Recover clean, runnable code from a reprex captured in the wild and write it to user's clipboard. The code is also returned invisibly and optionally written to file. Three different functions address various forms of wild-caught reprex.

reprex_invert(input = NULL, outfile = NULL, venue = c("gh", "r",
  "so", "ds"), comment = opt("#>"))

reprex_clean(input = NULL, outfile = NULL, comment = opt("#>"))

reprex_rescue(input = NULL, outfile = NULL,
  prompt = getOption("prompt"), continue = getOption("continue"))



Character. If has length one and lacks a terminating newline, interpreted as the path to a file containing reprex code. Otherwise, assumed to hold reprex code as character vector. If not provided, the clipboard is consulted for input.


Optional basename for output file. When NULL, no file is left behind. If outfile = "foo", expect an output file in current working directory named foo_clean.R. If outfile = NA, expect on output file in a location and with basename derived from input, if a path, or in current working directory with basename derived from tempfile() otherwise.


Character. Must be one of the following (case insensitive):

  • "gh" for GitHub-Flavored Markdown, the default

  • "r" for a runnable R script, with commented output interleaved

  • "rtf" for Rich Text Format (not supported for un-reprexing)

  • "html" for an HTML fragment suitable for inclusion in a larger HTML document (not supported for un-reprexing)

  • "so" for Stack Overflow Markdown. Note: this is just an alias for "gh", since Stack Overflow started to support CommonMark-style fenced code blocks in January 2019.

  • "ds" for Discourse, e.g., Note: this is currently just an alias for "gh".


regular expression that matches commented output lines


character, the prompt at the start of R commands


character, the prompt for continuation lines


Character vector holding just the clean R code, invisibly


  • reprex_invert: Attempts to reverse the effect of reprex(). When venue = "r", this just becomes a wrapper around reprex_clean().

  • reprex_clean: Assumes R code is top-level, possibly interleaved with commented output, e.g., a displayed reprex copied from GitHub or the direct output of reprex(..., venue = "R"). This function removes commented output.

  • reprex_rescue: Assumes R code lines start with a prompt and that printed output is top-level, e.g., what you'd get from copy/paste from the R Console. Removes lines of output and strips prompts from lines holding R commands.


## a rendered reprex can be inverted, at least approximately
tmp_in <- file.path(tempdir(), "roundtrip-input")
x <- reprex({
  #' Some text
  #+ chunk-label-and-options-cannot-be-recovered, message = TRUE
  (x <- 1:4)
  #' More text
  y <- 2:5
  x + y
}, show = FALSE, advertise = FALSE, outfile = tmp_in)
tmp_out <- file.path(tempdir(), "roundtrip-output")
x <- reprex_invert(x, outfile = tmp_out)

# clean up
  list.files(dirname(tmp_in), pattern = "roundtrip", full.names = TRUE)
# }# NOT RUN {
## a displayed reprex can be cleaned of commented output
tmp <- file.path(tempdir(), "commented-code")
x <- c(
  "## a regular comment, which is retained",
  "(x <- 1:4)",
  "#> [1] 1 2 3 4",
  "#> [1] 2.5"
out <- reprex_clean(x, outfile = tmp)

# clean up
  list.files(dirname(tmp), pattern = "commented-code", full.names = TRUE)

## round trip with reprex(..., venue = "R")
code_in <- c("x <- rnorm(2)", "min(x)")
res <- reprex(input = code_in, venue = "R", advertise = FALSE)
(code_out <- reprex_clean(res))
identical(code_in, code_out)
# }# NOT RUN {
## rescue a reprex that was copied from a live R session
tmp <- file.path(tempdir(), "live-transcript")
x <- c(
  "> ## a regular comment, which is retained",
  "> (x <- 1:4)",
  "[1] 1 2 3 4",
  "> median(x)",
  "[1] 2.5"
out <- reprex_rescue(x, outfile = tmp)

# clean up
  list.files(dirname(tmp),pattern = "live-transcript", full.names = TRUE)
# }