Skip to contents

Changelog

Source: NEWS.md

imuRUN 0.1.0

  • Initial release.
  • Exported the core CLI engine as ordinary R functions: run_fit(), read_inputs(), load_by_ext(), find_input_file(), and check_all_inputs().
  • Added install_cli() to put the bundled imurun command on the PATH.
  • Added an .xlsx workbook input mode: read_inputs() now accepts either a directory of CSV/RDS files or a single workbook with one sheet per input, via the new read_workbook(). Missing sheets are reported all at once.
  • Pinned the canonical input schema as IMURUN_SCHEMA and shipped a blank template workbook (inst/templates/imurun_template.xlsx) plus a small filled example (inst/extdata/imurun_example.xlsx) derived from imuGAP’s *_sim data; resolve them with imurun_template() / imurun_example().
  • Added imurun_init() and imurun_copy_example() (and imurun init / imurun example subcommands) to scaffold workbooks into a directory.
  • Added validate_inputs(), a friendly validation layer over the imuGAP canonicalizers that reports every problem at once in spreadsheet terms.
  • Observations now describe an inclusive age range (age_min/age_max, labeled Youngest age / Oldest age) rather than a single age. A count covering several ages expands to one imuGAP population row per age, weighted equally, so those ages form that observation’s coverage mixture. A sheet carrying a single age column is still read, as a one-age range.
  • Renamed the observations and target cohort column from Birth cohort to Reference cohort, and documented it as the cohort of the row’s oldest age: the cohorts of younger ages are derived so that age plus cohort stays constant, matching how the target sheet is already expanded. The former Birth cohort header is still accepted.
  • run_fit() now writes a human-readable result, not just fit.rds. The target sheet is expanded, predicted, and summarized into a results sheet in the workbook the user supplied, preserving its existing sheets and formatting. --results writes an amended copy instead, and --csv writes a results-only CSV. Existing results are not replaced unless --overwrite is passed. The target and configuration sheets are validated before fitting, so a bad row or sampler setting fails immediately rather than after the model runs.
  • Generated workbooks now include a configuration sheet for iter, chains, seed, and warmup. Command-line flags remain available as explicit automation/compatibility overrides.
  • Set up the development infrastructure: R CMD check, lint, test coverage, pkgdown, and spelling continuous-integration workflows.