dput() and an error popped up that I had not previously encountered:

Error: unexpected '<' in: " class = c("data.table", "data.frame"), .internal.selfref = <"

After some digging I found this is related specifically to data.tables and as advised in these answers, removing the internal.selfref = <pointer: 0x7fd60e036ce0>) did the trick to successfully assign their data.

However, I anticipate sharing these types of data frequently between novice users; I have not found a reasonable/sustainable solution to prevent this from exporting with the dput, only ad-hoc functions and/or removing it after receiving running it.

If I remove showAttributes from the control = c("keepNA", "keepInteger", "niceNames", "showAttributes") in dput the .internal.selfref is gone but so is everything else about the structure.

The questions and answers provided in the above linked questions were 5-9 years old; I was hoping that some improved functionality may be available (that I am obviously unaware of) that would tell dput to ignore this, or perhaps if there is something I can do to the data table itself before dput that would remove the .internal.selfref altogether.

Is there a way to provide the dput of a data.table object without producing the .internal.selfref?

Thanks in advance.

Example of the issue:

dattab <- data.table::data.table(a = 1:5, b = 6:10)
dput(dattab)

structure(list(a = 1:5, b = 6:10), row.names = c(NA, -5L), 
          class = c("data.table", "data.frame"), 
          .internal.selfref = <pointer: 0x7fd60e036ce0>)

Answer 1

Convert the data.table to a data.frame and then dput it.

dput(as.data.frame(dattab))
## structure(list(a = 1:5, b = 6:10), row.names = c(NA, -5L), class = "data.frame")

Answer 2

Sometimes I think it's useful to share it with the appropriate class, so I use this function:

#' dput into a single line
#'
#' @param x object
#' @param assign logical, whether to get the 'x' object's name and
#'   prepend it to the string
#' @param DT logical, whether to remove (and wrap with 'setDT')
#'   'internal.selfref' in the dput output
#' @return character length 1, invisibly
#' @export
r2dput <- function(x, assign = TRUE, DT = TRUE) {
  out <- deparse1(x)
  if (isTRUE(DT)) {
    out2 <- gsub(",\\s*\\.internal\\.selfref\\s*=\\s*<pointer:\\s*[0-9a-fx]+>", "", out)
    if (out != out2) {
      out <- paste0("data.table::as.data.table(", out2, ")")
    }
  }
  if (assign) {
    out <- paste(c(as.character(substitute(x)), "<-", out), collapse = " ")
  }
  if (R.version$os == "mingw32") {
    writeLines(out, con = "clipboard")
  } else {
    clipr::write_clip(out)
  }
  invisible(out)
}

As an example,

MT <- as.data.table(mtcars[1:3, 11:4])
MT
#     carb  gear    am    vs  qsec    wt  drat    hp
#    <num> <num> <num> <num> <num> <num> <num> <num>
# 1:     4     4     1     0 16.46 2.620  3.90   110
# 2:     4     4     1     0 17.02 2.875  3.90   110
# 3:     1     4     1     1 18.61 2.320  3.85    93

When I want to share that data in a question on SO (or with colleagues), I type in r2dput(MT) (because, you know, I'm r2evans ;-), I just paste (Ctrl-V) into the comment/question/answer and get this:

MT <- data.table::as.data.table(structure(list(carb = c(4, 4, 1), gear = c(4, 4, 4), am = c(1, 1, 1), vs = c(0, 0, 1), qsec = c(16.46, 17.02, 18.61), wt = c(2.62, 2.875, 2.32), drat = c(3.9, 3.9, 3.85), hp = c(110, 110, 93)), row.names = c(NA, -3L), class = c("data.table", "data.frame")))

This removes the pointer that doesn't work, it wraps in as.data.table(.) making it clear what class I'm really working with, and keeps it all on one line (because I dislike the default dput method of line-breaks in formatted code blocks).

(Edited to use clipr on non-windows. I'm not sure it works the same on macos.)