Do rolling on all dataframe rows-CodePudding

I would like to find in dataframe of prices all all-time highs in that were in history.

timestamp	close	ath
x	1234	0
x	2000	1
x	1956	0
x	1884	0
x	2234	1

So for specific row I would like to find a maximum for all previous rows and check if value in current row is greater than this maximum. I cannot do it without indexing for all rows and I know that it is inefficient.

CodePudding user response：

Try this:

df['ath'] = df.groupby('timestamp')['close'].shift(1).lt(df['close']).astype(int)

Output:

>>> df
  timestamp  close  ath
0         x   1234    0
1         x   2000    1
2         x   1956    0
3         x   1884    0
4         x   2234    1

CodePudding user response：

FWIW, here is an alternative:

df['ath'] = (df['close'] > df['close'].expanding().max().shift()).astype(int)

CodePudding user response：

Use cummax to keep the highest seen value then shift rows and compare to the current row:

df['ath'] = df['close'].gt(df['close'].cummax().shift()).astype(int)
print(df)

# Output
  timestamp  close  ath
0         x   1234    0
1         x   2000    1
2         x   1956    0
3         x   1884    0
4         x   2234    1