R/remove_all_NA_rowcols.R

In mjdufort/miscHelpers: Miscellaneous helper functions

#' Trim any rows and/or columns containing all NAs from a matrix or data frame.
#'
#' Trim any rows and/or columns containing all NAs from a matrix or data frame. Any rows or columns
#' composed entirely of NAs are removed.This is used primarily after reading in a spreadsheet, when excess
#' columns and/or rows may be included due to quirks in the file reading process. It can also remove cases
#' and/or variables with no data.
#' @param x data frame or matrix with rows and/or columns to be removed
#' @param cols,rows logical, whether to trim columns and rows, respectively. By default, both columns and rows are trimmed.
#' @param interpret_text whether to convert text "NA" to NA values. Default behavior is to set values that are read as "NA" to NA before filtering.
#' @export
#' @return an object of the same type as \code{x}, potentially with fewer rows and/or columns
#' @usage 
#' remove_all_NA_rowcols(
#'   x, cols = TRUE, rows = TRUE, interpret_text = TRUE)
remove_all_NA_rowcols <- function(x, cols = TRUE, rows = TRUE, interpret_text = TRUE) {

  if (!is.data.frame(x) & !is.matrix(x))
    stop("Input object type not recognized. This function is for use on data frames and matrices.")

  if (nrow(x) == 0 | ncol(x) == 0) {
    cat("Input object has 0 rows or 0 columns; returning as is\n")
    return(x)
  }

  # this step has been repeatedly modified to handle tibbles and data frames
  if (interpret_text)
    for (i in 1:ncol(x))
      for (i in 1:ncol(x)) x[[i]][x[[i]] %in% "NA"] <- NA

  if (cols) {
    keep_cols <-
      apply(x, MARGIN = 2, function(x) {sum(!is.na(x))}) > 0
    colnames.orig <- colnames(x)[keep_cols] # necessary to avoid deduplicating column names
    x <- x[,keep_cols]
    colnames(x) <- colnames.orig # necessary to avoid deduplicating column names
  }

  if (rows) {
    keep_rows <-
      apply(x, MARGIN = 1, function(x) {sum(!is.na(x))}) > 0
    x <- x[keep_rows,]
  }

  x
}