Pandas 根据具体情况映射多列

Question

我的组织对各种员工属性使用特殊代码。我们正在迁移到一个新系统，我必须根据一定的逻辑将这些代码映射到新代码。

这是我的映射 df

Mappings

:

State    Old_Mgmt    New_Mgmt   Old_ID   New_ID   New_Site
   01        A001        A100     0000     0101        123
   01        A002        A100     0000     0102       
   01        A003        A105     0000     0103        123       
   02        A001        A100     0000     0101        

这是

EmployeeData

：

State      Management     ID    Site
   01            A001   0000     456
   01            A002   0000     987
   02            A002   0000     987
....

映射的逻辑是遍历

EmployeeData

的每一行，如果存在

State

、

Management

和

ID

匹配，那么它将更新为相应的

New_

值。然而，对于

Site

，仅当

New_Site

不为空/NaN 时，它才会更新站点 ID。此映射将修改原始数据框。

根据上述映射，新的

EmployeeData

将是：

State      Management     ID    Site
   01            A100   0101     123 (modified this row)
   01            A100   0102     987 (modified this row)
   02            A002   0000     987
....

我最初的想法是做这样的事情：

for i,r in EmployeeData.iterrows(): # For each employee row
    # Create masks for the filters we are looking for
    mask_state = Mappings['State'] = r['State']
    mask_mgmt = Mappings['Old_Mgmt'] = r['Management']
    mask_id = Mappings['Old_ID'] = r['ID']

    # Filter mappings for the above 3 conditions
    MATCH = Mappings[mask_state & mask_mgmt & mask_id]

    if MATCH.empty: # No matches found
        print("No matches found in mapping. No need to update. Skipping.")
        continue
    
    MATCH = MATCH.iloc[0] # If a match is found, it will correspond to only 1 row
    EmployeeData.at[i, 'Management'] = MATCH['New_Mgmt']
    EmployeeData.at[i, 'ID'] = MATCH['New_ID']
    if pd.notna(MATCH['New_Site']):
        EmployeeData.at[i, 'Site'] = MATCH['New_Site']

然而，这似乎相当低效，因为我必须过滤每一行的映射。如果仅映射 1 列，我会执行以下操作：

# Make a dict mapping Old_Mgmt -> New_Mgmt
MGMT_MAPPING = pd.Series(Mappings['New_Mgmt'].values,index=Mappings['Old_Mgmt']).to_dict()
mask_state = Mappings['State'] = r['State']

EmployeeData.loc[mask_state, 'Management'] = EmployeeData.loc[mask_state, 'Management'].replace(MGMT_MAPPING)

但这不适用于我的情况，因为我需要映射多个值

Answer 1

尝试：

# merge mappings to EmployeeData
out = EmployeeData.merge(
    Mappings,
    left_on=["State", "Management"],
    right_on=["State", "Old_Mgmt"],
    how="left",
)

# fill NaN values with old values
out["New_ID"] = out["New_ID"].fillna(out["ID"])
out["New_Mgmt"] = out["New_Mgmt"].fillna(out["Management"])

# create final dataframe, rename columns
out = out[["State", "New_Mgmt", "New_ID", "Site"]].rename(
    columns={"New_Mgmt": "Management", "New_ID": "ID"}
)

print(out)

打印：

   State Management     ID  Site
0      1       A100  101.0   456
1      1       A100  102.0   987
2      2       A002    0.0   987