将文本格式矩阵转换为NumPy二维数组并导入NetworkX构建图的技术问题求助

Let's break down and fix your issue step by step:

1. Why the NetworkX Error Occurs #

Your current code converts the entire processed string into a 1D numpy array (each character is a single element in a flat array), but nx.from_numpy_matrix expects a 2D array (either an adjacency matrix or a grid structure). That's exactly why you're seeing the "Input array must be 2D, not 0" error.

2. Fix the Array Conversion #

First, we need to turn your processed text into a proper 2D numpy array—each line of the text becomes a row in the array, and each character gets converted to an integer.

3. Build the Graph Correctly #

Instead of using nx.from_numpy_matrix directly (it's designed for adjacency matrices, not grid data), we'll iterate through the 2D grid manually:

• Treat every position with value 1 (original "w") as a node (we'll use the (row, column) tuple as a unique node ID).
• For each node, check its four adjacent positions (up, down, left, right). If an adjacent position also has value 1, add an undirected edge between the two nodes.

Corrected Code #

import glob
import os
import numpy as np
import networkx as nx

def create_matrix_from_layout(folder):
    files = glob.glob(folder)
    file_graphs = {}  # Store graphs for each processed file
    
    for f in files:
        fname = os.path.basename(f)
        with open(f, "r") as tmp:
            tmp_data = tmp.read()
            
            # Step 1: Replace characters per your requirements
            tmp_data = tmp_data.replace("w", "1").replace("A", "2")
            # Replace all other characters (except newlines) with 0
            tmp_data = ''.join(['0' if c not in {'1','2','\n'} else c for c in tmp_data])
            
            # Step 2: Convert to a 2D numpy array of integers
            # Split into lines, skip empty lines from trailing newlines
            lines = [line.strip() for line in tmp_data.split('\n') if line.strip()]
            # Convert each line to a list of integers, then to numpy array
            grid = np.array([[int(c) for c in line] for line in lines], dtype=int)
            print(f"Processed grid for {fname}:\n{grid}\n")
            
            # Step 3: Build the NetworkX graph
            G = nx.Graph()
            rows, cols = grid.shape
            
            # Iterate over every cell in the grid
            for i in range(rows):
                for j in range(cols):
                    if grid[i][j] == 1:
                        # Add the grid position as a node
                        G.add_node((i,j))
                        
                        # Check adjacent cells for valid connections
                        # Up
                        if i > 0 and grid[i-1][j] == 1:
                            G.add_edge((i,j), (i-1,j))
                        # Down
                        if i < rows - 1 and grid[i+1][j] == 1:
                            G.add_edge((i,j), (i+1,j))
                        # Left
                        if j > 0 and grid[i][j-1] == 1:
                            G.add_edge((i,j), (i,j-1))
                        # Right
                        if j < cols - 1 and grid[i][j+1] == 1:
                            G.add_edge((i,j), (i,j+1))
            
            file_graphs[fname] = G
            print(f"Graph for {fname} has {G.number_of_nodes()} nodes and {G.number_of_edges()} edges\n")
    
    return file_graphs

# Run the function with your layout files
graphs = create_matrix_from_layout("layouts/*.txt")

Key Modifications Explained #

• 2D Array Conversion: We split the processed text into lines, convert each line to a list of integers, then build a 2D numpy array. This gives us the grid structure NetworkX can work with.
• Graph Construction: Manual iteration over the grid lets us precisely control which nodes and edges are added—only 1 positions become nodes, and edges only connect adjacent 1s.
• Node Identification: Using (row, column) tuples as node IDs makes it easy to map graph nodes back to their original positions in the text layout.

Verification #

To check the output, you can inspect any graph in the returned dictionary:

# Example: Inspect the first processed graph
first_fname = next(iter(graphs.keys()))
G = graphs[first_fname]
print(nx.info(G))

This should resolve the 2D array error and build the exact graph you need.

内容的提问来源于stack exchange，提问作者Pehy