regmatches（而不是strsplit）怎么样？

> n <- 3

> regmatches(s, gregexpr(sprintf("(\\W?\\w){1,%i}", n), s))
[[1]]
[1] "QW%E"  "RT%Z"  "U%I%O" "%P"

或tapply + strsplit

v <- unlist(strsplit(s, ""))
l <- which(grepl("\\w", v))
tapply(
    v,
    cumsum(seq_along(v) %in% (1 + by(l, ceiling(seq_along(l) / n), max))),
    paste0,
    collapse = ""
)

它给出了

0       1       2       3
 "QW%E"  "RT%Z" "U%I%O"    "%P"

比上面的简洁得多，但基本解决方案都是一样的：

# Function to only consider certain characters define by a regex
# and split a string scalar into seperate elements in a vector
split_string_to_vec <- function(s, n, consider_elements_pattern = "[[:alpha:]]"){
  # Ensure s is a character scalar:
  stopifnot(is.character(s) && length(s) == 1)
  # Ensure n is an integer scalar: 
  stopifnot(is.numeric(n) && length(n) == 1)
  # Split the string into separate elements:
  # str_vec => character vector
  str_vec <- unlist(strsplit(s, ""))
  # Assign an index to the string vector: 
  # idx => named integer vector
  idx <- setNames(seq_len(length(str_vec)), str_vec)
  # Resolve which values are to be considered (only alpha numerics):
  # considered_vals => named integer vector
  considered_vals <- idx[grepl(consider_elements_pattern, names(idx))]
  # Split the string vector into a list: 
  # grpd_strings => list of character vectors
  grpd_strings <- split(
    considered_vals,
    ceiling(seq_along(considered_vals) / n)
  )
  # For each string group, resolve the group with the 
  # appropriate characters in order: res_vec => character vector
  res_vec <- vapply(
    seq_along(grpd_strings),
    function(i){
      # Get current list element: 
      curr <- grpd_strings[[i]]
      # If its the first element: 
      if(i == 1){
        # Ignore previous element only focus on this 
        # one: ir => named integer vector
        ir <- sort(c(curr, idx[min(curr):max(curr)]))
      # Otherwise:
      }else{
        # Resolve the previous element: 
        prev <- grpd_strings[[(i-1)]]
        # ir => named integer vector
        ir <- sort(c(curr, idx[(max(prev)+1):max(curr)]))
      }
      # Flatten result into a unique (by idx) string: 
      # character scalar => env
      paste0(
        names(
          subset(
            ir,
            !(duplicated(ir))
          )
        ),
        collapse = ""
      )
    },
    character(1)
  )
  # Explicitly define the returned object:
  # character vector => env
  return(res_vec) 
}
# Input Data:
# s => string scalar
s <- "QW%ERT%ZU%I%O%P"
# n => integer scalar
n <- 3
# Apply the function: string scalar => stdout(console)
split_string_to_vec(s, n, consider_elements_pattern = "[[:alpha:]]")