【随笔】Python如何读取.x的DirectX的3D格式文件

Python如何读取.x的DirectX的3D格式文件

(其实这是一个陈年老Bug)之前2015年大三下学期看张若愚版本的<>发现上面的读取.x的DirectX的3D格式文件的程序有错，最近仔细读了读源码，发现他默认模型的点、面、颜色属性是按顺序排列的，其实不同3D建模工具产生的.x文件的排列是不一样的。

---

修改后的程序

还有就是作者当时用的那个VPython3D库其实真的好用，但是目前这个项目到0.6版就不再维护了，和作者交流他说：

#!/usr/bin/env python
# -*- coding: utf-8 -*-
import numpy as np
from visual import *

def search_line_startswith(f, pattern):
    while True:
        line = f.readline()
        if line == "": return False
        if line.strip().startswith(pattern):
            return True

def test_read(f,pattern):
    i=0;
    while True:
        line = f.readline();print i;i+=1;
        if line == "": print 'wtf';return False;
        if line.strip().startswith(pattern):
            return True
        

def read_directx_model(filename, double=False):
    f = open(filename)
    search_line_startswith(f, "Mesh") # 定位Mesh行
            
    point_num = int(f.readline().split(";")[0]) # 读入顶点数
    points = np.zeros((point_num, 3)) # 初始化保存顶点坐标的数组
    # 读入所有顶点坐标
    for i in xrange(point_num):
        points[i, :] = [float(x) for x in f.readline().split(";")[:3]]

    this_line =f.readline();#print this_line;
    face_num = int(this_line.split(";\r\n")[0]) # 读入面数
    
    # 初始化三角形的顶点坐标数组
    pfaces = zeros( (face_num*3 , 3)) # 正面的顶点下标
           
    # 建造所有的模型面的顶点坐标数组，将面的顶点下标替换为顶点坐标
    cnt = 0
    for i in xrange(face_num):
        p_index = [int(x) for x in f.readline().split(";")[1].split(",")]
        for j in xrange(3):
            pfaces[cnt, :] = points[p_index[j], :]
            cnt += 1
            
    search_line_startswith(f, "MeshMaterialList") # 定位材质
    color_num = int(f.readline().split(";")[0])  # 读入材质数
    face_color_num = int(f.readline().split(";")[0]) # 读出面数
    
    # 读入每个面的材质下标
    face_color_list=[]
    for i in xrange(face_color_num):
        sss = f.readline();#print sss;
        face_color_list.append( int(sss.strip(",;\n\r")) )
    
    colors = np.zeros((color_num, 3)) # 初始化储存材质颜色的数组
    # 读入所有材质的颜色
    for i in xrange(color_num):
        search_line_startswith(f, "Material")
        colors[i,:] = [float(x) for x in f.readline().split(";")[:3]]
   
    # 初始化保存每个顶点颜色的数组
    pcolors = np.zeros( (face_color_num*3, 3), np.float)

    # 建造顶点颜色数组，将面的材质下标替换为顶点的材质颜色
    cnt = 0
    for i in xrange(face_color_num):
        color_idx = face_color_list[i] 
        for j in xrange(3):
            pcolors[cnt, :] = colors[color_idx,:]
            cnt += 1
            
    f = open(filename);
    search_line_startswith(f, "MeshNormals") # 定位法线方向 
    #test_read(f,"MeshNormals");
    
    # 读入法线方向
    this_line = f.readline();#print [this_line];#print 'here';
    point_num = int(this_line.split(";")[0])
    npoints = np.zeros((point_num, 3))
    for i in xrange(point_num):
        npoints[i,:] = [float(x) for x in f.readline().split(";")[:3]]
            
    face_num = int(f.readline().split(";")[0]) # 读入面数
    
    # 初始化顶点法线方向数组
    nfaces = np.zeros( (face_num*3 , 3), np.float)
        
    # 建造顶点法线方向数组，将每个面的法线下标替换为实际的法线方向
    cnt = 0
    for i in xrange(face_num):
        p_index = [int(x) for x in f.readline().split(";")[1].split(",")]
        for j in xrange(3):
            nfaces[cnt, :] = npoints[p_index[j],:]
            cnt += 1

    f.close()
    
    if double:
       pfaces = np.vstack((pfaces, pfaces[::-1]))
       nfaces = np.vstack((nfaces, nfaces[::-1]))
       pcolors = np.vstack((pcolors, pcolors[::-1]))
    
    model_frame = frame()
    # 使用 顶点坐标数组，顶点颜色数组和顶点法线方向数组构造模型
    # print pfaces[0].shape();print pfaces[0].shape();
    #pcolors[:,0]=np.random.random(size = pcolors[:,0].shape)*0.7;
    #pcolors[:,1]=np.random.random(size = pcolors[:,1].shape)*0.7;
    #pcolors[:,2]=np.random.random(size = pcolors[:,2].shape)*0.7;
    #print pcolors[:,0];print pcolors[1];
    model = faces( pos = pfaces, normal = nfaces, color = np.random.random(3), frame = model_frame )
    return model_frame
    
if __name__ == "__main__":
    import sys
    scene.background = (12, 222, 222)
    #scene.title = sys.argv[1]
    read_directx_model("models_dx/truck.x", True)