问题
data.csv
1, 22, 3432
1, 23, \N
2, 24, 54335
2, 25, 3928
I have a csv file of data that is collected from a device. Every now and then the device doesn't relay information and outputs '\N'. I want to treat these as NaN and did this by doing
read_csv(data.csv, na_values=['\\N'])
which worked fine. However, I would prefer to have not only this string turned to NaN but any string that is in the csv file just in case the data I get in the future has a different string.
Is it possible to me to make any changes in the argument so it covers all strings?
回答1:
You have to manually pass all the keywords as a list or dict to na_values
na_values : list-like or dict, default None
Alternatively, use pd.to_numeric and set errors to coerce to convert all values to numeric after reading the csv file.
sample input df:
A B
0 1 2
1 0 \N
2 \N 8
3 11 5
4 11 Kud
df = df.apply(pd.to_numeric, errors='coerce')
output:
A B
0 1 2
1 0 NaN
2 NaN 8
3 11 5
4 11 NaN
来源：https://stackoverflow.com/questions/52229804/pandas-read-csv-and-setting-na-values-to-any-string-in-the-csv-file