使用numpy insert时未插入行反而替换行的问题排查

尝试编写向CSV文件插入行的程序，使用numpy.insert实现，但实际操作是替换行而非插入，最终输出的CSV行数未增加（预期11行，实际仅10行）。

当前代码 #

import numpy as np
import pandas as pd

#import the csv file
file = np.loadtxt("name.csv", skiprows=1, dtype='<U70', delimiter =',')
#get row and col count
fileShape = file.shape
rows = fileShape[0]
cols = fileShape[1]

#iterate over array
for row in range(rows):
    for col in range(cols):
        #the only column that matters to me is the 5th one of my csv
        #also blocking overflow with the row check
        if (col == 4 and row + 1 < rows):
            #if the current row does not equal the next row
            #then I want to insert a row to do some things in Excel with
            if (link[row][col] != link[row+1][col]):
                #grab the next row and store it
                temp = file[row+1]
                #replace the 6th column with nothing
                temp[5] = ""
                #insert the new row into the array
                np.insert(file, row, [temp], 0)

outfile = pd.DataFrame(file)
outfile.to_csv("OutFile.csv")

输入CSV内容 #

ccType,number,date,payee,total,indAmt,memo,category
mastercard,30,11/21/2022,Bluejam,287.24,44.33,,Sports
mastercard,30,11/23/2022,Fanoodle,287.24,95.95,,Health
mastercard,30,11/25/2022,Eazzy,287.24,1.2,,Automotive
mastercard,30,11/26/2022,Dabfeed,287.24,68.97,,Games
mastercard,30,11/30/2022,Jaloo,287.24,76.79,,Games
mastercard,50,7/4/2023,Shufflebeat,317.13,91.91,,Sports
mastercard,50,7/4/2023,Meembee,317.13,94.69,,Toys
mastercard,50,7/5/2023,Jabberbean,317.13,67.01,,Computers
mastercard,50,7/28/2023,Wikibox,317.13,33.18,,Movies
mastercard,50,7/29/2023,Shufflebeat,317.13,30.34,,Automotive

实际输出CSV内容 #

ccType,number,date,payee,total,indAmt,memo,category
mastercard,30,11/21/2022,Bluejam,287.24,44.33,,Sports
mastercard,30,11/23/2022,Fanoodle,287.24,95.95,,Health
mastercard,30,11/25/2022,Eazzy,287.24,1.2,,Automotive
mastercard,30,11/26/2022,Dabfeed,287.24,68.97,,Games
mastercard,30,11/30/2022,Jaloo,287.24,76.79,,Games
mastercard,50,7/4/2023,Shufflebeat,317.13,,,Sports
mastercard,50,7/4/2023,Meembee,317.13,94.69,,Toys
mastercard,50,7/5/2023,Jabberbean,317.13,67.01,,Computers
mastercard,50,7/28/2023,Wikibox,317.13,33.18,,Movies
mastercard,50,7/29/2023,Shufflebeat,317.13,30.34,,Automotive

注意：输出中mastercard,50,7/4/2023,Shufflebeat,317.13,,,Sports是替换了原行，而非插入，行数未增加。

核心问题点 #

• numpy.insert不修改原数组：np.insert()会返回一个插入后的新数组，不会直接修改原数组file，必须将返回值重新赋值给file才会生效。
• 变量名错误：代码中使用了未定义的link变量，应替换为实际的数组变量file。
• 插入位置错误：要在当前行和下一行之间插入新行，插入位置应为row+1，而非row。
• 循环遍历逻辑问题：正向遍历原数组行数时，插入行后数组长度增加会导致后续索引错位，需改为反向遍历。

修正后的代码 #

import numpy as np
import pandas as pd

# 单独读取表头，避免插入行时丢失表头
header = np.loadtxt("name.csv", max_rows=1, dtype='<U70', delimiter=',')
file = np.loadtxt("name.csv", skiprows=1, dtype='<U70', delimiter=',')
rows = file.shape[0]

# 反向遍历行索引，防止插入新行后原索引失效
for row in range(rows-2, -1, -1):
    # 检查第5列（索引4）当前行与下一行是否不同
    if file[row][4] != file[row+1][4]:
        # 复制下一行并清空第6列（索引5）
        temp = file[row+1].copy()
        temp[5] = ""
        # 插入新行到row+1位置，并更新原数组
        file = np.insert(file, row+1, temp, axis=0)

# 合并表头和处理后的数据
result = np.vstack([header, file])
# 保存为CSV，禁用索引和默认表头
outfile = pd.DataFrame(result)
outfile.to_csv("OutFile.csv", index=False, header=False)

修正说明 #

• 单独读取表头，确保输出CSV保留原始表头结构
• 反向遍历避免插入行后索引错位问题
• 将np.insert()的返回值赋值给file，确保原数组被更新
• 使用copy()避免修改原数组中的行数据
• 最终合并表头与数据，保证输出格式正确

内容的提问来源于stack exchange，提问作者C0deGyver