import json,re
from pathlib import Path

base=Path(r"C:\Apache24\htdocs\.codex_tmp")
main=json.loads((base/'ocr_candidates.json').read_text(encoding='utf-8'))
retry=json.loads((base/'ocr_retry_results.json').read_text(encoding='utf-8-sig'))
trans=str.maketrans('٠١٢٣٤٥٦٧٨٩۰۱۲۳۴۵۶۷۸۹','01234567890123456789')
retry_by_id={}
for r in retry:
    sid=int(r['file'].split('_')[1]); text=(r.get('text') or '').translate(trans)
    candidates=[]
    for idx,line in enumerate(text.split('|')):
        digits=re.sub(r'\D','',line)
        if 6<=len(digits)<=9: candidates.append({'value':digits,'line':idx})
    retry_by_id[sid]={'text':text,'candidates':candidates}
for r in main:
    if not r['chosen']:
        rr=retry_by_id.get(r['id'],{})
        c=rr.get('candidates',[])
        # The document number is printed at the top and normally appears in the first OCR lines.
        early=[x for x in c if x['line']<=2]
        if early:
            r['chosen']=early[0]['value']; r['source']='full-early'
        else: r['source']='unresolved'
        r['retry_candidates']=c
    else: r['source']='top-crop'
(base/'ocr_merged.json').write_text(json.dumps(main,ensure_ascii=False,indent=2),encoding='utf-8')
print(json.dumps({'resolved':sum(bool(r['chosen']) for r in main),'unresolved':sum(not r['chosen'] for r in main),'retry_any_candidates':sum(bool(retry_by_id.get(r['id'],{}).get('candidates')) for r in main if r['source']=='unresolved')},ensure_ascii=False))