Romanian Law Data

Zstd-compressed Parquet exports of the Romanian legal corpus (acts, articles, paragraphs) for use with DuckDB. Sourced from legislatie.just.ro (Ministry of Justice) via its public SOAP API.

Automated daily via GitHub Actions. Download from Releases.

Tables

Table	Content	Rows
acte	One row per act (LEGE, OUG, HG, ORDIN, DECIZIE, …) with metadata + full text	~187k
articole	One row per article (parsed from acte.content)	~993k
alineate	One row per paragraph — the finest citation unit (e.g. art. 188 alin. (1))	~1.96M

Tables use Romanian legal vocabulary (acte, articole, alineate); columns use English SQL convention (type, published_at, gazette_number, …) with Romanian COMMENT ON metadata in create_views.sql.

Subject lenses

create_views.sql also exposes pre-filtered views over acte for each canonical code and for jurisprudence:

View	Filters
constitutie	Constituția României (1991, republicată 2003)
cod_civil	Legea 287/2009
cod_penal	Legea 286/2009
cod_muncii	Legea 53/2003 (republicată)
cod_procedura_civila	Legea 134/2010 (republicată)
cod_procedura_penala	Legea 135/2010
cod_fiscal	Legea 227/2015
jurisprudenta	CCR + ÎCCJ decisions

Usage

# Download the latest bundle
gh release download -R scriptogre/romanian-law-data
tar xzf laws.tar.gz -C data/

import duckdb
conn = duckdb.connect()
conn.execute(open("data/create_views.sql").read())
conn.execute("""
    SELECT act_citation, link, article_citation, content
    FROM articole
    WHERE act_id IN (SELECT id FROM cod_penal)
      AND article_number = 188
""").fetchall()

Pipeline

collect.py    SOAP API → data/raw_acts.jsonl
normalize.py  fix encoding, dedup, extract dates + gazette → stdout
parse.py      extract articles + alineate                  → stdout
export.py     write parquet bundle + sha256                ← stdin

collect checkpoints raw_acts.jsonl (SOAP is slow + rate-limited, worth caching). normalize → parse → export is one pipe — no intermediate JSONL on disk.

uv sync
uv run python -m scripts.collect
uv run python -m scripts.normalize \
  | uv run python -m scripts.parse \
  | uv run python -m scripts.export

License

The corpus is published by the Romanian Ministry of Justice and is public information. This repository only provides format conversion + pipeline tooling.