将行追加到数据框

Question

I am stuck on a simple task. I want to create an empty DataFrame and append rows to it based on a query of another dataset. I have tried the answers here but I am missing something ..beginner Pythoner. Any help would be appreciated. I want to take the top 3 rows of each state and add them into a new dataframe for processing. I tried to append also..

def test():

    #get the list of states
    states_df = census_df.STNAME.unique()
    population_df = pd.DataFrame()

    for st in states_df:
        temp_df = pd.DataFrame(census_df[census_df['STNAME'] == st].nlargest(3,'CENSUS2010POP'))
        pd.concat([temp_df, population_df], ignore_index = True)

    return 1

Answer 1

I think I know what course you're doing, I had a great time with that a year ago, keep it up!

The simplest/fastest way I've found to concatenate a bunch of sliced dataframes is to append each df to a list, then at the end just concatenate that list. See the working code below (it does what I interpret you meant).

I agree with David's suggestion on sorting, easier to use sort and then just slice the first 3. As nlargest() works on and returns a Series I believe and not a dataframe, whereas you want to keep the whole dataframe structure (all the columns) for concatenation.

Also why is your function returning 1? Typo? I guess you want to return your desired output if you're putting it in a function, so I changed that too.

import pandas as pd
import numpy as np


#create fake data random numbers
data = np.random.randint(2,11,(40,3))
census_df = pd.DataFrame(index=range(40), columns=['Blah', 'Blah2','CENSUS2010POP'], data=data)
#create fake STNAME column
census_df['STNAME'] = list('aaaabbbbccccddddeeeeffffgggghhhhiiiijjjj')

#Function:
def test(census_df):
    states_list = census_df.STNAME.unique() #changed naming to _list as it's not a df.
    list_of_dfs = list() #more efficient to append each df to a list
    for st in states_list:
        temp_df = census_df[census_df['STNAME']==st]
        temp_df = temp_df.sort_values(by=['CENSUS2010POP'], ascending=False).iloc[:3]
        list_of_dfs.append(temp_df)
    population_df = pd.concat(list_of_dfs,ignore_index=True)
    return population_df

population_df = test(census_df)

Answer 2

Welcome to SO! Is your problem appending or the top three rows?

For append, try the df.append function. It could look something like:

#get the list of states
states_df = census_df.STNAME.unique()
population_df = pd.DataFrame()

for st in states_df:
    temp_df = pd.DataFrame(census_df[census_df['STNAME'] == st].nlargest(3,'CENSUS2010POP'))
    population_df = population_df.append(temp_df, ignore_index = True) #append the temp df to your main df, ignoring the index

For the top rows you could us df.sort_values(by=['column name'],ascending=False) and then select the top three rows:

population_df = population_df.append(temp_df[0:3], ignore_index = True)