Drop duplicate if one duplicate cell contains one value, and the other duplicate contains another-CodePudding

So I have this type of dataset while working in python with pandas:

    id   pos    result
0   1     1       AB
1   1     1       --
2   1     1       BC
3   1     1       AB
4   1     2       CA
5   2     1       CA
6   2     2       --
7   2     2       BA
8   3     1       --
9   3     1       --

Desired result:

    id   pos    result
0   1     1       AB

2   1     1       BC
3   1     1       AB
4   1     2       CA
5   2     1       CA

7   2     2       BA
8   3     1       --
9   3     1       --

I would like to drop duplicate rows on [id] AND [pos] where the result are '--', but only if there already exist a result on the same [id] AND [pos] that has a A, B or C combination.

If no A, B or C combination exist on the same [id] AND [pos], then I want to keep the result '--'.

I know how to drop duplicate rows, but this problem is really beyond my knowledge and I am hoping for your help.

CodePudding user response：

You can use masks and boolean indexing:

# is the result not a "--"?
m = df['result'].ne('--')

# is there at least a non "--" in the group?
m2 = (m
  .groupby([df['id'], df['pos']])
  .transform('max')
 )

# keep if both conditions are equal
out = df[m==m2]

Alternative for the last step:

# keep if not "--" or not a non "--" in group
out = df[m|~m2]

Output:

   id  pos result
0   1    1     AB
2   1    1     BC
3   1    1     AB
4   1    2     CA
5   2    1     CA
7   2    2     BA
8   3    1     --
9   3    1     --