How to delete specific rows from pandas data frame

问题:

I am dealing with a pandas data frame as shown below.

    id          x1          y1
 0  2           some_val    some_val
 1  2           some_val    some_val
 2  2           some_val    some_val
 3  2           some_val    some_val
 4  2           some_val    some_val
 5  0           0           0 
 6  3           some_val    some_val
 7  3           some_val    some_val
 8  0           0           0 
 9  5           some_val    some_val
10  5           some_val    some_val
11  5           some_val    some_val
12  0           0           0
13  6           some_val    some_val
14  6           some_val    some_val
15  6           some_val    some_val
16  6           some_val    some_val

My original data frame was the data frame without the rows with all '0' values. As per the project requirement I had to insert the rows with all 0's value whenever the "id" changes.

Now I want to delete all the rows of any "id" which has 3 and less than 3 rows. From the above data frame, I would want to delete all the respective rows of id- "3" and "5" . My resultant data frame should look like below:

   id          x1          y1
0  2           some_val    some_val
1  2           some_val    some_val
2  2           some_val    some_val
3  2           some_val    some_val
4  2           some_val    some_val
5  0           0           0
6  6           some_val    some_val
7  6           some_val    some_val
8  6           some_val    some_val
9  6           some_val    some_val

Kindly suggest me a way to obtain this result.

回答1:

The simplest answer is to remove the zero rows because they may get in the way of the calculation if you have more than 3 of them. then do a group by. then filter. then add back zeros like you did in other question/answer

d1 = df.query('ProjID != 0').groupby('ProjID').filter(lambda df: len(df) > 3)
d1

    ProjID     Xcoord    Ycoord
0        2  -7.863509  5.221327
1        2   some_val  some_val
2        2   some_val  some_val
3        2   some_val  some_val
4        2   some_val  some_val
13       6   some_val  some_val
14       6   some_val  some_val
15       6   some_val  some_val
16       6   some_val  some_val

Then add back

pidv = d1.ProjID.values
pid_chg = np.append(pidv[:-1] != pidv[1:], True)

i = d1.index.repeat(pid_chg + 1)

d2 = d1.loc[i, :].copy()

d2.loc[i.duplicated()] = 0

d2.reset_index(drop=True)

    ProjID     Xcoord    Ycoord
0        2  -7.863509  5.221327
1        2   some_val  some_val
2        2   some_val  some_val
3        2   some_val  some_val
4        2   some_val  some_val
5        0          0         0
6        6   some_val  some_val
7        6   some_val  some_val
8        6   some_val  some_val
9        6   some_val  some_val
10       0          0         0

回答2:

Say your DataFrame name is df, you need to do the following:

df = df[df['col'<>=condition]] 

Specifically to your case:

df = df[df['ProjID'!=3]] 

Same with 5. You can combine both filters with an 'and' for efficiency.

This is called DataFrame indexing filters.

回答3:

You can use groupby and filter the IDs with count less than three and use the resulting list to index the df.

filtered = df.groupby('ProjID').Xcoord.filter(lambda x: x.count() > 3)
df.iloc[filtered.index.tolist()]


    ProjID  Xcoord  Ycoord
0   2   -7.863509   5.221327
1   2   some_val    some_val
2   2   some_val    some_val
3   2   some_val    some_val
4   2   some_val    some_val
13  6   some_val    some_val
14  6   some_val    some_val
15  6   some_val    some_val
16  6   some_val    some_val