Python pandas出现Single Positional Index Out-of-Bounds报错排查

我在处理一个包含5个DataFrame的字典，需要提取每个DataFrame中符合条件的第一行数据。编写的筛选代码在a0、a1、a2上可正常运行返回预期结果，但在a3、a4上执行时触发Single Positional Index Out-of-Bounds索引越界报错。

a0 = {'dataset': {'0.05, 1.0, 175': 'CS:1', '0.05, 1.0, 150': 'CS:1'},
 'mean_score': {'0.05, 1.0, 175': -0.2820450154520415,
  '0.05, 1.0, 150': -0.28204501545204186},
 'rank_score': {'0.05, 1.0, 175': 1, '0.05, 1.0, 150': 2},
 'std_score': {'0.05, 1.0, 175': 0.11499605607995111,
  '0.05, 1.0, 150': 0.11499605607995127},
 'contamination': {'0.05, 1.0, 175': 0.05, '0.05, 1.0, 150': 0.05},
 'max_samples': {'0.05, 1.0, 175': 1.0, '0.05, 1.0, 150': 1.0},
 'n_estimators': {'0.05, 1.0, 175': 175, '0.05, 1.0, 150': 150}}

a1 = {'dataset': {'0.05, 1.0, 200': 'CK:-1', '0.05, 1.0, 175': 'CK:-1'},
 'mean_score': {'0.05, 1.0, 200': -0.20618057780261195,
  '0.05, 1.0, 175': -0.20618057780261267},
 'rank_score': {'0.05, 1.0, 200': 1, '0.05, 1.0, 175': 2},
 'std_score': {'0.05, 1.0, 200': 0.13109535628226052,
  '0.05, 1.0, 175': 0.13109535628226013},
 'contamination': {'0.05, 1.0, 200': 0.05, '0.05, 1.0, 175': 0.05},
 'max_samples': {'0.05, 1.0, 200': 1.0, '0.05, 1.0, 175': 1.0},
 'n_estimators': {'0.05, 1.0, 200': 200, '0.05, 1.0, 175': 175}}

a2 = {'dataset': {'0.05, 0.7, 125': 'PH:1',
  '0.05, 0.7749999999999999, 200': 'PH:1'},
 'mean_score': {'0.05, 0.7, 125': -0.22096885360666768,
  '0.05, 0.7749999999999999, 200': -0.22416479620828117},
 'rank_score': {'0.05, 0.7, 125': 1, '0.05, 0.7749999999999999, 200': 2},
 'std_score': {'0.05, 0.7, 125': 0.05228492731122392,
  '0.05, 0.7749999999999999, 200': 0.061897704957581456},
 'contamination': {'0.05, 0.7, 125': 0.05,
  '0.05, 0.7749999999999999, 200': 0.05},
 'max_samples': {'0.05, 0.7, 125': 0.7,
  '0.05, 0.7749999999999999, 200': 0.7749999999999999},
 'n_estimators': {'0.05, 0.7, 125': 125, '0.05, 0.7749999999999999, 200': 200}}

a3 = {'dataset': {'0.05, 0.85, 125': 'PRT:-1',
  '0.05, 0.85, 100': 'PRT:-1'},
 'mean_score': {'0.05, 0.85, 125': -0.12896828405478034,
  '0.05, 0.85, 100': -0.13141635454748085},
 'rank_score': {'0.05, 0.85, 125': 1, '0.05, 0.85, 100': 2},
 'std_score': {'0.05, 0.85, 125': 0.016228240324984843,
  '0.05, 0.85, 100': 0.013178168219693726},
 'contamination': {'0.05, 0.85, 125': 0.05, '0.05, 0.85, 100': 0.05},
 'max_samples': {'0.05, 0.85, 125': 0.85, '0.05, 0.85, 100': 0.85},
 'n_estimators': {'0.05, 0.85, 125': 125, '0.05, 0.85, 100': 100}}

a4 = {'dataset': {'0.05, 1.0, 200': 'PRT:1',
  '0.05, 1.0, 175': 'PRT:1'},
 'mean_score': {'0.05, 1.0, 200': -0.1694053747115974,
  '0.05, 1.0, 175': -0.1694053747115976},
 'rank_score': {'0.05, 1.0, 200': 1, '0.05, 1.0, 175': 2},
 'std_score': {'0.05, 1.0, 200': 0.006550547930259526,
  '0.05, 1.0, 175': 0.006550547930259387},
 'contamination': {'0.05, 1.0, 200': 0.05, '0.05, 1.0, 175': 0.05},
 'max_samples': {'0.05, 1.0, 200': 1.0, '0.05, 1.0, 175': 1.0},
 'n_estimators': {'0.05, 1.0, 200': 200, '0.05, 1.0, 175': 175}}

注意：以上是原生字典结构，实际使用时需要通过pd.DataFrame()转换为DataFrame对象。

import pandas as pd
# 先将字典转为DataFrame
a0 = pd.DataFrame(a0)
a1 = pd.DataFrame(a1)
a2 = pd.DataFrame(a2)
a3 = pd.DataFrame(a3)
a4 = pd.DataFrame(a4)

subsets = ['CS:1', 'CK:-1', 'PH:1', 'PRT:-1', 'PRT:1']  # 待筛选的目标值

# a0-a2执行正常
(a0.loc[a0['dataset'].isin([subsets[0]])].iloc[0][['contamination', 'max_samples', 'n_estimators']])
(a1.loc[a1['dataset'].isin([subsets[1]])].iloc[0][['contamination', 'max_samples', 'n_estimators']])
(a2.loc[a2['dataset'].isin([subsets[2]])].iloc[0][['contamination', 'max_samples', 'n_estimators']])

# a3、a4触发索引越界报错
(a3.loc[a3['dataset'].isin([subsets[3]])].iloc[0][['contamination', 'max_samples', 'n_estimators']])
(a4.loc[a4['dataset'].isin([subsets[4]])].iloc[0][['contamination', 'max_samples', 'n_estimators']])

iloc[0]触发越界只有一个可能：前面的loc筛选返回了空DataFrame（长度为0，没有匹配到任何行），和iloc本身的索引写法无关，本质是筛选条件没有命中任何数据。

结合场景，最常见的触发原因有3个：

• 字符串不匹配：dataset列的实际值和subsets里的目标值存在肉眼不可见的差异，比如全角/半角冒号、首尾多余空格、tab缩进、不可见特殊字符。你贴出来的样例数据本身是可以正常运行的，实际环境里的a3、a4的dataset列值大概率和示例不一致。
• 预处理逻辑问题：执行筛选前，你对a3、a4做过过滤、去重、索引重置等操作，意外筛掉了所有行，或者修改了dataset列的值。
• DataFrame构造错误：转换a3、a4的时候传入了错误的orient参数，导致dataset没有成为列，或者列值被错误转换。

按顺序执行以下操作可以快速定位问题：

1. 先确认筛选结果是否为空：

# 排查a3
tmp = a3.loc[a3['dataset'].isin([subsets[3]])]
print(len(tmp))  # 输出0就说明确实没匹配到行
print(a3['dataset'].unique())  # 打印a3的dataset列所有唯一值，和目标值做对比

2. 如果打印出来的值看起来完全一致，做逐字符ASCII码校验，隐藏差异会直接暴露：

target = subsets[3]
for v in a3['dataset'].unique():
    print([ord(c) for c in v], [ord(c) for c in target])

3. 检查筛选前a3、a4的结构，确认列名、数据行数符合预期：

print(a3.info())
print(a3.head())

1. 针对字符串不匹配问题，先对dataset列做统一清洗再筛选：

# 统一去除首尾空格、替换全角冒号为半角冒号
a3['dataset'] = a3['dataset'].str.strip().str.replace('：', ':')
a4['dataset'] = a4['dataset'].str.strip().str.replace('：', ':')

2. 优化筛选逻辑，不要依赖行顺序硬取iloc[0]。你本身要取的是rank_score=1的最优行，直接加这个筛选条件更准确，也能规避行顺序变动导致的问题：

res3 = a3.loc[(a3['dataset'] == subsets[3]) & (a3['rank_score'] == 1), ['contamination', 'max_samples', 'n_estimators']].squeeze()
res4 = a4.loc[(a4['dataset'] == subsets[4]) & (a4['rank_score'] == 1), ['contamination', 'max_samples', 'n_estimators']].squeeze()

3. 批量处理字典里的所有DataFrame，减少重复代码，增加空结果容错：

df_dict = {'CS:1':a0, 'CK:-1':a1, 'PH:1':a2, 'PRT:-1':a3, 'PRT:1':a4}
result = {}
for target_val, df in df_dict.items():
    # 统一清洗列值
    df['dataset'] = df['dataset'].astype(str).str.strip().str.replace('：', ':')
    # 匹配目标行
    match_row = df.loc[(df['dataset'] == target_val) & (df['rank_score'] == 1), ['contamination', 'max_samples', 'n_estimators']]
    if not match_row.empty:
        result[target_val] = match_row.iloc[0]
    else:
        print(f"未匹配到{target_val}对应的数据，请检查列值")

内容的提问来源于stack exchange，提问作者GSA