Question Details

No question body available.

Tags

python pandas dataframe merge ffill

Answers (1)

Accepted Answer Available
Accepted Answer
February 5, 2026 Score: 4 Rep: 267,303 Quality: Expert Completeness: 80%

You can use a simple mergeasof on time:

merge = pd.mergeasof(truth, pred, on='time')

NB. mergeasof requires keys in both inputs to be sorted, and it doesn't maintain the original index (like all merges).

Output:

   time  truth  pred
0   0.0      0   NaN
1  10.0      0   NaN
2  20.0      0   NaN
3  30.0      0   3.0
4  40.0      0   3.0
5  50.0      0   2.0
6  60.0      0   2.0
7  70.0      0   1.0
8  80.0      0   1.0
9  90.0      0   1.0

By default, it picks the previous or equal available right key, for the next one use direction='forward':

pd.mergeasof(truth, pred, on='time', direction='forward')
   time  truth  pred
0   0.0      0   3.0
1  10.0      0   3.0
2  20.0      0   3.0
3  30.0      0   2.0
4  40.0      0   2.0
5  50.0      0   1.0
6  60.0      0   1.0
7  70.0      0   1.0
8  80.0      0   NaN
9  90.0      0   NaN

And handle exact matches with allowexactmatches=False:

pd.mergeasof(truth, pred, on='time', allowexactmatches=False)
   time  truth  pred
0   0.0      0   NaN
1  10.0      0   NaN
2  20.0      0   NaN
3  30.0      0   3.0
4  40.0      0   3.0
5  50.0      0   2.0
6  60.0      0   2.0
7  70.0      0   2.0  # picked the previous time
8  80.0      0   1.0
9  90.0      0   1.0