Read metadata/measurement_type.csv, refusing a corrupted registry

The canonical measurement vocabulary, validated on the way in via check_registry_na_strings() so a write_csv(na = "NA") round trip fails here instead of silently reaching the release.

Usage

read_measurement_type(path, validate = TRUE)

Arguments

path: path to metadata/measurement_type.csv
validate: error on sentinel strings (default TRUE). Only set FALSE to inspect a file you already know is broken.

Value

A tibble of the registry, with empty cells as NA.

Details

The read is deliberately strict: na = "", so only genuinely empty cells become NA. read_csv()'s default is na = c("", "NA"), which converts the literal string "NA" back to NA — meaning a default read cannot see this corruption, and no validator downstream of one ever could. DuckDB, which is not so forgiving, is where the damage surfaces.