R/trim.R

In rorynolan/strex: Extra String Manipulation Functions

#' Trim something other than whitespace
#'
#' The `stringi` and `stringr` packages let you trim whitespace, but
#' what if you want to trim something else from either (or both) side(s) of a
#' string? This function lets you select which pattern to trim and from which
#' side(s).
#'
#' @inheritParams str_after_nth
#' @param side Which side do you want to trim from? `"both"` is the
#'   default, but you can also have just either `"left"` or `"right"`
#'   (or optionally the shortened `"b"`, `"l"` and `"r"`).
#'
#' @return A string.
#'
#' @examples
#' str_trim_anything("..abcd.", ".", "left")
#' str_trim_anything("..abcd.", coll("."), "left")
#' str_trim_anything("-ghi--", "-", "both")
#' str_trim_anything("-ghi--", "-")
#' str_trim_anything("-ghi--", "-", "right")
#' str_trim_anything("-ghi--", "--")
#' str_trim_anything("-ghi--", "i-+")
#' @family removers
#'
#' @export
str_trim_anything <- function(string, pattern, side = "both") {
  if (is_l0_char(string)) {
    return(character())
  }
  verify_string_pattern(string, pattern, boundary_allowed = FALSE)
  out <- string
  checkmate::assert_string(side)
  side <- match_arg(side, c("both", "left", "right"), ignore_case = TRUE)
  type <- "regex"
  if (inherits(pattern, "stringr_fixed")) {
    type <- "fixed"
  } else if (inherits(pattern, "stringr_coll")) {
    type <- "coll"
  } else {
    bad_starts <- str_starts(pattern, "\\(*\\^")
    bad_ends <- str_ends(pattern, "\\$\\)*")
    if (any(bad_starts)) {
      rlang::abort(
        c(
          paste(
            "In `str_trim_anything()`, don't start your regular expression",
            "patterns with '^' to match the start of the string.",
            "The trimming by definition is happening at the edges."
          ),
          x = str_glue_data(
            list(
              pattern = pattern,
              first_bad = which.max(bad_starts)
            ),
            "Element {first_bad} of your pattern, ",
            "'{pattern[first_bad]}' is the first offender.",
          )
        )
      )
    } else if (any(bad_ends)) {
      rlang::abort(
        c(
          paste(
            "In `str_trim_anything()`, don't end your regular expression",
            "patterns with '$' to match the end of the string.",
            "The trimming by definition is happening at the edges."
          ),
          x = str_glue_data(
            list(pattern = pattern, first_bad = which.max(bad_ends)),
            "Element {first_bad} of your pattern, '{pattern[first_bad]}' ",
            "is the first offender."
          )
        )
      )
    }
    pattern <- str_c("(", pattern, ")+")
    pattern <- switch(side,
      left = str_c("^", pattern),
      right = str_c(pattern, "$"),
      pattern
    )
  }
  if (side == "both") {
    out <- string %>%
      str_trim_anything(pattern, "left") %>%
      str_trim_anything(pattern, "right")
  } else if (type == "regex") {
    out <- str_replace(string, pattern, "")
  } else if (side == "left") {
    starts <- which(str_starts(out, pattern))
    while (any(starts)) {
      out[starts] <- switch(type,
        fixed = stringi::stri_replace_first_fixed(
          out[starts],
          pattern[ifelse(length(pattern) == 1, 1, starts)],
          ""
        ),
        coll = stringi::stri_replace_first_coll(
          out[starts],
          pattern[ifelse(length(pattern) == 1, 1, starts)],
          ""
        )
      )
      starts <- starts[str_starts(out[starts], pattern)]
    }
  } else if (side == "right") {
    ends <- which(str_ends(out, pattern))
    while (length(ends)) {
      out[ends] <- switch(type,
        fixed = stringi::stri_replace_last_fixed(
          out[ends],
          pattern[ifelse(length(pattern) == 1, 1, ends)],
          ""
        ),
        coll = stringi::stri_replace_last_coll(
          out[ends],
          pattern[ifelse(length(pattern) == 1, 1, ends)],
          ""
        )
      )
      ends <- ends[str_ends(out[ends], pattern)]
    }
  }
  out
}