scjn

Client for the SCJN's SCOW JSON API (https://legislacion.scjn.gob.mx) and reader for the scjn-leyes GitHub release: a Mexican federal law's reform-dated snapshots, each tagged with a fuente: scjn provenance header. The SCJN is not an official source of legal text — the Diario Oficial de la Federación (DOF), read through dofjson, remains that; this package's crawl is a convenience corpus SCJN's own consolidated view makes possible, never a replacement for it.

This package was extracted out of a downstream package's own modules (issue #206). Fase 1 (issue #207) moved the transport (scjn.api), the catalogue's own algebra (scjn.catalog), per-instrument crawl state (scjn.state), and the provenance header's reader (scjn.header) — no behaviour change, no release-format change. Fase 3 (issue #209) moved the scjn-leyes release's own readers here too (scjn.release: download_scjn_leyes_corpus, download_scjn_leyes_index, download_scjn_leyes_catalog, iter_current_federal_laws, markdown_de_snapshot, download_scjn_leyes_assets, local_slugs), now in the package's own public scjn.__all__, disk-first and with their own cache directory (scjn.cache, $SCJN_CACHE_DIR — a separate lifecycle from the downstream package's cache, migrated once automatically from wherever it used to live). codNota linking — matching a snapshot to the DOF codNota that produced it — stays one layer up: a codNota is a DOF concept, and this package's dependency direction is one way (see tests/test_boundary.py).

Since issue #215 every law also carries the SCJN's own materia (subject classification), vigencia (whether it is still in force — seven values, not a boolean) and resumen (a one-paragraph abstract). They are per law, not per reform, so they live in each law's own estado.json and in the release index, and download_scjn_leyes_catalog/iter_current_federal_laws return them with no extra request — which makes scjn-leyes a classified corpus. scripts/fetch_federal_law_metadata.py is what fills them in, one SCJN search per law, matched by idOrdenamiento so a wrong document can never be described as the right one (issue #115, Hallazgo C).

scjn download [--slug SLUG] [--cache-dir DIR] [--refrescar]

puts the release on disk; every reader in scjn.release then reads off it with no network request at all, raising scjn.release.AssetNotCached for whatever is not there yet.

Changelog

• 0.5.0 — the fourth id-keyed collection, scjn-tratados (issue #277): every international treaty the SCJN serves (~1,456 instruments, members by the SCJN's ámbito TRATADOS INTERNACIONALES, any category). download_scjn_tratados_index / _corpus / _assets, local_tratados_ids and iter_current_tratados are the same thin wrappers over the generic core the reglamentos/lineamientos ones are (same AssetNotCached/SinTextoEnSCJN raise order), scjn.cache.TRATADOS joins COLECCIONES_POR_ID, scjn download --coleccion tratados fetches it, and scjn.release.construye_indice_global_tratados is its index builder. The release is a numbered series of tags (scjn-tratados, scjn-tratados-2, ...). scjn.discovery gains an ambito keyword (discover/discover_by_category, default "FEDERAL", existing calls unchanged), a by-ámbito mode (categoria="": no reform-category rescue and no coverage audit) and report_outside_ambito.

• 0.4.0 — articulos_a_markdown recovers table rows from a blank-line-separated block of an article's contenido that carries a tab (issue #253): the SCJN's plain text uses a run of tabs as its only column separator, and a wrapped multi-line cell used to come back as one Markdown paragraph per source line instead of one per row. A block with no tab is unaffected, byte for byte. Measured on the cached scjn-leyes release: 35 of 315 laws, 567 of 3,707 snapshots — a snapshot already on disk only picks up the fix via scripts/fetch_scjn_legislacion.py --reintenta <slug>, never automatically.

• 0.3.0 — iter_current_reglamentos / iter_current_lineamientos (issue #227, Fase 2): iter_current_federal_laws' id-keyed siblings, the current text of every instrument of either collection, one tarball open at a time. No codNota and no abrev (neither collection is DOF-linked), the newest snapshot chosen from the file names since an id-keyed tarball ships no indice.json, and an instrument the index lists with no consolidated text at all is skipped rather than raising SinTextoEnSCJN.

• 0.2.0 — the package's first release: the SCOW transport (scjn.api), the catalogue algebra, the crawl state, the provenance header and the disk-first readers of the scjn-leyes/scjn-reglamentos/ scjn-lineamientos releases (issues #206-#212, #220, #222).

Development

pip install -e "packages/scjn[test]"
pytest packages/scjn

License

Apache License 2.0. See LICENSE.