所以我现在要做的事情是读取一个列表的列表,并将它们通过名为
目前的代码无法运行并导致Python崩溃。
我的担忧和问题:
checker
的函数,然后让log_result
处理checker
函数的结果。我正在尝试使用多线程来完成这个过程,因为实际上变量名rows_to_parse
有数百万行,所以使用多个核心应该可以大大加快这个过程。目前的代码无法运行并导致Python崩溃。
我的担忧和问题:
- 希望保持存在于变量
df
中的现有df在整个过程中保持索引不变,否则log_result
将会对需要更新的行感到困惑。 - 我相当确定
apply_async
不是执行此任务的适当多进程函数,因为我认为计算机读取和写入df的顺序可能会破坏它??? - 我认为可能需要设置一个队列来写入和读取
df
,但我不确定如何去做。
import pandas as pd
import multiprocessing
from functools import partial
def checker(a,b,c,d,e):
match = df[(df['a'] == a) & (df['b'] == b) & (df['c'] == c) & (df['d'] == d) & (df['e'] == e)]
index_of_match = match.index.tolist()
if len(index_of_match) == 1: #one match in df
return index_of_match
elif len(index_of_match) > 1: #not likely because duplicates will be removed prior to: if "__name__" == __main__:
return [index_of_match[0]]
else: #no match, returns a result which then gets processed by the else statement in log_result. this means that [a,b,c,d,e] get written to the df
return [a,b,c,d,e]
def log_result(result, dataf):
if len(result) == 1: #
dataf.loc[result[0]]['e'] += 1
else: #append new row to exisiting df
new_row = pd.DataFrame([result],columns=cols)
dataf = dataf.append(new_row,ignore_index=True)
def apply_async_with_callback(parsing_material, dfr):
pool = multiprocessing.Pool()
for var_a, var_b, var_c, var_d, var_e in parsing_material:
pool.apply_async(checker, args = (var_a, var_b, var_c, var_d, var_e), callback = partial(log_result,dataf=dfr))
pool.close()
pool.join()
if __name__ == '__main__':
#setting up main dataframe
cols = ['a','b','c','d','e']
existing_data = [["YES","A","16052011","13031999",3],
["NO","Q","11022003","15081999",3],
["YES","A","22082010","03012001",9]]
#main dataframe
df = pd.DataFrame(existing_data,columns=cols)
#new data
rows_to_parse = [['NO', 'A', '09061997', '06122003', 5],
['YES', 'W', '17061992', '26032012', 6],
['YES', 'G', '01122006', '07082014', 2],
['YES', 'N', '06081992', '21052008', 9],
['YES', 'Y', '18051995', '24011996', 6],
['NO', 'Q', '11022003', '15081999', 3],
['NO', 'O', '20112004', '28062008', 0],
['YES', 'R', '10071994', '03091996', 8],
['NO', 'C', '09091998', '22051992', 1],
['YES', 'Q', '01051995', '02012000', 3],
['YES', 'Q', '26022015', '26092007', 5],
['NO', 'F', '15072002', '17062001', 8],
['YES', 'I', '24092006', '03112003', 2],
['YES', 'A', '22082010', '03012001', 9],
['YES', 'I', '15072016', '30092005', 7],
['YES', 'Y', '08111999', '02022006', 3],
['NO', 'V', '04012016', '10061996', 1],
['NO', 'I', '21012003', '11022001', 6],
['NO', 'P', '06041992', '30111993', 6],
['NO', 'W', '30081992', '02012016', 6]]
apply_async_with_callback(rows_to_parse, df)
return [a, b, c, d, e]
,你的代码实际上会完成,但你会有其他问题,你也从未在任何地方使用dataf。 - Padraic Cunningham[a,b,c,d,e]
会被写入到log_result
函数中的 df 中。 - user3374113partial(log_result,dataf=dfr)
的签名与log_results
不匹配。 - mdurant