The problem

Often I need to suffix large numbers with k, M, G … e.g. for printing number of reads or bases in sequencing runs. While there are options in many languages for this, I needed a general soluiton which can be used in all contexts - shell, python, R, shiny apps, nextflow etc.

The solution

Use the korn shell!

It turned out that this shell has a variation of printf which has inbuilt abilities to do exactly this, using the %#d as format specification. Interestingly, other shells (bash and zsh tested) do not have this printf flavour. An example:

# switch to ksh
$ printf "%#d\n" 1234567
1.2M

# and even this
printf "%#d\n" 1234 1234567 1.23e9
1.2k
1.2M
1.2G

Might sound a bit strange at first, but it is a general solution which can easily be used in any context. For example, here is how I use it in R:

• write a small ksh script and put it in your path, I’ve called mine si_format.sh

#!/usr/bin/env ksh

printf "%#d\n" "$@" 

• in R, I use it like this

# system call to the ksh script
system2("bin/si_format.sh", c(12345, 12345678), stdout = T)
[1] "12k" "12M"

# pack it in a function
si_fmt <- function(x) { system2("bin/si_format.sh", x, stdout = TRUE) }

runif(10, 1e3, 1e7) %>% si_fmt()
 [1] "7.3M" "7.4M" "8.0M" "6.6M" "1.4M" "7.0M" "874k" "4.2M" "9.3M" "9.5M"

# use in tidy workflows
df <- data.frame(
     a = runif(10, 1e3, 1e6), 
     b = runif(10, 1e6, 1e8)
     )
df
         a        b
1 222739.1  1691506
2 492221.2 41578973
3 232525.8 66668424
4 871410.3 21227623
5 142999.3 91327876

df %>% 
mutate(a = si_fmt(a), b = si_fmt(b))
     a    b
1 223k 1.7M
2 492k  42M
3 233k  67M
4 871k  21M
5 143k  91M

Also, ksh is already available on most unix systems by default, so no need to use external stuff.

Enjoy!