而使用regex或其他来处理新数据的condition或extractall

我从一个文件中读取了一个数据集，我认为整个文件都是这样的（总共500-600行）：

所以我用这段代码生成了我需要的数据帧：

with open("dataset.txt", 'r') as infile:
    l = [x.replace(']', ',').replace("[", '').replace('"', '').replace('\n', '').strip().split(',') for x in infile]
    df = pd.DataFrame(l)
    df['A'] = list(range(len(df.index)))
    del df[2]
    df.rename(columns={1: 'nutrient'}, inplace=True)
    df[['amount_S']] = df['nutrient'].str.extract(pat=r'(?:\'\s\')(S|\d+\.\d+)', expand=True).fillna(0)
    df['nutrient'] = df['nutrient'].str.replace(pat=r'\'\s\'S|\d+',repl ='')
    df['nutrient'] = df['nutrient'].str.replace('\'', repl = '')
    df['nutrient'] = df['nutrient'].str.replace('.', repl='')

代码的dataframe输出如下所示（准备与另一个dataset和pivot连接）：

现在，我发现我的文件中还有一些条目如下所示：

0,['' '' '' '']
1,['Size' 'S' 'Size' 'M']
2,['Energy (kJ)' '351' 'Energy (kJ)' '617']
3,['Protein (g)' '2.3' 'Protein (g)' '4']
4,['Carbohydrates (g)' '15.4' 'Carbohydrates (g)' '26.9']
5,['Sugars (g)' '1.9' 'Sugars (g)' '3.3']
6,['Total Fat (g)' '0.6' 'Total Fat (g)' '1']
7,['Saturated Fat' '0.1' 'Saturated Fat' '0.1']
8,['Trans Fat (g)' '0' 'Trans Fat (g)' '0']
9,['Dietary Fibre (g)' '1.9' 'Dietary Fibre (g)' '3.4']
10,['Sodium (mg)' '2' 'Sodium (mg)' '4']
11,['Serving Size (g)' '75' 'Serving Size (g)' '125']
0,['' '' '' '' '' '' '' '']
1,['Size' 'S' 'Size' 'M' 'Size' 'L' 'Size' 'XL']
2,"['Energy (kJ)' '1431' 'Energy (kJ)' '2030' 'Energy (kJ)' '2863' 'Energy (kJ)' '3383']"
3,"['Protein (g)' '5.7' 'Protein (g)' '8.1' 'Protein (g)' '11.4' 'Protein (g)' '13.5']"
4,"['Carbohydrates (g)' '41.5' 'Carbohydrates (g)' '58.8' 'Carbohydrates (g)' '82.9' 'Carbohydrates (g)' '98']"
5,"['Sugars (g)' '1.2' 'Sugars (g)' '1.7' 'Sugars (g)' '2.4' 'Sugars (g)' '2.9']"
6,"['Total Fat (g)' '17.9' 'Total Fat (g)' '25.4' 'Total Fat (g)' '35.9' 'Total Fat (g)' '42.4']"
7,"['Saturated Fat' '7.9' 'Saturated Fat' '11.2' 'Saturated Fat' '15.8' 'Saturated Fat' '18.7']"
8,"['Trans Fat (g)' '0' 'Trans Fat (g)' '0' 'Trans Fat (g)' '0' 'Trans Fat (g)' '0']"
9,"['Dietary Fibre (g)' '3.7' 'Dietary Fibre (g)' '5.3' 'Dietary Fibre (g)' '7.5' 'Dietary Fibre (g)' '8.8']"
10,"['Sodium (mg)' '305' 'Sodium (mg)' '432.1' 'Sodium (mg)' '609' 'Sodium (mg)' '720']"
11,"['Serving Size (g)' '110' 'Serving Size (g)' '156' 'Serving Size (g)' '220' 'Serving Size (g)' '260']"

我想将数字数据移动到新的列（amount\u M、amount\u L、amount\u XL）。“营养素”栏无需重复。处理这些病例的最佳方法是什么？

使用：

import ast
# convert output fo 2 column dataframe
df = pd.read_csv('file5.csv', names=['a','b'])
#add comma to ' ', convert each row to lists 
df['b'] = df['b'].str.replace("'\s+'", "','").apply(ast.literal_eval)
#remove rows with 0 in a column
df =  df[df['a'] != 0]
#print (df)

fin = {}
#create dictionary of dataframes - groupby by helper Series - 
# necessary first value 1 for distinguish groups in a column
for i, x in dict(tuple(df.groupby(df['a'].eq(1).cumsum().sub(1)))).items():
#    print (x)

    #create DataFrame with column b, first row is header
    df2 = pd.DataFrame(x.b.values.tolist()[1:], columns=x.b.iloc[0])
    #remove duplicates columns names
    df2 = df2.loc[:, ~df2.columns.duplicated()]
#    print (df2)
    #convert output to dictionary (if necessary)
    fin[i] = df2

---

---

虽然您已经有了答案，但我想也提供我的方法。
这里，在使用正则表达式和生成器创建

DataFrame

时考虑不同的大小

(?:
    \G(?!\A)\D+              # match after the last match + non-digit
    (?P<value>\d+(?:\.\d+)?) # capture digits
    |                        # or
    '(?P<key>[A-Z][^']+)'    # match between 'A-Z...'
 )

---

产生

            nutrition     S      M     L    XL
0         Energy (kJ)  1644   None  None  None
1         Protein (g)  20.9   None  None  None
2   Carbohydrates (g)  33.6   None  None  None
3          Sugars (g)   1.8   None  None  None
4       Total Fat (g)  18.7   None  None  None
5       Saturated Fat   4.9   None  None  None
6       Trans Fat (g)     0   None  None  None
7   Dietary Fibre (g)   5.2   None  None  None
8         Sodium (mg)   845   None  None  None
9    Serving Size (g)   180   None  None  None
10        Energy (kJ)   351    617  None  None
11        Protein (g)   2.3      4  None  None
12  Carbohydrates (g)  15.4   26.9  None  None
13         Sugars (g)   1.9    3.3  None  None
14      Total Fat (g)   0.6      1  None  None
15      Saturated Fat   0.1    0.1  None  None
16      Trans Fat (g)     0      0  None  None
17  Dietary Fibre (g)   1.9    3.4  None  None
18        Sodium (mg)     2      4  None  None
19   Serving Size (g)    75    125  None  None
20        Energy (kJ)  1431   2030  2863  3383

---

非常感谢这项工作，但即使有评论，我也不能说代码对我来说是完全清楚的。您是否可以明确解释此处发生的步骤：“对于dict中的i，x（tuple（df.groupby（df['a'].eq（1.cumsum（）.sub（1）））））.items（）？

df['a'].eq（1.cumsum（）.sub（1）

意味着将列

a

与

1

first（

df['a'].eq（1）

相同）。然后获取

True

和

False

值，

Trues

用于

1

和

cumsum

创建组的行。因此

True-False-False

create

1122

和last

sub1

减去

1

计数表0-

0101

。最后一个带有dict的

元组
转换
groupby
对象到数据帧和
项的字典（）
用于循环。
print (fin[0])
                Size     S     M
0        Energy (kJ)   351   617
1        Protein (g)   2.3     4
2  Carbohydrates (g)  15.4  26.9
3         Sugars (g)   1.9   3.3
4      Total Fat (g)   0.6     1
5      Saturated Fat   0.1   0.1
6      Trans Fat (g)     0     0
7  Dietary Fibre (g)   1.9   3.4
8        Sodium (mg)     2     4
9   Serving Size (g)    75   125

print (fin[1])
                Size     S      M     L    XL
0        Energy (kJ)  1431   2030  2863  3383
1        Protein (g)   5.7    8.1  11.4  13.5
2  Carbohydrates (g)  41.5   58.8  82.9    98
3         Sugars (g)   1.2    1.7   2.4   2.9
4      Total Fat (g)  17.9   25.4  35.9  42.4
5      Saturated Fat   7.9   11.2  15.8  18.7
6      Trans Fat (g)     0      0     0     0
7  Dietary Fibre (g)   3.7    5.3   7.5   8.8
8        Sodium (mg)   305  432.1   609   720
9   Serving Size (g)   110    156   220   260

(?:
    \G(?!\A)\D+              # match after the last match + non-digit
    (?P<value>\d+(?:\.\d+)?) # capture digits
    |                        # or
    '(?P<key>[A-Z][^']+)'    # match between 'A-Z...'
 )

import regex as re
import pandas as pd

# construct the regular expression    
rx = re.compile(r'''(?:\G(?!\A)\D+(?P<value>\d+(?:\.\d+)?)|'(?P<key>[A-Z][^']+)')''')

# remove empty lines and lines without any number
valid = re.compile(r'\d+,\D+\d')

# generator comprehension
result = ([m.group('key') if m.group('key') is not None else m.group('value') 
            for m in rx.finditer(line)]
            for line in your_string_here.split('\n')
            if valid.match(line))

df = pd.DataFrame(result, columns = ['nutrition', 'S', 'M', 'L', 'XL'])
print(df)

            nutrition     S      M     L    XL
0         Energy (kJ)  1644   None  None  None
1         Protein (g)  20.9   None  None  None
2   Carbohydrates (g)  33.6   None  None  None
3          Sugars (g)   1.8   None  None  None
4       Total Fat (g)  18.7   None  None  None
5       Saturated Fat   4.9   None  None  None
6       Trans Fat (g)     0   None  None  None
7   Dietary Fibre (g)   5.2   None  None  None
8         Sodium (mg)   845   None  None  None
9    Serving Size (g)   180   None  None  None
10        Energy (kJ)   351    617  None  None
11        Protein (g)   2.3      4  None  None
12  Carbohydrates (g)  15.4   26.9  None  None
13         Sugars (g)   1.9    3.3  None  None
14      Total Fat (g)   0.6      1  None  None
15      Saturated Fat   0.1    0.1  None  None
16      Trans Fat (g)     0      0  None  None
17  Dietary Fibre (g)   1.9    3.4  None  None
18        Sodium (mg)     2      4  None  None
19   Serving Size (g)    75    125  None  None
20        Energy (kJ)  1431   2030  2863  3383