Extract and append multiple values embedded in rows

unembed.r

# Extract and append multiple values embedded in rows
#
# data: data.frame
# col: column name containing embedded values
# sep: regular expression to split column by
#
# df <- data.frame(key = c("a", "a;b", "a;b;c"), val = 1:3)
# unembed(df, "key", ";")

unembed <- function(data, col, sep, ...) {
  
  stopifnot(is.data.frame(data))
  col_i <- which(names(data) == col)
  
  data[[col]] <- as.character(data[[col]])
  pieces <- strsplit(data[[col]], sep, ...)
  ns <- vapply(pieces, length, integer(1))
  
  structure(data.frame(unlist(pieces), 
                       data[rep(seq_along(ns), ns), -col_i]), 
                       names = c(col, names(data)[-col_i]))
}

stringsAsFactors bites. Recommend adding:

structure(data.frame(unlist(pieces),
data[rep(seq_along(ns), ns), -col_i], stringsAsFactors = FALSE),
names = c(col, names(data)[-col_i]))

The only downside to that approach is the columns you might actually want to remain factors would silently be converted to characters. I usually add options(stringsAsFactors = FALSE) to the top of my script to prevent this.

Actually, better idea: I just forced col to be a character vector before splitting. Does that help?

It does help, and the function is great.