Bugfix: unequal allocation

[kamilsi]

What happened:

• Bug discovery: @aleksandraduda2 identified an issue with unequal allocation ratio handling in minimisation randomization.
• TDD first: I began with tests, adding 7 failing tests (failures: https://github.com/ttscience/unbiased/actions/runs/18165618890) to reproduce the bug (and added a few more that already passed to improve coverage).
• Maintenance: Performed routine GitHub Actions housekeeping to keep CI healthy.
• Fix + housekeeping: Implemented the fix, then did all versioning updates and documentation:
  • bumped package to 1.0.2
  • updated @apiversion, tests, and vignette references
  • added the 1.0.2 entry to NEWS.md
  • committed the changes together with the bugfix.

Summary by CodeRabbit

• Bug Fixes
  • Fixed minimisation randomization to correctly honor unequal allocation ratios (e.g., 3:1:1).
• Documentation
  • Updated release notes and references for version 1.0.2.
  • Enhanced API documentation metadata with new tags for easier navigation.
• Tests
  • Added end-to-end and unit tests validating ratio adherence, deterministic behavior, scaling invariance, and method variability.
• Chores
  • Updated CI to use environment restoration, pinned R 4.2.3, improved artifact upload, and refined linting.
  • Minor editor settings adjustments and version bumps across metadata.

[kamilsi]

@coderabbitai review

[coderabbitai]

✅ Actions performed

Review triggered.

Note: CodeRabbit is an incremental review system and does not re-review already reviewed commits. This command is applicable only when automatic reviews are paused.

[coderabbitai]

Caution

Review failed

The pull request is closed.

Walkthrough

CI workflows switch from setup-r-dependencies to setup-renv with explicit R 4.2.3, plus ad-hoc installs for needed R packages. A bug fix updates imbalance calculation in randomize_minimisation_pocock. Version bumped to 1.0.2 with NEWS, API annotation, and references updated. Additional tests validate ratio adherence and error handling.

Changes

Cohort / File(s)	Summary
CI workflows (renv migration, explicit R 4.2.3) .github/workflows/document.yaml, .github/workflows/lint.yaml, .github/workflows/pkgdown.yaml, .github/workflows/test-coverage.yaml	Replace setup-r-dependencies with setup-renv; pin R to 4.2.3; add on-demand installs for roxygen2/lintr/cyclocomp/pkgdown/covr; set Rscript shells; upgrade upload-artifact v3→v4 (coverage).
Developer tooling and lint config .lintr, .vscode/settings.json	Update lintr config: enable cyclocomp and line length, disable several linters, add exclusions for vignettes/; set VSCode r.lsp.promptToInstall to false.
Versioning and metadata DESCRIPTION, NEWS.md, inst/plumber/.../plumber.R, vignettes/articles/references.bib	Bump to 1.0.2; add NEWS entry describing minimisation ratio bug fix; update plumber API version to 1.0.2 with new tags (read, other); update BibTeX version note.
Core algorithm fix R/randomize-minimisation-pocock.R	In imbalance calculation, change operation from multiply by ratio[arm] to divide by ratio[arm].
Tests (unit and E2E) tests/testthat/test-randomize-minimisation-pocock.R, tests/testthat/test-E2E-study-minimisation-pocock.R, tests/testthat/test-error-handling.R	Add comprehensive tests for ratio adherence, determinism, scaling invariance, method variability, and E2E 3:1:1 allocation; update SENTRY_RELEASE expectation to 1.0.2.

Sequence Diagram(s)

sequenceDiagram
  autonumber
  actor Client
  participant API as Plumber API
  participant Core as randomize_minimisation_pocock()
  participant State as Study State

  Client->>API: POST /randomize (covariates, ratio, p)
  API->>State: Read current allocations
  API->>Core: Compute next arm (state, ratio, p, method)
  note over Core: Imbalance computation uses division by ratio[arm]
  Core-->>API: Selected arm
  API-->>State: Persist assignment
  API-->>Client: 200 OK (assigned arm)

Loading

Estimated code review effort

🎯 3 (Moderate) | ⏱️ ~25 minutes

Poem

I thump my paws at version two,
Ratios right, the math anew.
Pipelines hop through renv’s gate,
Linters sniff, then validate.
Tests now nibble every case—
3:1:1 falls into place.
Carrot-merge complete; ship with grace. 🥕✨

✨ Finishing touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Post copyable unit tests in a comment
• Commit unit tests in branch 106-validate-minimization-randomization-with-unequal-allocation-ratios

---

📜 Recent review details

Configuration used: CodeRabbit UI

Review profile: CHILL

Plan: Pro

📥 Commits

Reviewing files that changed from the base of the PR and between f570d20 and 951f8f3.

📒 Files selected for processing (14)

• .github/workflows/document.yaml (1 hunks)
• .github/workflows/lint.yaml (1 hunks)
• .github/workflows/pkgdown.yaml (1 hunks)
• .github/workflows/test-coverage.yaml (2 hunks)
• .lintr (1 hunks)
• .vscode/settings.json (1 hunks)
• DESCRIPTION (1 hunks)
• NEWS.md (1 hunks)
• R/randomize-minimisation-pocock.R (1 hunks)
• inst/plumber/unbiased_api/plumber.R (1 hunks)
• tests/testthat/test-E2E-study-minimisation-pocock.R (1 hunks)
• tests/testthat/test-error-handling.R (2 hunks)
• tests/testthat/test-randomize-minimisation-pocock.R (1 hunks)
• vignettes/articles/references.bib (1 hunks)

---

_{Comment @coderabbitai help to get the list of available commands and usage tips.}