/*!
# RustyTorch++ sklearn-compatible Python bindings

This crate provides Python bindings for RustyTorch++ classical ML algorithms
with full scikit-learn API compatibility for drop-in replacement.

## Features

- **Drop-in sklearn compatibility**: Same API, same behavior
- **High performance**: Rust implementation with optional GPU acceleration
- **Async support**: Non-blocking training for large datasets
- **Memory efficiency**: Optimized memory usage compared to sklearn
- **Parallel processing**: Built-in parallelization support

## Usage

```python
# Before (sklearn)
from sklearn.ensemble import RandomForestClassifier
clf = RandomForestClassifier(n_estimators=100).fit(X, y)

# After (rustytorch) - same code!
from rustytorch_ml import RandomForestClassifier
clf = RandomForestClassifier(n_estimators=100).fit(X, y)
```

## Performance Features

- GPU acceleration with `device='cuda:0'`
- Async training with `clf.fit_async(X, y)`
- Batch prediction with `clf.predict_batch(X, batch_size=1000)`
- Memory-mapped datasets for large data
*/

// Allow unsafe operations in unsafe functions generated by PyO3 macros
#![allow(unsafe_op_in_unsafe_fn)]

use pyo3::prelude::*;
use pyo3::types::PyModuleMethods;
use pyo3::types::{PyDict, PyTuple};
use pyo3::wrap_pyfunction;

pub mod error;
pub mod utils;
pub mod wrappers;

pub use wrappers::simple_model_selection::{cross_val_score, train_test_split};
use wrappers::{
    AgglomerativeClustering, DBSCAN, DecisionTreeClassifier, ElasticNet, GridSearchCV, KMeans,
    LabelEncoder, Lasso, LinearRegression, MinMaxScaler, OneHotEncoder, Ridge, StandardScaler,
};

/// Initialize the rustytorch_ml Python module
#[pymodule]
fn rustytorch_ml(m: &Bound<'_, PyModule>) -> PyResult<()> {
    let py = m.py();

    // Add module metadata
    m.add("__version__", env!("CARGO_PKG_VERSION"))?;
    m.add("__author__", "RustyTorch++ Team")?;
    m.add(
        "__description__",
        "High-performance sklearn-compatible ML library",
    )?;

    // Register exception classes
    m.add(
        "SklearnError",
        py.get_type::<pyo3::exceptions::PyRuntimeError>(),
    )?;

    // Classifiers
    m.add_class::<DecisionTreeClassifier>()?;

    // Regressors
    m.add_class::<LinearRegression>()?;
    m.add_class::<Ridge>()?;
    m.add_class::<Lasso>()?;
    m.add_class::<ElasticNet>()?;

    // Clustering
    m.add_class::<KMeans>()?;
    m.add_class::<DBSCAN>()?;
    m.add_class::<AgglomerativeClustering>()?;

    // Preprocessing
    m.add_class::<StandardScaler>()?;
    m.add_class::<MinMaxScaler>()?;
    m.add_class::<OneHotEncoder>()?;
    m.add_class::<LabelEncoder>()?;

    // Model Selection
    m.add_class::<GridSearchCV>()?;
    m.add_function(wrap_pyfunction!(cross_val_score, m)?)?;
    m.add_function(wrap_pyfunction!(train_test_split, m)?)?;

    // Utility functions
    m.add_function(wrap_pyfunction!(utils::set_random_state, m)?)?;
    m.add_function(wrap_pyfunction!(utils::get_device_info, m)?)?;
    m.add_function(wrap_pyfunction!(utils::enable_gpu_acceleration, m)?)?;

    // Performance monitoring
    m.add_function(wrap_pyfunction!(utils::get_memory_usage, m)?)?;
    m.add_function(wrap_pyfunction!(utils::benchmark_against_sklearn, m)?)?;

    // Module-level configuration
    setup_module_config(py, m)?;

    Ok(())
}

/// Setup module-level configuration and logging
fn setup_module_config(py: Python, m: &Bound<'_, PyModule>) -> PyResult<()> {
    // Set default configuration
    let config = PyDict::new(py);
    config.set_item("default_device", "cpu")?;
    config.set_item("enable_gpu", false)?;
    config.set_item("default_batch_size", 1000)?;
    config.set_item("enable_async", true)?;
    config.set_item("verbose", false)?;

    m.add("config", config)?;

    // Add version info
    let version_info = PyTuple::new(py, [0_u32, 1, 0])?;
    m.add("version_info", version_info)?;

    // Add sklearn compatibility info
    let sklearn_version = "1.3.0"; // Target sklearn compatibility version
    m.add("sklearn_compatible_version", sklearn_version)?;

    Ok(())
}

// Re-export commonly used types for internal use
pub use error::SklearnError;
pub use utils::{ArrayConverter, DeviceConfig, PyArray};

#[cfg(test)]
mod tests {
    use super::*;
    use pyo3::Python;

    #[test]
    fn test_module_creation() {
        Python::with_gil(|py| {
            let module = PyModule::new(py, "rustytorch_ml").unwrap();
            assert!(rustytorch_ml(&module).is_ok());

            // Check that classes are registered
            assert!(module.hasattr("DecisionTreeClassifier").unwrap());
            assert!(module.hasattr("LinearRegression").unwrap());
            assert!(module.hasattr("KMeans").unwrap());
            assert!(module.hasattr("StandardScaler").unwrap());
            assert!(module.hasattr("GridSearchCV").unwrap());

            // Check version info
            let version: &str = module.getattr("__version__").unwrap().extract().unwrap();
            assert_eq!(version, env!("CARGO_PKG_VERSION"));
        });
    }
}