background-zmq-ipython

Run some IPython ZMQ kernel in the background, without an interactive shell. You can connect to that kernel remotely via ZMQ.

Installation (package is on PyPI <https://pypi.org/project/background_zmq_ipython/>_)::

pip install background_zmq_ipython

Usage::

import background_zmq_ipython
background_zmq_ipython.init_ipython_kernel()

The kernel will then run in the background in its own thread. The init will print some message similar to this on stdout::

To connect another client to this IPython kernel, use:
jupyter console --existing kernel-1111.json

Now, after you connected, how to access globals from the main module? E.g. let's say there is a variable a.

• sys.modules["__main__"].a should work. Via sys.modules, you can access all the imported modules. Or just do import mod again.

• Get the stack of all/other frames via sys._current_frames(), and there you can access the globals from other threads as well.

• You can provide user_ns to init_ipython_kernel, e.g::

  init_ipython_kernel(user_ns=globals())

  or::

  init_ipython_kernel(user_ns={"main_globals": globals()})

  or::

  init_ipython_kernel(user_ns={"a": a})

Alternatives / related links:

• pydbattach <https://github.com/albertz/pydbattach>_
• Twisted SSH <https://crochet.readthedocs.io/en/stable/introduction.html#ssh-into-your-server>_
  (example code <https://github.com/msabramo/pyramid_ssh_crochet/blob/master/pyramid_ssh_crochet.py>__)
• IPython.embed_kernel in a background thread
  (example code <https://github.com/msabramo/pyramid_ipython_kernel/blob/master/pyramid_ipython_kernel.py>__).
  This has some issues
  (e.g. here <https://github.com/ipython/ipython/issues/4032>_;
  messing around with sys.stdout etc).
• This code was introduced in
  this StackOverflow question <https://stackoverflow.com/questions/29148319/provide-remote-shell-for-python-script>,
  and also discussed in this IPython GitHub issue #8097 <https://github.com/ipython/ipython/issues/8097>,
  but it has become outdated, so this project provides a rewrite / updated code,
  and the goal was also an easy to install pip package.