Skip to main content

Source format

The canonical source is the full CSV export linked from the Ministerio de Sanidad's Nomenclátor de facturación.

CSV format​

The inspected export is UTF-8, uses commas as separators, and contains 20 official columns. In the files examined so far:

  • Código Nacional arrives as text and is unique per inspected snapshot;
  • dates use DD/MM/YYYY;
  • decimal prices use a dot separator;
  • boolean-like fields use SI and NO;
  • blank cells are common and meaningful;
  • medicine-only fields may be blank for financed health-product rows.

The parser requires all known columns, and the build stops if any are missing. Populated unknown columns are kept in extra_source_fields and produce a warning. This prevents a Ministry format change from passing unnoticed.

The exact ordered header set and the source facts observed during initial analysis are maintained in DATA_NOTES.md. The executable contract lives in src/nomenclator_delta/schema.py.

Why old monthly files are hard to find​

The Ministry page links to the current export. When a new month is published, the same download address returns the new file; it does not continue to serve the previous one. There is no clear public catalogue of old CSV downloads on that page. This is why finding June after July was published was not as simple as changing a month in the URL.

The collector therefore stores each response immediately under:

sources/artifacts/<YYYY-MM>/<sha256>.<extension>

The accompanying metadata records the authority, requested URL, UTC retrieval time, nominal month, original filename, file type, SHA-256 checksum, byte size, and parser version. Downloading identical bytes again reuses the same path; a different file gets a different checksum and path.

Raw source files are not committed by default. Keep them in reliable storage controlled by the maintainer so that a generated release can be reproduced.

How the June 2026 comparison was built​

No preserved copy of the June CSV response was found. This is a consequence of the current-file download described above, not evidence that a June release did not exist. Internet Archive preserved two useful items: the Ministry page that identified the June release and the complete official June XLS file.

The parser input for June was produced from that XLS in a fixed, repeatable way: cell-edge whitespace was trimmed, three known older header names were mapped to their current names, and the rows were written using the Ministry's CSV format.

The conversion was checked against August: converting the official August XLS reproduced the official August CSV byte for byte. Even with that check, the June source record is explicit about what was recovered. It sets exact_original_csv_bytes_preserved: false, records the original XLS filename and SHA-256 checksum, and lists Internet Archive only as the transport. The project does not claim to have preserved the original June CSV bytes.

Historical transport​

When a file is recovered through a web archive, the archive is the transport, not the data authority. The Ministerio de Sanidad remains the authority. Archive-specific identifiers are stored in snapshot metadata and included in generated histories.

Parser policy​

The parser rejects invalid identifiers, missing required headers, malformed dates, and duplicate CNs. It does not guess the meaning of unfamiliar values: an unexpected flag value is preserved and reported.

Last updated on