Pandas中如何用isna()判断，仅对非全空列执行.str操作？

Great question! That error pops up because when your Merchant_old column is entirely filled with NaNs, pandas typically sets its data type to float instead of the object (string) dtype that the .str accessor requires. Trying to use .str on a float column throws the error you're seeing.

Here's a straightforward way to apply your string processing only when the column has at least one non-null value:

# Check if there's any non-NaN value in the column
if df1['Merchant_old'].notna().any():
    # Run the string operations only if non-NaNs exist
    df1['Merchant_old'] = df1['Merchant_old'].str.lower().str.strip()

Let's break this down: #

• df1['Merchant_old'].notna() generates a boolean Series where each entry is True if the value isn't NaN, and False otherwise.
• .any() checks if there's at least one True in that Series. If yes, it means the column isn't all NaNs, so it's safe to use .str.
• This conditional check skips the string processing entirely when the column is all NaNs, avoiding the dtype mismatch error.

Another option (for mixed NaN/string columns): #

If your column usually has a mix of strings and NaNs (and only rarely is all NaNs), you could convert the column to string dtype first. Just note this will turn NaNs into the literal string "nan", which you might want to convert back to actual NaNs afterward:

# Convert column to string dtype (works even with mixed NaNs/strings)
df1['Merchant_old'] = df1['Merchant_old'].astype(str)
# Apply your string transformations
df1['Merchant_old'] = df1['Merchant_old'].str.lower().str.strip()
# Optional: Turn "nan" strings back into proper NaN values
df1['Merchant_old'] = df1['Merchant_old'].replace('nan', pd.NA)

But the first approach is cleaner because it avoids unnecessary conversions when the column is all NaNs.

内容的提问来源于stack exchange，提问作者Manu Mohan