Python DataFrame中query结合assign使用报错问题排查

这是我正在使用的数据集：

运行以下代码时出现错误：

mast_df.query("(Time_Orderd != '0:-:0') & (Time_Order_picked != '0:-:0')").assign(pick_time = mast_df.Time_Order_picked.str.split(":",expand = True)[0].astype("int")*60+mast_df.Time_Order_picked.str.split(":",expand = True)[1].astype("float")-mast_df.Time_Orderd.str.split(":",expand = True)[0].astype("int")*60+mast_df.Time_Orderd.str.split(":",expand = True)[1].astype("float"))

错误信息：

ValueError: could not convert string to float: '-'

但拆分步骤运行就没有报错：

mast_df = mast_df.query("(Time_Orderd != '0:-:0') & (Time_Order_picked != '0:-:0')")
mast_df.assign(
    pick_time = mast_df.Time_Order_picked.str.split(":",expand = True)[0].astype("int")*60+mast_df.Time_Order_picked.str.split(":",expand = True)[1].astype("float")
    -mast_df.Time_Orderd.str.split(":",expand = True)[0].astype("int")*60+mast_df.Time_Orderd.str.split(":",expand = True)[1].astype("float"))

看起来第一段代码里的query筛选没生效——明明已经排除了值为'0:-:0'的行，却还是出现了转换失败的问题。之前我把空值替换成了'0:-:0'，后续会再处理这部分数据。

最终用lambda函数解决了问题，代码如下：

mast_df.query("(Time_Orderd != '0:-:0') & (Time_Order_picked != '0:-:0')").assign(
    pick_time = lambda df_ : df_.Time_Order_picked.str.split(":",expand = True)[0].astype("int")*60+
    df_.Time_Order_picked.str.split(":",expand = True)[1].astype("float")-
    df_.Time_Orderd.str.split(":",expand = True)[0].astype("int")*60+
    df_.Time_Orderd.str.split(":",expand = True)[1].astype("float"))

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