为什么这个脚本需要这么长时间才能运行？

Question

我有一个包含 200,000 行的 csv 文件。我已将其加载到数据框中，并希望通过以下脚本使用 faker 对其进行匿名化：

for i in range(MasterDE1.FirstName.size):
    MasterDE1.loc[(MasterDE1["Gender__pc"] == 'Female'), ['FirstName','LastName']] = fake.first_name_female(),fake.last_name_female()
    MasterDE1.loc[(MasterDE1["Gender__pc"] == 'Male'), ['FirstName','LastName']] = fake.first_name_male(),fake.last_name_male()
    MasterDE1.loc[(MasterDE1["Gender__pc"] == 'Unknown'), ['FirstName','LastName']] = fake.first_name(),fake.last_name()
    MasterDE1['Name'] = MasterDE1['FirstName'] + ' ' + MasterDE1['LastName']
    MasterDE1['EmailAddress'] = 'smithandthunder' + str(i+1) + '@gmail.com'

它从过去 20 分钟开始运行（我不认为内核已经死了）。

Answer 1

您可以先生成名称然后分配，而不是在每次迭代中更新 DataFrame：

df = pd.DataFrame({'Gender': np.random.choice(['Female', 'Male', 'Unknown'], p=[0.45, 0.45, 0.1], size=2*10**5), 
                   'First Name': np.nan, 'Last Name': np.nan})


df.head()
Out: 
   First Name  Gender  Last Name
0         NaN  Female        NaN
1         NaN    Male        NaN
2         NaN  Female        NaN
3         NaN    Male        NaN
4         NaN    Male        NaN

df.shape
Out: (200000, 3)

现在应该在几分钟内完成以下操作：

df.loc[df['Gender']=='Female', ('First Name', 'Last Name')] = [(fake.first_name_female(), fake.last_name_female()) for _ in range(df[df['Gender']=='Female'].shape[0])]

df.loc[df['Gender']=='Male', ('First Name', 'Last Name')] = [(fake.first_name_male(), fake.last_name_male()) for _ in range(df[df['Gender']=='Male'].shape[0])]

df.loc[df['Gender']=='Unknown', ('First Name', 'Last Name')] = [(fake.first_name(), fake.last_name()) for _ in range(df[df['Gender']=='Unknown'].shape[0])]

df.head()
Out: 
  First Name   Gender Last Name
0       Ruth   Female     Moore
1  Christina   Female     Jones
2    Lindsey   Female     Davis
3      Aaron  Unknown   Watkins
4     Joshua     Male     Henry

之后，df['Name'] = df['First Name'] + ' ' + df['Last Name'] 之类的东西应该会很快。

Answer 2

你可以省略循环：

MasterDE1 = pd.DataFrame({'Gender__pc':['Female','Male','Unknown'],
                         'FirstName':['s','d','f'],
                         'LastName': ['d','f','r']})
MasterDE1 = pd.concat([MasterDE1]*3).reset_index(drop=True)
print (MasterDE1)
  FirstName Gender__pc LastName
0         s     Female        d
1         d       Male        f
2         f    Unknown        r
3         s     Female        d
4         d       Male        f
5         f    Unknown        r
6         s     Female        d
7         d       Male        f
8         f    Unknown        r

def f1():
    return 'first_name_female' + str(np.random.randint(100))
def f2():
    return 'last_name_female' + str(np.random.randint(100))

maskfem = (MasterDE1["Gender__pc"] == 'Female')
a = pd.Series(((np.arange(len(MasterDE1.index))) + 1).astype(str))

MasterDE1.loc[maskfem, 'FirstName'] = [f1() for x in np.arange(maskfem.sum())]
MasterDE1.loc[maskfem, 'LastName'] =  [f2() for x in np.arange(maskfem.sum())]

MasterDE1['Name'] = MasterDE1['FirstName'] + ' ' + MasterDE1['LastName']
MasterDE1['EmailAddress'] = 'smithandthunder' + a + '@gmail.com'

print (MasterDE1)
             FirstName Gender__pc            LastName  \
0  first_name_female70     Female  last_name_female64   
1                    d       Male                   f   
2                    f    Unknown                   r   
3   first_name_female6     Female  last_name_female67   
4                    d       Male                   f   
5                    f    Unknown                   r   
6  first_name_female59     Female  last_name_female99   
7                    d       Male                   f   
8                    f    Unknown                   r   

                                     Name                EmailAddress  
0  first_name_female70 last_name_female64  smithandthunder1@gmail.com  
1                                     d f  smithandthunder2@gmail.com  
2                                     f r  smithandthunder3@gmail.com  
3   first_name_female6 last_name_female67  smithandthunder4@gmail.com  
4                                     d f  smithandthunder5@gmail.com  
5                                     f r  smithandthunder6@gmail.com  
6  first_name_female59 last_name_female99  smithandthunder7@gmail.com  
7                                     d f  smithandthunder8@gmail.com  
8                                     f r  smithandthunder9@gmail.com  

Answer 3

我不知道确切地告诉你为什么要花这么长时间，但这可能是因为文件的大小。

但是，您可以找到一种方法来监视该循环以了解它是否仍在工作：

signal = 0

for i in range(0,200000):
    ....
    # something going on in the loop
    ....
    # signal the loop
    signal += 1
    if signal == 50000 or signal == 100000 or signal == 150000:
        print('It\'s still going!')
    elif signal > 200000:
        print('It\'s over 200000 already!')
        break # or you can raise an error instead of break (raise RuntimeError)