0

I have developed a workflow for batch-export of org-mode files, and have come across an issue in which execution halts when using a python session. I'm using emacs 28.1 on MacOS 11.6.5 with Python 3.9.5.

Here's a minimal org-mode file (python-session.org)

#+PROPERTY: header-args:python :exports both :eval yes :session

#+begin_src python :results output
import sys
print(sys.executable)
print(sys.version)
var = "from the first block"
#+end_src

#+begin_src python :results output
print(var)
#+end_src

..and some minimal code to export on batch mode

(let* ((infile (car (last command-line-args)))
       (outfile (concat (file-name-sans-extension infile) ".html")))

  (find-file infile)
  (org-mode)
  (org-babel-do-load-languages
   'org-babel-load-languages
   '((emacs-lisp . t)
     (shell . t)
     (python . t)))

  (setq org-confirm-babel-evaluate nil)
  (setq org-babel-python-command "python3")
  (setq python-indent-offset 4)
  (setq python-shell-completion-native-enable nil)
  (setq python-indent-guess-indent-offset t)
  (setq python-indent-guess-indent-offset-verbose nil)

  (org-html-export-as-html)
  (write-file outfile))

Here's the command to perform the export (and the resulting message):

% emacs --batch --load minimal-export.el --kill -- python-session.org
executing Python code block...

At this point the process hangs and must be interrupted.

The above command results in an exported html file when the :session option is removed from the first line of the file:

% emacs --batch --load minimal-export.el --kill -- python-session.org
executing Python code block...
Code block evaluation complete.
executing Python code block...
Babel evaluation exited with code 1
Code block produced no output.
Cannot fontify source block (htmlize.el >= 1.34 required)
Cannot fontify source block (htmlize.el >= 1.34 required)

Any ideas on how to troubleshoot further?

6
  • FWIW, I have no problem doing the above with GNU Emacs 29.0.50 (build 1, x86_64-pc-linux-gnu, GTK+ Version 3.24.31, cairo version 1.17.4) of 2022-02-12, so it may be a bug in Emacs 28.1. BTW, when I say no problem, I mean there is no hang. But the output of the Org Babel source block is messy; in particular, it is not an exact copy of what you would get when you run the program from the command line. For that reason alone, I avoid python sessions in Org Babel.
    – NickD
    Apr 24, 2022 at 0:07
  • I also have no problem with Emacs 27.2 which is the version that comes with my Linux distro (Fedora 34): no hang; same output problem as described above.
    – NickD
    Apr 24, 2022 at 0:10
  • I misspoke about the output: it is exactly as would be expected in both cases.
    – NickD
    Apr 24, 2022 at 0:15
  • 1
    Thanks @NickD - it occurred to me to check on another OS, and I can't reproduce this on Ubuntu 20.04 with emacs 28.1... so apparently this is a MacOS-only problem. I wonder if anyone else can reproduce on a mac. Apr 24, 2022 at 2:38
  • I can confirm same behaviour with Emacs 28.1, MacOS 12.3.1 and Python 3.10.1
    – Ian
    Apr 25, 2022 at 13:24

2 Answers 2

0

Ok, after learning a great deal about how to (clumisly) trace elisp code, I identified this as a bug in prompt detection in python.el: apparently the startup message for the python interpreter is not being recognized. Launching the interpreter with python -q suppresses the prompt, but the variable python-shell-interpreter-args does not appear to be respected. So the brute force solution is to advise the function that sets up inferior-python-mode to add -q:

(defun advised-python-shell-make-comint (orig-fun &rest args)
  (setq args (append '("python3 -q") (cdr args)))
  (apply orig-fun args))

That's it. Not a great long term solution, and hopefully this will get fixed in python-mode.

0

I was running into the same problem.

macOS 10.15.7, GNU Emacs 28.1, Doom Core 3.0.0-dev, Doom Modules 22.07.0-dev

And using pyenv with the +pyenv Doom Emacs setting.

My process was to run org-babel-execute-buffer(C-c C-v b) which let me C-g out of the 'executing python codeblock...' status. And then all the following codeblock executions with RET and C-c C-c ran correctly.

I have #+PROPERTY: header-args-python :session in my header, which I always ran separately from the blocks. So in short, (C-c C-v b) fixed it for me.

Hope this helps.

EDIT: Forget about my original answer, I think it was a coincidence. This really fixes it for me: rename the session to the current python process' buffer. So if the buffer is named "Python", set the session name like :session Python

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.