ExplainMath

ExplainMath is a small Python library that catches invalid numeric operations (like division by zero, undefined results, or bad power operations) and explains why it went wrong in plain English — instead of silently giving NaN or inf.

This is useful in machine learning, simulations, finance, or any code where one wrong number can poison the whole pipeline without you noticing.

🔥 Why This Library Exists

In Python, many invalid numeric operations do not crash your program. Instead, they produce NaN, inf, or a complex number — and these values spread quietly through your code.

Example of a real problem:

```python result = model(data) print(result) # nan ... now what?

Traditional debugging tells you where the error happened,

but not why the math became invalid.

ExplainMath catches the failure at the moment it happens and tells you the reason.

```

✨ Features

• Tracks invalid math operations
• Supports addition, subtraction, multiplication, division, and power
• Marks invalid values explicitly (no hidden NaNs or silent errors)
• Preserves the reason for the failure
• Propagates invalid state safely through further calculations
• Optional .require() strict mode that raises an exception

🚀 Quick Start

📦 Examples

Basic addition

```python from explainmath import Value

a = Value(10) b = Value(5) print(a.add(b).value) # 15 ```

Invalid Division

```python from explainmath import Value

a = Value(10) b = Value(0) c = a.div(b) print(c.is_valid()) # False print(c.explanation) # "Division by zero while evaluating 10 / 0" ```

🔒 Strict Mode Example (Fail Fast)

```python from explainmath import Value, SemanticError

try: Value(10).div(Value(0)).require() except SemanticError as e: print("Error caught:", e) ```

🧪 Running Tests

```bash python -m unittest discover -v ```

📌 Project Status

This is version v0.1. It is intentionally small and focused:

• A single numeric type
• Basic arithmetic
• Error explanation
• Safe propagation

Future versions will include:

• Operation history
• Provenance tracking
• Better debugging reports
• Optional integration with NumPy/PyTorch

🗂 Folder Structure

```text core/ → implementation examples/ → usage demos tests/ → unit tests docs/ → (reserved for future) ```

License

This project is licensed under the MIT License.

Made with curiosity, logic, and a desire to reduce silent math bugs.