import torch
import numpy as np
import networkx as nx
import matplotlib.pyplot as plt


def adjacency_to_incidence(A, n_nodes, max_edges):
    """
    Converts an adjacency matrix A to an incidence matrix B.
    Pads the incidence matrix to have max_edges columns.
    
    Args:
        A (torch.Tensor): Adjacency matrix
        n_nodes (int): Number of nodes
        max_edges (int): Maximum number of edges to include
        
    Returns:
        B (torch.Tensor): Incidence matrix
        n_edges (int): Actual number of edges
    """
    edges = []
    for i in range(n_nodes):
        for j in range(i+1, n_nodes):
            if A[i, j] != 0:
                edges.append((i, j))
    
    n_edges = len(edges)
    B = torch.zeros(n_nodes, max_edges)
    
    for idx, (i, j) in enumerate(edges):
        if idx < max_edges:
            B[i, idx] = 1
            B[j, idx] = -1
    
    return B, n_edges


def get_laplacian(B):
    """
    Compute Laplacian matrix from incidence matrix B.
    
    Args:
        B (torch.Tensor): Incidence matrix [batch_size, n_nodes, n_edges]
        
    Returns:
        L (torch.Tensor): Laplacian matrix [batch_size, n_nodes, n_nodes]
    """
    return torch.einsum('ijk,ilk->ijl', B, B)


def compute_normalized_laplacian(adj):
    """
    Compute normalized Laplacian from adjacency matrix.
    
    Args:
        adj (torch.Tensor): Adjacency matrix [n_nodes, n_nodes]
        
    Returns:
        L (torch.Tensor): Normalized Laplacian matrix [n_nodes, n_nodes]
    """
    # Get degree matrix
    deg = adj.sum(dim=-1)
    
    # Avoid division by zero
    deg = torch.clamp(deg, min=1e-10)
    
    # Compute D^(-1/2)
    D_inv_sqrt = torch.diag(deg.pow(-0.5))
    
    # Compute normalized Laplacian: L = I - D^(-1/2) A D^(-1/2)
    I = torch.eye(adj.size(0), device=adj.device)
    L = I - D_inv_sqrt @ adj @ D_inv_sqrt
    
    return L


def compute_unnormalized_laplacian(adj):
    """
    Compute unnormalized Laplacian from adjacency matrix.
    
    Args:
        adj (torch.Tensor): Adjacency matrix [n_nodes, n_nodes]
        
    Returns:
        L (torch.Tensor): Unnormalized Laplacian matrix [n_nodes, n_nodes]
    """
    # Get degree matrix
    deg = adj.sum(dim=-1)
    D = torch.diag(deg)
    
    # Compute L = D - A
    L = D - adj
    
    return L


def visualize_graph(adj, pos=None, node_color=None, title="Graph Visualization"):
    """
    Visualize a graph from its adjacency matrix.
    
    Args:
        adj (np.ndarray or torch.Tensor): Adjacency matrix
        pos (dict, optional): Node positions for visualization
        node_color (list, optional): Node colors
        title (str): Title for the plot
    """
    # Convert to numpy if tensor
    if isinstance(adj, torch.Tensor):
        adj = adj.cpu().numpy()
    
    # Create networkx graph
    G = nx.from_numpy_array(adj)
    
    # Set up figure
    plt.figure(figsize=(10, 8))
    
    # Compute positions if not provided
    if pos is None:
        pos = nx.spring_layout(G)
    
    # Set default node colors if not provided
    if node_color is None:
        node_color = 'skyblue'
    
    # Draw the graph
    nx.draw(G, pos=pos, with_labels=True, node_color=node_color, 
            node_size=300, font_size=10, font_weight='bold',
            edge_color='gray', width=[G[u][v]['weight']*5 for u,v in G.edges()])
    
    plt.title(title)
    plt.tight_layout()
    plt.show()


def visualize_laplacian_eigenvectors(L, k=4, title="Laplacian Eigenvectors"):
    """
    Compute and visualize the first k eigenvectors of a Laplacian matrix.
    
    Args:
        L (torch.Tensor): Laplacian matrix
        k (int): Number of eigenvectors to visualize
        title (str): Title for the plot
    """
    # Convert to numpy if tensor
    if isinstance(L, torch.Tensor):
        L = L.cpu().numpy()
    
    # Compute eigendecomposition
    eigenvals, eigenvecs = np.linalg.eigh(L)
    
    # Sort by eigenvalues
    idx = eigenvals.argsort()
    eigenvals = eigenvals[idx]
    eigenvecs = eigenvecs[:, idx]
    
    # Plot the first k eigenvectors
    n_rows = (k + 1) // 2
    n_cols = 2
    
    plt.figure(figsize=(12, 3 * n_rows))
    
    for i in range(min(k, eigenvecs.shape[1])):
        plt.subplot(n_rows, n_cols, i+1)
        plt.plot(eigenvecs[:, i], 'o-')
        plt.grid(True)
        plt.title(f"Eigenvector {i+1}, λ = {eigenvals[i]:.6f}")
        
    plt.suptitle(title)
    plt.tight_layout(rect=[0, 0, 1, 0.95])  # Adjust for the super title
    plt.show()


def compare_eigenvectors(eigenvecs1, eigenvecs2, k=4, labels=None, title="Eigenvector Comparison"):
    """
    Compare two sets of eigenvectors side by side.
    
    Args:
        eigenvecs1 (numpy.ndarray): First set of eigenvectors [n_nodes, n_evecs]
        eigenvecs2 (numpy.ndarray): Second set of eigenvectors [n_nodes, n_evecs]
        k (int): Number of eigenvectors to compare
        labels (list): Optional labels for the legend
        title (str): Title for the plot
    """
    # Convert to numpy if tensor
    if isinstance(eigenvecs1, torch.Tensor):
        eigenvecs1 = eigenvecs1.cpu().numpy()
    if isinstance(eigenvecs2, torch.Tensor):
        eigenvecs2 = eigenvecs2.cpu().numpy()
    
    # Normalize eigenvectors
    eigenvecs1_norm = eigenvecs1 / np.linalg.norm(eigenvecs1, axis=0, keepdims=True)
    eigenvecs2_norm = eigenvecs2 / np.linalg.norm(eigenvecs2, axis=0, keepdims=True)
    
    # Set default labels
    if labels is None:
        labels = ["Eigenvector Set 1", "Eigenvector Set 2"]
    
    # Plot comparisons
    n_rows = (k + 1) // 2
    n_cols = 2
    
    plt.figure(figsize=(14, 3 * n_rows))
    
    for i in range(min(k, eigenvecs1.shape[1], eigenvecs2.shape[1])):
        plt.subplot(n_rows, n_cols, i+1)
        
        # Handle sign ambiguity: flip eigenvector if it better matches the flipped version
        flipped = np.linalg.norm(eigenvecs1_norm[:, i] + eigenvecs2_norm[:, i]) < np.linalg.norm(eigenvecs1_norm[:, i] - eigenvecs2_norm[:, i])
        ev2_to_use = -eigenvecs2_norm[:, i] if flipped else eigenvecs2_norm[:, i]
        
        plt.plot(eigenvecs1_norm[:, i], 'o-', label=labels[0])
        plt.plot(ev2_to_use, 's-', label=labels[1])
        plt.grid(True)
        plt.title(f"Eigenvector {i+1}")
        plt.legend()
        
    plt.suptitle(title)
    plt.tight_layout(rect=[0, 0, 1, 0.95])  # Adjust for the super title
    plt.show()