glam/scripts/regenerate_historical_ghcids.py

#!/usr/bin/env python3
"""
Regenerate GHCIDs for Historical Institutions using GeoNames-based lookup.

This script:
1. Loads historical institutions validation YAML
2. Uses coordinates to lookup modern location (country, region, city) from GeoNames
3. Generates proper GHCIDs using COUNTRY-REGION-CITY-TYPE-NAME format
4. Updates YAML with corrected GHCIDs matching production implementation

Key principle: Use coordinates to find MODERN (2025) location identifiers,
not coordinate hashes.

## GHCID Collision Resolution and Temporal Priority

This script handles GHCID collisions according to temporal priority rules:

**First Batch Collision** (multiple institutions discovered simultaneously):
    - ALL colliding institutions receive native language name suffixes in snake_case
    - Example: If two museums in Amsterdam both generate NL-NH-AMS-M-SM,
      both receive name suffixes: NL-NH-AMS-M-SM-stedelijk_museum_amsterdam,
      NL-NH-AMS-M-SM-science_museum_amsterdam
    - Rationale: Fair treatment when discovered at the same time

**Historical Addition** (new institution collides with published GHCID):
    - ONLY the newly discovered institution receives a name suffix
    - Previously published GHCID remains unchanged (PID stability)
    - Example: If NL-NH-AMS-M-HM was published 2025-11-01, and a new
      institution with the same GHCID is found 2025-11-15, only the new
      institution gets: NL-NH-AMS-M-HM-historical_museum_amsterdam
    - Rationale: "Cool URIs don't change" - preserve existing citations

**Implementation Notes**:
    - Track publication_date in provenance metadata
    - Detect collisions by comparing GHCID strings before name suffix addition
    - Decision logic: if existing GHCID has publication_date < current extraction,
      preserve it; only add name suffix to new institution
    - Log collision events in collision registry (future enhancement)

**References**:
    - docs/PERSISTENT_IDENTIFIERS.md - Section "Historical Collision Resolution"
    - docs/plan/global_glam/07-ghcid-collision-resolution.md - Temporal dimension
    - docs/GHCID_PID_SCHEME.md - Timeline examples

Usage:
    python scripts/regenerate_historical_ghcids.py
"""

import json
import sys
import yaml
from datetime import datetime, timezone
from pathlib import Path
from typing import Dict, List, Optional

# Add src to path for imports
sys.path.insert(0, str(Path(__file__).parent.parent / "src"))

from glam_extractor.identifiers.ghcid import (
    GHCIDGenerator,
    GHCIDComponents,
    InstitutionType,
    extract_abbreviation_from_name,
)
from glam_extractor.geocoding.geonames_lookup import GeoNamesDB


class HistoricalGHCIDGenerator:
    """
    Generate GHCIDs for historical institutions using modern geographic references.
    
    This generator creates BASE GHCIDs (without name suffixes) by:
    1. Reverse geocoding coordinates to modern location (city, region)
    2. Looking up city abbreviations from GeoNames database
    3. Mapping region names to ISO 3166-2 codes
    4. Generating institution abbreviations from names
    
    COLLISION HANDLING:
    -------------------
    This class generates deterministic base GHCIDs. Collision detection and
    name suffix assignment happen in a separate collision resolution stage.
    
    Collision scenarios:
    
    1. **First Batch Collision** (multiple institutions processed together):
       - Two institutions generate the same base GHCID
       - Both have the same publication_date (extraction timestamp)
       - Resolution: BOTH receive native language name suffixes
       - Example: NL-NH-AMS-M-SM-stedelijk_museum_amsterdam and
                  NL-NH-AMS-M-SM-science_museum_amsterdam
    
    2. **Historical Addition** (new institution added later):
       - New institution generates GHCID matching already-published identifier
       - publication_date is AFTER existing record's timestamp
       - Resolution: ONLY new institution receives name suffix
       - Existing GHCID preserved (maintains citation stability)
       - Example: Existing NL-NH-AMS-M-HM unchanged,
                  new becomes NL-NH-AMS-M-HM-historical_museum_amsterdam
    
    PID Stability Principle:
    ------------------------
    Once a GHCID is published (exported to RDF, JSON-LD, CSV), it becomes a
    persistent identifier that may be:
    - Cited in academic papers
    - Used in external databases and APIs
    - Embedded in linked data triples
    - Referenced in archival finding aids
    
    Per W3C "Cool URIs Don't Change" principle, published GHCIDs MUST NOT be
    modified retroactively. Historical additions get name suffixes; existing PIDs
    remain stable.
    
    References:
    -----------
    - docs/PERSISTENT_IDENTIFIERS.md (Historical Collision Resolution section)
    - docs/plan/global_glam/07-ghcid-collision-resolution.md (Temporal dimension)
    - docs/GHCID_PID_SCHEME.md (Timeline examples)
    """
    
    def __init__(self, geonames_db_path: Path):
        """
        Initialize with GeoNames database.
        
        Args:
            geonames_db_path: Path object to geonames.db SQLite database
        """
        self.geonames_db = GeoNamesDB(geonames_db_path)
        self.ghcid_gen = GHCIDGenerator()
        
        # Load ISO 3166-2 mappings for region codes
        self.reference_dir = Path(__file__).parent.parent / "data" / "reference"
        self.nl_mapping = self._load_region_mapping("iso_3166_2_nl.json")
        self.it_mapping = self._load_region_mapping("iso_3166_2_it.json")
        self.be_mapping = self._load_region_mapping("iso_3166_2_be.json")
        self.ru_mapping = self._load_region_mapping("iso_3166_2_ru.json")
        self.dk_mapping = self._load_region_mapping("iso_3166_2_dk.json")
        self.ar_mapping = self._load_region_mapping("iso_3166_2_ar.json")
        self.ru_mapping = self._load_region_mapping("iso_3166_2_ru.json")
        self.dk_mapping = self._load_region_mapping("iso_3166_2_dk.json")
        self.ar_mapping = self._load_region_mapping("iso_3166_2_ar.json")
        
        # Statistics
        self.stats = {
            'total': 0,
            'success': 0,
            'city_found': 0,
            'city_fallback': 0,
            'region_found': 0,
            'region_fallback': 0,
        }
    
    def _load_region_mapping(self, filename: str) -> Dict[str, str]:
        """Load ISO 3166-2 region code mapping."""
        filepath = self.reference_dir / filename
        if not filepath.exists():
            print(f"Warning: {filename} not found, region lookups may fail")
            return {}
        
        with open(filepath, 'r', encoding='utf-8') as f:
            data = json.load(f)
        
        # Extract provinces mapping (different structure per file)
        import unicodedata
        mapping = {}
        
        # Handle different JSON structures
        if 'provinces' in data:
            # NL/AR format: {"provinces": {"Drenthe": "DR", ...}}
            province_data = data['provinces']
        elif 'regions' in data:
            # IT/DK format: {"regions": {"Lombardy": "25", ...}}
            province_data = data['regions']
        elif 'federal_subjects' in data:
            # RU format: {"federal_subjects": {"Kaliningrad Oblast": "39", ...}}
            province_data = data['federal_subjects']
        else:
            # Simple format: {"DR": "Drenthe", ...} - need to reverse
            province_data = {v: k for k, v in data.items() if isinstance(v, str) and len(k) == 2}
        
        # Create normalized name -> code mapping
        for name, code in province_data.items():
            # Normalize name for lookup
            normalized = unicodedata.normalize('NFD', name.upper())
            normalized = ''.join(c for c in normalized if unicodedata.category(c) != 'Mn')
            mapping[normalized.strip()] = code
        
        return mapping
    
    def _get_region_code(self, country: str, region_name: Optional[str]) -> str:
        """
        Get ISO 3166-2 region code for a country/region.
        
        Args:
            country: ISO 3166-1 alpha-2 country code
            region_name: Region/province name (optional)
        
        Returns:
            2-letter region code or "00" if not found
        """
        if not region_name:
            self.stats['region_fallback'] += 1
            return "00"
        
        # Normalize region name
        import unicodedata
        normalized = unicodedata.normalize('NFD', region_name.upper())
        normalized = ''.join(c for c in normalized if unicodedata.category(c) != 'Mn')
        normalized = normalized.strip()
        
        # Lookup by country
        mapping = None
        if country == "NL":
            mapping = self.nl_mapping
        elif country == "IT":
            mapping = self.it_mapping
        elif country == "BE":
            mapping = self.be_mapping
        elif country == "RU":
            mapping = self.ru_mapping
        elif country == "DK":
            mapping = self.dk_mapping
        elif country == "AR":
            mapping = self.ar_mapping
        
        if mapping and normalized in mapping:
            self.stats['region_found'] += 1
            return mapping[normalized]
        
        self.stats['region_fallback'] += 1
        return "00"
    
    def _get_city_code_fallback(self, city_name: str) -> str:
        """
        Generate 3-letter city code from city name.
        Same logic as Latin America script.
        
        Args:
            city_name: City name
        
        Returns:
            3-letter uppercase code
        """
        import unicodedata
        
        # Remove accents
        normalized = unicodedata.normalize('NFD', city_name)
        normalized = ''.join(c for c in normalized if unicodedata.category(c) != 'Mn')
        
        # Split into words
        words = normalized.split()
        
        if len(words) == 1:
            # Single word: take first 3 letters
            code = words[0][:3].upper()
        elif words[0].lower() in ['la', 'el', 'los', 'las', 'o', 'a', 'de', 'den', 'het']:
            # City with article: first letter of article + first 2 of next word
            if len(words) > 1:
                code = (words[0][0] + words[1][:2]).upper()
            else:
                code = words[0][:3].upper()
        else:
            # Multi-word: take first letter of each word (up to 3)
            code = ''.join(w[0] for w in words[:3]).upper()
        
        # Ensure exactly 3 letters
        if len(code) < 3:
            code = code.ljust(3, 'X')
        elif len(code) > 3:
            code = code[:3]
        
        return code
    
    def _reverse_geocode_with_geonames(
        self,
        latitude: float,
        longitude: float,
        country: str
    ) -> Optional[Dict[str, str]]:
        """
        Reverse geocode coordinates to find modern location info from GeoNames.
        
        Args:
            latitude: Latitude coordinate
            longitude: Longitude coordinate
            country: ISO 3166-1 alpha-2 country code
        
        Returns:
            Dict with 'city', 'region', 'city_code' or None
        """
        # Query GeoNames database for nearby cities
        # Using a SQL query to find the closest city
        
        query = """
        SELECT 
            name,
            admin1_code,
            admin1_name,
            latitude,
            longitude,
            geonames_id,
            ((latitude - ?) * (latitude - ?) + (longitude - ?) * (longitude - ?)) as distance
        FROM cities
        WHERE country_code = ?
        ORDER BY distance
        LIMIT 1
        """
        
        import sqlite3
        conn = sqlite3.connect(self.geonames_db.db_path)
        cursor = conn.cursor()
        
        try:
            cursor.execute(query, (latitude, latitude, longitude, longitude, country))
            row = cursor.fetchone()
            
            if row:
                city_name, admin1, admin1_name, lat, lon, geonameid, distance = row
                
                # Lookup city abbreviation from GeoNames
                city_info = self.geonames_db.lookup_city(city_name, country)
                if city_info:
                    city_code = city_info.get_abbreviation()
                else:
                    # Fallback to 3-letter abbreviation from name
                    city_code = self._get_city_code_fallback(city_name)
                
                return {
                    'city': city_name,
                    'region': admin1_name,  # Use admin1_name for region lookup
                    'region_code': admin1,  # Also return the code
                    'city_code': city_code,
                    'geonames_id': geonameid,
                    'distance_deg': distance ** 0.5  # Approximate distance
                }
        finally:
            conn.close()
        
        return None
    
    def generate_ghcid_for_institution(self, record: dict) -> Optional[GHCIDComponents]:
        """
        Generate GHCID for a historical institution using modern geographic lookup.
        
        **GHCID Collision Handling**:
        
        This method generates a base GHCID (without name suffix). Collision detection
        and name suffix assignment happen at a later stage when comparing against
        previously published records.
        
        Decision logic (implemented in future collision detector):
        
        1. **Check if base GHCID already exists in published dataset**
           - If NO → Use base GHCID without name suffix
           - If YES → Check publication timestamp
        
        2. **Compare publication dates**:
           - Same batch (same publication_date) → ALL instances get name suffixes
           - Historical addition (current date > existing publication_date):
             → ONLY new institution gets name suffix
             → Existing GHCID preserved (PID stability)
        
        3. **Name suffix format**:
           - Native language institution name converted to snake_case
           - Example: "Stedelijk Museum Amsterdam" → "stedelijk_museum_amsterdam"
        
        Example scenarios:
        
        **Scenario 1: First Batch Collision**
        ```
        # 2025-11-01 batch import discovers two institutions:
        Institution A: "Stedelijk Museum Amsterdam"
          → NL-NH-AMS-M-SM-stedelijk_museum_amsterdam
        Institution B: "Science Museum Amsterdam"  
          → NL-NH-AMS-M-SM-science_museum_amsterdam
        # Both get name suffixes because discovered simultaneously
        ```
        
        **Scenario 2: Historical Addition**
        ```
        # 2025-11-01: "Hermitage Amsterdam" published
        Institution A: NL-NH-AMS-M-HM (publication_date: 2025-11-01T10:00:00Z)
        
        # 2025-11-15: "Historical Museum Amsterdam" discovered
        Institution B: Base GHCID = NL-NH-AMS-M-HM (collision!)
          → Check A.publication_date (2025-11-01) < current (2025-11-15)
          → A unchanged, B gets name suffix: NL-NH-AMS-M-HM-historical_museum_amsterdam
        ```
        
        Args:
            record: Institution record from YAML
        
        Returns:
            GHCIDComponents or None if generation fails
        """
        self.stats['total'] += 1
        
        # Extract required fields
        name = record.get('name', '')
        institution_type_str = record.get('institution_type', 'U')
        
        # Get location data
        locations = record.get('locations', [])
        if not locations:
            print(f"Warning: No location data for {name}")
            return None
        
        location = locations[0]  # Use primary location
        city = location.get('city', '')
        country = location.get('country', 'XX')
        latitude = location.get('latitude')
        longitude = location.get('longitude')
        
        if not latitude or not longitude:
            print(f"Warning: No coordinates for {name}")
            return None
        
        # Reverse geocode to get modern location info
        geo_info = self._reverse_geocode_with_geonames(latitude, longitude, country)
        
        if geo_info:
            city_code = geo_info['city_code']
            region = geo_info['region']
            print(f"  ✓ {name}: Found {geo_info['city']} (code: {city_code}, region: {region})")
            self.stats['city_found'] += 1
        else:
            # Fallback: use city name from record
            city_info = self.geonames_db.lookup_city(city, country)
            if city_info:
                city_code = city_info.get_abbreviation()
                region = None  # Will use fallback "00"
                print(f"  ○ {name}: Using city name lookup ({city} → {city_code})")
                self.stats['city_found'] += 1
            else:
                city_code = self._get_city_code_fallback(city)
                region = None
                print(f"  ○ {name}: Using fallback ({city} → {city_code})")
                self.stats['city_fallback'] += 1
        
        # Get region code
        region_code = self._get_region_code(country, region)
        
        # Map institution type
        if len(institution_type_str) == 1:
            # Single letter - map to InstitutionType
            type_map = {
                'G': InstitutionType.GALLERY,
                'L': InstitutionType.LIBRARY,
                'A': InstitutionType.ARCHIVE,
                'M': InstitutionType.MUSEUM,
                'O': InstitutionType.OFFICIAL_INSTITUTION,
                'R': InstitutionType.RESEARCH_CENTER,
                'C': InstitutionType.CORPORATION,
                'U': InstitutionType.UNKNOWN,
                'B': InstitutionType.BOTANICAL_ZOO,
                'E': InstitutionType.EDUCATION_PROVIDER,
                'P': InstitutionType.PERSONAL_COLLECTION,
                'S': InstitutionType.COLLECTING_SOCIETY,
            }
            inst_type = type_map.get(institution_type_str, InstitutionType.UNKNOWN)
        else:
            try:
                inst_type = InstitutionType[institution_type_str]
            except KeyError:
                inst_type = InstitutionType.UNKNOWN
        
        # Generate abbreviation from name
        abbreviation = extract_abbreviation_from_name(name)
        
        # Generate GHCID
        components = GHCIDComponents(
            country_code=country,
            region_code=region_code,
            city_locode=city_code,
            institution_type=inst_type,
            abbreviation=abbreviation
        )
        
        self.stats['success'] += 1
        return components
    
    def print_statistics(self):
        """Print generation statistics."""
        print("\n" + "="*70)
        print("HISTORICAL GHCID REGENERATION STATISTICS")
        print("="*70)
        print(f"Total institutions processed:     {self.stats['total']}")
        print(f"GHCIDs successfully generated:    {self.stats['success']}")
        print(f"City codes from GeoNames:         {self.stats['city_found']}")
        print(f"City codes from fallback:         {self.stats['city_fallback']}")
        print(f"Region codes found:               {self.stats['region_found']}")
        print(f"Region codes fallback (00):       {self.stats['region_fallback']}")
        print()


def main():
    """Main execution."""
    
    # Paths
    project_root = Path(__file__).parent.parent
    yaml_path = project_root / "data" / "instances" / "historical_institutions_validation.yaml"
    geonames_db_path = project_root / "data" / "reference" / "geonames.db"
    
    print("="*70)
    print("REGENERATE HISTORICAL INSTITUTION GHCIDs")
    print("="*70)
    print(f"Input:  {yaml_path}")
    print(f"GeoNames DB: {geonames_db_path}")
    print()
    
    # Create backup
    backup_dir = project_root / "data" / "instances" / "archive"
    backup_dir.mkdir(exist_ok=True)
    timestamp = datetime.now(timezone.utc).strftime("%Y%m%d_%H%M%S")
    backup_path = backup_dir / f"historical_institutions_pre_regenerate_{timestamp}.yaml"
    
    import shutil
    shutil.copy(yaml_path, backup_path)
    print(f"✓ Backup created: {backup_path}")
    print()
    
    # Load YAML
    with open(yaml_path, 'r', encoding='utf-8') as f:
        institutions = yaml.safe_load(f)
    
    print(f"Loaded {len(institutions)} historical institutions")
    print()
    
    # Initialize generator
    generator = HistoricalGHCIDGenerator(geonames_db_path)
    
    # Process each institution
    print("Generating GHCIDs using modern geographic lookup...")
    print()
    
    # COLLISION DETECTION NOTE:
    # This script regenerates GHCIDs for historical validation data.
    # In production, collision detection occurs when comparing against
    # previously published GHCIDs with temporal priority:
    #
    # 1. Build collision registry: group institutions by base GHCID (no name suffix)
    # 2. For each collision group:
    #    a) Check publication_date metadata (from provenance.extraction_date)
    #    b) If all have same publication_date → First Batch (all get name suffixes)
    #    c) If dates differ → Historical Addition (only new ones get name suffixes)
    # 3. Preserve existing PIDs: Never modify GHCIDs with earlier publication_date
    #
    # See docs/plan/global_glam/07-ghcid-collision-resolution.md for algorithm
    
    for record in institutions:
        name = record.get('name', 'Unknown')
        
        # Generate GHCID
        components = generator.generate_ghcid_for_institution(record)
        
        if not components:
            continue
        
        # Generate all identifier formats
        # 
        # NAME SUFFIX COLLISION HANDLING:
        # Currently generating base GHCID without name suffix.
        # In production collision detector:
        # 1. Check if this base GHCID exists in published dataset
        # 2. If collision detected:
        #    - First batch (same publication_date): Add name suffix to ALL
        #    - Historical addition (later date): Add name suffix to NEW only
        # 3. Name suffix generation:
        #    - Convert native language institution name to snake_case
        #    - Example: "Stedelijk Museum Amsterdam" → "stedelijk_museum_amsterdam"
        #
        ghcid_str = components.to_string()
        ghcid_uuid = components.to_uuid()
        ghcid_uuid_sha256 = components.to_uuid_sha256()
        ghcid_numeric = components.to_numeric()
        record_id = GHCIDComponents.generate_uuid_v7()
        
        # Update record
        record['ghcid_current'] = ghcid_str
        record['ghcid_original'] = ghcid_str  # Same as current for regeneration
        record['ghcid_uuid'] = str(ghcid_uuid)
        record['ghcid_uuid_sha256'] = str(ghcid_uuid_sha256)
        record['ghcid_numeric'] = ghcid_numeric
        record['record_id'] = str(record_id)
        
        # Update GHCID history (first entry)
        # 
        # GHCID HISTORY AND COLLISIONS:
        # ghcid_history tracks GHCID changes over time (see GHCIDHistoryEntry schema).
        # When collisions are resolved by adding name suffixes:
        #
        # 1. Create new history entry with valid_from timestamp
        # 2. Update previous entry with valid_to timestamp (if name suffix added)
        # 3. Record reason for change (e.g., "Name suffix added to resolve collision")
        #
        # Example history for collision resolution:
        # ghcid_history:
        #   - ghcid: NL-NH-AMS-M-HM-historical_museum_amsterdam  # Current (with name suffix)
        #     ghcid_numeric: 789012345678
        #     valid_from: "2025-11-15T10:00:00Z"
        #     valid_to: null
        #     reason: "Name suffix added to resolve collision with existing NL-NH-AMS-M-HM"
        #   - ghcid: NL-NH-AMS-M-HM  # Original (without name suffix) 
        #     ghcid_numeric: 123456789012
        #     valid_from: "2025-11-15T09:00:00Z"
        #     valid_to: "2025-11-15T10:00:00Z"
        #     reason: "Historical identifier based on modern geographic location"
        #
        if 'ghcid_history' in record and record['ghcid_history']:
            record['ghcid_history'][0]['ghcid'] = ghcid_str
            record['ghcid_history'][0]['ghcid_numeric'] = ghcid_numeric
            record['ghcid_history'][0]['reason'] = (
                "Historical identifier based on modern geographic location "
                "from GeoNames reverse geocoding"
            )
        
        # Update identifiers list
        if 'identifiers' not in record:
            record['identifiers'] = []
        
        # Remove old GHCID identifiers
        record['identifiers'] = [
            i for i in record['identifiers'] 
            if i.get('identifier_scheme') not in ['GHCID', 'GHCID_NUMERIC', 'GHCID_UUID', 'GHCID_UUID_SHA256', 'RECORD_ID']
        ]
        
        # Add new GHCID identifiers
        record['identifiers'].extend([
            {
                'identifier_scheme': 'GHCID',
                'identifier_value': ghcid_str
            },
            {
                'identifier_scheme': 'GHCID_NUMERIC',
                'identifier_value': str(ghcid_numeric)
            },
            {
                'identifier_scheme': 'GHCID_UUID',
                'identifier_value': str(ghcid_uuid),
                'identifier_url': f'urn:uuid:{ghcid_uuid}'
            },
            {
                'identifier_scheme': 'GHCID_UUID_SHA256',
                'identifier_value': str(ghcid_uuid_sha256),
                'identifier_url': f'urn:uuid:{ghcid_uuid_sha256}'
            },
            {
                'identifier_scheme': 'RECORD_ID',
                'identifier_value': str(record_id),
                'identifier_url': f'urn:uuid:{record_id}'
            }
        ])
    
    # Print statistics
    generator.print_statistics()
    
    # NEXT STEPS: COLLISION DETECTION AND RESOLUTION
    # ================================================
    # This script generates base GHCIDs. For production deployment:
    #
    # 1. Build collision detector module:
    #    - Group institutions by base GHCID (without name suffix)
    #    - Identify collision sets (2+ institutions with same base GHCID)
    #
    # 2. Implement temporal priority resolver:
    #    - Extract publication_date from provenance.extraction_date
    #    - Compare timestamps within each collision set
    #    - Apply decision logic:
    #      * Same timestamp → First Batch → ALL get name suffixes
    #      * Different timestamps → Historical Addition → NEW get name suffixes
    #
    # 3. Name suffix generation:
    #    - Convert native language institution name to snake_case
    #    - Example: "Stedelijk Museum Amsterdam" → "stedelijk_museum_amsterdam"
    #    - Append to base GHCID: NL-NH-AMS-M-HM → NL-NH-AMS-M-HM-historical_museum_amsterdam
    #
    # 4. Update GHCID history:
    #    - Create new GHCIDHistoryEntry when name suffix added
    #    - Set valid_from to current timestamp
    #    - Set valid_to on previous entry (if exists)
    #    - Document reason: "Name suffix added to resolve collision with [other institution]"
    #
    # 5. Generate collision registry report:
    #    - CSV with columns: base_ghcid, collision_type, resolution_strategy,
    #      affected_institutions, publication_dates, timestamp
    #    - Store in data/reports/ghcid_collision_registry.csv
    #
    # 6. Validation:
    #    - Verify all GHCIDs are unique (base or with name suffix)
    #    - Check no published GHCIDs were modified (PID stability test)
    #    - Validate GHCID history entries have correct temporal ordering
    #
    # See docs/plan/global_glam/07-ghcid-collision-resolution.md for algorithm
    # See docs/GHCID_PID_SCHEME.md for timeline examples
    
    # Write updated YAML
    print(f"Writing updated YAML to: {yaml_path}")
    with open(yaml_path, 'w', encoding='utf-8') as f:
        # Write header
        f.write("---\n")
        f.write("# Historical Heritage Institutions - GHCID Validation Examples\n")
        f.write("# Generated from Wikidata SPARQL query\n")
        f.write("# GHCIDs regenerated using modern GeoNames-based location lookup\n")
        f.write(f"# Last updated: {datetime.now(timezone.utc).isoformat()}\n")
        f.write("#\n")
        f.write("# GHCID Format: COUNTRY-REGION-CITY-TYPE-NAME\n")
        f.write("# - City codes from GeoNames reverse geocoding (coordinates → city)\n")
        f.write("# - Region codes from ISO 3166-2 mappings\n")
        f.write("# - Same format as production implementation\n")
        f.write("\n")
        
        # Write institutions
        yaml.dump(institutions, f, default_flow_style=False, allow_unicode=True, sort_keys=False)
    
    print("✓ Done!")
    print()
    print("="*70)
    print("VALIDATION COMPLETE")
    print("="*70)


if __name__ == "__main__":
    main()