如何提取Pandas DataFrame每行首个非零列值？

给定如下结构的Pandas DataFrame：

import pandas as pd

data = {"Column1":["0", "1", "0", "0", "1"], "Column2":["2","0","2", "0", "2"], "Column3":["3","0","3", "3", "3"]}
df = pd.DataFrame(data)
print(df)

输出：

Column1 Column2 Column3
0       0       2       3
1       1       0       0
2       0       2       3
3       0       0       3
4       1       2       3

需要生成新的DataFrame，每行保留原DataFrame中从左到右出现的首个非零值，目标结果如下：

data1 = {"Column1":["2", "1", "2", "3", "1"]}
df1 = pd.DataFrame(data1)
print(df1)

输出：

Column1
0       2
1       1
2       2
3       3
4       1

尝试使用np.argmax(y, axis=1)未成功，求解决方案。

方法一：利用NaN填充提取首个非零值 #

先将字符串类型的"0"替换为NaN，再通过行方向填充提取首个非空值：

import pandas as pd
import numpy as np

# 原始数据
data = {"Column1":["0", "1", "0", "0", "1"], "Column2":["2","0","2", "0", "2"], "Column3":["3","0","3", "3", "3"]}
df = pd.DataFrame(data)

# 将"0"替换为NaN并转换为数值类型
df_numeric = df.replace("0", np.nan).astype(float)

# 提取每行第一个非NaN值，生成新DataFrame
df1 = df_numeric.bfill(axis=1).iloc[:, 0].to_frame(name="Column1")

# 转换为字符串类型匹配目标结果
df1 = df1.astype(int).astype(str)

print(df1)

方法二：用stack+分组取首值 #

通过堆叠列转行，再按原行号分组取第一个非零值：

import pandas as pd
import numpy as np

df = pd.DataFrame({"Column1":["0", "1", "0", "0", "1"], "Column2":["2","0","2", "0", "2"], "Column3":["3","0","3", "3", "3"]})

# 替换"0"为NaN后堆叠，分组取首值
df1 = df.replace("0", np.nan).stack().groupby(level=0).first().to_frame(name="Column1")

# 转换类型
df1 = df1.astype(int).astype(str)

print(df1)

方法三：修正np.argmax的用法 #

原问题中np.argmax失效是因为数据类型和逻辑错误，修正后可以用索引取值：

import pandas as pd
import numpy as np

df = pd.DataFrame({"Column1":["0", "1", "0", "0", "1"], "Column2":["2","0","2", "0", "2"], "Column3":["3","0","3", "3", "3"]})

# 转换为数值类型
df_num = df.astype(int)

# 找到每行第一个非零值的列索引
indices = np.argmax(df_num != 0, axis=1)

# 根据索引提取对应值并生成DataFrame
result = df_num.values[range(len(df_num)), indices]
df1 = pd.DataFrame({"Column1": result.astype(str)})

print(df1)

内容的提问来源于stack exchange，提问作者Elisa