如何用循环将多客户广告数据合并为单个DataFrame？

我有一个包含客户ID的列表：

clients = ['act_1078577379193828', 'act_1503882476478990', 'act_232830897389702']

我通过Python循环执行异步任务，获取每个客户的广告活动数据：

for advertiser in clients:
    async_job = AdAccount(advertiser).get_insights(fields=fields, params=params, is_async=True)
    async_job.api_get()
    while async_job[AdReportRun.Field.async_status]!= 'Job Completed':
        time.sleep(1)
        async_job.api_get()
    time.sleep(1)
    df = pd.DataFrame(async_job.get_result())

但当前代码每次循环都会覆盖df，最终只能得到最后一个客户的数据。我曾用if-elif逐个创建DataFrame：

for advertiser in clients:
    async_job = AdAccount(advertiser).get_insights(fields=fields, params=params, is_async=True)
    async_job.api_get()
    while async_job[AdReportRun.Field.async_status]!= 'Job Completed':
        time.sleep(1)
        async_job.api_get()
    time.sleep(1)
    if advertiser == 'act_1078577379193828':
        df1 = pd.DataFrame(async_job.get_result())
    elif advertiser == 'act_1503882476478990':
        df2 = pd.DataFrame(async_job.get_result())
    elif advertiser == 'act_232830897389702':
        df3 = pd.DataFrame(async_job.get_result())

但后续客户ID会超过44个，这种方式完全不可行。求一种无需大量分支判断，就能将所有客户数据合并到单个DataFrame的方法。

*注：async_job.get_result()是数据返回的位置，其输出示例如下：

[<AdsInsights> {
    "account_id": "232830897389702",
    "account_name": "Advertiser_Account_Name",
    "actions": [
        {
            "action_type": "link_click",
            "value": "4"
        },
        {
            "action_type": "post_engagement",
            "value": "4"
        },
        {
            "action_type": "page_engagement",
            "value": "4"
        }
    ],
    "ad_name": "***",
    "adset_name": "***",
    "campaign_name": "***",
    "clicks": "15",
    "cpc": "0.568",
    "cpm": "15.160142",
    "ctr": "2.669039",
    "date_start": "2022-08-26",
    "date_stop": "2022-08-26",
    "frequency": "1.05838",
    "impressions": "562",
    "inline_post_engagement": "4",
    "objective": "LEAD_GENERATION",
    "reach": "531",
    "spend": "8.52"
}, 

可以通过列表存储单个客户DataFrame，最后合并的方式实现，完全不需要分支判断，代码简洁且易于扩展：

优化后代码 #

import pandas as pd
import time
# 假设AdAccount、AdReportRun已提前导入

clients = ['act_1078577379193828', 'act_1503882476478990', 'act_232830897389702']
# 初始化空列表，用于存放每个客户的DataFrame
df_list = []

for advertiser in clients:
    # 执行异步任务并等待完成
    async_job = AdAccount(advertiser).get_insights(fields=fields, params=params, is_async=True)
    async_job.api_get()
    while async_job[AdReportRun.Field.async_status] != 'Job Completed':
        time.sleep(1)
        async_job.api_get()
    time.sleep(1)
    
    # 将当前客户的数据转为DataFrame，添加到列表中
    current_df = pd.DataFrame(async_job.get_result())
    df_list.append(current_df)

# 合并所有客户的DataFrame
final_df = pd.concat(df_list, ignore_index=True)

关键步骤说明 #

1. 初始化空列表：df_list用来临时存储每个客户的DataFrame，避免变量覆盖或大量命名变量的问题。
2. 循环添加DataFrame：每次处理完一个客户，就把生成的DataFrame追加到列表中。
3. 合并所有数据：用pd.concat()将列表中的所有DataFrame纵向合并，ignore_index=True重置合并后的索引，避免索引重复。

可选优化 #

如果担心某个客户返回空数据导致合并失败，可以在添加前做判断：

current_df = pd.DataFrame(async_job.get_result())
if not current_df.empty:
    df_list.append(current_df)

内容的提问来源于stack exchange，提问作者André Filho