Derive measurement_type → contributing datasets from the data

Scans every measurement-bearing table (those with a measurement_type column) for its distinct measurement types and unions in the dataset(s) that own the table, yielding the true one-to-many map (e.g. temperature reported by both calcofi_bottle and calcofi_ctd-cast). Feed the result to merge_metadata_json() via its measurement_datasets argument so the schema site's per-measurement dataset filter reflects reality rather than the single "first-defined" dataset.

Usage

derive_measurement_type_datasets(con, table_datasets, from_fn = identity)

Arguments

con

A DBI connection (e.g. from get_duckdb_con()). Reads happen via from_fn(table) so the same helper works against in-DB tables or remote parquet over httpfs.

table_datasets

Named list, table -> character vector of provider_datasets. Typically built from the ingest YAML tables_owned or the metadata contributions block.

from_fn

Function mapping a table name to a SQL FROM source. Defaults to the identity (the table exists in con); for remote parquet pass e.g. function(t) sprintf("read_parquet('%s/%s.parquet')", base, t).

Value

Named list, measurement_type -> sorted unique character vector of provider_datasets.

Details

Tables in table_datasets that lack a measurement_type column are skipped silently. Exclude the measurement_type vocabulary/lookup table from table_datasets — its measurement_type column is the term list, not measured rows, so including it would mis-attribute every term to every contributing dataset.

Examples

if (FALSE) { # \dontrun{
td <- list(bottle_measurement = "calcofi_bottle",
           ctd_thin           = "calcofi_ctd-cast")
derive_measurement_type_datasets(con, td)
} # }