Continuous Integration and Continuous Delivery (CICD) tool library for integration testing, deployment and releases of R and Python codebases for clinical reporting
In 2023 PHUSE released the Open Source E2E guidance, providing high level guidance and talking points for using and collaborating on open source code, with a focus on R packages. This guidance referenced several tools, but stops short of creating an annotated library of Continuous Integration and Continuous Delivery (CICD) tools to make developers more efficient and code more robust.
We believe there remains a gap between the numerous tools across gh-actions (a CICD syntax used with Github), R packages and general recommendations, and what is most relevant to late stage drug development packages. Work has been done within the pharmaverse to bring the CICD tooling used in Nest to support other package families like admiral, but this is largely driven based on principles defined and deployed initially by Roche. There is a fundamental gap in terms of having an inclusive company-independent set of guidance and off the shelf tooling to apply CICD in the specific context of R and python packages used in clinical reporting. This is an area where homogeneity would not only improve the reliability of our code and productivity of our developers, but would also help set the industry on a trajectory that would allow better integration with pan- company efforts to validate packages used for clinical reporting.
Landscape of what CICD is used today across the pharmaverse, and key pre-named reference packages outside the pharmaverse used in clinical reporting (e.g. ggplot2, survminer, etc)
- Summarise integration deployment and release pain points within these R packages
- Connect with the R Validation Hub and vendor tooling (e.g. Atorus openvalidate) to understand what key devops tooling could be applied to packages to accelerate validation
- Deliverable 1: Summarise the state of DevOps in the pharmaverse today, with proposals on key tools used already, and where there are gaps.
- Deliverable 2: Develop the existing CICD library used in some pharmaverse R packages (referenced earlier in this proposal) to:
- Deliver a recommended suite of CICD tools for R and Python packages that improve developer efficiency, and ensure code is well tested and consistent when merged to branch.
- Provide off the shelf configurations to use tools like managed docker images locally or via Github codespaces to simplify environment management
- 0-6 months - Prepare Deliverable 1, aiming for abstract ready for PHUSE US Connect 2024 (assumption close is September 2023)
- At 8 months – V1.0 of suite/library of CICD tools released
- 8-12 months - Support adoption across pharmaverse and biostatistics packages where maintainers are willing to collaborate
- 12 months - Close our V1.0 with hackathon in lead up to PHUSE CSS 2024
- Dinakar Kulkarni (@cicdguy)
- Ben Straub (@bms63)
- James Black (@epijim)