R/melt.R

In xavier-gilbert/sabre: Spruce up Addresse and Business REcords

#' Separate a collapsed column record into multiple rows.
#'
#' Wrapper around `tidyr::separate_rows()`. Performs the operation for a given
#' list of dividers (delimiters).
#' Iterates through the set of possible dividers.
#'
#' If a feature/variable in a record contains multiple delimited values, this
#' separates the values, placing each on in its own row (creating a new record).
#' The transformation is similar to wide to long, but only (feature, record)
#' having delimited values will create new rows.
#'
#' @param df A `DataFrame`.
#' @param column The name of the column/feature to perfom the separation on.
#' @param dividers A vector of delimiters to
#' @param trim_ws A boolean; determines if trailing white spaces (potentially
#'   generated by the split), are trimmed or not.
#'
#' @return A `DataFrame` with the additional rows (as many as performed splits).
#' @export
#'
#' @examples
#' melt_rows(
#'   businesses[4:8, 3:4],
#'   "trading_name",
#'   dividers = c("\\|", " trading as ", "t/a", "\\/")
#'   )
#'
#' @importFrom stringr str_c str_squish
#' @importFrom tidyr separate_rows
#' @importFrom dplyr distinct mutate_at filter
#' @importFrom purrr when
#' @importFrom rlang sym
melt_rows <- function(df, column, dividers = c("\\|"), trim_ws = TRUE){
  dividers_ <- str_c(dividers, collapse = "|")
  # quo_column <- sym(column)

  df %>%
    separate_rows(column, sep = dividers_) %>%
    when(trim_ws ~ mutate_at(., column, str_trim),
         ~.
    ) %>%
    # FIXME will also filter out pre-existing NA in the column
    # filter((!! quo_column) != "") %>%
    distinct()
}