based on dataframe column result all following rows equal a repetitive value until result changes and new repetitive value occurs

问题

My simplified dataframe is as follows:

df = pd.DataFrame()
df['A'] = ('IGNORE','IGNORE','IGNORE','YES','IGNORE','YES','YES','YES','IGNORE','IGNORE','IGNORE','YES','IGNORE','IGNORE','IGNORE','IGNORE','IGNORE','IGNORE','IGNORE','IGNORE','IGNORE', 'NO','IGNORE','IGNORE','IGNORE','IGNORE')

I need to reverse dataframe (which I know I can do via df = df[::-1]) then make column B as follows.

if 'YES' occurs then following rows result in 'GOOD' until a 'YES' or 'NO' occurs again and via versa for 'NO' occurring except 'BAD' will replace 'GOOD'

Desire output is as follows:

df['B'] = ('GOOD','GOOD','GOOD','YES','IGNORE','YES','YES','YES','GOOD','GOOD','GOOD','YES','BAD','BAD','BAD','BAD','BAD','BAD','BAD','BAD','BAD', 'NO','IGNORE','IGNORE','IGNORE','IGNORE')

回答1:

Idea is use Series.map dy dictioanry first with back filling missing values and replace last group by fillna to Series, which is used for replace IGNORE consecutive values - 2 or more:

s = df['A'].map({'IGNORE': np.nan, 'YES':'GOOD', 'NO':'BAD'}).bfill().fillna(df['A'])
m1 = df.groupby(df['A'].ne(df['A'].shift()).cumsum())['A'].transform('size').ne(1)
m2 = df['A'].eq('IGNORE')

df['C'] = np.where(m1 & m2, s, df['A'])
print(df)
         A       B       C
0   IGNORE    GOOD    GOOD
1   IGNORE    GOOD    GOOD
2   IGNORE    GOOD    GOOD
3      YES     YES     YES
4   IGNORE  IGNORE  IGNORE
5      YES     YES     YES
6      YES     YES     YES
7      YES     YES     YES
8   IGNORE    GOOD    GOOD
9   IGNORE    GOOD    GOOD
10  IGNORE    GOOD    GOOD
11     YES     YES     YES
12  IGNORE     BAD     BAD
13  IGNORE     BAD     BAD
14  IGNORE     BAD     BAD
15  IGNORE     BAD     BAD
16  IGNORE     BAD     BAD
17  IGNORE     BAD     BAD
18  IGNORE     BAD     BAD
19  IGNORE     BAD     BAD
20  IGNORE     BAD     BAD
21      NO      NO      NO
22  IGNORE  IGNORE  IGNORE
23  IGNORE  IGNORE  IGNORE
24  IGNORE  IGNORE  IGNORE
25  IGNORE  IGNORE  IGNORE

来源：https://stackoverflow.com/questions/56772211/based-on-dataframe-column-result-all-following-rows-equal-a-repetitive-value-unt

标签

pandas

numpy

dataframe

for-loop