Skip to content

Latest commit

 

History

History
113 lines (79 loc) · 2.66 KB

README.md

File metadata and controls

113 lines (79 loc) · 2.66 KB
Raw

comma-python

A persistent Jupyter python kernel for your shell

Usage:

Execute command (and start a kernel if one isn't running):

,python [command]

Kill running kernel:

,python --kill

New kernel (and kill any existing):

,python --new

New kernel (and kill any existing) and execute command:

,python --new [command]

Jump into an console of the current kernel:

# for an ipython-like console in your terminal
,python --console

# for a graphical qt cosnole
,python --qtconsole

Notes:

If stdin is provided, it is assigned to stdin, along with lines = stdin.splitlines().

You can further customize your kernel, e.g. run specific imports or magic commands on init:

if is_new_kernel:
    exec_km('%pylab')

For an experience a little bit closer to a Notebook cell or QtConsole prompt, try using fish shell with Alt+E to edit your ,python command:

  • Try using micro for mouse text selection and Ctrl+C Ctrl+V
  • set -U EDITOR "micro -wordwrap true -softwrap true -autosave 9999999" where Ctrl+Q exits with the command changes saved automatically

Examples:

,python "%pylab"
,python "print(arange(10))"

,python "import numpy as np"

ls -1t | ,python "
jpgs = [x for x in lines if x.endswith('.jpg')]
pairs = np.array(jpgs[:len(jpgs) // 2 * 2]).reshape((-1, 2))
for i, (a, b) in enumerate(pairs):
    print(f'{i=}, {a=}, {b=}')
"

,python "print(f'{i=}, {a=}, {b=}')"

,python "for i, (a, b) in enumerate(pairs):
    print(f'{i=}, {a=}, {b=}')
" | grep "i=3"

,python "x=1"
,python "print(x)"
# This also shows the repr():
,python "x"

# also ipython magic commands work, e.g.
,python "files = !ls"

Bugs:

Sometimes after creating a new kernel, calling km.execute(to_execute) immediately never results in a km.iopub_channel.msg_ready() returing True.

  • This can be hard to debug because it tends to only get stuck NOT using the debugger.
  • A reproducer that seems to work with a debugger sometimes is ,python --new "import numpy as np"
  • The current workaround is a sleep(1) after creating a new kernel
  • Another workaround is to only create kernels with ,python --new without another argument for a command.

Some discussion here jupyter/jupyter_client#941 (comment)

Requirements (tested with):

  • jupyter_client == 8.0.3

Development

Here's a snippet to access the kernel manager from python

import jupyter_client
cf = jupyter_client.find_connection_file('shell_kernel.json')
km = jupyter_client.BlockingKernelClient(connection_file=cf)
km.load_connection_file()

It may be handy to jump into a REPL of the kernel, e.g.

jupyter console --existing shell_kernel.json