使用另一个事件的 dataframe 屏蔽日期的 pandas dataframe

Question

I have a dataframe as follows

df = pd.DataFrame({"date": pd.date_range(start="2012-03-01", end="2012-03-05"),
                       "date+1": pd.date_range(start="2012-03-02", end="2012-03-06"),
                       "date+2": pd.date_range(start="2012-03-03", end="2012-03-07")})

I also have another dataframe representing events with start date and end date as follows.

event = pd.DataFrame({"event": ["A", "B"],
                      "start": ["2012-03-02", "2012-03-04"],
                      "end": ["2012-03-03", "2012-03-06"]})
event["start"] = pd.to_datetime(event["start"])
event["end"] = pd.to_datetime(event["end"])

I want to create a mask dataframe that return True if any date in df is between start date and end date of any event in the event dataframe. The expected output should be

0,  1,  1
1,  1,  1
1,  1,  1
1,  1,  1
1,  1,  0

This expected output correspond to the df

2012-03-01,  2012-03-02,  2012-03-03
2012-03-02,  2012-03-03,  2012-03-04
2012-03-03,  2012-03-04,  2012-03-05
2012-03-04,  2012-03-05,  2012-03-06
2012-03-05,  2012-03-06,  2012-03-07

As you can see that only 2012-03-01 and 2012-03-07 are not between any event in the event dataframe. Looping could be computational expensive. May I have your suggestions how to minimize looping?

Answer 1

You can use cartesian join, then check that the date is between start and end of an interval, and aggregate:

# cartesian join
z = (df
    .stack().reset_index().assign(k=1)
    .merge(event.assign(k=1)))

# check if date between start and end
z['mask'] = z[0].between(z['start'], z['end'])

# aggregate
df_m = z.groupby(['level_0', 'level_1'])['mask'].max().unstack().astype(int)
df_m

Output:

level_1  date  date+1  date+2
level_0                      
0           0       1       1
1           1       1       1
2           1       1       1
3           1       1       1
4           1       1       0

PS Instead of that trick with assigning k=1 to both frames before merging, if you're on a newer version of pandas (1.2.0+), you can use merge(how='cross') directly

Answer 2

Create an interval index from events :

intervals = pd.IntervalIndex.from_tuples([*zip(event.start, event.end)], 
                                         closed = 'both')

IntervalIndex([[2012-03-02, 2012-03-03], [2012-03-04, 2012-03-06]],
              closed='both',
              dtype='interval[datetime64[ns]]')

Run applymap on df :

df.applymap(lambda df: intervals.contains(df).any()).astype(int)
 
   date  date+1  date+2
0     0       1       1
1     1       1       1
2     1       1       1
3     1       1       1
4     1       1       0