Readline

From ArchWiki

Readline is a library by the GNU Project, used by Bash and other CLI programs to edit and interact with the command line. See readline(3) for more information.

Installation

The readline package is most likely already installed as a dependency of Bash.

Editing mode

By default Readline uses Emacs style shortcuts for interacting with command line. However, vi style editing interface is also supported by adding the following to ~/.inputrc:

~/.inputrc
set editing-mode vi

Alternatively, to set it only for Bash by adding the following line to ~/.bashrc:

~/.bashrc
set -o vi

Mode indicator in prompt

Vi-style editing has two modes: command and insert. You can display which one is currently active by adding the following option:

~/.inputrc
set show-mode-in-prompt on

This will print a string in your prompt ((cmd)/(ins) by default) that can be customized with the vi-ins-mode-string and vi-cmd-mode-string variables.

Different cursor shapes for each mode

You can set a different cursor shape for each mode by using "\1 .. \2" escapes:

~/.inputrc
set vi-ins-mode-string \1\e[6 q\2
set vi-cmd-mode-string \1\e[2 q\2

This will set a block shaped cursor when in command mode and a pipe cursor when in insert mode. Note that you must have the mode indicator enabled for this to work (see #Mode indicator in prompt).

The Virtual Console uses different escape codes, so you should check first which term is being used:

~/.inputrc
$if term=linux
	set vi-ins-mode-string \1\e[?0c\2
	set vi-cmd-mode-string \1\e[?8c\2
$else
	set vi-ins-mode-string \1\e[6 q\2
	set vi-cmd-mode-string \1\e[2 q\2
$endif

See software cursor for VGA for further details.

Fast word movement

Xterm supports moving between words with Ctrl+Left and Ctrl+Right by default. To achieve this effect with other terminal emulators, find the correct terminal codes, and bind them to backward-word and forward-word in ~/.inputrc.

For example, for urxvt:

~/.inputrc
"\e[1;5D": backward-word
"\e[1;5C": forward-word

Another example for macOS style (Alt+Left and Alt+Right) word moving:

~/.inputrc or /etc/inputrc
"\e[1;9D": backward-word
"\e[1;9C": forward-word

History

The history configuration is split into:

  • controlling its operation, which is configured shell-specific, for example see Bash#History customization and
  • influencing the interaction, which may be configured via a shell-specific configuration, or its native inputrc configuration applicable for all shells, as explained in the following examples.

Usually, pressing the up arrow key will cause the last command to be shown regardless of the command that has been typed so far. However, users may find it more practical to list only past commands that match the current input.

For example, if the user has typed the following commands:

  • ls /usr/src/linux-2.6.15-ARCH/kernel/power/Kconfig
  • who
  • mount
  • man mount

In this situation, when typing ls and pressing the up arrow key, current input will be replaced with man mount, the last performed command. If the history search has been enabled, only past commands beginning with ls (the current input) will be shown, in this case ls /usr/src/linux-2.6.15-ARCH/kernel/power/Kconfig.

You can enable the history search mode by adding the following lines to /etc/inputrc or ~/.inputrc:

"\e[A": history-search-backward
"\e[B": history-search-forward

If you are using vi mode, you can add the following lines to the readline configuration file (from BBS#54972):

~/.inputrc
set editing-mode vi
$if mode=vi
set keymap vi-command
# these are for vi-command mode
"\e[A": history-search-backward
"\e[B": history-search-forward
j: history-search-forward
k: history-search-backward
set keymap vi-insert
# these are for vi-insert mode
"\e[A": history-search-backward
"\e[B": history-search-forward
$endif

If you chose to add these lines to ~/.inputrc, it is recommended that you also add the following line at the beginning of this file to avoid strange things like BBS#112537:

$include /etc/inputrc

Alternatively, one can use reverse-search-history (incremental search) by pressing Ctrl+r, which does not search based on previous input but instead jumps backwards in the history buffer as commands are typed in a search term. Pressing Ctrl+r again during this mode will display the previous line in the buffer that matches the current search term, while pressing Ctrl+g (abort) will cancel the search and restore the current input line. So in order to search through all previous mount commands, press Ctrl+r, type 'mount' and keep pressing Ctrl+r until the desired line is found.

The forward equivalent to this mode is called forward-search-history and is bound to Ctrl+s by default. Beware that most terminals override Ctrl+s to suspend execution until Ctrl+q is entered. (This is called XON/XOFF flow control). For activating forward-search-history, either disable flow control by issuing:

$ stty -ixon

or use a different key in inputrc. For example, to use Alt+s which is not bound by default:

"\es": forward-search-history

Faster completion

When performing tab completion, a single tab attempts to partially complete the current word. If no partial completions are possible, a double tab shows all possible completions.

The double tab can be changed to a single tab by setting:

~/.inputrc
set show-all-if-unmodified on

Or you can set it such that a single tab will perform both steps: partially complete the word and show all possible completions if it is still ambiguous:

~/.inputrc
set show-all-if-ambiguous on

Colorized completion

You can enable coloring of completion of filenames with the colored-stats option. You can also color the identical prefix of completion-lists with colored-completion-prefix. For example:

~/.inputrc
# Color files by types
# Note that this may cause completion text blink in some terminals (e.g. xterm).
set colored-stats On
# Append char to indicate type
set visible-stats On
# Mark symlinked directories
set mark-symlinked-directories On
# Color the common prefix
set colored-completion-prefix On
# Color the common prefix in menu-complete
set menu-complete-display-prefix On

Macros

Readline also supports binding keys to keyboard macros, for example:

$ bind '"\ew": "\C-e # macro"'

To keep it permanently, add the part within single quotes to readline's configuration file:

~/.inputrc
"\ew": "\C-e # macro"

Now type a line and press Alt+w. Readline will act as though Ctrl+e (end-of-line) had been pressed, appended with ' # macro'.

Use any of the existing keybindings within a readline macro, which can be quite useful to automate frequently used idioms. For example, this one makes Ctrl+Alt+l append | less to the line and run it (Ctrl+m is equivalent to Enter):

"\e\C-l": "\C-e | less\C-m"

The next one prefixes the line with yes | when pressing Ctrl+Alt+y, confirming any yes/no question the command might ask:

"\e\C-y": "\C-ayes | \C-m"

This example wraps the line in su -c '' and runs it, if Alt+s is pressed:

"\es": "\C-a su -c '\C-e'\C-m"

This example prefixes the line with sudo , if Alt+s is pressed. It lets you review the result and will not input the Enter key.

"\es": "\C-asudo \C-e"

As a last example, quickly send a command in the background with Ctrl+Alt+b, discarding all of its output:

"\e\C-b": "\C-e > /dev/null 2>&1 &\C-m"

Disabling control echo

Readline causes the terminal to echo ^C after Ctrl+c is pressed. For users who wish to disable this, simply add the following to ~/.inputrc:

set echo-control-characters off

Bracketed paste

By default, bracketed paste mode is on. It can be set manually in ~/.inputrc:

set enable-bracketed-paste on

This ensures that pasting into Readline inserts the clipboard as single string of characters, instead of inserting characters as if they were entered from the keyboard. This is a safety measure to prevent Readline from automatically modifying and running pasted commands.

See also