extensions configured twice?

[flying-sheep]

Hi, in my logs I get:

nbextension 'highlight_selected_word/main' has duplicate listings in both
'$HOME/.local/share/jupyter/nbextensions/highlight_selected_word/configurator.yaml' and
'$HOME/.local/share/jupyter/nbextensions/highlight_selected_word/configurator.yaml'

for each extension. (i added linebreaks for clarity)

but i have no configuration that could add another copy of that path AFAIK.

$ ag nbextensions ~/.jupyter
$HOME/.jupyter/jupyter_notebook_config.json
4:      "jupyter_nbextensions_configurator": true

$HOME/.jupyter/jupyter_notebook_config.py
224:# c.NotebookApp.extra_nbextensions_path = traitlets.Undefined

$HOME/.jupyter/jupyter_nbconvert_config.json
5:      "jupyter_contrib_nbextensions.nbconvert_support.CodeFoldingPreprocessor",
6:      "jupyter_contrib_nbextensions.nbconvert_support.PyMarkdownPreprocessor"
11:      "$HOME/.local/lib/python3.5/site-packages/jupyter_contrib_nbextensions/templates"
15:    "postprocessor_class": "jupyter_contrib_nbextensions.nbconvert_support.EmbedPostProcessor"

$HOME/.jupyter/nbconfig/tree.json
3:    "nbextensions_configurator/tree_tab/main": true

$HOME/.jupyter/nbconfig/notebook.json
6:    "nbextensions_configurator/config_menu/main": true,

[jcb91]

Hmm, strange, particularly since they are both the same path. Seems that somehow the jupyter path contains duplicate entries, something which presumably happens from a jupyter config file somewhere else in the path. I'd suggest trying

jupyter --paths

to find the full list of config directories on your system, then look in each of those for .py or .json config files

[jcb91]

did you make any progress on this, @flying-sheep?

[paulperry]

I have this problem, but I really do have duplicate paths:
nbextension 'autosavetime/main' has duplicate listings in both '/Users/paulperry/anaconda3/share/jupyter/nbextensions/autosavetime/autosavetime.yaml' and '/Users/paulperry/Library/Jupyter/nbextensions/autosavetime/autosavetime.yaml'
my paths look like:
(root) ~>jupyter --paths

config:
    /Users/paulperry/.jupyter
    /Users/paulperry/anaconda3/etc/jupyter
    /usr/local/etc/jupyter
    /etc/jupyter
data:
    /Users/paulperry/Library/Jupyter
    /Users/paulperry/anaconda3/share/jupyter
    /usr/local/share/jupyter
    /usr/share/jupyter
runtime:
    /Users/paulperry/Library/Jupyter/runtime
(root) ~>

I see nbextension directories both under Library and anaconda3 . How do I resolve this conflict? How do I control the jupyter paths?

[jcb91]

@paulperry so, the reason this can happen is that there are multiple places that nbextensions can be installed, as noted in your jupyter --paths output (the server looks for them in any nbextensions subdirectory of any of the data directories). Essentially there's a per-user directory /Users/paulperry/Library/Jupyter, a per-python (so per-environment) sys-prefix directory /Users/paulperry/anaconda3/share/jupyter, and one or more system-wide directories (/usr/local/share/jupyter, /usr/share/jupyter).

How do I control the jupyter paths?

If you installed autosametime using the jupyter_contrib_nbextensions conda package (as opposed to using pip, running python setup.py, or jupyter nbextension ...), then the installation is performed into the conda sys-prefix directory automatically as part of a post-link script, so you don't get a choice. If you use another install method, then you need to pick where to install yourself, using one or none of the flags --user, --sys-prefix, or --system (see step two of the install readme for details). The install docs currently advise the user directory (--user flag) by default, which is where your /Users/paulperry/anaconda3/share/jupyter copy came from.

How do I resolve this conflict?

To fix this, you can just remove one of the installations. If you're mostly using anaconda, and run your notebook server from the same environment each time, I would advise using the sys-prefix installation over the user one, particularly since any conda package installation will go there. However, you can keep either, it doesn't matter. To remove the jupyter_contrib_nbextensions install from the user directory, you can use

jupyter contrib nbextensions uninstall --user

or equivalently to remove the sys-prefix version,

jupyter contrib nbextensions uninstall --sys-prefix

If you didn't use the jupyter_contrib_nbextensions package, you can uninstall each duplicated nbextension individually, using commands of the form

jupyter nbextension uninstall --user path/url path/url/entrypoint

e.g.

jupyter nbextension uninstall --user autosavetime autosavetime/main

Does that make sense?

[flying-sheep]

yeah, for me it’s:

config:
    /home/$me/.jupyter
    /home/$me/.local/etc/jupyter
    /usr/local/etc/jupyter
    /etc/jupyter
data:
    /home/$me/.local/share/jupyter
    /home/$me/.local/share/jupyter
    /usr/local/share/jupyter
    /usr/share/jupyter
runtime:
    /run/user/127365/jupyter

note the duplicate ~/.local/share/jupyter

[jcb91]

note the duplicate ~/.local/share/jupyter

Yes, strange, and it seems you're missing the sys-prefix location too 🤔

[jcb91]

@flying-sheep it looks suspiciously like your sys.prefix is set to /home/$me/.local/. Is this perhaps related to ipython-contrib/jupyter_contrib_nbextensions#710? Not sure. Could you test

import sys
print(sys.prefix)

[flying-sheep]

you’re right. that’s how i compiled python! ./configure --prefix="$HOME/.local"

i’m doing this because it puts things into the correct user-wide places.

[jcb91]

ah! Wait, so --prefix=$HOME/.local, but sys.prefix == '/usr'? I'm confused 😕

[flying-sheep]

no i deleted that comment as i was wrong (forgot to ssh into the right machine).

--prefix=$HOME/.local → Path(sys.prefix) == Path.home() / '.local'

[jcb91]

ah, ok, I didn't refresh the page, so got both for a while!

So in that case, it might be worth the configurator only checking paths which it hasn't already checked, to cater for duplicates from jupyter_core.paths...

[jcb91]

I've made PR #35 to hopefully deal with this occurrence @flying-sheep. If you get a chance to try it, let me know if it works ok for you 😄

[flying-sheep]

i bet it’ll work just fine. i just think that stackoverflow code is a bit too clever to easily understand.

[jcb91]

🤔 yes, it's not that clear. I'd go for a set, but the ordering can be significant. Perhaps I'll rethink it later...