pandas take last consecutive value-CodePudding

I have pandas DataFrame with column:

index, min_or_max
0,     np.nan
1,     np.nan
2,     min
3,     np.nan
4,     np.nan
5,     max
6,     np.nan
7,     np.nan
8,     np.nan
9,     max
10,    np.nan
11,    np.nan
12,    min

I want to create col2 such that:

index, min_or_max, col2
0,     np.nan      np.nan
1,     np.nan      np.nan
2,     min         min
3,     np.nan      np.nan
4,     np.nan      np.nan
5,     max         np.nan
6,     np.nan      np.nan
7,     np.nan      np.nan
8,     np.nan      np.nan
9,     max         max
10,    np.nan      np.nan
11,    np.nan      np.nan
12,    min         min

how can I check for consecutive values in the column and take the last one?

I can have multiple consecutive value max or min but repetition is max 10 in a row
I can have repetitions of either in or max value

EDIT:

I tried this:

df1 = df[df['min_or_max'].ne(df['min_or_max'].shift(-1))]
df1["col2"] = df1["min_or_max"]
df1 = df1.reset_index()
df1 = df1[["index", "col2"]]


df = df.reset_index()
df = df[["index"]]

df = df.merge(df1, on="index", how="left")

EDIT:

my proposed solution:

df1 = df.dropna(subset=['min_or_max'], how='all', inplace=False)
df1 = df1[df1['min_or_max'].ne(df1['min_or_max'].shift(-1))]


df = df.reset_index()
df = df[["index", "min_or_max"]]

df1 = df1.reset_index()
df.columns = ["index", "col2"]

df1 = df1[["index", "col2"]]
df = df.merge(df1, on="index", how="left")

CodePudding user response：

Example:

s = pd.Series(['A', np.nan, 'A', 'bb', 'bb', 'A', 'A', 'bb', 'A', 'bb', 'bb'])

0       A
1     NaN
2       A
3      bb
4      bb
5       A
6       A
7      bb
8       A
9      bb
10     bb
dtype: object

Solution:

In each group: leave only the last value. (NaNs are considered parts of the groups.)

s = s.bfill()
to_delete = s == s.shift(-1)
s[to_delete] = np.nan

Result:

0     NaN
1     NaN
2       A
3     NaN
4      bb
5     NaN
6       A
7      bb
8       A
9     NaN
10     bb
dtype: object