R/xgx_breaks_log10.R

In Novartis/xgxr: Exploratory Graphics for Pharmacometrics

#' Sets the default breaks for log10
#' 
#' \code{xgx_breaks_log10} sets nice breaks for log10 scale.
#' it's better than the default function because it ensures there is at least 
#' 2 breaks
#' and also, it will try to go by 3s (i.e. 1,3,10,30,100) if it makes sense
#'
#' for the extended breaks function, weights is a set of 4 weights for
#' \enumerate{
#' \item simplicity - how early in the Q order are you
#' \item coverage - labelings that don't extend outside the data: 
#' range(data) / range(labels)
#' \item density (previously granularity) - how close to the number of 
#' ticks do you get (default is 5)
#' \item legibility - has to do with fontsize and formatting to prevent 
#' label overlap
#' }
#' 
#' @references Talbot, Justin, Sharon Lin, and Pat Hanrahan. 
#' "An extension of Wilkinson’s 
#' algorithm for positioning tick labels on axes." IEEE Transactions 
#' on visualization and computer graphics 16.6 (2010): 1036-1043.
#' 
#' @param data_range range of the data
#' 
#' @return numeric vector of breaks
#' 
#' @examples 
#' xgx_breaks_log10(c(1, 1000))
#' xgx_breaks_log10(c(0.001, 100))
#' xgx_breaks_log10(c(1e-4, 1e4))
#' xgx_breaks_log10(c(1e-9, 1e9))
#' xgx_breaks_log10(c(1, 2))
#' xgx_breaks_log10(c(1, 5))
#' xgx_breaks_log10(c(1, 10))
#' xgx_breaks_log10(c(1, 100))
#' xgx_breaks_log10(c(1, 1.01))
#' xgx_breaks_log10(c(1, 1.0001))
#' print(xgx_breaks_log10(c(1, 1.000001)), digits = 10)
#' 
#' @importFrom labeling extended
#' @export
xgx_breaks_log10 <-  function(data_range) {
  data_min <- min(log10(data_range))
  data_max <- max(log10(data_range))
  
  if (is.infinite(data_min) & is.infinite(data_max)) {
    breaks_out = data_range
  } else {
    n_breaks <- 5   # number of breaks to aim for
    # preferred breaks, in log10-space
    preferred_increment <- c(1, 0.5)
  
    breaks <- labeling::extended(data_min, data_max, n_breaks, Q = preferred_increment)
    breaks <- 10^breaks
  
    # ensure that there are at least 2 breaks
    # but also try to present "nice" breaks with only one significant digit
    breaks1 <- unique(signif(breaks, 1))
    breaks2 <- unique(signif(breaks, 2))
    breaks3 <- unique(signif(breaks, 3))
    if (length(breaks1) >= 2) {
      breaks_out <- breaks1
    } else if (length(breaks2) >= 2) {
      breaks_out <- breaks2
    } else if (length(breaks3) >= 2) {
      breaks_out <- breaks3
    } else {
      breaks_out <- unique(breaks)
    }
  }
  return(breaks_out)
}