如何通过循环将文件夹中的1D numpy数组堆叠为2D数组？

需要读取文件夹内296个.txt文件中的1D数组（每个数组含30个元素），将每个1D数组作为行组成一个2D数组。此前手动逐个读取文件并命名为array1、array2……后用np.vstack可完成任务，但使用循环实现时，尝试了vstack、stack、append、concatenate等方法均未成功，创建空2D数组B逐个赋值也无法正常运行。

原代码 #

import os
  
# Folder Path
path = "C:\\2022-07-25 second result more iteration"
 
# Change the directory
os.chdir(path)

B = np.empty([296, 30], dtype=float)
 
 # iterate through all file
for file in os.listdir():
    # Check whether file is in text format or not
    if file.endswith(".txt"):
        file_path = f"{path}\{file}"
  
        
        # open the files
        A = open(f"{path}\{file}", "r")
        A = np.loadtxt(A, delimiter="\t")
        A = np.asarray(A)
        A = A[:, 0] # because of some reasons the txt files have two similar columns that I just take one of them
        A = np.subtract(A, n) # just some math (n is another 1D array same size as A)
        A = np.divide(A, n) # just some math
        
        #A = A.T
        #print("A = ", A)
        #B = np.vstack (A)
        #B = np.stack(A)
        #B = np.stack([A])
        #B = np.concatenate([A])
        # B = np.append(B, A, axis=0)
        # print("B in loop = ", B)
        B[i] = A
     
             
print("B= ", B)

im=plt.imshow(B, origin='lower', extent=[650, 850, 1, 296], aspect='auto', cmap=cm.seismic,  norm=colors.CenteredNorm(), interpolation='None')

其中一个txt文件内容（仅取一列，两列内容一致） #

6563.64300386213    6563.64300386213
7627.5296220466 7627.5296220466
8922.36588941225    8922.36588941225
10515.1799846774    10515.1799846774
12493.117633046 12493.117633046
14970.0858566   14970.0858566
18090.6895050039    18090.6895050039
22026.1642649415    22026.1642649415
26946.5638244996    26946.5638244996
32926.5439947365    32926.5439947365
39699.096077692 39699.096077692
46279.3014208503    46279.3014208503
50755.8724722489    50755.8724722489
51032.9873391785    51032.9873391785
46565.0683371707    46565.0683371707
39022.807682766 39022.807682766
30795.2702323368    30795.2702323368
23449.1094708325    23449.1094708325
17518.9650853553    17518.9650853553
12968.5269428127    12968.5269428127
9591.79370475353    9591.79370475353
7215.20723743423    7215.20723743423
5791.11928730885    5791.11928730885
5452.1471216442 5452.1471216442
6480.10226555175    6480.10226555175
8910.98447577286    8910.98447577286
11691.5414159922    11691.5414159922
13052.3996945771    13052.3996945771
12588.0161703823    12588.0161703823
11198.2182945086    11198.2182945086

核心错误 #

原代码中未定义并递增索引变量i，导致执行B[i] = A时直接报错。此外还有几个可优化的细节：

1. 文件路径拼接用f"{path}\{file}"在Windows下可能因转义字符出现问题，建议用os.path.join更安全
2. np.loadtxt可直接接收文件路径，无需手动调用open
3. os.listdir()返回的文件顺序不固定，若需要按特定顺序读取，建议对文件列表排序

修正后的代码 #

import os
import numpy as np
import matplotlib.pyplot as plt
from matplotlib import cm, colors

# 文件夹路径
path = "C:\\2022-07-25 second result more iteration"

# 初始化索引
i = 0
# 创建空数组
B = np.empty([296, 30], dtype=float)

# 获取所有txt文件并排序（可选，保证读取顺序稳定）
txt_files = sorted([f for f in os.listdir(path) if f.endswith(".txt")])

for file in txt_files:
    file_path = os.path.join(path, file)
    # 直接读取文件，取第一列
    A = np.loadtxt(file_path, delimiter="\t")[:, 0]
    # 执行数学运算（确保n已提前定义且形状与A一致）
    A = (A - n) / n
    # 赋值到B的对应行
    B[i] = A
    # 索引递增
    i += 1

print("B= ", B)

im = plt.imshow(B, origin='lower', extent=[650, 850, 1, 296], 
                aspect='auto', cmap=cm.seismic,  
                norm=colors.CenteredNorm(), interpolation='None')
plt.show()

另一种更灵活的方案（无需提前指定数组大小） #

如果不确定文件数量，或者不想硬编码296，可以先把所有数组存入列表，最后再拼接成2D数组：

import os
import numpy as np
import matplotlib.pyplot as plt
from matplotlib import cm, colors

path = "C:\\2022-07-25 second result more iteration"
data_list = []

txt_files = sorted([f for f in os.listdir(path) if f.endswith(".txt")])

for file in txt_files:
    file_path = os.path.join(path, file)
    A = np.loadtxt(file_path, delimiter="\t")[:, 0]
    A = (A - n) / n
    data_list.append(A)

# 拼接成2D数组
B = np.vstack(data_list)

print("B= ", B)
print("B的形状：", B.shape)

im = plt.imshow(B, origin='lower', extent=[650, 850, 1, len(B)], 
                aspect='auto', cmap=cm.seismic,  
                norm=colors.CenteredNorm(), interpolation='None')
plt.show()

内容的提问来源于stack exchange，提问作者Nasrin Asgari