Good K/M/G abbreviation of y scale labels in ggplot2

Question

Question

How can we easily have Kilo/Mega/Giga labels without "bits" or "Bytes" unit symbols?

Example

data.frame(x = LETTERS[1:5], n = c(0, 5000, 10000, 15000, 20000)) %>% 
  ggplot(aes(x, n)) + 
  geom_point() +
  scale_y_continuous(labels = scales::number_bytes_format(units = "si"))

For the y scale I expect the labels 0K, 5K, 10K, 15K, 20K. No Kb!

Bonus question

Is there any available solution to get 0, 1K, 1M, 1G labels? I.e., the most appropriate abbreviation of the values?

Answer 1

The scales package added a helper function that can provide SI prefixes and optionally add a unit. From what I can tell, this was introduced in #339 (March 2022) and is documented here (see the scale_cut parameter).

Using the original example:

library(ggplot2)
library(scales)
data.frame(x = LETTERS[1:5], n = c(0, 5000, 10000, 15000, 20000)) |> 
    ggplot(aes(x, n)) + 
    geom_point() +
    scale_y_continuous(labels = label_number(scale_cut = cut_si("")))

By changing the empty quotes inside cut_si(), you can append a unit such as "m" or "g".

Example from the documenation:

demo_log10(
  c(1e-9, 1),
  breaks = log_breaks(10),
  labels = label_number(scale_cut = cut_si("g"))
)

Answer 2

gdata::humanReadable doesn't work with negative numbers, it adds the suffix B to numbers below 1000, and it adds one digit after the decimal point to all numbers, and it uses 1024-byte kilobytes if you specify standard="Unix" instead of standard="SI":

> gdata::humanReadable(-10,standard="Unix",sep="")
Error in gdata::humanReadable(-10, standard = "Unix", sep = "") :
  'x' must be positive
> gdata::humanReadable(2500,standard="Unix",sep="")
[1] "2.4K"
> gdata::humanReadable(2500,standard="SI",sep="")
[1] "2.5kB"
> gdata::humanReadable(10,standard="SI",sep="")
[1] "10.0B"

You can use scale_y_continuous(labels=kimi) with this function:

kimi=\(x){
  e=floor(log10(abs(x)))
  e2=pmax(e,0)%/%3+1
  suf=c("","k","M","B","T")
  x[]=ifelse(abs(x)<1,x,paste0(round(x/1e3^(e2-1),ifelse(e%%3==0,1,0)),suf[e2]))
  x
}

kimi(c(-2.3e9,-12345,-12,0,50,200,2235,12345,126583,125467080))

 [1] "-2B"    "-12.3k" "-12"    "0"      "50"     "200"    "2k"     "12.3k"
 [9] "127k"   "125M"

Answer 3

Try gdata::humanReadable:

library(ggplot2)
library(gdata)

myDat <- data.frame(x = LETTERS[1:5], n = c(0, 5000, 10000, 15000, 20000))

ggplot(myDat, aes(x, n)) + 
  geom_point() +
  scale_y_continuous(breaks = myDat$n, 
                     labels = humanReadable(myDat$n, standard = "Unix", sep = ""))

---

Edit:

We could customise the function:

humanReadableCustom <- function (x, units = "auto", standard = c("IEC", "SI", "Unix"), 
                                 digits = 1, width = NULL, sep = " ", justify = c("right", 
                                                                                  "left")) 
{
  #suffix.SI <- c("B", "kB", "MB", "GB", "TB", "PB", "EB", "ZB", "YB")
  # custom
  suffix.SI <- c("", "K", "M", "G", "T", "P", "E", "Z", "Y")

  suffix.IEC <- c("B", "KiB", "MiB", "GiB", "TiB", "PiB", "EiB", "ZiB", "YiB")
  suffix.Unix <- c("B", "K", "M", "G", "T", "P", "E", "Z", "Y")
  standard <- match.arg(standard)
  if (length(justify) == 1) 
    justify <- c(justify, justify)
  .applyHuman <- function(x, base, suffix, digits, width, 
                          sep) {
    n <- length(suffix)
    i <- pmax(pmin(floor(log(x, base)), n - 1), 0)
    if (!is.finite(i)) 
      i <- 0
    x <- x/base^i
    if (is.null(width)) 
      x <- format(round(x = x, digits = digits), nsmall = digits)
    else {
      lenX <- nchar(x)
      if (lenX > width) {
        digits <- pmax(width - nchar(round(x)) - 1, 
                       0)
      }
      if (i == 0) 
        digits <- 0
      x <- round(x, digits = digits)
    }
    c(x, suffix[i + 1])
  }
  if (any(x < 0)) 
    stop("'x' must be positive")
  if (standard == "SI") {
    suffix <- suffix.SI
    base <- 10^3
  }
  else if (standard == "IEC") {
    suffix <- suffix.IEC
    base <- 2^10
  }
  else {
    suffix <- suffix.Unix
    base <- 2^10
  }
  if (!missing(units) && units == "bytes") {
    retval <- rbind(x, "bytes")
  }
  else if (!missing(units) && units != "auto") {
    units <- suffix[match(toupper(units), toupper(suffix))]
    power <- match(units, suffix) - 1
    X <- x/(base^power)
    X <- format.default(x = X, digits = digits, nsmall = digits)
    retval <- rbind(X, rep(units, length(X)))
  }
  else retval <- sapply(X = x, FUN = ".applyHuman", base = base, 
                        suffix = suffix, digits = digits, width = width, sep = sep)
  if (all(justify == "none")) 
    paste(trim(retval[1, ]), trim(retval[2, ]), sep = sep)
  else paste(format(trim(retval[1, ]), justify = justify[1]), 
             format(trim(retval[2, ]), justify = justify[2]), sep = sep)
}

Then plot

library(ggplot2)
library(gdata)

myDat <- data.frame(x = LETTERS[1:5], n = c(0, 5000, 10000, 15000, 20000))

ggplot(myDat, aes(x, n)) + 
  geom_point() +
  scale_y_continuous(breaks = myDat$n, 
                     labels = humanReadableCustom(myDat$n,
                                                  standard = "SI", sep = ""))