# ksh

OpenBSD ksh (sometimes called opdksh or oksh) is the default shell on
OpenBSD, and is generally my go-to choince on other systems too.  It has
a good ratio of features and simplicity

	if [ "$TERM" = dumb ]; then
		PS1='$ '
		return
	fi

Enable emacs-like command editing regardless of $EDITOR and csh-like
history expansion with !

	set -o emacs
	set -o csh-history

Talking about history, by default ksh won't store any, which is
unfortunate.  I can't live without my C-r working!

	HISTCONTROL=ignoredups:ignorespace
	HISTFILE=$HOME/.history
	HISTSIZE=10000

OpenBSD ksh has a limited support for programmed completions through
static lists.  The completions are provided via an array called
complete_$progname; or complete_$progname_$nth for the nth argument.

Here's the completions for ssh and scp:

	HOST_LIST=$(awk '/Host / {print $2}' ~/.ssh/config | xargs echo)

	set -A complete_ssh -- $HOST_LIST
	set -A complete_scp -- $HOST_LIST

and for kill(1) and pkill(1)

	set -A complete_kill_1 -- -9 -HUP -INFO -KILL -TERM
	set -A complete_pkill_2 -- -SIGHUP -SIGUSR1 -SIGUSR2 -SIGTERM -SIGKILL

and for vmd(8) if available

	if pgrep -fq /usr/sbin/vmd; then
		set -A complete_vmctl_1 -- console load reload start stop \
			reset status send receive
		set -A complete_vmctl -- \
			$(vmctl status | awk '!/NAME/{printf "%s ", $NF}')
	fi

and for ifconfig(8)

	set -A complete_ifconfig_1 -- $(ifconfig | grep ^[a-z] | cut -d: -f1)

and for got(1)

	set -A complete_got_1 --	\
		init			\
		import im		\
		clone cl		\
		fetch fe		\
		checkout co		\
		update up		\
		status st		\
		log			\
		diff di			\
		blame bl		\
		tree tr			\
		ref			\
		branch br		\
		tag			\
		add			\
		remove rm		\
		revert rv		\
		commit ci		\
		send se			\
		cherrypick cy		\
		backout bo		\
		rebase rb		\
		histedit he		\
		integrate ig		\
		merge mg		\
		stage sg		\
		unstage ug		\
		cat			\
		info

Tweak the output of ls

	alias ls='ls -F'

reset(1) doesn't work as expected inside tmux: the old output can still
be consulted when scrolling.  If I, lazy as I am, bother to type "reset"
I want to be sure that the history was cleared!

	if [ -n "$TMUX" ]; then
		alias reset='reset && tmux clear-history'
	fi

CDPATH is super useful!  I even wrote a post about it:
https://www.omarpolo.com/post/enjoying-cdpath.html

	export CDPATH=".:$HOME/w:/usr/ports:/usr/ports/mystuff:$HOME/quicklisp/local-projects"

I love to hate gpg!  It needs some special treatments to work and this
should also (finger crossed!) fix pinentry over ssh.  I'm not sure it
works though, it's been a while since I've connected remotely to my
desktop.

	export GPG_TTY=$(tty)
	if [ -n "$SSH_CONNECTION" ]; then
		export PINENTRY_USER_DATA="USE_CURSES=1"
	fi

The BSDs have this incredibly useful signal available, it's a shame not
to use it!

	stty status ^T

I really like my prompt to be as minimal as possible.  For some time
I've used a single colon `;' as prompt, it's really nice!  At the moment
thought I'm usign a more plan9-esque percent sign:

	PS1='% '

I got tired of trying to remember the set of flags for nc to walk to
Gemini servers, so here we are:

	# "post" stdin to the gemini server
	# usage: gem host [port]
	gem()
	{
		host="${1:?missing host}"
		port="${2:-1965}"
		nc -c -Tnoverify "${host}" "${port}"
	}

I think I've stolen these from someone.  It makes a copy of the file and
launch an editor on the original file, incledibly useful when working
with ports (that's why doas!)

	mgdiff()
	{
		if [ -z "$1" ]; then
			printf "%s\n" "USAGE: mgdiff file" >&2
			return
		fi
		doas cp -p "$1" "$1.orig"
		doas mg "$1"
	}

hist is a quick wrapper around history and grep to quickly search for a
previous command:

	hist()
	{
		if [ -z "$1" ]; then
			printf "%s\n" "USAGE: hist pattern" >&2
			return 1
		fi
		history 0 | grep "$1"
	}

Some aliases I use when working with the OpenBSD port tree:

	alias m="make"
	alias mup="make update-patches"
	alias mupl="make update-plist"
	alias build="MAKE_JOBS=5 time make 2>&1 | tee build"
	alias pclear='make clean="package plist"'

find(1) is an invaluable tool and I use it all the time.  walk is an
attempt to build a wrapper around some common usages of find that is a
little bit less verbose to use.  The name is stolen from 9front, but the
implementation is completely different.

	# usage: walk [dir] [type] [name regexp] [! command to execute]
	walk()
	{
		if [ $# -eq 0 ]; then
			find .
			return
		fi

		local dir=.
		local type=
		local name=\*

		if [ -n "$1" -a -d "$1" ]; then
			dir="$1"
			shift
		fi

		case "$1" in
		b|c|d|f|l|p|s)
			type="-type $1"
			shift
		esac

		if [ -n "$1" -a "x$1" != "x!" ]; then
			name="$1"
			shift
		fi

		if [ "x$1" = x! ]; then
			shift
		fi

		if [ $# -eq 0 ]; then
			find "$dir" $type -iname "$name"
		else
			find "$dir" $type -iname "$name" -exec "$@" {} +
		fi
	}