python - 如何使用并发将数据帧附加到空数据帧

Question

我想在 Python 中使用 concurrent 运行一个函数。这是我的功能:

import concurrent.futures
import pandas as pd
import time

def putIndf(file):
    listSel = getline(file)
    datFram = savetoDataFrame(listSel)
    return datFram #datatype : dataframe

def main():
    newData = pd.DataFrame()
    with concurrent.futures.ProcessPoolExecutor(max_workers=30) as executor:
        for i,file in zip(fileList, executor.map(dp.putIndf, fileList)):
            df = newData.append(file, ignore_index=True)
    return df

if __name__ == '__main__':
    main()

我想将数据帧连接为一个数据帧newData，但结果只是该函数的最后一个数据帧

Answer 1

本质上，您是在每次迭代时重新分配df，并且从不增加它。您可能的意思(不明智)是初始化一个空的 df 并迭代追加:

df = pd.DataFrame()
...
df = df.append(file, ignore_index=True)

尽管如此，首选方法是构建一个数据帧集合，在循环外部一次附加在一起，并避免在循环内生成任何复杂的对象，例如数据帧。

def main():
    with concurrent.futures.ProcessPoolExecutor(max_workers=30) as executor:
        # LIST COMPREHENSION
        df_list = [file for i,file in zip(fileList, executor.map(dp.putIndf, fileList))]

        # DICTIONARY COMPREHENSION
        # df_dict = {i:file for i,file in zip(fileList, executor.map(dp.putIndf, fileList))}

    df = pd.concat(df_list, ignore_index=True)
    return df

或者，由于池进程，将数据帧附加到列表中，在循环外仍然连接:

def main():
    df_list = []      # df_dict = {}
    with concurrent.futures.ProcessPoolExecutor(max_workers=30) as executor:
        for i,file in zip(fileList, executor.map(dp.putIndf, fileList)):
            df_list.append(file)
            # df_dict[i] = file

    df = pd.concat(df_list, ignore_index=True)
    return df