我有一个Excel工作表,其中的标题(TradeDate,Value)相互堆叠,并以类型(ABS,MBS)分开,格式如下:
ABS,
TradeDate,Value
2019-01-21,21
2019-01-22,22
MBS,
TradeDate,Value
2019-01-21,11
2019-01-22,12
2019-01-23,13
如何将其加载到python中,最好是pandas或其他程序包中,分别在其中加载每个标题?标题是唯一的/每种类型都相同,并且标题索引可以更改。使用上面的示例,我想返回两个单独的数据帧或对象。
df_abs
TradeDate,Value
2019-01-21,21
2019-01-22,22
和df_mbs
TradeDate,Value
2019-01-21,11
2019-01-22,12
2019-01-23,13
答案 0 :(得分:1)
这可能有点过大,但是找不到更简单的解决方案:
# Mask all the rows which have a date
m = df[0].str.match('([12]\d{3}-(0[1-9]|1[0-2])-(0[1-9]|[12]\d|3[01]))')
# Create an equal indicator on each row which has a date, but stops when value changes so we can groupby
df['ind'] = np.where(~m, 0, (m != m.shift(1)).astype(int).cumsum())
# Extract seperate dataframe into a list
dfs = [d for _, d in df[df.ind.ne(0)].groupby('ind')]
# Rename columns to expected output
dfs = [df.reset_index(drop=True).rename(columns={0:'TradeDate', 1:'Value'}) for df in dfs]
输出
for d in dfs:
print(d,'\n')
TradeDate Value ind
0 2019-01-21 21 2
1 2019-01-22 22 2
TradeDate Value ind
0 2019-01-21 11 4
1 2019-01-22 12 4
2 2019-01-23 13 4
可复制的示例
from io import StringIO
a = StringIO('''
ABS,
TradeDate,Value
2019-01-21,21
2019-01-22,22
MBS,
TradeDate,Value
2019-01-21,11
2019-01-22,12
2019-01-23,13
''')
df = pd.read_csv(a, header=None)
# Out
0 1
0 ABS NaN
1 TradeDate Value
2 2019-01-21 21
3 2019-01-22 22
4 MBS NaN
5 TradeDate Value
6 2019-01-21 11
7 2019-01-22 12
8 2019-01-23 13