#!/usr/bin/env python3
"""Find Brazilian bonus institutions from main dataset."""

import yaml
import re

# Load main dataset
print("Loading main dataset...")
with open("data/instances/all/globalglam-20251111.yaml", 'r', encoding='utf-8') as f:
    institutions = list(yaml.safe_load_all(f))
    if len(institutions) == 1 and isinstance(institutions[0], list):
        institutions = institutions[0]

print(f"Loaded {len(institutions)} institutions")

# Bonus targets
bonus_targets = {
    "Museu do Estado de Pernambuco": ["pernambuco", "estado"],
    "Museu Histórico Nacional": ["histórico nacional", "historico nacional"],
    "Fundação Cultural Palmares": ["palmares", "fundação cultural"],
    "Museu Imperial": ["imperial", "petrópolis"]
}

print("\n" + "="*80)
print("SEARCHING FOR BONUS INSTITUTIONS")
print("="*80)

for inst in institutions:
    if not isinstance(inst, dict):
        continue
    
    locations = inst.get('locations', [])
    is_brazil = any(loc.get('country') == 'BR' for loc in locations if isinstance(loc, dict))
    
    if not is_brazil:
        continue
    
    name = inst.get('name', '').lower()
    description = str(inst.get('description', '')).lower()
    inst_id = inst.get('id', 'NO_ID')
    
    for target_name, keywords in bonus_targets.items():
        # Check if ALL keywords appear in name or description
        if all(kw.lower() in name or kw.lower() in description for kw in keywords):
            print(f"\n✓ MATCH: {target_name}")
            print(f"  ID: {inst_id}")
            print(f"  Name: {inst.get('name', 'N/A')}")
            if description:
                desc_short = description[:100] + "..." if len(description) > 100 else description
                print(f"  Description: {desc_short}")

# Also check if already enriched with these Q-numbers
print("\n" + "="*80)
print("CHECKING IF Q-NUMBERS ALREADY EXIST")
print("="*80)

target_qids = ["Q6940628", "Q510993", "Q10286282", "Q1887049"]

for inst in institutions:
    if not isinstance(inst, dict):
        continue
    
    identifiers = inst.get('identifiers', [])
    for ident in identifiers:
        if not isinstance(ident, dict):
            continue
        if ident.get('identifier_value') in target_qids:
            print(f"\n✓ Q-NUMBER EXISTS: {ident.get('identifier_value')}")
            print(f"  ID: {inst.get('id', 'NO_ID')}")
            print(f"  Name: {inst.get('name', 'N/A')}")

print("\n" + "="*80)