writing Python in Emacs

[Richard Szopa]

Hi All,

I am a devoted Emacs user and I write a lot in Python. However, I
never managed to get my Emacs configuration right for this purpose.
There were some discussions on this, but the threads that show if I
search the group are either old or not so relevant.

I need the following features:

0) Of course, syntax coloring and so on... But this works good enough
ootb in the two most popular python-modes.

1) Tab completion, ideally Slime like. That is, when there's not
enough letters to unambiguously complete a symbol, I want it to show a
buffer (w/o taking the focus) w/ the possible completions. In an ideal
world, it would be able to complete fo.ba<TABto foo.bar. I imagine
this would require quite tight Emacs-Python integration.

2) Sending the toplevel definition (class or function) to the Python
buffer.

3) Hints on function/method arguments. IDLE has this done nearly
right, but the hints are a bit too intrusive for me. I would like to
see them in the minibuffer.

4) (optional) I would like to see the definition of a function
function or class by hitting M-. on its name. (I understand that this
may be impossible for methods, as Emacs would have to automagically
infer the type of the object).

I have tried a couple of times both python-modes (the one shipped w/
Python and the one shipped w/ Emacs), pymacs and stuff like that...
And, as I said, never got it right. But, maybe I just cannot find the
way to configure it, and some configuration hints will be enough...

As for other editors, I have tried Eclipse and Komodo... But I cannot
get used to them. As for non-emacs stuff, the most comfortable for me
has been IDLE.

Cheers and thanks in advance,

-- Richard

[Terry Jones]

>>>>"Richard" == Richard Szopa <ry***********@ gmail.comwrites :

RichardI am a devoted Emacs user and I write a lot in Python.

Me too.

RichardI need the following features:

Richard1) Tab completion, ideally Slime like. That is, when there's not
Richardenough letters to unambiguously complete a symbol, I want it to
Richardshow a buffer (w/o taking the focus) w/ the possible
Richardcompleti ons. In an ideal world, it would be able to complete
Richardfo.ba<TA Bto foo.bar. I imagine this would require quite tight
RichardEmacs-Python integration.

I know this is not what you want, but I use hippie expand (M-/) to cycle
through possible completions. It's not Python aware, but it is of some use.

Richard2) Sending the toplevel definition (class or function) to the Python
Richardbuffer.

I switched to IPython to have better interaction with a spawned Python.

To use IPython you need to use the Python mode that is NOT the one from
(endorsed?) by the FSF. It gives you some completion (at least in the
*Python* buffer) and you can send pieces of the buffer to the python
process, via py-send-region (C-c |), py-execute-def-or-class (M-C-x), etc.

Richard3) Hints on function/method arguments. IDLE has this done nearly
Richardright, but the hints are a bit too intrusive for me. I would like to
Richardsee them in the minibuffer.

I don't have this.

Richard4) (optional) I would like to see the definition of a function
Richardfunction or class by hitting M-. on its name. (I understand that
Richardthis may be impossible for methods, as Emacs would have to
Richardautomagi cally infer the type of the object).

This is just an emacs tag file need. Have you googled for something like
emacs tags python? The issue of methods might be overcome by just moving
through tags with the same name. Yes, that requires _you_ to know when
you've hit the right thing. That's not optimal, but it's better than
nothing. Ideally you could send the definition to IPython, ask it for the
class info, and use that to jump to the right tag.

RichardI have tried a couple of times both python-modes (the one shipped w/
RichardPython and the one shipped w/ Emacs), pymacs and stuff like that...
RichardAnd, as I said, never got it right. But, maybe I just cannot find the
Richardway to configure it, and some configuration hints will be enough...

If you have the time, please summarize your findings. The emacs/python
world has always seemed quite amorphous to me too.

Terry

[Thierry Volpiatto]

I add just a note about ipython:
if you use a version 0.6.15 may be you will have a bad output
on error like:

== " ":
instead of:
if __name__ == "__main__":

all the characters are missing.

To avoid that, run in ipython:
%upgrade -nolegacy

uncomment in ~/.ipython/ipy_user_config .py:
import ipy_defaults

restart emacs and try a .py with some syntax errors.
It should be ok now.

Terry Jones <te***@jon.eswr ites:

>>>>>"Richard " == Richard Szopa <ry***********@ gmail.comwrites :

RichardI am a devoted Emacs user and I write a lot in Python.

Me too.

RichardI need the following features:

Richard1) Tab completion, ideally Slime like. That is, when there's not
Richardenough letters to unambiguously complete a symbol, I want it to
Richardshow a buffer (w/o taking the focus) w/ the possible
Richardcompleti ons. In an ideal world, it would be able to complete
Richardfo.ba<TA Bto foo.bar. I imagine this would require quite tight
RichardEmacs-Python integration.

I know this is not what you want, but I use hippie expand (M-/) to cycle
through possible completions. It's not Python aware, but it is of some use.

Richard2) Sending the toplevel definition (class or function) to the Python
Richardbuffer.

I switched to IPython to have better interaction with a spawned Python.

To use IPython you need to use the Python mode that is NOT the one from
(endorsed?) by the FSF. It gives you some completion (at least in the
*Python* buffer) and you can send pieces of the buffer to the python
process, via py-send-region (C-c |), py-execute-def-or-class (M-C-x), etc.

Richard3) Hints on function/method arguments. IDLE has this done nearly
Richardright, but the hints are a bit too intrusive for me. I would like to
Richardsee them in the minibuffer.

I don't have this.

Richard4) (optional) I would like to see the definition of a function
Richardfunction or class by hitting M-. on its name. (I understand that
Richardthis may be impossible for methods, as Emacs would have to
Richardautomagi cally infer the type of the object).

This is just an emacs tag file need. Have you googled for something like
emacs tags python? The issue of methods might be overcome by just moving
through tags with the same name. Yes, that requires _you_ to know when
you've hit the right thing. That's not optimal, but it's better than
nothing. Ideally you could send the definition to IPython, ask it for the
class info, and use that to jump to the right tag.

RichardI have tried a couple of times both python-modes (the one shipped w/
RichardPython and the one shipped w/ Emacs), pymacs and stuff like that...
RichardAnd, as I said, never got it right. But, maybe I just cannot find the
Richardway to configure it, and some configuration hints will be enough...

If you have the time, please summarize your findings. The emacs/python
world has always seemed quite amorphous to me too.

Terry
_______________ _______________ _______________ __
help-gnu-emacs mailing list
he************@ gnu.org
http://lists.gnu.org/mailman/listinfo/help-gnu-emacs

--
A + Thierry
Pub key: http://pgp.mit.edu

[Ryszard Szopa]

Thanks Rob. Your code should basically do the trick.

-- Richard

[alitosis]

Rob Wolfe wrote:

The good news is that I managed to configure completion for Python
in Emacs using pymacs, python-mode.el, pycomplete.el and pycomplete.py.
For contents of my pycomplete.el, pycomplete.py and necessary
settings in .emacs see below.

Thanks for that! I've been hoping something like this landed on my
lap for years.