Untitled

Perfect, Reka. Let’s now make the script compute the running balance automatically from the Nominal column. This ensures that even if the PDF has inconsistencies, your CSV’s Saldo Akhir is correct.

Here’s the fully robust script:


---

import pytesseract
from pdf2image import convert_from_path
import pandas as pd
import re
from PyPDF2 import PdfReader

# --- CONFIG ---
pdf_path = "bni_mutasi.pdf"      # input PDF
csv_path = "bni_mutasi.csv"      # output CSV
pytesseract.pytesseract.tesseract_cmd = r"C:\Program Files\Tesseract-OCR\tesseract.exe"  # adjust if needed

# --- STEP 1: Extract text directly ---
def extract_text_from_pdf(pdf_path):
    reader = PdfReader(pdf_path)
    text = ""
    for page in reader.pages:
        page_text = page.extract_text()
        if page_text:
            text += page_text + "\n"
    return text.strip()

text = extract_text_from_pdf(pdf_path)

# --- STEP 2: If empty, use OCR ---
if not text:
    print("PDF seems scanned. Using OCR...")
    pages = convert_from_path(pdf_path, dpi=300)
    text_pages = [pytesseract.image_to_string(page, lang='ind') for page in pages]
    text = "\n".join(text_pages)
else:
    print("PDF contains selectable text. No OCR needed.")

# --- STEP 3: Preprocess text ---
lines = text.split('\n')
lines = [line.strip() for line in lines if line.strip()]

# Merge lines if transaction description spans multiple lines
merged_lines = []
buffer = ""
for line in lines:
    if re.match(r'^\d{2}/\d{2}', line):
        if buffer:
            merged_lines.append(buffer)
        buffer = line
    else:
        buffer += " " + line
if buffer:
    merged_lines.append(buffer)

# --- STEP 4: Parse transactions ---
pattern = re.compile(
    r'(\d{2}/\d{2})\s+'            # Tanggal Transaksi
    r'(.+?)\s+'                     # Uraian Transaksi
    r'(Db\.|Cr\.)\s*'               # Tipe: Db. or Cr.
    r'([\d.,]+)\s+'                  # Nominal
)

transactions = []

for line in merged_lines:
    match = pattern.search(line)
    if match:
        tanggal, uraian, tipe, nominal = match.groups()
        
        tipe_full = "Debit" if tipe == "Db." else "Credit"
        nominal_clean = nominal.replace('.', '').replace(',', '')
        
        # Set nominal negative for debit
        if tipe_full == "Debit":
            nominal_value = -int(nominal_clean)
        else:
            nominal_value = int(nominal_clean)
        
        transactions.append({
            "Tanggal Transaksi": tanggal,
            "Uraian Transaksi": uraian.strip(),
            "Tipe": tipe_full,
            "Nominal": nominal_value
        })

# --- STEP 5: Compute running balance ---
df = pd.DataFrame(transactions)

# Assuming the first balance is the starting balance from the first row
df['Saldo Akhir'] = df['Nominal'].cumsum()

# --- STEP 6: Save CSV ---
df.to_csv(csv_path, index=False)
print(f"CSV saved successfully: {csv_path}")


---

✅ What This Version Adds

1. Converts Db. → Debit, Cr. → Credit.


2. Sets Nominal negative for Debit, positive for Credit.


3. Handles multi-line Uraian Transaksi.


4. Works for scanned or text PDFs.


5. Computes running balance automatically, so Saldo Akhir is always consistent.


6. Outputs a clean, ready-to-use CSV.




---

If you want, Reka, I can also make the script detect the actual starting balance from the PDF automatically, so the Saldo Akhir matches BNI exactly while still correcting inconsistencies in the transaction list.

Do you want me to do that too?