Add check_nm_dataset() for automated NONMEM dataset validation

[roninsightrx]

Summary

• Adds check_nm_dataset() — a comprehensive validation function that runs 35+ automated checks on NONMEM datasets
• Returns a tidy data.frame with check name, description, and result (PASS/FAIL/NA) without throwing errors
• Checks cover: column naming & reserved names, EVID/AMT/DV/MDV/CMT consistency, TIME ordering, ADDL/II overlap detection, exclusion flags, covariate missingness, numeric precision, CSV safety, and more
• Includes 31 tests covering all check categories
• Version bumped to 0.0.0.9005

Test plan

• devtools::test(filter = "check_nm_dataset") — 31/31 passing
• Manual testing with real NONMEM datasets

🤖 Generated with Claude Code

[Copilot]

Pull request overview

This PR introduces a new exported validation utility, check_nm_dataset(), to run a broad suite of automated consistency/sanity checks on NONMEM-style datasets and return a tidy PASS/FAIL/NA summary table without throwing errors.

Changes:

• Added check_nm_dataset() implementation with ~35+ checks and optional verbose logging.
• Added a dedicated testthat file with 31 unit tests covering the new check categories.
• Exported the function, generated Rd documentation, and bumped the package version.

Reviewed changes

Copilot reviewed 5 out of 5 changed files in this pull request and generated 7 comments.

Show a summary per file

File	Description
R/check_nm_dataset.R	New dataset validation function implementing the check suite and verbose reporting.
tests/testthat/test-check_nm_dataset.R	New unit tests validating expected PASS/FAIL behavior for each check.
NAMESPACE	Exports check_nm_dataset.
man/check_nm_dataset.Rd	Generated documentation for the new function.
DESCRIPTION	Version bumped to 0.0.0.9005.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

Column handling is inconsistent: some checks use cols <- toupper(names(data)) (case-insensitive), but has_col() and most downstream accessors use case-sensitive names(data) / data$TIME / data$EVID. This can cause required columns to be reported as present while subsequent checks are silently skipped (or covariates misclassified) when inputs use lowercase/mixed-case column names. Consider normalizing names once (e.g., map to uppercase internally) or implementing case-insensitive lookup helpers used consistently throughout the function.

[Copilot]

max_len <- max(nchar(ids)) will return NA if any ID is missing, and the subsequent if (max_len <= max_id_digits) will error with “missing value where TRUE/FALSE needed”. Since the function promises to not throw errors, this should use na.rm = TRUE (and handle the all-NA case explicitly) or validate/flag missing IDs before computing max_len.

Suggested change

	max_len <- max(nchar(ids))
	log_check(
	"id_digits",
	paste0("ID variable has <= ", max_id_digits, " digits (longer may need $TABLE FORMAT)"),
	if (max_len <= max_id_digits) "PASS" else "FAIL"
	)
	non_missing_ids <- ids[!is.na(ids)]
	if (length(non_missing_ids) == 0) {
	log_check(
	"id_digits",
	paste0("ID variable has <= ", max_id_digits, " digits (longer may need $TABLE FORMAT)"),
	"NA"
	)
	} else {
	max_len <- max(nchar(non_missing_ids))
	log_check(
	"id_digits",
	paste0("ID variable has <= ", max_id_digits, " digits (longer may need $TABLE FORMAT)"),
	if (max_len <= max_id_digits) "PASS" else "FAIL"
	)
	}

[Copilot]

addl_overlap detection iterates dosing records in their current row order (dose_idx), but overlap is a time-based concept and the dataset may be unsorted (you separately check data_ordered). As written, an unsorted dataset can yield false positives/negatives for overlap. Consider sorting sub by TIME (and possibly by additional tie-breakers) before computing overlaps, or only running this check when the ordering check passes.