samizdat-shell-help.bash

#!/bin/bash
###
### my-script — does one thing well
###
### Usage:
###   my-script <input> <output>
###
### Options:
###   <input>   Input file to read.
###   <output>  Output file to write. Use '-' for stdout.
###   -h        Show this message.

help() {
    sed -rn 's/^### ?//;T;p' "$0"
}

if [[ $# == 0 ]] || [[ "$1" == "-h" ]]; then
    help
    exit 1
fi

echo Hello World

This doesn't actually work cross-platform, the -r option does not exist on macOS. You can do the same with awk though, just use:

Install GNU coreutils and gnu-sed on MacOS using brew install coreutils gnu-sed if you want to get closer to Linux.

awk '/^###/' "$0"

awk -F'### ' '/^###/' "$0"

That's cool but I think I'll stick with here documents.

#!/bin/bash

Help=$(cat <<-"HELP"

 my-script — does one thing well

 Usage:
   my-script <input> <output>

 Options:
   <input>   Input file to read.
   <output>  Output file to write. Use '-' for stdout.
   -h        Show this message.

HELP
)

help() {
    echo "$Help"
}

if [[ $# == 0 ]] || [[ "$1" == "-h" ]]; then
    help
    exit 1
fi

echo Hello World

This is a really nifty approach, thanks for sharing. It's fairly obvious in retrospect but it hadn't occurred to me before!

Slight tweak, this awk solution will only print the help text, not the leading ### :

    awk -F'### ' '/^###/ { print $2 }' "$0"

Compacted:
[[ $# == 0 || "$1" == "-h" ]]

This doesn't actually work cross-platform, the -r option does not exist on macOS. You can do the same with awk though, just use:

awk '/^###/' "$0"

Complete script:

#!/bin/bash
###
### my-script — does one thing well
###
### Usage:
###   my-script <input> <output>
###
### Options:
###   <input>   Input file to read.
###   <output>  Output file to write. Use '-' for stdout.
###   -h        Show this message.

help() {
    awk '/^###/' "$0"
}

if [[ $# == 0 ]] || [[ "$1" == "-h" ]]; then
    help
    exit 1
fi

echo Hello World

or you can use gsed, if you don't want to pollute the macos sed namespace

A sh compatible version, with portable sed usage

#!/bin/sh
###
### my-script — does one thing well
###
### Usage:
###   my-script <input> <output>
###
### Options:
###   <input>   Input file to read.
###   <output>  Output file to write. Use '-' for stdout.
###   -h        Show this message.

help() {
    sed -n 's/^### \?//p' "$0"
}

if [ $# -eq 0 ] || [ "$1" = "-h" ]; then
    help
    exit 1
fi

echo Hello World

Complete awk version:

#!/usr/bin/env bash
###
### my-script — does one thing well
###
### Usage:
###   my-script <input> <output>
###
### Options:
###   <input>   Input file to read.
###   <output>  Output file to write. Use '-' for stdout.
###   -h        Show this message.

help() {
    awk -F'### ' '/^###/ { print $2 }' "$0"
}

if [[ $# == 0 || "$1" == "-h" ]]; then
    help
    exit 1
fi

echo Hello World

I personally use macros to add help messages to my zsh functions. It's as easy as:

mdoc-test() {
	# uses the global alias MAGIC
    mdoc Usage: sth MAGIC
    echo no
}
mdoc-test2() {
	# without global alias
    magic mdoc Usage: sth ; mret
    echo no
}

Awesome idea, thanks very much!

Very cool.

That's a good idea. Awesome!

@brucewoodward 🙏

Glad someone proved this is a good strategy to write helper message, I also did it in my commonly-used script:
https://gist.github.com/typebrook/b0d2e7e67aa50298fdf8111ae7466b56#file-gist

what I did is simply print the comment at the top(except the first line to # --):
sed -Ene "/^#/ !q; 1,/^# --/ d; s/^# //p

@kovetskiy
@benhutchins

This doesn't actually work cross-platform, the -r option does not exist on macOS. You can do the same with awk though, just use:

-E works on BSD sed, so you can make it portable with sed -Ene 's/^### ?//;T;p' "$0" both on mac(BSD sed) or Linux(GNU sed).

666

Why even use the extended version?

sed -n 's/^###//p' achieves the same. Just keep the leading ### format to the help only.

sed -n '/^###/ s///p' this one's faster but a tiny bit more verbose.

and sed '/^###/!d' if you don't mind hashes. Or grep "^###"

My version, that doesn't require multiple # (but allows more than one if you want it for stylistic reasons):

#!/bin/bash
# Usage:
#   test_usage.sh ...

print_usage_and_exit() {
  # The sed script below:
  # - Deletes the first line (shebang).
  # - Quits on the first line that doesn't start with #.
  # - Removes the '#+ ?' prefix from the string.
  sed -En '1d; /^[^#]/ q; s/^#+ ?//p;' < "$0" >&2
  exit 1
}