Skip to contents

The consent gate: no dataset reaches the public database before its provider accepts

Source: R/consent.R
check_dataset_consent.Rd

Reads each released dataset's standing consent question — the row of metadata/{provider}/{dataset}/questions.csv whose related_field is consent — and classifies it:

none

no questions.csv, or no consent row (ok: the dataset predates the consent process)

pending

the row is open or proposed: the ask is out

declined

the row is wontfix, or answered with an answer that is not a yes

given

the row is answered and the answer begins with yes, or with no ask needed (the 2026-10-22 convention for a dataset openly licensed at its public source, who = "Ben Best (licence at source)"); it is ok only when who names who gave it

Usage

check_dataset_consent(
  workflow_dir,
  datasets = NULL,
  exempt = character(),
  legacy = NULL,
  halt = TRUE
)

Arguments

workflow_dir

the workflows checkout: holds the ingest_*.qmd notebooks and metadata/{provider}/{dataset}/questions.csv

datasets

dataset keys ({provider}_{dataset}) to check; default: the datasets the release will include (in_release not false) plus exempt

exempt

dataset keys reported but never failed (a preview run's CALCOFI_RELEASE_INCLUDE)

legacy

dataset keys released before the gate existed, for which a pending ask is reported but does not fail; default: every notebook with no in_release key. Pass character() to require a given consent everywhere.

halt

stop when any dataset is not ok (default TRUE), naming each failing dataset and its state

Value

A tibble, one row per dataset: dataset_key, consent (none | pending | declined | given), label, status, answer, who, answered_date, ok, plus exempt, legacy and why (the basis of ok).

Details

A dataset is ok when its consent is none, or given with a who. A pending ask is ok for a dataset in exempt (a private preview run passes CALCOFI_RELEASE_INCLUDE: the preview is where the decision is made) and for a dataset in legacy (one that entered the public database before the gate existed; by default a notebook with no in_release key at all — un-holding a dataset leaves in_release: true in its notebook, so a dataset that came in through the preview pathway is never legacy). declined is ok only for an exempt dataset: a provider's no withdraws the data. A public run has an empty exempt, so every dataset un-held through the preview must carry a given consent before release_database.qmd proceeds.

See also

read_questions(), release_excluded_datasets()

Examples

if (FALSE) { # \dontrun{
# release_database.qmd, after ds_excluded is settled
check_dataset_consent(here::here(), exempt = release_include, halt = TRUE)
} # }