Python中合并DataFrame后按EID_STAVBA合并行的实现方法

我原本合并两个DataFrame的代码是：

GES1 = pd.merge(db1, db2, on = 'KO-STA', how = 'left')

其中db1、db2的数据结构如下：

db1数据： #

,EID_STAVBA,ST_STAVBE,PLIN,LETO_IZGRA,LETO_OBNOV,LETO_OBNO0,BRUTO_TLOR,EID_OBCINA,SIFKO,EID_DEL_STAVBE,VRSTA_DEJANSKE_RABE_DEL_ST_ID,LETO_OBNOVE_INSTALACIJ,LETO_OBNOVE_OKEN,POVRSINA,ST_DELA_STAVBE,VRSTA_STANOVANJA_ID,UPORABNA_POVRSINA,KO-STA,EID_PARCEL,KO_ID,ST_PARCELE,KO-PAR
0,100200000250868938,4645,0.0,1980.0,,,,110200000110277170,1722,100200000250868958,1,,,37.8,1,,37.8,1722-4645,100100000210682339,1722,903,1722-903

db2数据： #

,KO-STA,STA-PAR,KO-PAR,Ukrep_Opis,Ukrep_Kolicina,Ukrep_Enota,UkrepES_Fasada,UkrepES_Streha,UkrepES_Okna,UkrepES_TC,UkrepES_TC_tip,UkrepES_Biomasa,UkrepES_Biomasa_tip,UkrepES_Prezracevanje,UkrepES_PV,UkrepES_SSE
0,1722-4645,0.0,1722-903,Zamenjava oken,0.0,0.0,0.0,0.0,2022.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0
1,1722-4645,0.0,1722-903,TI fasade,0.0,0.0,2022.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0
2,1722-4645,0.0,1722-903,SSE,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,2022.0
3,1722-4645,0.0,1722-903,PV,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,2022.0,0.0

合并后的输出为：

EID_STAVBA  ST_STAVBE  ...  UkrepES_PV  UkrepES_SSE
0  100200000250868938       4645  ...         0.0          0.0
1  100200000250868938       4645  ...         0.0       2022.0
2  100200000250868938       4645  ...         0.0          0.0
3  100200000250868938       4645  ...      2022.0          0.0

其中同一EID_STAVBA对应多行数据，我希望将同一EID_STAVBA的所有数据合并到一行，示例如下：

,EID_STAVBA,ST_STAVBE,PLIN,LETO_IZGRA,LETO_OBNOV,LETO_OBNO0,BRUTO_TLOR,EID_OBCINA,SIFKO,EID_DEL_STAVBE,VRSTA_DEJANSKE_RABE_DEL_ST_ID,LETO_OBNOVE_INSTALACIJ,LETO_OBNOVE_OKEN,POVRSINA,ST_DELA_STAVBE,VRSTA_STANOVANJA_ID,UPORABNA_POVRSINA,KO-STA,EID_PARCEL,KO_ID,ST_PARCELE,index,STA-PAR,KO-PAR,Ukrep_Opis,Ukrep_Kolicina,Ukrep_Enota,UkrepES_Fasada,UkrepES_Streha,UkrepES_Okna,UkrepES_TC,UkrepES_TC_tip,UkrepES_Biomasa,UkrepES_Biomasa_tip,UkrepES_Prezracevanje,UkrepES_PV,UkrepES_SSE
0,100200000250868938,4645,0.0,1980.0,,,,110200000110277170,1722,100200000250868958,1,,,37.8,1,,37.8,1722-4645,100100000210682339,1722,903,0,0.0,1722-903,Zamenjava oken,0.0,0.0,0.0,0.0,2022.0,0.0,2022.0,0.0,0.0,0.0,2022.0,2022.0

请问应如何修改代码实现该需求？是否需要使用groupby方法？

需要使用groupby方法实现，核心逻辑是先完成合并，再按EID_STAVBA分组，针对不同类型的列设置聚合规则：

1. 原db1中的基础列（如EID_STAVBA、ST_STAVBE等），同一EID_STAVBA下值完全一致，直接取唯一值即可；
2. db2中的数值型列（如UkrepES_Fasada、UkrepES_Okna等），需要保留分散在各行的有效年份（非0值），用max聚合就能自动提取有效数值；
3. 文本列Ukrep_Opis，可以用字符串拼接的方式合并所有操作描述。

具体代码如下：

# 执行原合并操作
GES1 = pd.merge(db1, db2, on='KO-STA', how='left')

# 定义各列的聚合规则
agg_rules = {}
for col in GES1.columns:
    if col in db1.columns:
        # 原db1的列取第一个值（同一分组下值相同）
        agg_rules[col] = 'first'
    elif col == 'Ukrep_Opis':
        # 拼接所有操作描述
        agg_rules[col] = lambda x: ','.join(x.astype(str))
    else:
        # 数值列取最大值，保留有效年份
        agg_rules[col] = 'max'

# 按EID_STAVBA分组聚合，合并为单行
result = GES1.groupby('EID_STAVBA', as_index=False).agg(agg_rules)

执行后即可得到同一EID_STAVBA对应一行的结果，其中数值列会合并分散的有效年份，文本列会汇总所有操作描述。

内容的提问来源于stack exchange，提问作者energyMax