Build the CF/ACDD global attributes for a published dataset

Derives the file's self-description from the ingest notebook's calcofi.dataset_meta YAML — the same block that feeds the release dataset table — so the netCDF, the database and the schema site cannot describe the same dataset differently. Per-file overrides go in calcofi.netcdf.

Usage

nc_global_atts(
  dataset_key,
  dataset_meta = list(),
  release,
  shape = c("profile", "groups"),
  cf_scope = NULL,
  workflow_url = NULL,
  date_created = NULL,
  extra = list()
)

Arguments

dataset_key

Provenance stamp, e.g. "calcofi_ctd-cast".

dataset_meta

The ingest's dataset_meta list (dataset_name, description, citation_main, coverage_temporal, coverage_spatial, license, …). Missing keys simply omit their attribute. A title or description here wins, so a caller overrides the derived text by merging its calcofi.netcdf block over dataset_meta before passing it.

release

Database release the file was built from, e.g. "v2026.07.30".

shape

"profile" or "groups", from plan_dataset_netcdf().

cf_scope

Override for the honesty statement about CF coverage; defaults to text appropriate to shape.

workflow_url

Rendered notebook URL, written as references.

date_created

ACDD creation date. Defaults to the release date parsed from release, deliberately: see Details.

extra

Named list of additional or overriding global attributes.

Value

Named list of global attributes, ready for ncdf4::ncatt_put(nc, 0, …).

Details

Why date_created is the release date, not Sys.time(). The publisher skips re-uploading a file whose sha256 matches an earlier release (release-named paths, bytes written once). A wall-clock date_created puts a fresh timestamp inside every build, so no rebuild is ever byte-identical and that check can never fire — it silently degrades to "always re-upload". Tying the attribute to the release makes a rebuild of the same release reproducible, which is also the more useful claim: it dates the data product, not the run.