Ask quick dev questions about mybinder.org, BinderHub, and repo2docker. Use discourse.jupyter.org for user questions, support, and discussion.
Aha, I know what the problem is
It was the build cache. Instead of
COPY ., the Dockerfile uses git clone repo_url..., which wouldn't invalidate the build cache when there are pushes to the repo
another reason to avoid writing a
Dockerfile by hand when ever possible :)
Has anyone watched the video or looked at the slides from this webinar from Tuesday? It was really good. https://www.dataone.org/webinars/developing-packaging-and-sharing-reproducible-research-objects-whole-tale-approach
Hi! Is it possible to use jupyterlab >= 1.0 in binder? I came across jupyter/repo2docker#724 which says it is possible to choose, but even with a
requirements.txt with jupyterlab==1.0.0 (or >= 1.0) the container is still using 0.35 (as checked by importing jupyterlab and printing its version from the launched notebook). Please tell me if I need to provide other information.
I guess that you could create your own Dockerfile in the root of your repo, or in a binder/.binder folder. But that seems like an elephant gun. this seems to suggest what you tried.
Let me dig up examples that at least were working and we can check.
Or it might be better to point us at yours that isn't working? I have mostly converted all my repos and they run with 1.2 I think it is.
Example at: https://github.com/fomightez/muscle-binder . That is one of my latest updates.
Sorry it is
Version 1.1.4
@fomightez Thank you for the very detailed response! Using
python run runs the script through but doesn't publish the web app itself. That's an interesting note about the proxying required for other apps. I'll look more into the comments and it seems like the best next step for me would be to follow up with an issue on streamlit to figure out more about JupyterHub compatibility.
Thanks @fomightez I'll give it a try with
environment.yml. My repo is https://github.com/plotly/customdata-app.
Oh, you need
>=1.0.0 I think.
Ohh @emmanuelle . I see the problem with your repository.
You have two sets of instructions to Binder there and it is only using the one in the
Binder directory. Where you didn't specify JupyterLab>1.0.
You need to put
requirements.txt in binder directory. Or simply move postBuild to main level and delete binder directory.
@pdurbin Thanks for posting that link. There is some interesting conversation at the bottom, too.
Unfortunately it was the example I could get that I knew was working the fastest. Sorry.
hum broken line @fm75 (or I miss the credentials).
It is in the documentation for repo2docker as well if anyone doesn't want to read the code --> https://repo2docker.readthedocs.io/en/latest/usage.html#where-to-put-configuration-files .
great. Thanks a lot for the explanations. It's amazing to get such top-quality user-support!
https://github.com/jupyter/repo2docker/blob/0dba6111c6b1f0a5f0208193ca9b16891d4469e6/repo2docker/buildpacks/base.py#L503 . Sorry - was looking at our internal implementation.
does someone know what bagit-ro is? saw it mentioned in the slides for the webinar @pdurbin linked
https://github.com/whole-tale/tale_serialization_formats links to https://github.com/ResearchObject/bagit-ro so I assume that's it. A "nascent standard" according to the webinar (around the 40:00 mark). @betatim
you'll have to install
master for that
Last workshop this year https://twitter.com/arnimb/status/1184054363456704513 but first to be cloud-only ... crossing fingers