Getting “complete” and “menu-complete” to work together

Question

I found out that the Bash shell supports a type of autocompletion that is different from the "traditional" autocompletion, where all possibilities get listed on the following line.

With the "traditional" autocompletion, if I type ch and then press the Tab key, I get something like:

$ ch
chacl  chgrp  chmod  chown  chvt

But if I add the following line to my /etc/inputrc (which remaps the Tab key to the built-in menu-complete function):

Tab: menu-complete

then the behavior of the shell changes: the word to be completed is replaced "inline" with a single match from the list of possible completions, and if I press the Tab key again, the word gets replaced with the next match.

I found this useful, but I still wanted to keep the traditional autocompletion and have it bound to the key combination Ctrl + Tab. So I added the following line to my /etc/inputrc file, according to what the readline library documentation suggests:

Ctrl-Tab: complete

However, adding this line only seems to make both Tab and Ctrl-Tab call the traditional complete function.

Does anyone know what I am doing wrong?

Thanks in advance!

Answer 1

To start with, I'm not a massive expert in this area, but I think I can answer your question. First of all, while you are using Bash, Bash is a shell which interprets keyboard commands that it receives from a terminal / console. While you are informing Bash how to react to specific key combinations in the inputrc file, your Terminal determines precisely which character is 'sent' to the Shell before the inputrc file even enters the equation.

Unfortunately, on my system (granted, it's OSX - but I don't think this is strange behaviour when compared to Linux), both Tab and Ctrl-Tab send the same keyboard input to the shell. Infact, both Tab and Ctrl-Tab send a Ctrl-I command to the shell, and indeed, if I enter Ctrl-I when using the terminal, it performs the completion as if I hit Tab.

The software (installed on most Linux systems by default), showkey will tell you what keys the shell is receiving when you press specific keyboard inputs as you push them.

Anyway, my suggestion to you is to use Shift-Tab, which does appear to send it's own key-code to the shell. Shift-Tab on my computer shows up (using showkey) as '<ESC>[Z', which I think is pretty standard across the board. As such, your inputrc file with the following bindings should allow you to use shift-tab instead of ctrl-tab to achieve what you desire:

Tab: menu-complete
"\e[Z": complete

The \e in the second binding represents the escape character, and the [Z are simply the characters as shown using showkey. You can get a similar effect on OSX by simply using cat, running cat from within a terminal and pressing Shift-Tab will show you "^[[Z", where ^[ represents the escape character and the other characters are as before.

I know this doesn't resolve your question precisely, but as I don't think you are able to use Ctrl-Tab as a key combination, without re-mapping Ctrl-Tab to another keybinding within your terminal (more likely to be easier if you are using a GUI terminal), this is likely as close as you can get without significant effort!

Answer 2

I have ShiftTab bound to menu-complete-backward, so it goes back one step if I skipped the right completion, and I've mapped Ctrlq to complete, so if there are several possible completions I hit Ctrlq to list them without having to cycle through them.

# Make Tab cycle between possible completions
# Cycle forward: Tab
# Cycle backward: Shift-Tab

TAB: menu-complete
"\e[Z": menu-complete-backward

# Make C-q display the list of possible completions

Control-q: complete

# Display the list of matches when no further completion is possible

set show-all-if-unmodified on

Edit: Ctrlq is bound to quoted-insert by default, that is, it tells the shell to take the next key literally. quoted-insert is also bound to Ctrlv, so you don't lose that functionality if you rebind Ctrlq. Anyway, I've found that AltESC also works, by default, for showing the possible completions (as far as I can tell it is equivalent to TAB); note that it may be seized by Gnome, then either double press ESC or rebind "Switch windows directly" in Settings → Devices → Keyboard → Navigation.

Answer 3

The following should achieve what you're looking for (if I understand correctly!)

In your .inputrc

# display all possible matches for an ambiguous pattern at first tab
set show-all-if-ambiguous on

# next tab(s) will cycle through matches
TAB: menu-complete
# shift tab cycles backward
"\e[Z": menu-complete-backward

Answer 4

I'm not sure Ctrl-Tab is a real character; my terminal, for instance, ignores the combination. I think the only way to use Ctrl-Tab is to use your terminal emulator to map it to some otherwise unused escape sequence, then bind that sequence to complete.

Answer 5

Where to start, if you can or cant do this is dependent your keyboard and your drivers and there isn't one catch all answer. Each key press and release generates a sequenced key pair (key down and release) (scancode) these codes are then translated by the kernel into keycodes for example on my laptop keyboard 0x3a 0xba are translated to keycode 15 (down and up) these are then translated into actions such as return letter c a / you can assign actions to keysyms using the keycode/hex/binary/octal notation which codes match which letters is determined by the kernel translation table which is fairly standardized, however the first part signal that's translated to keycodes is different for most keyboards.

Continuing with the earlier example for me shift tab (and alt and control and any combination) produce keycode 15 however in hex it produces 0x2a 0x2a 0x2a 0x2a 0x2a 0x2a 0x2a 0x2a 0x2a 0x2a 0x2a and this is because shift alt and control are special keys (modifiers) these multiply out against the keycodes and fill out the dumpkeys table the kernel is limited to the number of assignments as well this is determined by your choice of keymap and shares resources with your terminal colors (if your char set its defined above the threshold it limits your terminal color scope). And this all goes out the window if your in an xserver and has a whole new system. Most of these things can be changed,modified and manipulated by the user and programs installed. My point to all this is to emphasize that there is no catch all for the mapping of the tab key and its going to vary keyboard drivers to kbd drivers (now if you find a solution that happens to work for you excellent :)) but chances are it won't be portable and might not work if you change keyboards and might not translate between xserver and tui. What i recommend is learning the steps to modify your kbd on the go.

will give you the decimal octal hex notation for a key press on the same line

--full-table -1 >> keytable 

will give you a documented with your full list of keycode->keysym pairing in a format that will give you a better picture of your layout and from there you can either use loadkey to change a keys value or ad an entry in .inputrc or your main rc file. You can also create a custom key.map file. Further escape sequence translation is determined by the "$TERM" variable and each virtual terminal emulator can be different

infocmp "$TERM"

will give you a list of your terminal escape sequences

Resources:

https://man7.org/linux/man-pages/man4/console_codes.4.html

https://www.gnu.org/software/screen/manual/html_node/Input-Translation.html

http://kbd-project.org/docs/scancodes/scancodes.html

https://www.vt100.net/

So to sum up. Your keyboard drivers Your kemap choice Your virtual terminal emulator and your kernel form the backbone of remapping dificult keys (tab/s-tab/a-tab)