IRC channel logs

<PurpleSym>Ah no, we cannot use esbuild, because it does not bundle css into the same JavaScript file, which RStudio depends on.

<rekado>so then the upgrade is ready to merge?

<PurpleSym>I want to make sure the desktop version works as well.

<PurpleSym>And then I guess then we can merge the change to node-build-system (CI looks good) and the new RStudio.

<PurpleSym>Strange that it rebuilt ungoogled-chromium 🤨

<rekado>when you upgraded all the node packages, did you use some kind of automation?

<rekado>or was that literally just the result of running the npm importer a few times and tweaking its output?

<PurpleSym>I just ran the importer on every dependency in package.json.

<PurpleSym>Sorry, no magic ☹️

<PurpleSym>The desktop version is showing an annoying update prompt.

<rekado>we can patch that out

<PurpleSym>Yeah, looking into that.

<rekado>I only just noticed this huge pull request: https://github.com/guix-science/guix-science/pull/25

<PurpleSym>I’ve seen it, but it doesn’t look complete. There’s lots of todo’s in there.

<PurpleSym>Alright, I finally pushed the RStudio update.

<PurpleSym>Thanks rekado !

<rekado>thank you!

<rekado>sometimes I’m thinking about starting a fork of RStudio that doesn’t use libR.so but talks to an independent R REPL like ESS does.

<rekado>reusing all of that Java stuff for the UI

<rekado>but replacing all the weird session handling

<rekado>and then add a feature to launch a new R session within a Guix container.

<civodul>o/

<civodul>is there a preprint of https://ashpublications.org/blood/article-abstract/140/23/2500/486351/Azithromycin-promotes-relapse-by-disrupting-immune?redirectedFrom=fulltext somewhere?

<zimoun>civodul: no. Or I am not aware, that’s not the spirit of this medical field, sadly.

<zimoun>I do not remember which paper corresponds to which Git repo. Well, for instance: https://gitlab.com/nivall/azimut-blood

<civodul>ok

<rekado>nothing on biorxiv

<civodul>so i guess we won't add it to the "Examples" section of reproducible-paper-guide.md

<civodul>(i believe Inserm, CNRS, etc. all require that papers be first posted on a preprint server)

<zimoun> https://www.sciencedirect.com/science/article/pii/S0006497122010734

<zimoun>Is it not available for you?

<zimoun>(My colleagues tell me that it’s not a requirement but only a strong recommandation.)

<civodul>ah yes, this one works, thanks!

<civodul>it only has excerpts though, IIUC

<civodul>i was looking for an "availability" or "reproducibility" section, which i cannot see here

<zimoun>ahah! you are so naive about “medical paper” :-)

<zimoun>it’s the complete paper.

<zimoun>And what you are looking for is the “supplemental Methods” which is not part of the paper.

<zimoun>if you download the compressed archive containing this “supplemental Methods”, you have a section “Data and code Sharing”

<zimoun>And this section points to https://archive.softwareheritage.org/browse/directory/16a829f98f9d8d707343b3b48bc13f5d641998d9/

<zimoun> https://archive.softwareheritage.org/browse/directory/cc8aec0e0c85d7ea70979581e766353d3e2fec22/

<zimoun> https://archive.softwareheritage.org/browse/directory/68ec62f3d8f0cc9f42a15449f75844df12f65934/

<zimoun>which are the description of “availability” and “reproducibility”, somehow.

<civodul>hmm ok!

<civodul>not what i expected :-)

<zimoun>yeah, how make stuff complicted when they could be simpler. :-)

<civodul>hey

<civodul>i'm looking for an example of a paper with a long list of package name/version pairs

<civodul>anyone has something like that at hand?

<civodul>rekado, zimoun maybe?

<zimoun>no, I do not have at hand. And often, people do not publish that. That just publish a couple of name/version pairs; letting the rest as an excercise for the reader. :-)

<rekado>civodul: here’s one with a table containing just the names: https://www.sciencedirect.com/science/article/pii/S1046202313004180#t0015

<rekado>I don’t think this is terribly common

<civodul>you mean it's less common than providing nothing?

<rekado>in bioinfo papers they often just point to a git repo, mention the programming language, and the dependencies are not explicitly mentioned

<civodul>ok

<rekado>this one has a long list of dependencies (names, versions, method of installation) in the supplementary files: https://academic.oup.com/bioinformatics/article/39/4/btad152/7100339#supplementary-data

<rekado>method of installation is “Conda”, “Manual Download”, or “Manual Registration and Download” :)

<rekado>this one has a terrifying “Installation” section with LD_LIBRARY_PATH and conditionals: https://www.frontiersin.org/articles/10.3389/fchem.2021.722087/full

<rekado>e.g. “If after installation of pyopenms, an error message comes up informing that a library cannot be loaded, ‘openssl’ library needs to be installed as well.”

<rekado>it’s wild out there.

<rekado>most often I encounter just a link to a repo on github, without commit. The claim is that the repo describes all installation instructions and dependencies in sufficient detail.

<civodul>looks fun :-)

<rekado>in reality these instructions are often absent or relegated to “pip install”

<civodul>starting to look good: https://gitlab.inria.fr/guix-hpc/website/-/blob/master/drafts/reproducible-paper-guide.md

IRC channel logs

2023-06-22.log