Skip to contents

Reads the manifest.json provenance files of the healthbr-data R2 mirror and returns, for each published partition, when it was processed, from which Ministry source file, and how many records it holds. Use it to see which years/months are actually available before calling sipni_data(), and to audit the provenance of the mirror.

Usage

sipni_status(dataset = c("microdados", "doses", "cobertura"), cache_dir = NULL)

Arguments

dataset

Character. Which dataset(s) to report: "microdados" (individual-level 2020+, monthly partitions), "doses" (DPNI aggregates 1994–2019, year x UF partitions), "cobertura" (CPNI aggregates 1994–2019). Default: all three.

cache_dir

Character. Cache directory (manifests are cached locally and revalidated by ETag). Default: tools::R_user_dir("healthbR", "cache").

Value

A tibble with columns: dataset, year, month (NA for aggregates), uf (NA for microdata – partitions are monthly and national), records, processing_timestamp (as recorded by the pipeline, no timezone), source_url (the Ministry file the partition was derived from).

Details

The manifests are the mirror's source of truth: they record the source file's URL, size and hash, the processing timestamp, and the SHA-256 of every published Parquet. Data on the mirror are byte-identical to the Ministry's files; see the healthbr-data reproducibility policy for the full audit recipe.

Examples

if (FALSE) { # interactive()
# everything the mirror currently holds
sipni_status()

# which 2026 microdata months are published?
dplyr::filter(sipni_status("microdados"), year == 2026)
}