Skip to content

CLAUDE_CHANGELOG.md

Latest commit

 

History

History
37 lines (26 loc) · 4.25 KB

CLAUDE_CHANGELOG.md

File metadata and controls

37 lines (26 loc) · 4.25 KB
Raw

Claude Changelog

Record of all changes made by Claude Code in this repository.

2026-04-21

  • Added SIF auto-download to tests/test.sh and validation/validate.sh — both scripts now check for the SIF file at $CONTAINER_PATH/bioc_cosia_1.10.0.sif before running. If missing, they pull it via singularity pull from docker://lizzyr/bioc_cosia:1.10.0. The source URI can be overridden with the COSIA_SIF_URI environment variable.

2026-05-05

  • Bumped version DESCRIPTION: 1.11.1 → 1.11.2. 1.11.1 is the current Bioconductor production version (with the CV_Tissue bug); 1.11.2 is the fixed release.
  • Revised validation scripts — clarified that 1.11.1 is the production version, not a Bioconductor release from a separate branch. Rewrote all three scripts to cover all six metric types. 01_run_bioc.R: tests all metrics against 1.11.1, expects CV_Tissue to crash, captures reference values for working metrics. 02_run_dev.R: tests all metrics against 1.11.2, all should succeed. 03_compare.R: three-part strategy — (1) confirm CV_Tissue crashes in 1.11.1 and succeeds in 1.11.2 with structural checks + cross-species consistency check; (2) exact value match for unchanged metrics (CV_Species, DS_*); (3) manual spot-check instructions.

2026-04-23

  • Added GitHub Actions workflow (.github/workflows/test.yml) — runs the test suite on every push and pull request to devel and main. Uses the official bioconductor/bioconductor_docker:RELEASE_3_22 container. Caches R packages (keyed to DESCRIPTION) and ExperimentHub data (static key since datasets EH7858–EH7863 are versioned) to minimise run time.

  • Added validation scripts (validation/) — three scripts for comparing Bioconductor release vs dev outputs to verify correctness of the CV_Tissue merge rewrite: 01_run_bioc.R, 02_run_dev.R, 03_compare.R. Tests single-species (must match exactly), two-species, and three-species cases. Includes manual spot-check instructions.

  • Updated .Rbuildignore — added validation/, CLAUDE.md, and CLAUDE_CHANGELOG.md so they are excluded from the package tarball.

2026-04-22

  • Created CLAUDE.md — initial guidance file documenting project overview, common commands, architecture, key files, and conventions.
  • Removed deprecation markers — CoSIA was listed on a Bioconductor deprecation candidate list but was never actually removed; it remains active in Bioconductor 3.22 at v1.10.1. Removed the .Deprecated() call from R/zzz.R and PackageStatus: Deprecated from DESCRIPTION.
  • Created unit test suite — added tests/testthat.R and five test files covering the full public API:
    • tests/testthat/helper-CoSIA.R — shared fixtures and skip_if_no_experimenthub() helper
    • tests/testthat/test-CoSIAn-constructor.R — 11 tests for constructor validation (offline)
    • tests/testthat/test-viewCoSIAn.R — 4 tests for the viewCoSIAn accessor (offline)
    • tests/testthat/test-getConversion.R — 5 tests for gene ID/ortholog conversion; 4 run offline via annotationDBI + HomoloGene, 1 skips without network (NCBIOrtho)
    • tests/testthat/test-getGEx.R — 5 tests for ExperimentHub expression data retrieval (skip if offline)
    • tests/testthat/test-getGExMetrics.R — 9 tests covering all 6 metric types (skip if offline)
  • Updated DESCRIPTION — added Config/testthat/edition: 3 to declare testthat edition 3.
  • Updated CLAUDE.md — added Claude Changelog section requiring this file to be updated with every Claude session; corrected stale note about zzz.R.
  • Fixed viewCoSIAn bug (R/CoSIA-utility-method.R) — replaced ifelse() with if/else. ifelse() is vectorized and strips S3 attributes, causing it to return a bare scalar instead of the data frame slot value.
  • Fixed CV_Tissue multi-species bug (R/CoSIA-getGExMetrics-method.R) — rewrote the merge loop that accumulated the growing result back into itself on each iteration, creating a Cartesian product with cascading .x/.y duplicate columns and a broken Anatomical_entity_name column (also contained a typo "Anatomical_enti_name.y"). Replaced with a per-species independent left join against the original id_dataframe, combining results by joining on all ID columns + Anatomical_entity_name.