numrs2 0.3.0

//! Array operations for Python bindings
//!
//! Provides comprehensive array creation, manipulation, and operations.

use crate::array::Array;
use crate::math;
use crate::NumRs2Error;
use pyo3::exceptions::{PyIndexError, PyTypeError, PyValueError};
use pyo3::prelude::*;
use pyo3::types::{PyList, PyTuple};
use scirs2_numpy::{
    PyArrayDyn, PyArrayMethods, PyReadonlyArrayDyn, PyUntypedArrayMethods, ToPyArray,
};

/// Python wrapper for NumRS2 Array
#[pyclass(name = "Array")]
#[derive(Clone)]
pub struct PyArray {
    pub(crate) inner: Array<f64>,
}

#[pymethods]
impl PyArray {
    /// Create a new array from a Python sequence
    #[new]
    fn new(data: &Bound<'_, PyAny>) -> PyResult<Self> {
        // Try to extract as NumPy array first
        if let Ok(np_arr) = data.extract::<PyReadonlyArrayDyn<f64>>() {
            let shape = np_arr.shape().to_vec();
            let data_vec: Vec<f64> = np_arr
                .as_slice()
                .map_err(|_| {
                    PyValueError::new_err("Cannot convert NumPy array to contiguous slice")
                })?
                .to_vec();
            let array = Array::from_vec(data_vec).reshape(&shape);
            return Ok(PyArray { inner: array });
        }

        // Try as list
        if let Ok(list) = data.downcast::<PyList>() {
            let vec: Vec<f64> = list.extract()?;
            return Ok(PyArray {
                inner: Array::from_vec(vec),
            });
        }

        // Try as tuple
        if let Ok(tuple) = data.downcast::<PyTuple>() {
            let vec: Vec<f64> = tuple.extract()?;
            return Ok(PyArray {
                inner: Array::from_vec(vec),
            });
        }

        Err(PyTypeError::new_err("Expected list, tuple, or NumPy array"))
    }

    /// Get the shape of the array
    #[getter]
    fn shape(&self) -> Vec<usize> {
        self.inner.shape()
    }

    /// Get the number of dimensions
    #[getter]
    fn ndim(&self) -> usize {
        self.inner.ndim()
    }

    /// Get the total number of elements
    #[getter]
    fn size(&self) -> usize {
        self.inner.size()
    }

    /// Get the data type
    #[getter]
    fn dtype(&self) -> &str {
        "float64"
    }

    /// Reshape the array to a new shape
    fn reshape(&self, shape: Vec<usize>) -> PyResult<Self> {
        Ok(PyArray {
            inner: self.inner.clone().reshape(&shape),
        })
    }

    /// Transpose the array
    fn transpose(&self) -> PyResult<Self> {
        Ok(PyArray {
            inner: self.inner.transpose(),
        })
    }

    /// Flatten the array to 1D
    fn flatten(&self) -> PyResult<Self> {
        let size = self.inner.size();
        Ok(PyArray {
            inner: self.inner.clone().reshape(&[size]),
        })
    }

    /// Squeeze - remove dimensions of size 1
    fn squeeze(&self) -> PyResult<Self> {
        let shape: Vec<usize> = self
            .inner
            .shape()
            .iter()
            .copied()
            .filter(|&s| s != 1)
            .collect();
        if shape.is_empty() {
            Ok(PyArray {
                inner: self.inner.clone().reshape(&[1]),
            })
        } else {
            Ok(PyArray {
                inner: self.inner.clone().reshape(&shape),
            })
        }
    }

    /// Convert to a flat Python list
    pub fn tolist(&self) -> Vec<f64> {
        self.inner.to_vec()
    }

    /// Convert to NumPy array
    fn to_numpy<'py>(&self, py: Python<'py>) -> Bound<'py, PyArrayDyn<f64>> {
        let vec = self.inner.to_vec();
        let shape: Vec<usize> = self.inner.shape();

        // Convert to NumPy array with proper shape
        let arr = vec.to_pyarray(py);
        arr.reshape(shape)
            .expect("reshape to array shape should not fail since shape is valid")
    }

    /// Copy the array
    fn copy(&self) -> Self {
        self.clone()
    }

    /// Sum of all elements
    fn sum(&self) -> f64 {
        self.inner.to_vec().iter().sum()
    }

    /// Mean of all elements
    pub fn mean(&self) -> f64 {
        let vec = self.inner.to_vec();
        if vec.is_empty() {
            0.0
        } else {
            vec.iter().sum::<f64>() / vec.len() as f64
        }
    }

    /// Minimum value
    fn min(&self) -> PyResult<f64> {
        self.inner
            .to_vec()
            .iter()
            .copied()
            .min_by(|a, b| a.partial_cmp(b).unwrap_or(std::cmp::Ordering::Equal))
            .ok_or_else(|| PyValueError::new_err("Array is empty"))
    }

    /// Maximum value
    fn max(&self) -> PyResult<f64> {
        self.inner
            .to_vec()
            .iter()
            .copied()
            .max_by(|a, b| a.partial_cmp(b).unwrap_or(std::cmp::Ordering::Equal))
            .ok_or_else(|| PyValueError::new_err("Array is empty"))
    }

    /// Element-wise addition
    fn __add__(&self, other: &PyArray) -> PyResult<Self> {
        Ok(PyArray {
            inner: &self.inner + &other.inner,
        })
    }

    /// Element-wise subtraction
    fn __sub__(&self, other: &PyArray) -> PyResult<Self> {
        Ok(PyArray {
            inner: &self.inner - &other.inner,
        })
    }

    /// Element-wise multiplication
    fn __mul__(&self, other: &PyArray) -> PyResult<Self> {
        Ok(PyArray {
            inner: &self.inner * &other.inner,
        })
    }

    /// Element-wise division
    fn __truediv__(&self, other: &PyArray) -> PyResult<Self> {
        Ok(PyArray {
            inner: &self.inner / &other.inner,
        })
    }

    /// Negation
    fn __neg__(&self) -> PyResult<Self> {
        Ok(PyArray {
            inner: -self.inner.clone(),
        })
    }

    /// String representation
    fn __repr__(&self) -> String {
        format!(
            "Array(shape={:?}, dtype={}, size={})",
            self.shape(),
            self.dtype(),
            self.size()
        )
    }

    /// String conversion
    fn __str__(&self) -> String {
        self.__repr__()
    }

    /// Length for len() function
    fn __len__(&self) -> usize {
        self.inner.shape().first().copied().unwrap_or(0)
    }
}

/// Create an array from a Python sequence
#[pyfunction]
fn array(data: &Bound<'_, PyAny>) -> PyResult<PyArray> {
    PyArray::new(data)
}

/// Create an array of zeros
#[pyfunction]
fn zeros(shape: Vec<usize>) -> PyArray {
    PyArray {
        inner: Array::zeros(&shape),
    }
}

/// Create an array of ones
#[pyfunction]
fn ones(shape: Vec<usize>) -> PyArray {
    PyArray {
        inner: Array::ones(&shape),
    }
}

/// Create a 2D array with ones on the diagonal and zeros elsewhere
#[pyfunction]
fn eye(n: usize, m: Option<usize>, k: Option<isize>) -> PyArray {
    let m = m.unwrap_or(n);
    let k = k.unwrap_or(0);
    PyArray {
        inner: Array::eye(n, m, k),
    }
}

/// Create an identity matrix
#[pyfunction]
fn identity(n: usize) -> PyArray {
    PyArray {
        inner: Array::eye(n, n, 0),
    }
}

/// Create an array with evenly spaced values
#[pyfunction]
fn linspace(start: f64, stop: f64, num: usize, endpoint: Option<bool>) -> PyArray {
    let endpoint = endpoint.unwrap_or(true);
    if endpoint {
        PyArray {
            inner: math::linspace(start, stop, num),
        }
    } else {
        // Without endpoint, spacing is (stop - start) / num
        let step = (stop - start) / num as f64;
        let values: Vec<f64> = (0..num).map(|i| start + i as f64 * step).collect();
        PyArray {
            inner: Array::from_vec(values),
        }
    }
}

/// Create an array with values in a range
#[pyfunction]
fn arange(start: f64, stop: f64, step: Option<f64>) -> PyArray {
    let step = step.unwrap_or(1.0);
    PyArray {
        inner: math::arange(start, stop, step),
    }
}

/// Create an array filled with a constant value
#[pyfunction]
fn full(shape: Vec<usize>, fill_value: f64) -> PyArray {
    let size: usize = shape.iter().product();
    let data = vec![fill_value; size];
    PyArray {
        inner: Array::from_vec(data).reshape(&shape),
    }
}

/// Create an array filled with zeros (like another array)
#[pyfunction]
fn zeros_like(a: &PyArray) -> PyArray {
    zeros(a.shape())
}

/// Create an array filled with ones (like another array)
#[pyfunction]
fn ones_like(a: &PyArray) -> PyArray {
    ones(a.shape())
}

/// Concatenate arrays along an axis
#[pyfunction]
fn concatenate(arrays: Vec<PyArray>, axis: Option<usize>) -> PyResult<PyArray> {
    let _axis = axis.unwrap_or(0);
    // Simplified implementation - just concatenate flattened arrays
    let mut result = Vec::new();
    for arr in arrays {
        result.extend(arr.tolist());
    }
    Ok(PyArray {
        inner: Array::from_vec(result),
    })
}

/// Register array module functions and classes
pub fn register(m: &Bound<'_, PyModule>) -> PyResult<()> {
    // Add class
    m.add_class::<PyArray>()?;

    // Add creation functions
    m.add_function(wrap_pyfunction!(array, m)?)?;
    m.add_function(wrap_pyfunction!(zeros, m)?)?;
    m.add_function(wrap_pyfunction!(ones, m)?)?;
    m.add_function(wrap_pyfunction!(eye, m)?)?;
    m.add_function(wrap_pyfunction!(identity, m)?)?;
    m.add_function(wrap_pyfunction!(linspace, m)?)?;
    m.add_function(wrap_pyfunction!(arange, m)?)?;
    m.add_function(wrap_pyfunction!(full, m)?)?;
    m.add_function(wrap_pyfunction!(zeros_like, m)?)?;
    m.add_function(wrap_pyfunction!(ones_like, m)?)?;
    m.add_function(wrap_pyfunction!(concatenate, m)?)?;

    Ok(())
}