endomorphosis/world_municipal_law

World Municipal Law

Local (municipal / county / water-board / regional) law worldwide, normalized to one schema. Companion to endomorphosis/american_municipal_law, which stays unchanged.

Split: national law lives in world_national_law

Since 2026-10-11, national, subnational (first-level: state / Land / province / région) and supranational law, i.e. every iso:* place (iso_FR, iso_ES-CT, iso_EU, …), is published in endomorphosis/world_national_law, which uses the same schema and layout. This repo keeps municipal and other local law (municipalities, counties, water boards, …). The iso_* files were copied there first, verified (sha256 and row counts), and then deleted here in a separate commit; the repo history still has them.

Layout

• world_law/data/{CC}/{place_key}_html.parquet: one file per jurisdiction (ZSTD); large places are sharded as {place_key}.s0001_html.parquet, .s0002_… (≤20k rows / ~200 MB each)
• world_law/metadata/{CC}/{place_key}.json
• Countries with more than 5,000 places (DE, FR, IT) are hash-bucketed, because the Hub allows at most 10,000 files per folder: world_law/data/{CC}/{b}/{place_key}_html.parquet and world_law/metadata/{CC}/{b}/{place_key}.json, where b = first 2 hex characters of sha256(place_key) (256 buckets). places.parquet always holds each place's exact html_path.
• world_law/manifests/*.jsonl: per-upload manifests with sha256
• world_law/places.parquet: crosswalk of place_id, Wikidata QID, GNIS and platform ids

place_key is the GNIS for US places, otherwise place_id with : replaced by _ (wd_Q90 = Paris, 1008538 = a US place by GNIS, KG-kg_cbd_<organ id> = a Kyrgyz aiyl aimak council without a Wikidata item, place_id KG-kg_cbd:<organ id> keyed by its CBD (cbd.minjust.gov.kg) issuing-organ id; a QID found later is added as qid, the place_id stays).

Columns (*_html.parquet)

cid, doc_id, doc_order, html_title, html (same as american_municipal_law), then place_id, country_code, gnis, source_platform, source_url, doc_type, language, title_native, number, adopted_date, published_date, effective_date, in_force, consolidated, content_format, text, retrieved_at, then optional provenance (v1.1.0): archive_source, archive_crawl_id, archive_timestamp, archive_url, original_url, content_complete, possibly_stale, source_policy, source_official, source_org, license_note, extra_metadata (JSON string of other source fields), then (v1.2.0) pdf_ref: for documents whose source is an image-only scanned PDF, the PDF in endomorphosis/world_law_scanned_pdfs as endomorphosis/world_law_scanned_pdfs:pdfs/<CC>/[<b>/]<place_key>/<sha256>.pdf (loose file) or, for PDFs uploaded since 2026-10-18, endomorphosis/world_law_scanned_pdfs:shards/<CC>/<YYYY>/<MM>/<CC>_<seq>.tar#member=<place_key>/<sha256>.pdf&offset=<o>&size=<n>&sha256=<h> (one member of a per-country tar shard, fetchable with one HTTP Range request); null otherwise. Files written before v1.2.0 have no pdf_ref column — treat it as null. world_law_loader.fetch_pdf(row) downloads either form (fetch_pdf_bytes returns the bytes) and checks the sha256. Access: world_law_scanned_pdfs is a private dataset, so resolving a pdf_ref needs a Hugging Face token that is authorized for it (set HF_TOKEN, or pass token= to the loader). The text, metadata and every other column in this dataset are public and need no token.

Related datasets: world_municipal_law (local places) · world_national_law (iso:* places) · world_law_scanned_pdfs (scanned source PDFs) · code: world_law_pipeline. content_complete=false marks metadata-only / partial records.

doc_type ∈ code_section | ordinance | regulation | statute | gazette_act | decision. consolidated=false marks as-published gazette acts.

cid = IPFS CIDv1 (raw, sha2-256, base32) of {place_key}_{doc_id}.json, which is identical to american_municipal_law for US places.

Provenance

• Documents: 3,284,470 in 48,764 places.
• From a web archive (Wayback / Common Crawl; archive_source set): 124,870 (3.8%).
• Of those, archived copies of sites whose robots.txt disallows crawling or whose pages carry a robots noarchive/noindex meta tag (source_policy = archived_robots_disallowed / noarchive_meta / noindex_meta): 8,563.
• content_complete = false (metadata-only or partial text): 394,215.
• source_official = false (non-government mirror/aggregator): 8,334.
• Rows whose live fetch was not permitted (robots.txt unreachable or disallowing) are not published.

country	places	docs	archived	archived share	archived (restricted sites)	content_complete=false	source_official=false
AE	4	270	247	91.5%	0	5	0
AL	34	1,950	1,950	100.0%	0	1,099	0
AM	17	797	797	100.0%	0	324	0
AR	201	5,691	5,691	100.0%	0	2,411	0
AT	241	4,146	2,006	48.4%	0	0	0
AU	353	4,167	1,125	27.0%	0	680	0
BA	33	1,562	1,562	100.0%	0	494	0
BE	223	173,504	2	0.0%	2	0	0
BJ	3	43	43	100.0%	0	25	0
BN	3	39	0	0.0%	0	0	0
BO	23	929	929	100.0%	0	659	0
BR	4,161	435,183	0	0.0%	0	290,060	0
BW	2	6	3	50.0%	0	2	0
CA	1,395	58,245	0	0.0%	0	0	0
CI	2	5	5	100.0%	0	0	0
CL	337	9,816	0	0.0%	0	5	0
CM	14	43	43	100.0%	0	5	0
CO	761	5,597	5,597	100.0%	0	5,312	0
CR	84	2,406	631	26.2%	0	109	0
CU	1	2	2	100.0%	0	0	0
CV	13	43	18	41.9%	0	2	0
CZ	1,206	10,658	337	3.2%	0	0	0
DE	6,038	219,340	6,149	2.8%	4,624	774	0
DO	16	710	710	100.0%	0	360	0
DZ	14	99	99	100.0%	0	4	0
EC	111	1,644	1,644	100.0%	0	691	0
EE	78	1,970	0	0.0%	0	0	0
EG	18	220	220	100.0%	0	161	0
ES	2,015	74,368	1	0.0%	0	1	0
ET	4	19	19	100.0%	0	7	0
FJ	13	142	0	0.0%	0	0	0
FR	10,535	329,142	1	0.0%	1	3,840	0
GB	411	2,837	26	0.9%	26	0	0
GE	21	546	546	100.0%	0	99	0
GH	58	122	120	98.4%	0	55	0
GM	2	2	2	100.0%	0	0	0
GR	336	29,747	0	0.0%	0	0	0
GT	246	1,499	83	5.5%	0	1,466	0
HN	4	11	11	100.0%	0	2	0
HR	361	13,875	13,875	100.0%	0	5,231	0
ID	437	18,686	2,540	13.6%	0	1,701	0
IE	31	845	1	0.1%	1	0	0
IL	7	85	0	0.0%	0	14	0
IN	5	7	0	0.0%	0	0	0
IQ	14	63	63	100.0%	0	18	0
IR	120	2,721	2,711	99.6%	0	865	0
IT	7,615	66,286	0	0.0%	0	2	0
JM	3	70	70	100.0%	0	6	0
JO	12	88	88	100.0%	0	37	0
JP	1,080	493,306	3,995	0.8%	3,091	0	0
KE	45	983	0	0.0%	0	12	0
KG	530	91,597	0	0.0%	0	0	0
KR	132	609	609	100.0%	603	0	0
KZ	36	737	737	100.0%	0	15	0
LB	17	54	54	100.0%	0	9	0
LR	5	96	4	4.2%	0	6	0
LT	30	86	86	100.0%	0	0	0
LV	35	3,590	0	0.0%	0	0	0
MA	56	393	393	100.0%	0	156	0
MD	75	4,612	4,612	100.0%	0	933	0
ME	22	548	548	100.0%	0	287	0
MK	34	249	249	100.0%	0	117	0
MN	314	1,268	0	0.0%	0	0	0
MW	9	23	6	26.1%	0	17	21
MX	388	3,426	319	9.3%	0	0	0
MY	3	80	0	0.0%	0	38	0
MZ	2	4	0	0.0%	0	0	0
NG	13	87	16	18.4%	0	40	0
NI	3	29	29	100.0%	0	9	0
NL	366	121,429	0	0.0%	0	0	0
NP	66	1,753	0	0.0%	0	0	0
NZ	29	323	0	0.0%	0	24	0
OM	1	2	2	100.0%	0	1	0
PA	21	37	37	100.0%	0	7	0
PE	1,128	46,026	4	0.0%	0	46,010	0
PH	200	12,856	808	6.3%	0	8,744	87
PL	2,418	781,403	1,155	0.1%	0	0	0
PS	11	99	99	100.0%	0	35	0
PT	306	6,670	0	0.0%	0	0	0
PY	42	2,015	2,015	100.0%	0	1,029	0
QA	1	40	40	100.0%	0	0	0
RO	659	104,644	113	0.1%	0	3,951	0
RS	161	5,564	5,564	100.0%	0	1,734	0
RU	110	437	22	5.0%	0	0	0
SA	6	56	53	94.6%	0	23	0
SB	1	35	0	0.0%	0	0	0
SI	203	20,030	0	0.0%	0	0	0
SK	931	19,637	0	0.0%	0	0	0
SL	5	6	6	100.0%	0	0	0
SN	2	44	44	100.0%	0	24	0
SO	3	4	4	100.0%	0	0	0
SS	1	24	0	0.0%	0	0	0
SV	9	116	116	100.0%	0	55	0
SY	9	265	265	100.0%	0	240	0
TC	1	19	0	0.0%	0	0	0
TG	1	20	20	100.0%	0	19	0
TL	5	9	0	0.0%	0	0	0
TN	69	985	985	100.0%	0	425	0
TR	481	13,134	5,485	41.8%	41	2,471	0
TW	9	11,487	2,748	23.9%	0	2	0
TZ	90	274	84	30.7%	0	218	190
UA	457	41,095	41,095	100.0%	0	10,691	0
UG	57	159	106	66.7%	0	54	23
US	76	7,590	0	0.0%	0	70	7,552
UY	19	143	97	67.8%	0	35	0
UZ	178	1,171	0	0.0%	0	0	0
VE	14	163	163	100.0%	0	22	0
VN	7	174	174	100.0%	174	0	0
VU	3	31	0	0.0%	0	5	0
ZA	75	2,321	1,888	81.3%	0	51	335
ZM	59	132	29	22.0%	0	105	126
ZW	9	25	25	100.0%	0	5	0

Configs

default covers every country; there is also one config per ISO 3166-1 country code (AE, AL, AM, AR, AT, AU, BA, BE, BJ, BN, BO, BR, BW, CA, CI, CL, CM, CO, CR, CU, CV, CZ, DE, DO, DZ, EC, EE, EG, ES, ET, FJ, FR, GB, GE, GH, GM, GR, GT, HN, HR, HU, ID, IE, IL, IN, IQ, IR, IT, JM, JO, JP, KE, KG, KR, KZ, LB, LR, LT, LV, MA, MD, ME, MK, MN, MW, MX, MY, MZ, NG, NI, NL, NP, NZ, OM, PA, PE, PH, PL, PS, PT, PY, QA, RO, RS, RU, SA, SB, SI, SK, SL, SN, SO, SS, SV, SY, TC, TG, TL, TN, TR, TW, TZ, UA, UG, US, UY, UZ, VE, VN, VU, ZA, ZM, ZW). A country config can show as empty until its first files are uploaded.

Code

The scrapers, normalizer/uploader, schema and place-code crosswalk that build this dataset are published at endomorphosis/world_law_pipeline. Rows whose live fetch was not permitted (robots.txt unreachable or disallowing) are held back and not published.

Licence

The texts come from official government sources under various open licences (CC0, CC-BY, OGL, NLOD, public domain) and from codifier platforms, following the existing US practice. Each row's source_platform and source_url identify its origin, so check the source's terms before reuse.

Combined loader (world + US)

US municipal codes stay in endomorphosis/american_municipal_law (not copied). endomorphosis/world_municipal_law world_law/places.parquet holds one pointer row per US GNIS place (place_id gnis:N, country_code US, has_local_data false, source_repo, source_paths, total_docs). world_law_loader.py reads both and maps the US 5-column files (cid, doc_id, doc_order, html_title, html) onto the 35-column schema (other columns null):

from huggingface_hub import hf_hub_download
import importlib.util
spec = importlib.util.spec_from_file_location("wll", hf_hub_download("endomorphosis/world_law_pipeline", "world_law_loader.py"))
wll = importlib.util.module_from_spec(spec); spec.loader.exec_module(wll)
wll.places(countries=["US", "FR"])                          # crosswalk (pandas)
t = wll.load(places=["gnis:1085809", "wd:Q10001"])              # pyarrow Table, 35 columns
t = wll.load(countries=["DE"], include_national=True, max_places=10)