Oracle SQL 查询根据值排除一组记录

Question

My post title might be awkward, but let me try to explain what I am trying to achieve with an oracle sql query.

My table is as below.

Requirement is, when the seqNo is shared by "apple", then I Do Not want the record to be selected.

Only if Apple alone is having a sequence number(Not shared by any other Category), only then select the record, as shown in the below table.

Any help is highly appreciated.

I tried to partion based on seqNo, and ranking it on category, but since the category is string, I could not achieve it.

|SeqNo   |   Category|
|------ |---------|
|1    |   Banana|
|3    |   Apple ---> *do not select as sequence number is shared*|
|3    |   Orange ---> *do not select as sequence number is shared*|
|4    |   Mango|
|4    |   Banana|
|7    |   Orange---> *do not select as sequence number is shared*|
|7    |   Apple---> *do not select as sequence number is shared*|
|7    |   Mango---> *do not select as sequence number is shared*|
|9    |   Apple-->**Only Select this**|
|11   |   Banana|
|13   |   Mango|
|14   |   Apple-->**Only Select this**|
|18   |   Apple-->**Only Select this**|
|42   |   Mango|

Answer 1

Schema:

 create table mytable (SeqNo int,   Category varchar(505));
 insert into mytable values(1    ,'Banana');
 insert into mytable values(3    ,'Apple');
 insert into mytable values(3    ,'Orange');
 insert into mytable values(4    ,'Mango');
 insert into mytable values(4    ,'Banana');
 insert into mytable values(7    ,'Orange');
 insert into mytable values(7    ,'Apple');
 insert into mytable values(7    ,'Mango');
 insert into mytable values(9    ,'Apple');
 insert into mytable values(11   ,'Banana');
 insert into mytable values(13   ,'Mango');
 insert into mytable values(14   ,'Apple');
 insert into mytable values(18   ,'Apple');
 insert into mytable values(42   ,'Mango');

Query#1 (to find rows for seqno with single record and which is 'Apple')

  with cte as ( select seqno,category,count(*)over(partition by seqno ) cnt from mytable )
  select * from cte where cnt=1 and trim(category)='Apple'

Output:

seqno	category	cnt
9	Apple	1
14	Apple	1
18	Apple	1

Query#2 (to remove the rows which have multiple rows for a sequence number and one have category 'Apple')

     with cte as 
     (
         select seqno,category,count(*)over(partition by seqno ) cnt from mytable
     )
     select * from mytable 
     where not exists
                    (
                     select * from cte where cnt=2 and trim(Category)='Apple' and 
                     mytable.seqno=cte.seqno
                    );

Output:

SeqNo	Category
1	Banana
4	Mango
4	Banana
9	Apple
11	Banana
13	Mango
14	Apple
18	Apple
42	Mango

db<fiddle here

Answer 2

I would use window functions to count the number of apples and total rows for a seq_no and then use filtering logic:

select t.*
from (select t.*,
             count(*) over (partition by seq_no) as cnt,
             sum(case when category = 'Apple' then 1 else 0 end) over (partition by seq_no) as num_apple
      from t
     ) t
where num_apple = 0 or
      (num_apple = cnt and num_apple = 1);

The last condition is the one that is ambiguous in the question. If there are exactly two apples for the seq_no , I don't know if it should be returned or not. If you want such cases, then remove the num_apple = 1 .