R/Pharma 2023

[jmanitz]

The R/Pharma Program Committee has invited us to give a 10-minute talk on "R validation hub" at the sixth annual 2023 R/Pharma Conference to be held virtually October 24th-26th 2023.

First brainstroming resulted in the following possible topics:

• updates from each workstream.
• {riskscore}: https://github.com/pharmaR/riskscore
• the possibility to collect more data from the survey (https://bit.ly/risk_survey)
• position on python packages

[AARON-CLARK]

• {riskscore}: https://github.com/pharmaR/riskscore

@emilliman5 & @dgkf, when do you foresee us officially launching/publicizing this repo? I know we've had a few "soft openings" by teasing it at userR! Conf in Basel, and we'll no doubt be mentioning it next week at the all-hands meetings, but could/should we target Posit Conf or R/Pharma for the "Grand Opening"?

I think the package is usable in it's current format, but it only holds placeholder data for now. Ideally, it'd be nice to add the assessment data.frame and build out an infrastructure to generate for a the current and previous versions of riskmetric. And oh, we need a hex logo. What are your thoughts?

Do we need more talent contributing to this workstream? I could see if one or two people would be interested.

[AARON-CLARK]

• updates from each workstream.

I think the single most exciting update from {riskassessment} is that we now allow users to explore source code in the app! I think there'll be more updates on this closer to the conference, so we we can wait to gather screenshots & details, but here is a simple screenshot of the app (as of today) exploring the {options} pkg source code. If you want to play with it yourself, here is a link to a deployed 'dev' version: https://rinpharma.shinyapps.io/riskassessment_v100_dev/

[dgkf]

Regarding riskscore, I'd be fine using this to publicize that work and potentially recruit for it.

@AARON-CLARK - I'll also pick up on the questions you posed, but I think maybe a longer discussion around these might be better served in issues in the riskscore repo.

@emilliman5 & @dgkf, when do you foresee us officially launching/publicizing this repo?

Good question! I don't think there's anything wrong with letting people know where we're at with any necessary disclaimers. If we're looking to release as a polished product where we're fully confident in the data, though, I think there may be a few things we want to touch up. I think we briefly mentioned possible effects of API rate limits and uncaught failures being reported as NAs. If we still want to investigate these things, or do more polish passes on the package, we may consider kicking off a task backlog so we know what still needs to happen.

Do we need more talent contributing to this workstream? I could see if one or two people would be interested.

I think that this work leans heavily on knowledge of riskmetric and its update cycle, so perhaps it would make sense to find a few analytically-minded folks among the riskmetric contributors to try to get involved? We can always try to recruit more folks to riskmetric if we feel we're stretched too thin between the two.

[jmanitz]

Update on RvalHub contributions to R pharma:

Combined 20min presentation slot shared by Juliane and Coline 1) Brief updates from all workstreams and 2) More detail on happenings of the Repo group

No RvalHub workshop this year

[jmanitz]

Updates:

Title: “Updates from the R Validation Hub: Toward a Pharma Repository” Authors: Juliane Manitz and Coline Zeballos Time: Tuesday Oct 25, 11.40-12 ET

[jmanitz]

github space to create slides added: https://github.com/pharmaR/events/tree/main/Rpharma2023

[AARON-CLARK]

Hi @jmanitz, the previous image used for positconf showed the app browsing test files which is OKAY, but the app has something better now: a "Function Explorer" that allows users to effortlessly track exactly when & where that function is tested in the package. Screenshot below for function options::define_option(). Users can use this same module to browse Help documentation and source code on the same function. It's a one-stop shop.

As for the new feature we were discussing during today's exec call, I've included a screenshot below. Users can now write their own custom code to automatically categorize packages into "Low", "Medium", or "High" risk, based on {riskmetric} assessments. For example, below, we are dictating that a package should be considered "High Risk" in any of the following scenarios:

• It's testing coverage is below 60%
• It's bug closure rate is below 25%
• It has more than 30 dependencies
• Or the overall package score is above 67%

Thus, the user can upload hundreds of packages at one time, and any package that violate these rules is automatically tagged as "high risk". Pretty handy!

[jmanitz]

@AARON-CLARK, I am integrating your comments into the slides. I also will add the shiny conf award and a link to the demo. Is it possible that the function explorer has not been updated yet in the demo?

[AARON-CLARK]

Is it possible that the function explorer has not been updated yet in the demo?

Yes. We cut the new release to master that contains the Function Explorer on Wednesday but we're having some issues with it when deployed to shinyapps.io. Feel free to share the link. Once we've figured it out the issue, we'll re-deploy. I assume it will be ready for your talk on Wednesday.

[jmanitz]

Great, please check out the new slides. I hope I summarized the new features correctly

[jmanitz]

@dgkf can you deploy the slides at the shortlink https://pharmar.github.io/events-Rpharma2023 ?

[dgkf]

@jmanitz Sure, will do. I'll spin up a new repo and copy the slides and gh actions over to publish them.

[dgkf]

Done! Hosted at https://github.com/pharmaR/events-rpharma2023 (I made it all lowercase to match the other repos, also updated the link the README and in the slides. At least according to grep I got all of the references).

It automatically rebuilds when you push changes, so you should be able to make any last-minute changes and it will automatically redeploy.

[AARON-CLARK]

Great job everyone!