Python '；数据帧构造函数未正确调用'；带for循环的听写理解

编辑问题以更好地呈现问题

我正在学习数据分析，不知道这里有什么问题

我通过API获取数据，并对其进行df，其中行表示一个匹配，其中一列以嵌套dict列表的形式包含dota匹配中所有玩家的各种信息（原始dict有点大，所以如果需要的话，我不知道如何将其包含在这里）

我想做的是为每场比赛的特定玩家创建一个带有详细统计数据的df。为此，我正在尝试：

循环浏览原始df中“玩家”列中的每一行（每一行代表一个游戏）

为每个玩家创建dfs，并将其存储在dict中（现在我们有了一个dfs dict，每个由10行组成，其中10行代表游戏中的10名玩家，列代表他们的统计数据）

循环遍历这些存储的df，在其中找到所需的行（按player_id），并将其附加到最终df

现在问题来了：

所以

它本身工作并创建df

{i: pd.DataFrame(in_df.players[i]) for i in range(10)}

也可以按预期工作。 但是这个：

names_for_dfs = [i for i in range(len(in_df))]
{name: pd.DataFrame(in_df.players[name]) for name in names_for_dfs}

不起作用。有关职能：

def get_player_stats(in_df, cols_to_keep, player_id):
#create a df from 'players' column for each game (row) - it contains 10 rows for 10 players
#find a row with player_id for player in each game (each df) and append it to out_df
out_df = pd.DataFrame()

names_for_dfs = [row for row in range(len(in_df))]
     
dfs = {
name : pd.DataFrame(in_df.loc[name, 'players'])
for name in names_for_dfs
} 

for name, df in dfs.items():
    out_df = out_df.append(df[df.account_id.isin([player_id])], ignore_index=True)  # get a row by id and append to final df
return out_df[cols_to_keep]

我得到一个错误：

    ---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
<ipython-input-27-1a40ba2737e6> in <module>
      7     return dfs
      8 
----> 9 dfs = get_player_stats(matches_data, core_stats, 34505203)
     10 dfs

<ipython-input-27-1a40ba2737e6> in get_player_stats(in_df, cols_to_keep, player_id)
      3     dfs = {
      4     name : pd.DataFrame(in_df.loc[name, 'players'])
----> 5     for name in names_for_dfs
      6     }
      7     return dfs

<ipython-input-27-1a40ba2737e6> in <dictcomp>(.0)
      3     dfs = {
      4     name : pd.DataFrame(in_df.loc[name, 'players'])
----> 5     for name in names_for_dfs
      6     }
      7     return dfs

~\miniconda3\lib\site-packages\pandas\core\frame.py in __init__(self, data, index, columns, dtype, copy)
    507                 )
    508             else:
--> 509                 raise ValueError("DataFrame constructor not properly called!")
    510 
    511         NDFrame.__init__(self, mgr, fastpath=True)

ValueError: DataFrame constructor not properly called!

然后我从中创建了df，它看起来有点像原创的

stats = pd.DataFrame(data = data)

然后，我使用与上面相同的步骤来确保一切正常，但事情进展顺利，没有错误

in_df = stats
names_for_dfs = [i for i in range(len(in_df))]
dfs = {name: pd.DataFrame(in_df.loc[name, 'players']) for name in names_for_dfs}

打印出这个

{0:      match_id  stat1      stat2                       stat3
 0  5490791923    101  [1, 2, 3]  {1: 1, 2: 2, 3: [1, 2, 3]},
 1:      match_id  stat1      stat2                       stat3
 0  5490791923    101  [1, 2, 3]  {1: 1, 2: 2, 3: [1, 2, 3]},
 2:      match_id  stat1      stat2                       stat3
 0  5490791923    101  [1, 2, 3]  {1: 1, 2: 2, 3: [1, 2, 3]}}

所以现在我开始思考，有什么区别会阻止解决方案最初起作用？ 获取原始数据的代码：

def get_player_ids(team_id: int):
    players = requests.get(f'https://api.opendota.com/api/teams/{team_id}/players').json()
    ids = []
    keys = ['account_id', 'name']
    for player in players:
        for k, v in player.items():
            if k in keys:
                ids.append({k: v})
    print(ids)
    return ids

def get_team_id(team_name: str):
    teams = pd.DataFrame(requests.get('https://api.opendota.com/api/teams').json())
    team_id = int(teams.team_id[teams.name.str.lower() == team_name.lower()])
    get_player_ids(team_id)
    return team_id

columns = ['match_id', 'duration', 'radiant_score', 'dire_score', 'radiant_gold_adv',
           'radiant_xp_adv', 'radiant_team', 'dire_team', 'players', 'league', 'patch', 'start_time']
def get_match_data_for_team(team_id: int):    
    l = requests.get(f'https://api.opendota.com/api/teams/{team_id}/matches').json()
    match_ids = [d['match_id'] for d in l]
    matches_data = []
    for m_id in match_ids:
        matches_data.append(requests.get('http://api.opendota.com/api/matches/' + f'{m_id}').json())
    
    return pd.DataFrame(matches_data)[columns]

matches_data = get_match_data_for_team(get_team_id('nigma'))

编辑： 已修复，以下代码现在起作用：

def get_player_stats(in_df, cols_to_keep, player_id):
    #create a df from 'players' column for each game (row) - it contains 10 rows for 10 players
    #find a row with player_id for MC in each game (df) and append it to out_df
    out_df = pd.DataFrame()
    dfs = {}

    names_for_dfs = [row for row in range(len(in_df))]
    for name in names_for_dfs:
        for player_dict in in_df.players[name]:
            if isinstance(player_dict, dict) and player_dict['account_id'] == player_id:
                df = pd.DataFrame({key: [value] for key, value in player_dict.items()})
                dfs.update({name: df})

    for name, df in dfs.items():
        out_df = out_df.append(df)
        
    return out_df[cols_to_keep]

但我错过了几排

if isinstance(player_dict, dict) and player_dict['account_id'] == player_id:

在这种情况下，似乎是这样的，因为

匹配的数据有193行，但
out\u df
只有143行。
这边

out_df = pd.DataFrame()
dfs = {}
for match_number in range(len(matches_data)):
    for player_dict in matches_data.players[match_number]:
        if isinstance(player_dict, dict):
            df = pd.DataFrame({key: [value] for key, value in player_dict.items()})
            dfs.update({match_number: df})
for name, df in dfs.items():
    out_df = out_df.append(df[df.account_id.isin([34505203])], ignore_index=True)

我得到的更少——138行。如何在这些巢穴中正确搜索所需的玩家？
我会尝试：


将您的函数简化为下面的内容，这满足了您的目标，即在所有比赛中生成给定球员编译的统计数据的
DataFrame
当从
DataFrame
中嵌套的
dict
中提取数据时，字典逻辑有助于减轻一些索引复杂性，因此此函数将
DataFrame
作为\u df

中的

，但使用
DataFrame.to\u dict（）
方法将其更改为
dict

代码：

输出：

         champ     cs
match1     Zoe  700.0
match2  Syndra  800.0

我想出来了。有两个问题：

Pandas在尝试从每个玩家的
dict
中生成
数据帧时引发了一个错误（存储在“玩家”列
系列
中的
列表中的十个
索引
），因为所有玩家的
dict
都具有不同数组长度的值。例如，“match_id”值只是一个数字，而“ability_upgrades_arr”值是许多数字的列表。这很容易通过将值放入列表中来解决，本质上使每个值的数组长度等于1


当你在比赛中反复寻找球员
dicts
，有时他们不在那里（检查数据）。因此熊猫正试图用
str
或
float
（
np.nan
）制作
DataFrame
）。使用内置Python函数
isinstance（）
修复此问题


这是我的代码版本，它只是将每个播放机的
dict
作为
DataFrame
获取为任意键的值，正如您在上面尝试的那样。从这里开始，只需通过迭代进入
dict
，查找一个玩家姓名的数据

import requests
import pandas as pd


def get_team_matches_from_api(team: str) -> pd.DataFrame:
    "Get data from all matches played by :arg:`team`"
    
    # First, pull all teams from OpenDOTA so that we can...
    teams = pd.DataFrame(requests.get('https://api.opendota.com/api/teams').json())

    # ...get Team ID of :arg:`team`
    team_id = int(teams.team_id[teams.name.str.lower() == team.lower()])

    # Second, pull all games played by :arg:`team` so that we can...
    team_matches = requests.get(f'https://api.opendota.com/api/teams/{team_id}/matches').json()

    # ...get match IDs for each match played
    match_ids = [team_match['match_id'] for team_match in team_matches]

    # Third, go back to OpenDOTA and pull each match played by :arg:`team` using our match IDs above
    matches_data = []
    for match_id in match_ids:
        matches_data.append(requests.get(f'http://api.opendota.com/api/matches/{match_id}').json())

    # Fourth, put the data from the pulled matches into a `DataFrame`
    columns = ['match_id', 'duration', 'radiant_score', 'dire_score', 'radiant_gold_adv', 'radiant_xp_adv', 'radiant_team', 'dire_team', 'players', 'league', 'patch', 'start_time']
    df = pd.DataFrame(matches_data)[columns]

    return df.fillna('N/A')


def get_player_data_from_team_matches(team_matches: pd.DataFrame) -> dict:
    "Pull player data from all games in :arg:`team_matches`"
    
    players_from_team_matches = dict()

    for match_number in range(team_matches.players.shape[0]):

        for player_dict in team_matches.players[match_number]:
            if isinstance(player_dict, dict):
                df = pd.DataFrame({key: [value] for key, value in player_dict.items()})
                players_from_team_matches.update({match_number: df})

    return players_from_team_matches


# DataFrame of all Nigma matches
nigma = get_team_matches_from_api('nigma')

# Dictionary of all player data from every Nigma match
nigma_players = get_player_data_from_team_matches(nigma)

非常感谢您的投入！我应该添加数据的外观。但是你知道为什么我的方法不起作用吗？嗯，我最好的猜测是熊猫们对你在df中使用
的
.loc[]
方法感到难过。您得到了
ValueError:DataFrame构造函数未正确调用你的
dict
理解中的异常。我将查看，看看您是否可以重新表述
DataFrame
结构。作为参考，我能够使用我上面给出的代码成功地使用
DataFrame
construction
dict
comp：
dfs={I:pd.DataFrame（faker.loc['match1']）来表示范围（5）}
。实际上，我今天刚刚回到这里，它工作了。我的意思是没有任何变化。我还应该提到，它以前曾在某个时候起过作用，而且形式也完全相同。因此，我完全不知道发生了什么，事实上，事情并不顺利，所以我编辑了这个问题，并提供了更多的细节和步骤，以重现我的想法，谢谢你！将最后一次编辑添加到问题中-不知何故，我们在过程中丢失了行
import pandas as pd

def get_player_stats(in_df, player_id):
    
    df = pd.DataFrame()

    for match, players in in_df.to_dict()['players'].items():
        # {match1: {players}}

        for player, info in players.items():
            # {player1: {info}}

            if info['account_id'] == player_id:
                # {player1: {'account_id': player_id}}

                df = df.append(pd.Series(data=info, name=match))

    cols_to_keep = [col for col in df.columns if col != 'account_id']

    return df[cols_to_keep]

# I assume your data looks something like this:
matches_2020 = {

    'date': {
        'match1': '2020-06-01',
        'match2': '2020-06-02'
    },
    'players': {
        'match1': {
            'player1': {'account_id': 'FAKER', 'cs': 700, 'champ': 'Zoe'},
            'player2': {'account_id': 'BJERGSON', 'cs': 500, 'champ': 'Talon'}
        },
        'match2': {
            'player1': {'account_id': 'FAKER', 'cs': 800, 'champ': 'Syndra'},
            'player2': {'account_id': 'REDMERCY', 'cs': 500, 'champ': 'Zed'}
        }
    }
}

in_df = pd.DataFrame(matches_2020)

# Let's pull Faker's stats:
faker = get_player_stats(in_df, 'FAKER')
print(faker)

         champ     cs
match1     Zoe  700.0
match2  Syndra  800.0

import requests
import pandas as pd


def get_team_matches_from_api(team: str) -> pd.DataFrame:
    "Get data from all matches played by :arg:`team`"
    
    # First, pull all teams from OpenDOTA so that we can...
    teams = pd.DataFrame(requests.get('https://api.opendota.com/api/teams').json())

    # ...get Team ID of :arg:`team`
    team_id = int(teams.team_id[teams.name.str.lower() == team.lower()])

    # Second, pull all games played by :arg:`team` so that we can...
    team_matches = requests.get(f'https://api.opendota.com/api/teams/{team_id}/matches').json()

    # ...get match IDs for each match played
    match_ids = [team_match['match_id'] for team_match in team_matches]

    # Third, go back to OpenDOTA and pull each match played by :arg:`team` using our match IDs above
    matches_data = []
    for match_id in match_ids:
        matches_data.append(requests.get(f'http://api.opendota.com/api/matches/{match_id}').json())

    # Fourth, put the data from the pulled matches into a `DataFrame`
    columns = ['match_id', 'duration', 'radiant_score', 'dire_score', 'radiant_gold_adv', 'radiant_xp_adv', 'radiant_team', 'dire_team', 'players', 'league', 'patch', 'start_time']
    df = pd.DataFrame(matches_data)[columns]

    return df.fillna('N/A')


def get_player_data_from_team_matches(team_matches: pd.DataFrame) -> dict:
    "Pull player data from all games in :arg:`team_matches`"
    
    players_from_team_matches = dict()

    for match_number in range(team_matches.players.shape[0]):

        for player_dict in team_matches.players[match_number]:
            if isinstance(player_dict, dict):
                df = pd.DataFrame({key: [value] for key, value in player_dict.items()})
                players_from_team_matches.update({match_number: df})

    return players_from_team_matches


# DataFrame of all Nigma matches
nigma = get_team_matches_from_api('nigma')

# Dictionary of all player data from every Nigma match
nigma_players = get_player_data_from_team_matches(nigma)