get the original file name from the temp file name

[lrlunin]

Hello,

first of all I would like to thank you for brining so much effort to build such a useful and customizable tool for python notebooks!

I have faced a little issue using the --pipe option. While I convert the .py file to .ipynb I want to append a "open in colab" badge which refers to the notebook in the GH repository.

To do so I use the following pipeline:

  - repo: https://github.com/mwouts/jupytext
    rev: v1.16.4
    hooks:
     - id: jupytext
       name: convert .py to .ipynb
       args:
         - --update
         - --pipe
         - "python scripts/insert_badge.py {}"
         - --to
         - "../notebooks//ipynb"
         - examples/scripts/*.py

while in insert_badge.py I have something like:

import sys

temp_filename = sys.argv[1]
...

The temp filenames look like original_filename_randomsuffix.py and I thought I can just rstrip the the temp_filename and take the substring left from last _ to get the basename. However, once I had a suffix of tempfile which contained another underscore in a way the tempfile was like: original_filename_randomsuffix_randomsuffix.py.

Is there any workarounds to get the original name? A possible sollution would be to basically set the suffix for the tempfiles to be constant or either restrict an underscore (really dirty mess).

[lrlunin]

Okay, this was much more interesting thing then I have expected.

I have expected that python relies on the standard libc and creates the random file with the posix mkstemp() function which generates a temp file with the template filenameXXXXXX where X can be an any alphanumeric (A-Za-z0-9) character.

However, in the reality they created their own posix-like random filename generator in the tempfile https://github.com/python/cpython/blob/8b24d042b4d23700f873c23f78cee0972ad24f62/Lib/tempfile.py#L140

This contains only lowcase letters, also numbers but also includes an "_" in the list for the allowed characters. This is really not clever to make the behaviour expected to be simillar to mkstemp but then make it not posix conform.

What would you think about to change "_" to any other character ("@", "^" or even ".") which cannot be possibly be within the tempfile name?

jupytext/src/jupytext/cli.py

Line 1181 in 940fce4

prefix = prefix + (" " if " " in prefix else "_")