Add Overture Maps POI enrichment layer for place details

Ingests 20.9M North America places from Overture Maps Foundation
(release 2026-04-15.0) into PostgreSQL. Enriches /api/place responses
with phone, website, and brand data via spatial + fuzzy name matching
when OSM extratags are sparse.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
This commit is contained in:
Matt 2026-04-21 16:51:25 +00:00
commit 65693d15aa
6 changed files with 597 additions and 0 deletions

350
scripts/overture_import.py Normal file
View file

@ -0,0 +1,350 @@
#!/usr/bin/env python3
"""Overture Maps Places → PostgreSQL import script (v2).
Downloads Overture Places Parquet from S3 via DuckDB (public bucket, no credentials),
filters to North America bounding box, and inserts into local PostgreSQL with PostGIS.
Usage:
cd /opt/recon && venv/bin/python scripts/overture_import.py
Re-runnable (idempotent via UPSERT).
"""
import json
import logging
import os
import re
import sys
import time
import duckdb
import psycopg2
import psycopg2.extras
logging.basicConfig(
level=logging.INFO,
format='%(asctime)s %(levelname)s %(message)s',
datefmt='%H:%M:%S'
)
log = logging.getLogger('overture_import')
# --- Config ---
OVERTURE_RELEASE = '2026-04-15.0'
S3_PATH = f's3://overturemaps-us-west-2/release/{OVERTURE_RELEASE}/theme=places/type=place/*'
# North America bounding box (generous — includes Hawaii, Puerto Rico, Canada)
BBOX = {
'xmin': -170.0,
'xmax': -50.0,
'ymin': 15.0,
'ymax': 85.0,
}
BATCH_SIZE = 50_000
OSM_RECORD_RE = re.compile(r'^([nwr])(\d+)@\d+$')
DB_CONFIG = {
'host': os.environ.get('OVERTURE_DB_HOST', 'localhost'),
'port': int(os.environ.get('OVERTURE_DB_PORT', '5432')),
'dbname': os.environ.get('OVERTURE_DB_NAME', 'overture'),
'user': os.environ.get('OVERTURE_DB_USER', 'overture'),
'password': os.environ.get('OVERTURE_DB_PASSWORD', ''),
}
def create_table(conn):
"""Create places table and indexes if they don't exist."""
with conn.cursor() as cur:
cur.execute("""
CREATE TABLE IF NOT EXISTS places (
id TEXT PRIMARY KEY,
geometry GEOMETRY(Point, 4326),
name TEXT,
basic_category TEXT,
confidence REAL,
phone TEXT,
website TEXT,
socials JSONB,
brand_name TEXT,
brand_wikidata TEXT,
osm_type CHAR(1),
osm_id BIGINT,
source_record_id TEXT,
raw_sources JSONB
);
""")
cur.execute("""
CREATE INDEX IF NOT EXISTS idx_places_osm
ON places(osm_type, osm_id) WHERE osm_type IS NOT NULL;
""")
cur.execute("""
CREATE INDEX IF NOT EXISTS idx_places_geom
ON places USING GIST(geometry);
""")
cur.execute("""
CREATE INDEX IF NOT EXISTS idx_places_name_trgm
ON places USING GIN(name gin_trgm_ops);
""")
conn.commit()
log.info('Table and indexes ready')
def parse_osm_ref(sources):
"""Extract OSM type letter and ID from Overture sources array."""
if not sources:
return None, None, None
for src in sources:
record_id = None
if isinstance(src, dict):
record_id = src.get('record_id', '')
elif hasattr(src, '__getitem__'):
# DuckDB struct — try attribute access
try:
record_id = src['record_id']
except (KeyError, TypeError, IndexError):
pass
if not record_id:
continue
m = OSM_RECORD_RE.match(str(record_id))
if m:
return m.group(1), int(m.group(2)), str(record_id)
return None, None, None
def run_import():
"""Main import: DuckDB reads S3 Parquet → PostgreSQL via chunked OFFSET/LIMIT."""
log.info(f'Overture release: {OVERTURE_RELEASE}')
log.info(f'S3 path: {S3_PATH}')
log.info(f'Bounding box: {BBOX}')
# Connect to PostgreSQL
conn = psycopg2.connect(**DB_CONFIG)
conn.autocommit = False
create_table(conn)
# Set up DuckDB with httpfs and spatial for S3 access
duck = duckdb.connect()
duck.execute("INSTALL httpfs; LOAD httpfs;")
duck.execute("INSTALL spatial; LOAD spatial;")
duck.execute("SET s3_region='us-west-2';")
# Use a materialized approach: DuckDB query → Arrow → iterate in Python
query = f"""
SELECT
id,
ST_X(geometry) AS lon,
ST_Y(geometry) AS lat,
names.primary AS name,
basic_category,
confidence,
phones,
websites,
socials,
brand,
sources
FROM read_parquet('{S3_PATH}', hive_partitioning=true)
WHERE bbox.xmin >= {BBOX['xmin']}
AND bbox.xmax <= {BBOX['xmax']}
AND bbox.ymin >= {BBOX['ymin']}
AND bbox.ymax <= {BBOX['ymax']}
"""
log.info('Starting DuckDB query against S3 (this will take several minutes)...')
t_start = time.time()
# Execute and fetch all as Arrow for efficient iteration
result_rel = duck.sql(query)
upsert_sql = """
INSERT INTO places (id, geometry, name, basic_category, confidence,
phone, website, socials, brand_name, brand_wikidata,
osm_type, osm_id, source_record_id, raw_sources)
VALUES %s
ON CONFLICT (id) DO UPDATE SET
geometry = EXCLUDED.geometry,
name = EXCLUDED.name,
basic_category = EXCLUDED.basic_category,
confidence = EXCLUDED.confidence,
phone = EXCLUDED.phone,
website = EXCLUDED.website,
socials = EXCLUDED.socials,
brand_name = EXCLUDED.brand_name,
brand_wikidata = EXCLUDED.brand_wikidata,
osm_type = EXCLUDED.osm_type,
osm_id = EXCLUDED.osm_id,
source_record_id = EXCLUDED.source_record_id,
raw_sources = EXCLUDED.raw_sources
"""
template = """(
%(id)s,
ST_SetSRID(ST_MakePoint(%(lon)s, %(lat)s), 4326),
%(name)s,
%(basic_category)s,
%(confidence)s,
%(phone)s,
%(website)s,
%(socials)s::jsonb,
%(brand_name)s,
%(brand_wikidata)s,
%(osm_type)s,
%(osm_id)s,
%(source_record_id)s,
%(raw_sources)s::jsonb
)"""
total = 0
osm_refs = 0
batch = []
log.info('DuckDB query executing, fetching results in chunks...')
# Fetch in chunks using fetchmany on the relation
chunk_size = BATCH_SIZE
while True:
chunk = result_rel.fetchmany(chunk_size)
if not chunk:
break
for row in chunk:
row_id = row[0]
lon = row[1]
lat = row[2]
name = row[3]
basic_cat = row[4]
conf = row[5]
phones = row[6]
websites = row[7]
socials_raw = row[8]
brand_raw = row[9]
sources_raw = row[10]
if lon is None or lat is None:
continue
# Phone: first element of VARCHAR[]
phone = None
if phones and len(phones) > 0:
phone = str(phones[0]) if phones[0] else None
# Website: first element of VARCHAR[]
website = None
if websites and len(websites) > 0:
website = str(websites[0]) if websites[0] else None
# Socials: VARCHAR[] → JSON array of strings
socials_json = None
if socials_raw and len(socials_raw) > 0:
socials_json = json.dumps([str(s) for s in socials_raw if s])
# Brand: struct with wikidata and names.primary
brand_name = None
brand_wikidata = None
if brand_raw:
try:
if isinstance(brand_raw, dict):
brand_wikidata = brand_raw.get('wikidata')
names_struct = brand_raw.get('names')
if names_struct and isinstance(names_struct, dict):
brand_name = names_struct.get('primary')
else:
# DuckDB struct — access by key
brand_wikidata = brand_raw['wikidata'] if 'wikidata' in dir(brand_raw) else None
try:
brand_wikidata = brand_raw[0] # wikidata is first field
names_struct = brand_raw[1] # names is second field
if names_struct:
brand_name = names_struct[0] # primary is first field
except (IndexError, TypeError):
pass
except Exception:
pass
# Sources: parse OSM cross-reference
sources_list = None
if sources_raw:
if isinstance(sources_raw, (list, tuple)):
sources_list = []
for s in sources_raw:
if isinstance(s, dict):
sources_list.append(s)
else:
# DuckDB struct tuple — convert
try:
sources_list.append({
'dataset': s[1] if len(s) > 1 else None,
'record_id': s[3] if len(s) > 3 else None,
})
except (TypeError, IndexError):
pass
osm_type_letter, osm_id_val, source_record_id = parse_osm_ref(sources_list)
if osm_type_letter:
osm_refs += 1
raw_sources_json = json.dumps(sources_list) if sources_list else None
batch.append({
'id': row_id,
'lon': float(lon),
'lat': float(lat),
'name': name,
'basic_category': basic_cat,
'confidence': float(conf) if conf is not None else None,
'phone': phone,
'website': website,
'socials': socials_json,
'brand_name': brand_name,
'brand_wikidata': brand_wikidata,
'osm_type': osm_type_letter,
'osm_id': osm_id_val,
'source_record_id': source_record_id,
'raw_sources': raw_sources_json,
})
if len(batch) >= BATCH_SIZE:
with conn.cursor() as cur:
psycopg2.extras.execute_values(
cur, upsert_sql, batch,
template=template,
page_size=BATCH_SIZE
)
conn.commit()
total += len(batch)
elapsed = time.time() - t_start
rate = total / elapsed if elapsed > 0 else 0
log.info(f'Inserted {total:,} rows ({osm_refs:,} OSM xrefs) '
f'[{rate:.0f} rows/sec, {elapsed:.0f}s elapsed]')
batch = []
# Flush remaining
if batch:
with conn.cursor() as cur:
psycopg2.extras.execute_values(
cur, upsert_sql, batch,
template=template,
page_size=BATCH_SIZE
)
conn.commit()
total += len(batch)
duck.close()
# Final stats
elapsed = time.time() - t_start
log.info(f'Import complete: {total:,} rows, {osm_refs:,} OSM cross-refs, '
f'{elapsed:.0f}s total ({total/elapsed:.0f} rows/sec)')
# Verify
with conn.cursor() as cur:
cur.execute("SELECT count(*) FROM places")
count = cur.fetchone()[0]
cur.execute("SELECT count(*) FROM places WHERE osm_type IS NOT NULL")
osm_count = cur.fetchone()[0]
log.info(f'Final table: {count:,} total rows, {osm_count:,} with OSM cross-references')
conn.close()
if __name__ == '__main__':
run_import()