[英]Pandas pivot\transpose rows to column headings
嘗試使用pd.wide_to_long
和unstack
:
df = pd.DataFrame({
'Name': ['Person1', 'Person1'],
'SetCode1': ['L6A', 'L6A'],
'SetDetail1': ['B', 'C'],
'SetCode2': ['G2G', 'G2G'],
'SetDetail2': ['B', 'B'],
})
df_melt = pd.wide_to_long(df.reset_index(),
['SetCode', 'SetDetail'],
['index', 'Name'],
'No')
df_out = df_melt.set_index('SetCode', append=True)\
.reset_index(level=2, drop=True)['SetDetail']\
.unstack()
df_out
輸出:
SetCode G2G L6A
index Name
0 Person1 B B
1 Person1 B C
我認為這更像是列重命名而不是旋轉。 這是我的代碼
code_cols = list(filter(lambda s: s.startswith('SetCode'), df.columns))
det_cols = list(filter(lambda s: s.startswith('SetDetail'), df.columns))
codes = [df[s][0] for s in code_cols]
df.rename(columns = dict(zip(det_cols, codes)), inplace=True)
df.drop(columns = code_cols, inplace=True)
df
產生
Name L6A G2G
0 Person1 B B
1 Person1 C B
感謝@Sander van den Oord 輸入數據框!
使用pandas.wide_to_long
是正確的解決方案,但必須謹慎對待某些列中的NaN
值。
因此,下面是對 Scott Boston 的回答的改編:
import pandas as pd
# I just allowed myself to write 'Person2' instead of 'Person1' at the second row
# of the DataFrame, as I imagine this is what was originally intended in the data,
# but this does not change the method
df = pd.DataFrame({
'Name': ['Person1', 'Person2'],
'SetCode1': ['L6A', 'L6A'],
'SetDetail1': ['B', 'C'],
'SetCode6': ['U2H', None],
'SetDetail6': ['B', None],
})
print(df)
Name SetCode1 SetDetail1 SetCode6 SetDetail6
0 Person1 L6A B U2H B
1 Person2 L6A C None None
# You will need to use reset_index to keep the original index moving forward only if
# the 'Name' column does not have unique values
df_melt = pd.wide_to_long(df, ['SetCode', 'SetDetail'], ['Name'], 'No')
df_out = df_melt[df_melt['SetCode'].notnull()]\
.set_index('SetCode', append=True)\
.reset_index(level=1, drop=True)['SetDetail']\
.unstack()
print(df_out)
SetCode L6A U2H
Name
Person1 B B
Person2 C NaN
聲明:本站的技術帖子網頁,遵循CC BY-SA 4.0協議,如果您需要轉載,請注明本站網址或者原文地址。任何問題請咨詢:yoyou2525@163.com.