如何使用pandas.read_csv读取csv文件时将pandas.dataframe中的元素转换为np.float？

Question

I have a .csv file, which is an exported output from a software. This .csv file contains a lot of NaNs. I need to analyze the data by reading it into a dataframe, and use dataframe.fillna(0) to replace all the NaNs with 0. However, when I use pandas.read_csv() to import this .csv file, the type of element in the dataframe is 'str', so dataframe.fillna(0) cannot be used. So my question is: how to convert the element as np.float while reading the .csv file?

There is an argument dtype for pandas.read_csv , here is the explantion:

dtype : Type name or dict of column -> type, default None
Data type for data or columns. E.g. {‘a’: np.float64, ‘b’: np.int32} Use str 
or object to preserve and not interpret dtype. If converters are specified, 
they will be applied INSTEAD of dtype conversion.

Any examples of how to use it?

Thank you very much!

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

UPDATE:

Here are several solutions proposed by the answerers:

(1) from @Jakub. Setting the na_values=NaN in pandas.read_csv() , so all of the elements in the .csv file can be converted to np.float as being read into dataframe.

(2) from @André Christoffer Andersen. After read the .csv file as dataframe. use pandas.to_numeric to convert a column of the dataframe into np.float . Use a for loop to convert all columns into numeric.

(3) from @ThisGuyCantEven. Use numpy.loadtxt to read .csv file into a numpy.ndarray . Use the skiprows argument to skip the rows with unequal elements. Then use numpy.nan_to_num() to convert nan to zeros.

Hopefully, it will help following readers!

Answer 1

If you have a new enough pandas version you can also use pd.to_numeric(...) for this:

df['mycol'] = pd.to_numeric(df['mycol'], errors='coerce')

And, here is a trick to convert the whole dataframe:

for col in df.columns:
    df[col] = pd.to_numeric(df[col], errors='coerce')

Answer 2

Why not just use numpy.loadtxt ? If you want to use pandas because, say, you have mixed data and you want a numeric column as a numpy array, you can always use df['column'].as_matrix() , or you can convert the whole data frame if you want.