使用.loc[row_indexer，col_indexer] =值更新列时收到警告

Question

我想对我收到的数据做些清理。

守则如下：

import pandas as pd

def cleanup(df: pd.DataFrame) -> pd.DataFrame:
    # Remove entries from the IT dept
    mask = (df['dept'] != 'IT')
    df = df[mask]

    # Rename the dept from marketing to comms for the remaining rows
    mask = df['dept'] == 'marketing'
    df.loc[mask, 'dept'] = "comms"
    # The warning occurs here...

    # Rename the dept from accounting to finance for the remaining rows
    mask = df['dept'] == 'accounting'
    df.loc[mask, 'dept'] = 'finance'

    return df

data = [[1,"marketing"],[2,"accounting"],[3,"marketing"],[4,"IT"],[5,"IT"],[6,"board"]]
df = pd.DataFrame(data, columns = ['id', 'dept'])
df=cleanup(df)

我收到以下警告：

/opt/conda/lib/python3.7/site-packages/pandas/core/indexing.py:480: SettingWithCopyWarning:一个值试图在来自DataFrame的片的副本上设置。尝试使用.locrow_indexer，col_indexer = value代替 请参阅文档中的注意事项：guide/indexing.html#returning-a-view-versus-a-copy self.objitem =s

我有点担心这个警告，因为返回的数据是正确的，而且文档似乎不适用于这种情况。

我的密码有什么问题吗？或者我能安全地忽略这个警告吗？

Answer 1

错误说出来了。您的df是df=df[mask]的另一个框架的一部分。尝试更新原始帧，而不是片：

def cleanup(df: pd.DataFrame) -> pd.DataFrame:
    # Remove entries from the IT dept
    mask1 = (df['dept'] != 'IT')

    # Rename the dept from marketing to comms for the remaining rows
    mask2 = df['dept'] == 'marketing'
    df.loc[mask1 & mask2, 'dept'] = "comms"
    # The warning occurs here...

    # Rename the dept from accounting to finance for the remaining rows
    mask2 = df['dept'] == 'accounting'
    df.loc[mask1&mask2, 'dept'] = 'finance'

    return df

data = [[1,"marketing"],[2,"accounting"],[3,"marketing"],[4,"IT"],[5,"IT"],[6,"board"]]
df = pd.DataFrame(data, columns = ['id', 'dept'])
df=cleanup(df)

修改后的函数返回新的df ，其中 IT值在dept中。实际上，如果您不想要这些记录，您可以复制并更新：

def cleanup(df: pd.DataFrame) -> pd.DataFrame:
    # Remove entries from the IT dept
    mask = (df['dept'] != 'IT')

    # we copy the data frame here so it's no longer a slice
    df = df[mask].copy()  

    # Rename the dept from marketing to comms for the remaining rows
    mask = df['dept'] == 'marketing'
    df.loc[mask, 'dept'] = "comms"
    # The warning occurs here...

    # Rename the dept from accounting to finance for the remaining rows
    mask = df['dept'] == 'accounting'
    df.loc[mask, 'dept'] = 'finance'

    return df