Casing arrow keys in bash

Question

Is it possible to case arrow keys in a bash script to run a certain set of commands if the up/left arrow keys are pressed, and a certain set if the down/right arrow keys are pressed? I am trying to have a way to switch between users quickly by using the arrow keys while it is displaying the data, using this script to read the data from.

function main()  # The main function that controls the execution of all other functions
{
  mkdir -p ~/usertmp  # Make a new temporary user directory if it doesn't exist
  touch ~/last_seen_output.txt  # Create the output file if it doesn't exist
  cat /dev/null > ~/last_seen_output.txt  # Make sure that the output file is empty
  gather  # Call the "gather" function
  total=$((`wc -l ~/usertmp/user_list.txt|awk '{print $1}'`-1))  # Calculate the total amount of lines and subtract 1 from the result
  echo Current Time: `date +%s` > ~/last_seen_output.txt  # Print the current time to the output file for later reference
  echo "" > ~/last_seen_output.txt  # Print a blank line to the output file
    if [ $log -eq 1 ]
      then
        # If it is enabled, then delete the old backups to prevent errors
        while [ $line_number -le $total ]
          do

            line_number=$((line_number+1))  # Add 1 to the current line number
            calculate # Call the "calculate" function
            hms  # Call the "hms" function to convert the time in seconds to normal time
            log
        done
      else
        while [ $line_number -le $total ]
          do
            line_number=$((line_number+1))  # Add 1 to the current line number
            calculate # Call the "calculate" function
            hms  # Call the "hms" function to convert the time in seconds to normal time
            echo "Displaying, please hit enter to view the users one by one."
            read  # Wait for user input
            if [ "$log_while_displaying" ]
              then
                log
                display
              else
                display
            fi
        done
    fi
}

https://github.com/jbondhus/last-seen/blob/master/last-seen.sh is the complete script.

The read command commented as "wait for user input" is the command that you hit enter to go to the next user for. Basically, what this script does it list users and the time that has passed since each user logged in. I am trying to switch between each user being displayed by using the arrow keys. I figured that it might be possible to use a case statement to case the key input. To reiterate my point, I'm not sure if this is possible. If it isn't, can anyone think of another way to do this?

Answer 1

You can read arrow keys as well as other keys without any unusual commands; you just need to conditionally add a second read call:

escape_char=$(printf "\u1b")
read -rsn1 mode # get 1 character
if [[ $mode == $escape_char ]]; then
    read -rsn2 mode # read 2 more chars
fi
case $mode in
    'q') echo QUITTING ; exit ;;
    '[A') echo UP ;;
    '[B') echo DN ;;
    '[D') echo LEFT ;;
    '[C') echo RIGHT ;;
    *) >&2 echo 'ERR bad input'; return ;;
esac

Answer 2

As mentioned before, the cursor keys generate three bytes - and keys like home/end even generate four! A solution I saw somewhere was to let the initial one-char read() follow three subsequent one-char reads with a very short timeout. Most common key sequences can be shown like this f.e.:

#!/bin/bash
for term in vt100 linux screen xterm
  { echo "$term:"
    infocmp -L1 $term|egrep 'key_(left|right|up|down|home|end)'
  }

Also, /etc/inputrc contains some of these with readline mappings.. So, answering original question, here's a snip from that bash menu i'm just hacking away at:

while read -sN1 key # 1 char (not delimiter), silent
do
  # catch multi-char special key sequences
  read -sN1 -t 0.0001 k1
  read -sN1 -t 0.0001 k2
  read -sN1 -t 0.0001 k3
  key+=${k1}${k2}${k3}

  case "$key" in
    i|j|$'\e[A'|$'\e0A'|$'\e[D'|$'\e0D')  # cursor up, left: previous item
      ((cur > 1)) && ((cur--));;

    k|l|$'\e[B'|$'\e0B'|$'\e[C'|$'\e0C')  # cursor down, right: next item
      ((cur < $#-1)) && ((cur++));;

    $'\e[1~'|$'\e0H'|$'\e[H')  # home: first item
      cur=0;;

    $'\e[4~'|$'\e0F'|$'\e[F')  # end: last item
      ((cur=$#-1));;

    ' ')  # space: mark/unmark item
      array_contains ${cur} "${sel[@]}" && \
      sel=($(array_remove $cur "${sel[@]}")) \
      || sel+=($cur);;

    q|'') # q, carriage return: quit
      echo "${sel[@]}" && return;;
  esac                  

  draw_menu $cur "${#sel[@]}" "${sel[@]}" "$@" >/dev/tty
  cursor_up $#
done

Answer 3

You can use read -n 1 to read one character then use a case statement to choose an action to take based on the key.

On problem is that arrow keys output more than one character and the sequence (and its length) varies from terminal to terminal.

For example, on the terminal I'm using, the right arrow outputs ^[[C. You can see what sequence your terminal outputs by pressing Ctrl-V Right Arrow. The same is true for other cursor-control keys such as Page Up and End.

I would recommend, instead, to use single-character keys like < and >. Handling them in your script will be much simpler.

read -n 1 key

case "$key" in
    '<') go_left;;
    '>') go_right;;
esac

Answer 4

# This will bind the arrow keys

while true # Forever = until ctrl+c
do
    # r: backslash is not for escape
    # s: silent
    # "n x": read x chars before returning
    read -rsn 1 t
    case $t in
        A) echo up ;;
        B) echo down ;;
        C) echo right ;;
        D) echo left ;;
    esac
done

Answer 5

Not sure if this answer the question directly, but I think it's related - I was wandering where do those codes come from, and I finally found:

• Linux Keycode Table (comptechdoc.org)

It's a bit difficult to read at first; for left arrow, lookup "LEFT 4" in the "Key" column, and for the sequence that bash sees, look up the 5th ("keymap" - "normal") column, where it is written as "[D 1b 5b 44" - which are the three bytes (27, 91, 68) representing this key.

Finding the thread How to read arrow keys on really old bash? - The UNIX and Linux Forums, inspired me to write a short one-liner which dumps the key codes of keys pressed. Basically, you press a key, then Enter (to trigger ending of read), and then use hexdump to output what read has saved (and finally hit Ctrl-C to exit the loop):

$ while true; do read -p?; echo -n $REPLY | hexdump -C; done
?^[[D     
00000000  1b 5b 44                                          |.[D| # left arrow
00000003
?^[[C
00000000  1b 5b 43                                          |.[C| # right arrow
00000003
?^[[1;2D
00000000  1b 5b 31 3b 32 44                                 |.[1;2D| # Shift+left arrow
00000006
?^[[1;2C
00000000  1b 5b 31 3b 32 43                                 |.[1;2C| # Shift+right arrow
00000006
?^C

So, while arrow keys require 3 bytes - Shift+arrow keys require 6! However, seemingly all these sequence start with 0x1b (27), so one could possibly check for this value for read -n1, before reading any more bytes; also 5b remains a second byte in multi-byte sequence for the "normal" and "shift/NUM-Lock" columns of the table above.

---

Edit: much easier and proper way to scan for terminal codes of pressed keys in Linux is via showkey:

$ showkey 
Couldn't get a file descriptor referring to the console

$ showkey -h
showkey version 1.15

usage: showkey [options...]

valid options are:

    -h --help   display this help text
    -a --ascii  display the decimal/octal/hex values of the keys
    -s --scancodes  display only the raw scan-codes
    -k --keycodes   display only the interpreted keycodes (default)

$ sudo showkey -a

Press any keys - Ctrl-D will terminate this program

^[[A     27 0033 0x1b
         91 0133 0x5b
         65 0101 0x41
^[[B     27 0033 0x1b
         91 0133 0x5b
         66 0102 0x42
^[[A     27 0033 0x1b
         91 0133 0x5b
         65 0101 0x41
^[[D     27 0033 0x1b
         91 0133 0x5b
         68 0104 0x44
^[[C     27 0033 0x1b
         91 0133 0x5b
         67 0103 0x43
^C       3 0003 0x03
^M       13 0015 0x0d
^D       4 0004 0x04

Answer 6

Using eMPee584 answer I think I came up a good solution for you. Its output is much the same as user3229933 answer but will not be triggered by shift keys and will work in most terminals.

It has UP DOWN LEFT RIGHT HOME and END Keys Press 'q' to quit Most of this is thanks to eMPee584

you may need to change '-sn1' to '-sN1' if you get an error like illegal option n.

#!/bin/bash

while read -sn1 key # 1 char (not delimiter), silent
do

  read -sn1 -t 0.0001 k1 # This grabs all three symbols 
  read -sn1 -t 0.0001 k2 # and puts them together
  read -sn1 -t 0.0001 k3 # so you can case their entire input.

   key+=${k1}${k2}${k3} 

  case "$key" in
    $'\e[A'|$'\e0A')  # up arrow
        ((cur > 1)) && ((cur--))
        echo up;;

    $'\e[D'|$'\e0D') # left arrow
        ((cur > 1)) && ((cur--))
        echo left;;

    $'\e[B'|$'\e0B')  # down arrow
        ((cur < $#-1)) && ((cur++))
        echo down;;

    $'\e[C'|$'\e0C')  # right arrow
        ((cur < $#-1)) && ((cur++))
        echo right;;

    $'\e[1~'|$'\e0H'|$'\e[H')  # home key:
        cur=0
        echo home;;

    $'\e[4~'|$'\e0F'|$'\e[F')  # end key:
        ((cur=$#-1))
        echo end;;

    q) # q: quit
        echo Bye!
        exit;;

   esac                  

done