熊猫：如果满足条件删除行

Question

I have a dataframe like:

date                airport_id  plane_type    runway
2020-01-01            11        333           3
2020-01-01            11        222           3
2020-01-02            11        333           3
2020-01-02            11        222           3
2020-01-03            11        333           3
2020-01-04            11        222           3
2020-01-01            12        222           3
2020-01-01            12        345           4

On a given date, no two types of plane ( plane_type ) can be present if they have a same runway , removing the row with bigger plane_type

Expected output:

date                airport_id  plane_type    runway
2020-01-01            11        222           3
2020-01-02            11        222           3
2020-01-03            11        333           3
2020-01-04            11        222           3
2020-01-01            12        222           3
2020-01-01            12        345           4

Any help would be very much appreciated! Thank you

Answer 1

It appears that you want to take the smallest plane_type for a given date , airport_id and plane_type . You could do that via a groupby statement as follows:

result = (
    df.groupby(['date', 'airport_id', 'runway'], as_index=False)['plane_type'].min()
   .sort_values(['airport_id', 'runway'])
)
>>> result
         date  airport_id  runway  plane_type
0  2020-01-01          11       3         222
3  2020-01-02          11       3         222
4  2020-01-03          11       3         333
5  2020-01-04          11       3         222
1  2020-01-01          12       3         222
2  2020-01-01          12       4         345

You can then merge additional columns (eg city and country ) back to this result, assuming that the values are unique for the given merge key.

result.merge(df, on=['date', 'airport_id', 'runway', 'plane_type'])

Answer 2

From your expected output I see that there should be added a requirement concerning the airport_id :

• no two types of plane...,
• in any given airport_id (this is the part to add),
• if they have a same runway .

To generate this result, run:

result = df.groupby(['date', 'airport_id', 'runway'], as_index=False,
    sort=False).apply(lambda grp: grp[grp.plane_type == grp.plane_type.min()])\
    .reset_index(level=0, drop=True)

The result is:

        date  airport_id  plane_type  runway
1 2020-01-01          11         222       3
3 2020-01-02          11         222       3
4 2020-01-03          11         333       3
5 2020-01-04          11         222       3
6 2020-01-01          12         222       3
7 2020-01-01          12         345       4

Try also another concept, ie:

• first set the grouping columns as the index,
• then groupby ,
• and the last point - change the index columns back to "normal" columns.

The code to do it is:

result = df.set_index(['date', 'airport_id', 'runway'])\
    .groupby(['date', 'airport_id', 'runway'], as_index=False)\
    .apply(lambda grp: grp[grp.plane_type == grp.plane_type.min()])\
    .reset_index(level=[1,2,3])

Access by the index should be considerably faster, so if the execution speed is the problem, this may be a better approach.