Using arrow keys in shell scripts

I recently needed to collect arrow keys (and function keys etc.) in a shell script so that I could run a text graphics-style data entry system (with text entry fields, drop-down list boxes, progress bars and the like). Yes you can do all this in shell, and portably too if you're careful.

I've seen others asking how to capture keypresses in shell scripts in the past, with a variety of responses from other people, so thought you might all like to see a little mock-up program that catches keypresses and reports the key that was pressed. This has been tested in Solaris, Linux (SuSE) and AIX. It's written in ksh just because I like it, but would equally work in bash if you changed the 'print' commands to 'echo'. The script reports 10 keypresses in octal and then exits. Note the use of 'tput' to determine the correct terminal codes.

Enjoy.

#!/bin/ksh
AWK=gawk
[ -x /bin/nawk ] && AWK=nawk
ECHO="print"
ECHO_N="print -n"

tty_save=$(stty -g)

function Get_odx
{
    od -t o1 | $AWK '{ for (i=2; i<=NF; i++)
                        printf("%s%s", i==2 ? "" : " ", $i)
                        exit }'
}

# Grab terminal capabilities
tty_cuu1=$(tput cuu1 2>&1 | Get_odx)            # up arrow
tty_kcuu1=$(tput kcuu1 2>&1 | Get_odx)
tty_cud1=$(tput cud1 2>&1 | Get_odx)            # down arrow
tty_kcud1=$(tput kcud1 2>&1 | Get_odx)
tty_cub1=$(tput cub1 2>&1 | Get_odx)            # left arrow
tty_kcub1=$(tput kcud1 2>&1 | Get_odx)
tty_cuf1=$(tput cuf1 2>&1 | Get_odx)            # right arrow
tty_kcuf1=$(tput kcud1 2>&1 | Get_odx)
tty_ent=$($ECHO | Get_odx)                      # Enter key
tty_kent=$(tput kent 2>&1 | Get_odx)
tty_bs=$($ECHO_N "\b" | Get_odx)                # Backspace key
tty_kbs=$(tput kbs 2>&1 | Get_odx)

# Some terminals (e.g. PuTTY) send the wrong code for certain arrow keys
if [ "$tty_cuu1" = "033 133 101" -o "$tty_kcuu1" = "033 133 101" ]; then
    tty_cudx="033 133 102"
    tty_cufx="033 133 103"
    tty_cubx="033 133 104"
fi

stty cs8 -icanon -echo min 10 time 1
stty intr '' susp ''

trap "stty $tty_save; exit"  INT HUP TERM

count=0
while :; do
    [ $count -eq 10 ] && break
    count=$((count+1))

    keypress=$(dd bs=10 count=1 2> /dev/null | Get_odx)

    $ECHO_N "keypress=\"$keypress\""

    case "$keypress" in
        "$tty_ent"|"$tty_kent") $ECHO " -- ENTER";;
        "$tty_bs"|"$tty_kbs") $ECHO " -- BACKSPACE";;
        "$tty_cuu1"|"$tty_kcuu1") $ECHO " -- KEY_UP";;
        "$tty_cud1"|"$tty_kcud1"|"$tty_cudx") $ECHO " -- KEY_DOWN";;
        "$tty_cub1"|"$tty_kcub1"|"$tty_cubx") $ECHO " -- KEY_LEFT";;
        "$tty_cuf1"|"$tty_kcuf1"|"$tty_cufx") $ECHO " -- KEY_RIGHT";;
        *) $ECHO;;
    esac
done

stty $tty_save

Pretty cool. Uh just a comment: ksh also works with "echo". No need to anonymize this command.

The 'echo' command is not normally a ksh built-in. This means that not only are the options system-dependent (and I needed code that worked reliably on numerous platforms), but it will also spawn a new process which will slow your code down. The 'print' command is built-in to the shell and does not suffer these problems. I would always recommend people use 'print' instead of 'echo' when using ksh.

The man pages say that echo is indeed a builtin:

Further evidence:

root@sf8 # uname -a
SunOS sf8 5.10 Generic_127127-11 sun4u sparc SUNW,SPARC-Enterprise
root@sf8 # echo $0
/usr/bin/ksh
root@sf8 # type echo
echo is a shell builtin

Yes, but it doesn't have to be a built-in. From the ksh man page on SuSE:

If I want to write a ksh script that has maximum portability, especially when I want to use certain options (e.g. -e or -r), I do not want to be using 'echo'.