tab completion stops working

[jseabold]

This has gotten to be enough of a headache that I'd like to get to the bottom of it, using the regular IPython interpreter

|21 $ ipython --version
2.1.0

Tab-completion randomly stops working for me. The other symptom is that the up arrow just scrolls through the history instead of filtering on what's already typed at the prompt.

Unfortunately, I can't reproduce it deterministically. Often it happens after I %paste some code in, but that's not the only time. I've noticed this (maybe) after I changed my .inputrc, so perhaps I'm hitting some odd key combination without realizing it? .inputrc in-lined below.

Is there something I can instrument or somewhere I can insert a debug statement to help figure this out?

set editing-mode vi
set keymap vi-command

set bell-style none
$if mode=vi
    set keymap vi-command
    "gg": beginning-of-history
    "G": end-of-history
    set keymap vi-insert
    "ii": vi-movement-mode
    "\C-w": backward-kill-word
    "\C-p": history-search-backward
    "\C-e": end-of-line
    "\C-a": beginning-of-line
$endif

[takluyver]

I think this must be something in readline itself, if up and down arrows stop behaving as expected. There's no callback we get for going through history with the up and down arrows, that's all handled by readline using history we've previously given it.

[jseabold]

Well, I figured out one way I can replicate this behavior. Type any character then 'e' and hit up.

E.g., if I wanted to autocomplete an np.random.seed call and got to e but wanted to re-use one

np.random.se<Up>

Then tab completion stops. Same if I do.

xe<Up>

It doesn't seem to happen if e is the first character.

[jseabold]

r too. I guess this is something with readline and my config.

[jseabold]

As another data point, it doesn't happen when I comment out vi-insert in my inputrc, but I don't see/know of any keybinding that would cause the current behavior. Pretty frustrating.

[jseabold]

For anyone who comes across this, by some miracle I finally found a "fix" to move from emacs mode back to vi-mode in IPython. I don't yet understand how I get dropped in to emacs mode in the first place and this only happens in ipython, but a "M-C-j" drops me back in to vi-mode (on OSX). Link