bash-completion.el defines dynamic completion hooks for shell-mode and shell-command prompts that is based on bash completion.

Bash completion for Emacs:

• is aware of bash builtins, aliases and functions
• does file expansion inside of colon-separated variables and after redirections (> or <)
• escapes special characters when expanding file names
• is configurable through programmable bash completion
• works on remote shells, through TRAMP.

A simpler and more complete alternative to bash-completion.el is to run a bash shell in a buffer in term mode (M-x ansi-term). Unfortunately, many Emacs editing features are not available when running in term mode. Also, term mode is not available in shell-command prompts.

Copy bash-completion.el into a directory that's on Emacs load-path. You can do that manually, or by installing it from MELPA.

To enable bash completion in shell buffers as well as in command prompts, such as the prompt started by compile, add the hook bash-completion-dynamic-complete to shell-dynamic-complete-functions.

For example:

        (autoload 'bash-completion-dynamic-complete
          "bash-completion"
          "BASH completion hook")
        (add-hook 'shell-dynamic-complete-functions
          'bash-completion-dynamic-complete)

or simpler, but forces you to load bash-completion at startup:

        (require 'bash-completion)
        (bash-completion-setup)

After that reload your .emacs (M-x eval-buffer) or restart.

When called from a bash shell buffer, bash-completion-dynamic-complete communicates with the current shell to reproduce, as closely as possible the normal bash auto-completion, available on full terminals.

When called from non-shell buffers, such as the prompt of M-x compile, bash-completion-dynamic-complete creates a separate bash process just for doing completion. Such processes have the environment variable EMACS_BASH_COMPLETE set to t, to help distinguish them from normal shell processes.

Additionally, you can enable bash completion in any buffer that contains bash commands. To do that, call

(bash-completion-dynamic-complete-nocomint COMP-START COMP-POS DYNAMIC-TABLE)

from a function added to completion-at-point-functions.

The trickiest part is setting COMP-START to where the bash command starts; It depends on the mode of the calling buffer and might, in some cases, span multiple lines.

COMP-POS is usually the current position of the cursor.

When calling from completion-at-point, make sure to pass a non-nil value to the DYNAMIC-TABLE argument so it returns a function instead of a list of strings. This isn't just an optimization: returning a function instead of a list tells Emacs it should avoids post-filtering the results and possibly discarding useful completion from bash.

For example, here's a function to to do bash completion from an eshell buffer. To try it out, add the function below to your init file and bind bash-completion-from-eshell to a custom shortcut.

(defun bash-completion-from-eshell ()
  (interactive)
  (let ((completion-at-point-functions
         '(bash-completion-eshell-capf)))
    (completion-at-point)))

(defun bash-completion-eshell-capf ()
  (bash-completion-dynamic-complete-nocomint
   (save-excursion (eshell-bol) (point))
   (point) t))

To report bugs, features or even to ask questions, please open an issue. To contribute code or documentation, please open a pull request.

See CONTRIBUTING.md for more details.

bash-completion.el is known to work with Bash 3, 4 and 5, on Emacs, starting with version 24.3, under Linux and OSX. It does not work on XEmacs.