[BUG] Binder link fails with 500 error "could not start rstudio in time" on server

[jim-smith]

Describe the bug
if you click on the link to run the demos on my binder.org -- which I think we should publicise more -- it takes a while ot build the docker image (fine) but when it finally starts you get a 500 error "could not start rstudio in time"

To Reproduce
click on the link in the readme

Expected behavior
should start on my binder.org.
I assume the building part will get faster as it is use a few times

Screenshots

Desktop (please complete the following information):

• macos, safari or chrome

Additional context

I went in to my binder.org and used their gui to build a docker that went into the containerised version of the repo but didn;t specify a file or url.
that opened in a version of jupyterlab as expected, but when you click on 'Rstudio' from the launcher it fils with the same 500 error.
sort of wondering if the repo2docker part is maybe broken?- maybe need to duo another launch and then capture the loges/output from the build process

would be good to document what has been done ot fix it -maybe in contribute.md?

[jim-smith]

@rpreen I don;t know if you kept any notes last time you fixed the binder thing (10 months ago).

Can you let me know if you think it might be relatively quick fix, and if so I will delay the emails announcing new versions of acro and acro-r

[rpreen]

I had to create a .binder folder which contains files: apt.txt which are packages the Ubuntu image needs (for here, that is Python), an install.R file which specifies which packages to install (so acro, and reticulate and the farff package for reading the charities dataset) and runtime.txt which specifies a particular R version for stability. I don't see why any of those would need changing and I'm not sure there is anything wrong on our end.

If you load it without?urlpath=rstudio added to the url (so it doesn't automatically try to start RStudio) then I see the environment load fine. But if I click on the RStudio button it gives the server error you found. If I click on the R console then I can run commands fine (library("acro") acro_init(suppress = TRUE)) so I think it's a problem with Mybinder's servers.

[rpreen]

For reference, the mybinder R example repo is: https://github.com/binder-examples/r

Looks like if we do the R notebook as an ipynb we can link straight to the notebook which is a bit nicer than starting RStudio and then having to find/click it.

[jim-smith]

tldr: we may need to wait a day or two

done some digging in CRAN policies, bold is my emphasis

Policies for when a (Windows or macOS) binary package will be distributed:

• all its package dependencies on CRAN are available for that platform. Dependencies from other repositories will be installed at CRAN’s discretion.
• any external software needed can easily be installed on the build machine for all the sub-architectures: here “easily” includes not depending on specific versions, nor should the installed binary depend on specific versions.
• it passes R CMD check without error for all the available sub-architectures, or at CRAN’s discretion, for the most important sub-architecture(s).
• Binary packages are not accepted from maintainers: CRAN will only host binary packages prepared by those responsible for the binary areas. Their packages are made automatically by batch jobs and can take a day or two to appear on the CRAN master site (maybe longer to reach CRAN mirrors).

Binary packages are built for the current version of R: they may also be built for the last version in the previous series (e.g. R 3.5.3 when R 3.6.x is current) or for R-devel.

Questions about binary packages should be addressed to those responsible for building them: Simon Urbanek (macOS) and Uwe Ligges (Windows); email addresses ‘First.Lastname@R-project.org’.

[rpreen]

I noticed that mybinder was using an old version of acro (v0.1.4) because it caches/snapshots CRAN. I have changed this so it now installs acro from the current main branch (now it will install v1.0.0). However, this still does not fix the RStudio crashing - I don't know how to fix this and RStudio is garbage anyway.

I think the best solution is to make an .ipynb notebook and then we can link directly to that which is a nicer UX and github can also render the content if viewed within the repo.

To start mybinder just run it without RStudio: https://mybinder.org/v2/gh/AI-SDC/ACRO-R/main then click on "R" for notebooks (top row) and acro can be used with library("acro") So what is left to do is to port the existing .Rmd notebook into there and save/download it and commit it here. Then we can simply add the notebook to the URL like we do for Python using the ? operator https://mybinder.org/v2/gh/AI-SDC/ACRO/main?urlpath=/lab/tree/notebooks/acro_demo_2026.ipynb

[jim-smith]

However, the point about acro being out of date doesn't explain why RStudio won't start

[rpreen]

It must be a problem on their end - I get the 500 server error with their own example

As I said above, I think linking to an ipynb notebook is better anyway since it's one click and the notebook renders; even when RStudio did work, you had to locate the notebook file first.