将pandas单元格中靠近第n个字符位置的长字符串拆分为多个单元格而不拆分单词

Question

由于 MS Excel 将单元格中的字符数限制为 32767，因此我必须将 pandas 数据框中的较长字符串拆分为多个单元格。

有没有办法将 pandas 列“Text”的字符串拆分为几列“Text_1”、“Text_2”、“Text_3”……来划分？文本块不能在一个单词中分隔也很重要，所以我认为需要正则表达式。

一个示例数据框：

df_test = pd.DataFrame({'Text' : ['This should be the first very long string','This is the second very long string','This is the third very long string','This is the last string which is very long'],
               'Date' : [2019, 2018, 2019, 2018],
               'Source' : ["FAZ", "SZ" , "HB", "HB"],
               'ID' : ["ID_1", "ID_2", "ID_3", "ID_4"]})
df_test

    Text                                        Date    Source  ID
0   This should be the first very long string   2019    FAZ     ID_1
1   This is the second very long string         2018    SZ      ID_2
2   This is the third very long string          2019    HB      ID_3
3   This is the last string which is very long  2018    HB      ID_4

假设此示例中的剪切发生在 n=15 而不是 n=32767，我想将 Text 列相应地拆分为以下内容：

    Text_1          Text_2          Text_3         Text_4      Date   Source    ID
0   This should be  the first very  long string                2019   FAZ       ID_1
1   This is the     second very     long string                2018   SZ        ID_2
2   This is the     third very long  string                    2019   HB        ID_3
3   This is the     last string     which is very  long        2018   HB        ID_4

最终，该方法应可扩展到n=32767 和至少十个新列"Text_1"、"Text_2" 等。

到目前为止，我已经创建了一个新列 "n"，指示每行 df_text["Text"] 字符串的长度：

df_test['n'] = df_test['Text'].str.split("").str.len()

Answer 1

这是大意。

# find longest long string, then divide the text 
# into the number of new cols you want, adding a | at
# the division and then later splitting by that |

longest = ""
for x in df_test['Text']:
    if len(x) > len(longest):
        longest = x
    continue

import math 

num_cols = math.floor(len(longest.split(' ')) / 3) # shoot for 3 words per row
for index,row in df_test.iterrows():

    word_str = row['Text']
    word_char_len = len(word_str)
    word_as_list = word_str.split(' ')
    num_words = len(word_as_list)

    col_index = math.ceil(len(word_as_list) / num_cols)

    for _ in range(num_cols - 1):
        word_as_list.insert(col_index,'|')
        col_index += col_index
    new = ' '.join(word_as_list)
    df_test.at[index,'Text'] = new

cols = ['Text'+str(i) for i in range(1,num_cols+1)]
df_test[cols] = df_test.Text.str.split('|',expand=True)
del df_test['Text']                                                                                                                   
print(df_test)

输出

   Date Source    ID            Text1          Text2                Text3
0  2019    FAZ  ID_1  This should be      the first      very long string
1  2018     SZ  ID_2     This is the    second very           long string
2  2019     HB  ID_3     This is the     third very           long string
3  2018     HB  ID_4     This is the    last string    which is very long

完成后我会上传一个完整的。如果您不喜欢这种方式或有其他建议，请发表评论。

Answer 2

是的 - 单个 pandas 单元格最多应包含 32767 个字符。因此 df_test[“Text”] 中的字符串应相应拆分。