Canonical real-data validation contracts
canonical_real_data.RdDefines licensed, checksum-pinned canonical datasets; materializes them only from explicit sources; verifies their bytes deterministically; compares numerical results against versioned external tools; and writes machine-readable validation evidence.
Usage
new_canonical_dataset(id, version, title, license, citation, files,
organism = "unspecified", analyses = character(), metadata = list())
validate_canonical_dataset(x)
verify_canonical_dataset(descriptor, directory, strict_size = TRUE)
materialize_canonical_dataset(descriptor, destination, source_dir = NULL,
allow_download = FALSE, quiet = TRUE)
compare_external_results(observed, reference, id_cols, value_cols,
tolerance = 1e-8, tool, tool_version)
write_canonical_validation_evidence(descriptor, directory, output_dir,
comparisons = NULL)Arguments
- id
Stable dataset identifier.
- version
Dataset release version.
- title
Human-readable dataset title.
- license
SPDX identifier or explicit license label.
- citation
Dataset citation.
- files
Data frame containing
filename,sha256, and optionalsize_bytesandsource.- organism
Organism label.
- analyses
Supported analysis identifiers.
- metadata
Additional named metadata.
- x,descriptor
A canonical dataset descriptor.
- directory
Directory containing materialized files.
- strict_size
Whether declared file sizes are enforced.
- destination
Destination directory.
- source_dir
Optional local mirror directory.
- allow_download
Whether explicit network downloads are permitted.
- quiet
Suppress download progress.
- observed,reference
Observed and external-reference result tables.
- id_cols
Columns used for deterministic alignment.
- value_cols
Numeric columns to compare.
- tolerance
Absolute numerical tolerance.
- tool,tool_version
External-tool provenance.
- output_dir
Evidence output directory.
- comparisons
Optional external-comparison table.