Feature request: open a notebook without running kernel

[ibobak]

We are working with notebooks which use PySpark kernel. Very often we need just to look inside the notebook, but we don't need to run it. But if I click on the notebook, it starts a kernel as well, but PySpark kernel is very heavy and it consumes resources: after waiting for several minutes until the notebook starts, I am doing "shutdown kernel".

It would be very fine if there was a button in the notebook list "View without running kernel".

[gppeixoto]

+1 for this feature, looks like it hasn't gotten much love for almost a year. Any updates on this?

[slidenerd]

I triple this!

[grisaitis]

Use cases

• browsing notebooks with no intent of running them
• editing non-kernel-related content, like markdown
• running notebooks with different kernels (e.g. kernel versions (e.g. py 3.7 vs 3.6), package versions)
• running multi-kernel notebooks
• when you're browsing notebooks on a non-dev machine or environment, and using a kernel would be doomed anyway

UX ideas:

• a "read only" option when right clicking on a notebook on the main page (file browser view)
• a "open without kernel" option

I'm surprised this isn't already possible. Maybe a plug-in implements it?

[rocconnick]

+1

[Zaky7]

+1

[thincal]

+1, editing / viewing without a runtime just like the Google Colab.

[nareshsankapelly]

+1, this will be a useful feature.

[amangarg96]

+1 Would help a lot, specially when using PySpark kernels as it takes some time to launch

[yucai]

+1, surprise it is not ready yet. Sometimes, just want to view the content.

[JimHBeam]

+1 would be usefull

[prgx-erodri02]

Agree with @grisaitis There are so many use cases for opening a notebook without the intent to run any cell. It would be great to have an "Open with no Kernel" option in the context menu

[JimHBeam]

How about always open a notebook without a kernel until you try to run it, then fire up the kernel? This could be optional setting for opening behaviour, not sure how this would affect multi-kernal notebooks mentioned above, didnt know that was a thing

[thavidu]

+1 would love this to be a setting which we could customise the default for

[ostrokach]

One option for editing a notebook without starting a kernel is to use nteract. This does not fully address this issue, but may be helpful to some people.

[LukaPitamic]

+1

my case: workspace with 3 notebooks, where only one is really run, two others are imported in the first one as code

[ibobak]

[ostrokach]

Another option is nbviewer.

I have found that the easiest way to start a local server is using docker:

docker run -p 8080:8080 -v $(realpath .):/localfiles jupyter/nbviewer python -m nbviewer --port=8080 --localfiles=/localfiles

We can then navigate to http://localhost:8080/localfile and click on the name of a notebook in the current working directory to open a preview.

[Zsailer]

Thanks @ibobak! This is a great idea.

Today in JupyterLab and in the coming Notebook v7, there is an option to switch from a running kernel to "No Kernel"; however it does not currently allow you to open a notebook with "No Kernel" directly. We could add an option to open with "No Kernel" directly from the tree-view/file-browser.

We think we will be able to solve this issue with the Notebook v7 release, but it won't be a blocker for this release.

[LukaPitamic]

funally, that would be super awesome. Now loading a set workspaces with a lot of notebooks all kernels are getting initiated.

[rickwierenga]

Has anyone built this yet?

[jtpio]

This should work in the latest Notebook 7 pre-release, using the --LabServerApp.notebook_starts_kernel=False option:

1. pip install notebook --pre
2. jupyter notebook --LabServerApp.notebook_starts_kernel=False

https://github.com/jupyter/notebook/assets/591645/1c7ef194-62e2-4ec3-a39c-dbfd0ec3ffa7