从现有 dataframe 的某些列创建新的 pandas dataframe

Question

I have read a csv file into a pandas dataframe and want to do some simple manipulations on the dataframe. I can not figure out how to create a new dataframe based on selected columns from my original dataframe. My attempt:

names = ['A','B','C','D']
dataset = pandas.read_csv('file.csv', names=names)
new_dataset = dataset['A','D']

I would like to create a new dataframe with the columns A and D from the original dataframe.

Answer 1

It is called subset - passed list of columns in [] :

dataset = pandas.read_csv('file.csv', names=names)

new_dataset = dataset[['A','D']]

what is same as:

new_dataset = dataset.loc[:, ['A','D']]

If need only filtered output add parameter usecols to read_csv :

new_dataset = pandas.read_csv('file.csv', names=names, usecols=['A','D'])

EDIT:

If use only:

new_dataset = dataset[['A','D']]

and use some data manipulation, obviously get:

A value is trying to be set on a copy of a slice from a DataFrame.
Try using .loc[row_indexer,col_indexer] = value instead

If you modify values in new_dataset later you will find that the modifications do not propagate back to the original data ( dataset ), and that Pandas does warning.

As pointed EdChum add copy for remove warning:

new_dataset = dataset[['A','D']].copy()

Answer 2

You must pass a list of column names to select columns. Otherwise, it will be interpreted as MultiIndex; df['A','D'] would work if df.columns was MultiIndex.

The most obvious way is df.loc[:, ['A', 'B']] but there are other ways (note how all of them take lists):

df1 = df.filter(items=['A', 'D'])

df1 = df.reindex(columns=['A', 'D'])

df1 = df.get(['A', 'D']).copy()

NB items is the first positional argument, so df.filter(['A', 'D']) also works.

Note that filter() and reindex() return a copy as well, so you don't need to worry about getting SettingWithCopyWarning later.