TAB completion not working

[Zeitlinie]

TAB completion does not work for me in the IPython shell.
My system details are:
 GNU Emacs 24.3.1
 IPython 1.0.0
 python-mode-6.1.3
 Python 2.7.6

I've boiled down my init.el to this torso

;;;;;;;;;;;;;;;;;
(require 'package)
(package-initialize)
(require 'python-mode)
(setq-default py-shell-name "ipython")
;;;;;;;;;;;;;;;;;

With that, I do
    M-x ipython RET
to start an IPython shell. At the IPython shell's prompt I type
    In[x]: pr
    TAB
and ... nothing happens ... ?! ... except that in the command buffer
    End of buffer
pops up
I have also tried ESC-TAB and C-M-i instead of TAB only.
In that case I get (instead of 'End of buffer')
'ipython-complete: Wrong type argument: stringp, nil'
in the command buffer.

When I run an IPython shell from a terminal session TAB completion on my system leads to
    In[x]: pr
    TAB
%%prun %precision %profile %prun print property

Additionally, I have exchanged the python-mode-6.1.3 version of
python-mode.elc with that obtained from 'bzr branch lp:python-mode'.
Also with that TAB completion remains non-functional. The only difference is, that the use of ESC-TAB and C-M-i instead of TAB now returns
'ipython-complete: Symbol's function definition is void: position'
in the command buffer.

Any help on how to get TAB completion to work would be most welcome.
Thank you.
Mark

[Andreas Roehler (a-roehler)]

What do you get from M-TAB, where completion in Python-shell is set to?

TAB just indents. BTW may change that, as many users seems to expect TAB to complete.

[Zeitlinie (zeitlinie)]

Well, for me, M-TAB does what it probably does on most window-managed systems: it switches between tasks - independent of Emacs - and I would be very reluctant to change that 'only' because of Emacs IPython TAB-completion ;)

Should I just bind completion to a different key? How? Actually within Emacs I cannot even describe the key M-TAB with C-h k, because it will immediately invoke the systems task switcher.

Thanks for any additional help,
Mark

> TAB just indents.
Yep, I realized that also while writing this report. That is why I went to http://www.gnu.org/software/emacs/manual/html_node/emacs/Symbol-Completion.html, where they claimed, that ESC-TAB and C-M-i would serve as replacement for just TAB alone (which I find awkward). At that place they also point out that M-TAB may not be a good choice on window-managed systems.

[Andreas Roehler (a-roehler)]

On 02.06.2014 10:37, Zeitlinie wrote:
> Well, for me, M-TAB does what it probably does on most window-managed
> systems: it switches between tasks - independent of Emacs - and I would
> be very reluctant to change that 'only' because of Emacs IPython TAB-
> completion ;)
>
> Should I just bind completion to a different key? How? Actually within
> Emacs I cannot even describe the key M-TAB with C-h k, because it will
> immediately invoke the systems task switcher.
>
> Thanks for any additional help,
> Mark
>
>> TAB just indents.
> Yep, I realized that also while writing this report. That is why I went to http://www.gnu.org/software/emacs/manual/html_node/emacs/Symbol-Completion.html, where they claimed, that ESC-TAB and C-M-i would serve as replacement for just TAB alone (which I find awkward). At that place they also point out that M-TAB may not be a good choice on window-managed systems.
>

Okay, understand. Will change the default for windows-systems

For the moment applying the patch attached will change the key-setting of py-shell-map.
After re-loading TAB should complete.

There is no key for `py-indent-line' than, however C-j should DTRT.

[Zeitlinie (zeitlinie)]

ehem, just to make sure: by "..window-managed .." systems I did not imply M$-Windows. Eg., mine runs X11/KDE
Will apply the patch now and get back later,
Thanks,
Mark

[Andreas Roehler (a-roehler)]

On 02.06.2014 11:39, Zeitlinie wrote:
> ehem, just to make sure: by "..window-managed .." systems I did not imply M$-Windows.

Thanks clarifying this. Was heading towards Windows indeed.
Well, most Emacs/Linux-users are used to re-define that keys.

Why not make it easier?

Being inclined to change that for all

Let's see what the Python-Rock users say :)

  Eg., mine runs X11/KDE
> Will apply the patch now and get back later,
> Thanks,
> Mark
>

[Zeitlinie (zeitlinie)]

No success yet.

Now, with the patch applied, and on a single TAB, Emacs's command buffer says to me:
'Symbol's function definition is void: position'
and the *Messages* buffer contains
'ipython-complete: Symbol's function definition is void: position'

That is exactly what I got previously, using ESC-TAB and C-M-i instead of TAB with launchpad's most recent head version 'bzr branch lp:python-mode'.

Any idea?
Mark

PS.: maybe I should mention, that.for installing the head of the devel branch I byte-compile my local (and now patched) version of the repo and then pick the python-mode.el(c) from there and replace only my systems version of python-model.el(c) with them ... nothing else ... I hope this is ok?

[Andreas Roehler (a-roehler)]

On 02.06.2014 12:57, Zeitlinie wrote:
> No success yet.
>
> Now, with the patch applied, and on a single TAB, Emacs's command buffer says to me:
> 'Symbol's function definition is void: position'
> and the *Messages* buffer contains
> 'ipython-complete:

so it got something useful to complete already

Symbol's function definition is void: position'
>

that's provived by cl

Please look into python-mode.el at the spot "(require 'cl)"
Remove the eval-when-compile, so it is loaded in any case.

> That is exactly what I got previously, using ESC-TAB and C-M-i instead
> of TAB with launchpad's most recent head version 'bzr branch lp:python-
> mode'.
>
> Any idea?
> Mark
>
> PS.: maybe I should mention, that.for installing the head of the devel
> branch I byte-compile my local (and now patched) version of the repo and
> then pick the python-mode.el(c) from there and replace only my systems
> version of python-model.el(c) with them ... nothing else ... I hope this
> is ok?
>

The gain from compiling is just nothing IMO at a single file, while it creates confusion easily.
Suggest to remove all "*.elc" files here and not to compile personal stuff. That makes sure always the latest version is loaded.

[Zeitlinie (zeitlinie)]

Yeyh! That did it.
(Wish I'd understand LISP at some point ... or maybe better not ;) )
Great job - thanks!

Should I expect this to be part of an upcoming version, or will this remain specific to just my private version of python-mode?

Mark

[Andreas Roehler (a-roehler)]

On 02.06.2014 13:55, Zeitlinie wrote:
> Yeyh! That did it.
> (Wish I'd understand LISP at some point ... or maybe better not ;) )
> Great job - thanks!
>
> Should I expect this to be part of an upcoming version, or will this
> remain specific to just my private version of python-mode?
>
> Mark
>

Will change the default so it's convenient without tweaking.
Please re-load and tell when fix is committed.

Cheers,

Andreas

[Andreas Roehler (a-roehler)]

On 02.06.2014 13:55, Zeitlinie wrote:
> Yeyh! That did it.
> (Wish I'd understand LISP at some point ... or maybe better not ;) )
> Great job - thanks!
>
> Should I expect this to be part of an upcoming version, or will this
> remain specific to just my private version of python-mode?
>
> Mark
>

As we may need some indentation also,
   introduced combined command `py-shell-complete-or-indent':

   Complete or indent depending on the context.

   If cursor is at current-indentation and further indent
   seems reasonable, indent. Otherwise try to complete.

Let's see if its better now,

Andreas