glam/scripts/migrate_egyptian_from_ch.py

#!/usr/bin/env python3
"""
Migrate Egyptian institutions incorrectly placed under CH (Switzerland) to EG (Egypt).
"""

import re
import sqlite3
import unicodedata
from datetime import datetime, timezone
from pathlib import Path

# Egyptian city mapping
EGYPTIAN_CITIES = {
    'Cairo': {'region': 'C', 'city_code': 'CAI'},
    'Alexandria': {'region': 'ALX', 'city_code': 'ALX'},
    'Giza': {'region': 'GZ', 'city_code': 'GIZ'},
    'Assiut': {'region': 'AST', 'city_code': 'ASS'},
    'Helwan': {'region': 'C', 'city_code': 'HEL'},
    '6th of October City': {'region': 'GZ', 'city_code': 'OCT'},
    'Ain Shams': {'region': 'C', 'city_code': 'ASH'},
    'Maadi': {'region': 'C', 'city_code': 'MAA'},
    'New Cairo': {'region': 'C', 'city_code': 'NCA'},
}

def extract_city_from_name(name):
    """Extract Egyptian city from institution name."""
    name_lower = name.lower()
    
    if 'cairo' in name_lower or 'ain shams' in name_lower or 'helwan' in name_lower:
        return 'Cairo'
    if 'alexandria' in name_lower:
        return 'Alexandria'
    if 'assiut' in name_lower or 'asyut' in name_lower:
        return 'Assiut'
    if 'giza' in name_lower or 'october' in name_lower:
        return 'Giza'
    if 'nile' in name_lower or 'maadi' in name_lower:
        return 'Cairo'  # Most Egyptian institutions without city are in Cairo
    if 'egypt' in name_lower or 'egyptian' in name_lower:
        return 'Cairo'  # Default for national institutions
    
    return 'Cairo'  # Default

def update_file(file_path, city_name, dry_run=False):
    """Update file from CH to EG namespace."""
    with open(file_path, 'r', encoding='utf-8') as f:
        content = f.read()
    
    city_info = EGYPTIAN_CITIES.get(city_name, {'region': 'C', 'city_code': 'CAI'})
    region_code = city_info['region']
    city_code = city_info['city_code']
    
    # Get current GHCID
    old_ghcid_match = re.search(r'ghcid_current:\s*([^\n]+)', content)
    if not old_ghcid_match:
        return False, None
    
    old_ghcid = old_ghcid_match.group(1).strip()
    
    # Create new GHCID with EG namespace
    new_ghcid = re.sub(r'^CH-XX-XXX-', f'EG-{region_code}-{city_code}-', old_ghcid)
    
    if dry_run:
        return True, (old_ghcid, new_ghcid)
    
    # Update all GHCID references
    content = content.replace(f'ghcid_current: {old_ghcid}', f'ghcid_current: {new_ghcid}')
    content = content.replace(f'ghcid_original: {old_ghcid}', f'ghcid_original: {new_ghcid}')
    content = content.replace(f'identifier_value: {old_ghcid}', f'identifier_value: {new_ghcid}')
    content = content.replace(f"ghcid: {old_ghcid}", f"ghcid: {new_ghcid}")
    
    # Update country code
    content = re.sub(r'country:\s*CH', 'country: EG', content)
    content = re.sub(r'country_code:\s*CH', 'country_code: EG', content)
    
    # Update location_resolution
    content = re.sub(r'region_code:\s*XX', f'region_code: {region_code}', content)
    content = re.sub(r'city_code:\s*XXX', f'city_code: {city_code}', content)
    
    # Add history entry
    timestamp = datetime.now(timezone.utc).isoformat()
    history_entry = f"""
  - ghcid: {new_ghcid}
    valid_from: '{timestamp}'
    reason: Migrated from CH to EG namespace - {city_name}"""
    
    history_match = re.search(r'(ghcid_history:\s*\n)', content)
    if history_match:
        insert_pos = history_match.end()
        content = content[:insert_pos] + history_entry + content[insert_pos:]
    
    with open(file_path, 'w', encoding='utf-8') as f:
        f.write(content)
    
    # Rename file
    old_filename = file_path.name
    new_filename = old_filename.replace('CH-XX-XXX-', f'EG-{region_code}-{city_code}-')
    if new_filename != old_filename:
        new_path = file_path.parent / new_filename
        file_path.rename(new_path)
    
    return True, (old_ghcid, new_ghcid)

def main():
    import sys
    dry_run = '--dry-run' in sys.argv
    
    base_dir = Path(__file__).parent.parent
    custodian_dir = base_dir / 'data' / 'custodian'
    
    print("Egyptian Institution Migration (CH → EG)")
    print("=" * 50)
    if dry_run:
        print("DRY RUN MODE\n")
    
    # Find CH-XX-XXX files that are actually Egyptian
    xxx_files = list(custodian_dir.glob('CH-XX-XXX-*.yaml'))
    print(f"Found {len(xxx_files)} CH-XX-XXX files\n")
    
    migrated = 0
    egyptian_keywords = ['egypt', 'cairo', 'alexandria', 'ain shams', 'helwan', 'assiut', 
                        'giza', 'nile', 'al-azhar', 'dar al-kutub', 'guc', 'auc', 'bue']
    
    for file_path in xxx_files:
        with open(file_path, 'r', encoding='utf-8') as f:
            content = f.read()
        
        # Check if this is an Egyptian institution
        name_match = re.search(r'claim_value:\s*(.+)', content)
        if not name_match:
            continue
        
        inst_name = name_match.group(1).strip().lower()
        
        is_egyptian = any(keyword in inst_name for keyword in egyptian_keywords)
        if not is_egyptian:
            continue
        
        city = extract_city_from_name(inst_name)
        success, ghcid_change = update_file(file_path, city, dry_run)
        
        if success:
            if dry_run:
                print(f"  {file_path.name}")
                print(f"    → {ghcid_change[0]} → {ghcid_change[1]}")
            else:
                print(f"✓ Migrated: {file_path.name} → {city}")
            migrated += 1
    
    print(f"\n{'=' * 50}")
    print(f"Migrated: {migrated}")

if __name__ == '__main__':
    main()