import camelot
import pandas as pd
import re

def extract_bank_data(file_path):
    print(f"Reading {file_path}... shifting Payment Details parsing one row lower.")
    
    # We use 'stream' flavor for the NLB report layout
    tables = camelot.read_pdf(file_path, pages='1-end', flavor='stream')
    
    final_transactions = []
    current_tx = None
    
    # Patterns
    iban_pattern = r'SI56\s?\d{4}\s?\d{4}\s?\d{4}\s?\d{3}'
    amount_pattern = r'=(\d+[\.,]\d{2})'
    date_pattern = r'(\d{2}\.\d{2}\.\d{2})'
    bic_pattern = r'\b[A-Z0-9]{8,11}\b'

    for table in tables:
        df = table.df
        for _, row in df.iterrows():
            cells = [str(c).strip() for c in row]
            row_text = " ".join(cells)
            
            amount_match = re.search(amount_pattern, row_text)
            date_match = re.search(date_pattern, row_text)
            
            # TRIGGER: New transaction starts
            if amount_match and date_match:
                if current_tx:
                    final_transactions.append(current_tx)
                
                amt_val = amount_match.group(1)
                
                # Debit/Credit alignment logic
                is_credit = False
                for i, c in enumerate(cells):
                    if '=' in c and i >= 2:
                        is_credit = True
                
                current_tx = {
                    "value date": date_match.group(1),
                    "Creditor / Debtor": cells[0].split('\n')[0].strip(),
                    "account": "",
                    "payment details": "",
                    "debit": "" if is_credit else amt_val,
                    "credit": amt_val if is_credit else "",
                    "skip_rows": 0  # Counter to shift parsing lower
                }
            
            elif current_tx:
                text_line = cells[0]
                if not text_line: continue

                # 1. Identify IBAN
                iban_search = re.search(iban_pattern, text_line)
                if iban_search:
                    current_tx["account"] = iban_search.group(0).replace(" ", "")
                    # When IBAN is found, the line below it is usually BIC. 
                    # We set skip_rows to 1 to skip that BIC line.
                    current_tx["skip_rows"] = 1
                    continue

                # 2. Skip the "BIC" line (one row lower logic)
                if current_tx["skip_rows"] > 0:
                    current_tx["skip_rows"] -= 1
                    continue

                # 3. Capture Payment Details (after skipping)
                # Filter out obvious address noise
                is_address = any(k in text_line.upper() for k in ["CESTA", "ULICA", "TRG", "LJUBLJANA"]) or re.search(r'\b\d{4}\b', text_line)
                
                if not is_address and len(text_line) > 2:
                    # Scrub labels and internal bank noise
                    clean_text = re.sub(r'^(purpose|payment details|reference)', '', text_line, flags=re.I).strip()
                    clean_text = re.sub(r'\b[CD]R\d+\b|\bNRC\b', '', clean_text).strip()
                    
                    if clean_text:
                        # Append to details
                        if not current_tx["payment details"]:
                            current_tx["payment details"] = clean_text
                        else:
                            current_tx["payment details"] += " " + clean_text

    if current_tx:
        final_transactions.append(current_tx)

    # Export
    df_final = pd.DataFrame(final_transactions).drop(columns=['skip_rows'])
    cols = ["value date", "Creditor / Debtor", "account", "payment details", "debit", "credit"]
    df_final = df_final[cols]
    
    # Final cleanup
    for col in df_final.columns:
        df_final[col] = df_final[col].astype(str).str.replace(r'\s+', ' ', regex=True).str.strip()
        df_final[col] = df_final[col].replace('nan', '')

    output_file = "bank_export_final_fixed.csv"
    df_final.to_csv(output_file, sep=';', index=False, encoding='utf-8-sig')
    print(f"Success! {len(df_final)} transactions exported.")

if __name__ == "__main__":
    extract_bank_data("MSDataReport (1).pdf")