从原始文本文件创建 pandas df

Question

我有一个文本文件，我想将其格式化为 pandas 数据框。它被读取为以下形式的字符串：
print(text)=

product: 1
description: product 1 desc
rating: 7.8
review: product 1 review

product: 2
description: product 2 desc
rating: 4.5
review: product 2 review

product: 3
description: product 3 desc
rating: 8.5
review: product 3 review

我想我会通过使用 text.split('\n\n') 将它们分组到列表中来拆分它们。我会假设将每个迭代到一个 dict 中，然后加载到 pandas df 将是一个很好的路线，但我在这样做时遇到了麻烦。这是最好的路线吗，有人可以帮我把它变成熊猫 df 吗？

Answer 1

您可以将read_csv 与通过product 字符串和pivot 比较第一列来创建组一起使用：

df = pd.read_csv('file.txt', header=None, sep=': ', engine='python')
df = df.assign(g = df[0].eq('product').cumsum()).pivot('g',0,1)
print (df)
0      description product rating             review
g                                                   
1   product 1 desc       1    7.8   product 1 review
2   product 2 desc       2    4.5   product 2 review
3   product 3 desc       3    8.5   product 3 review

或创建字典列表：

#https://stackoverflow.com/a/18970794/2901002
data = []
current = {}
with open('file.txt') as f:
    for line in f:
        pair = line.split(':', 1)
        if len(pair) == 2:
            if pair[0] == 'product' and current:
                # start of a new block
                data.append(current)
                current = {}
            current[pair[0]] = pair[1].strip()
    if current:
        data.append(current)
        
df = pd.DataFrame(data)
print (df)
  product     description rating            review
0       1  product 1 desc    7.8  product 1 review
1       2  product 2 desc    4.5  product 2 review
2       3  product 3 desc    8.5  product 3 review

或者将每 4 个值重新整形为 2d numpy 数组并传递给 DataFrame 构造函数：

df = pd.read_csv('file.txt', header=None, sep=': ', engine='python')

df = pd.DataFrame(df[1].to_numpy().reshape(-1, 4), columns=df[0].iloc[:4].tolist())
print (df)
  product     description rating            review
0       1  product 1 desc    7.8  product 1 review
1       2  product 2 desc    4.5  product 2 review
2       3  product 3 desc    8.5  product 3 review