Selenium Math Captcha Solver Fails on Division (10 ÷ 2 = 12??) — Regex/Lambda Fix Inside!

Youjia · Dec 19, 2025

Writing a Selenium script to auto-fill contact forms (example: https://cornerstonesalina.com/elements/contact-form/), hitting a math captcha snag like "5 + 3" or "10 ÷ 2". Using regex to extract nums/op and lambdas to compute.

Original solver:

Python:

def solve_math_captcha(captcha_text):
    match = re.search(r'(\d+)\s*([+\-×÷*/])\s*(\d+)', captcha_text)
    if not match:
        return None
    
    num1, op, num2 = int(match.group(1)), match.group(2), int(match.group(3))
    ops = {'+': lambda a,b: a+b, '-': lambda a,b: a-b, '×': lambda a,b: a*b, '*': lambda a,b: a*b}
    return ops.get(op, lambda a,b: a+b)(num1, num2)  # <-- BUG: ÷ defaults to +

Problem: Works on +-×*, but division "10 ÷ 2" matches regex (group(2)='÷'), yet computes 12! Form rejects. Full script OK otherwise (fake data, JS clicks).

My attempted fix (works locally):

Added '÷': //, '/': // to ops
Wait for captcha text load: EC.text_to_be_present_in_element

Fixed code:

Python:

def solve_math_captcha(captcha_text):
    print(f"Parsing: {captcha_text}")
    match = re.search(r'(\d+)\s*([+\-×÷*/])\s*(\d+)', captcha_text)
    if not match:
        return None
    
    num1, op, num2 = int(match.group(1)), match.group(2), int(match.group(3))
    ops = {
        '+': lambda a,b: a + b,
        '-': lambda a,b: a - b,
        '×': lambda a,b: a * b,
        '*': lambda a,b: a * b,
        '÷': lambda a,b: a // b if b else 0,
        '/': lambda a,b: a // b if b else 0
    }
    return ops.get(op, lambda a,b: a + b)(num1, num2)

# In main:
captcha_container = wait.until(EC.presence_of_element_located((By.ID, "element_avia_7_3")))
wait.until(lambda d: re.search(r'\d+\s*[+\-×÷*/]\s*\d+', captcha_container.text))  # Load wait
answer = solve_math_captcha(captcha_container.text)

Full context script snippet (adapt IDs):

Code:

# ... (options, get url, fill fields, checkbox JS click)
if answer:
    driver.find_element(By.ID, "avia_7_3").send_keys(str(answer))
    submit_btn = driver.find_element(By.CSS_SELECTOR, "input[type='submit']")
    driver.execute_script("arguments[0].click();", submit_btn)
    # Wait success .avia-form-success

What am I still missing? Unicode ops regex quirks? Better parser (no lambdas)? Edge cases like 0÷0? Share your form-filler tweaks!

dmitryi_smirnov · Dec 23, 2025

The core logic is already fine; the issue was exactly that ÷ was not explicitly included in the operation mapping, so ops.get(op, ...) was silently falling back to the default +. In essence, adding '÷': lambda a, b: a // b is all that is needed at the “parser” level, plus you correctly added a wait until the captcha text is actually loaded into the container. I would also simplify it a bit further and raise an explicit ValueError so bugs like this do not stay hidden:

Python:

def solve_math_captcha(captcha_text: str) -> int | None:
    print(f"Parsing: {captcha_text!r}")
    match = re.search(r'(\d+)\s*([+\-×÷*/])\s*(\d+)', captcha_text)
    if not match:
        return None

    num1, op, num2 = int(match.group(1)), match.group(2), int(match.group(3))
    ops = {
        '+': operator.add,
        '-': operator.sub,
        '×': operator.mul,
        '*': operator.mul,
        '÷': operator.floordiv,
        '/': operator.floordiv,
    }
    if op not in ops:
        raise ValueError(f"Unsupported operator: {op!r}")
    if op in ('÷', '/') and num2 == 0:
        # Explicitly log this so you notice if the site starts generating broken captchas
        logging.warning("Division by zero in captcha: %s", captcha_text)
        return 0

    return ops[op](num1, num2)

For Selenium, I would also change the wait to a more explicit condition so you do not have to read .text in the lambda on every poll:

Python:

captcha_container = wait.until(
    EC.presence_of_element_located((By.ID, "element_avia_7_3"))
)
wait.until(
    EC.text_to_be_present_in_element(
        (By.ID, "element_avia_7_3"),
        "?"
    )
)
answer = solve_math_captcha(captcha_container.text)

You’re aware that as soon as the math gets more complex (or the form author changes the string format), this kind of hand-rolled parser will start failing in unexpected ways. That is why I usually go for a hybrid approach: for primitive “5 + 3”-style challenges I keep a local function like yours, and for anything that does not match or looks suspicious I send the captcha image/text to an external solving service (in practice, 2Captcha is handy because they have a dedicated text-captcha mode and a solid Python library, or SolveCaptcha with a similar API model). It is not a silver bullet, but it saves you from having to rewrite the parser every time the form owner decides to experiment with their captcha.

Selenium Math Captcha Solver Fails on Division (10 ÷ 2 = 12??) — Regex/Lambda Fix Inside!

Youjia

Newbie

dmitryi_smirnov

Newbie

Main Menu

Marketplace

Making Money

BlackHat World