Quick Start

Hello World

Create a file called hello.ch:

print("Hello, World!")

Run it:

charl run hello.ch

Output:

Hello, World!

Variables and Types

// Basic types
let x = 42              // int
let y = 3.14            // float
let name = "Charl"      // string
let active = true       // bool

// Arrays
let numbers = [1, 2, 3, 4, 5]

// Print values
print("x = " + str(x))
print("y = " + str(y))
print("name = " + name)

Basic Math with Tensors

// Create tensors
let a = tensor([1.0, 2.0, 3.0])
let b = tensor([4.0, 5.0, 6.0])

// Element-wise operations
let sum = tensor_add(a, b)
let product = tensor_mul(a, b)

// Reduction operations
let total = tensor_sum(a)
let average = tensor_mean(a)

print("Sum: " + str(total))
print("Mean: " + str(average))

Simple Neural Network

// Initialize weights
let w = tensor_randn([2, 1])
let b = tensor_zeros([1])

// Input
let x = tensor([1.0, 0.5])

// Forward pass
let z = nn_linear(x, w, b)
let output = nn_sigmoid(z)

print("Output: " + str(tensor_sum(output)))

Control Flow

// If-else
let score = 85
if score >= 90 {
    print("Grade: A")
} else if score >= 80 {
    print("Grade: B")
} else {
    print("Grade: C")
}

// While loop
let i = 0
while i < 5 {
    print("Iteration: " + str(i))
    i = i + 1
}

// Match expression
let status = match score {
    100 => "Perfect"
    90 => "Excellent"
    _ => "Good"
}
print(status)

Functions

// Define a function
fn square(x: float) -> float {
    return x * x
}

// Call the function
let result = square(5.0)
print("5 squared = " + str(result))

// Function with multiple parameters
fn greet(name: string, age: int) -> string {
    return "Hello, " + name + "! You are " + str(age) + " years old."
}

print(greet("Alice", 25))

Training a Model

Complete example of gradient descent:

// Initialize parameter
let x = tensor([0.0])
let lr = 0.1
let epochs = 50
let epoch = 0

// Optimize f(x) = (x - 5)^2 to find x = 5
while epoch < epochs {
    // Forward
    let x_val = tensor_item(x)
    let loss = (x_val - 5.0) * (x_val - 5.0)

    // Gradient: d/dx[(x-5)^2] = 2(x-5)
    let grad = 2.0 * (x_val - 5.0)

    // Update: x = x - lr * grad
    let update = x_val - lr * grad
    let x = tensor([update], [1])

    // Print progress every 10 epochs
    let mod_check = epoch % 10
    if mod_check == 0 {
        print("Epoch " + str(epoch) + ": x = " + str(x_val) + ", loss = " + str(loss))
    }

    epoch = epoch + 1
}

print("Final x: " + str(tensor_sum(x)))

Next Steps