Return to Answer

To understand how the PIVOT works and why it solves the question, lets take a better example for our employee table sample:

To understand how the PIVOT works and why it solves the question, lets take a better example for our employee sample table:

with

employee(id, firstname, lastname, hobbies) as
(
  select 1, 'a', 'b', '1' from dual union 
  select 2, 'a', 'b', '2' from dual union 
  select 3, 'a', 'b', '3' from dual union 
  select 4, 'c', 'd', '3' from dual union 
  select 5, 'e', 'f', '2' from dual  
)

select * 
from employee 
pivot
( 
  max(1) fake  
  for (hobbies) -- put the undesired columns here
  IN () 
) 
where 1=1 -- and anything more...
order by id

What it does - the PIVOT - is group all rows through all columns except those declared in the PIVOT clause. The declared columns will be replaced by new columns defined by the aggregations before the FOR clause for each filter specified inside the IN clause.

A better example would be the following query:

That's the result:
 

All columns except the id and hobbies will be grouped, so it will group by firstname and lastname, which will result in three groups ('c', 'd') | ('e', 'f') | ('a', 'b'). Following, four columns will be created... two aggregation columns (as there are two MAX clause specified) for each filter in the IN clause and the aggregations are applied in the resulting groups.

Well, returning to the first query, it does works for two reasons:
1- you will not lose any row in the grouping process because the id column is unique and no columns were specified for aggregations;
2- as the pivot generates N * M new columns, where N = "number of filters" and M = "number of aggregations", having no filters and that single "harmless" aggregation means in 1 * 0 = 0 new columns, also removing all columns specified in the PIVOT clause, which is just the hobbies.

The first line of this question says: "... without having to specify the fields you want". In all other positive answers the proposed queries specifies the desired fields in the SELECT clause, except in mine, actually.

• a list of undesired columns;
• an IN clause;
• a three characters clause - FOR and NOT;

It means that you need to write a bit more in my approach as you need a fake aggregation (you don't need to give it an alias as I did, just write MAX(1)) and the PIVOT clause... It's too much for me if I spend two seconds to write these keywords and also, I have no problem to remember how to write this query - it is as easy to remember as the probable "standard". I can't really see a writer's cramp here.

with

employee(id, firstname, lastname, hobbies) as
(
  select 1, 'a', 'b', '1' from dual union 
  select 2, 'a', 'b', '2' from dual union 
  select 3, 'a', 'b', '3' from dual union 
  select 4, 'c', 'd', '3' from dual union 
  select 5, 'e', 'f', '2' from dual  
)

select * 
from employee 
pivot
( 
  max(1) -- fake  
  for (hobbies) -- put the undesired columns here
  IN () -- no values here...
) 
where 1=1 -- and your filters here...
order by id

To understand how the PIVOT works and why it solves the question, lets take a better example for our employee table sample:

The result here is:

The exact same output can be achieved using this easier to understand query:

select 
  firstname,
  lastname,
  max(case when hobbies = '2' then id end) two_foo,
  max(case when hobbies = '2' then 1  end) two_bar,
  max(case when hobbies = '3' then id end) three_foo,
  max(case when hobbies = '3' then 1  end) three_bar
from employee 
group by
  firstname,
  lastname

So, the column hobbies is never selected, just as the column id, both specified inside the PIVOT clause. All other columns are grouped and selected.

Well, returning to the first query, it works for two reasons:
1- you will not lose any row in the grouping process because the id column is unique and no columns were specified for aggregations;
2- as the pivot generates N * M new columns, where N = number of values of the IN clause and M = number of aggregations specified, so having no filters and that single harmless aggregation will produce 0 * 1 = 0 new columns and will remove the ones specified in the PIVOT clause, which is just the hobbies.

The first line of this question says: "... without having to specify the fields you want". In all other answers the proposed queries specifies the desired fields in the SELECT clause, except in mine, actually.

• a list of undesired columns;
• an IN clause;
• a three characters clause - FOR and NOT;

It means that you need to write a bit more in my approach as you need a fake aggregation and the PIVOT clause... but it's really few characters more...

The first line of this question says: "... without having to specify the fields you want". In all other positive answers the proposed queries specifies the desired fields in the SELECT clause, except in mine, actually.

Also, in the question title says "... without writer's cramp". Well, what's the correct measure to identify a writer's cramp? My best effort would be to foresee a good SQL standard to this problem and compare with my answer. Actually, I think this "standard" could be something like SELECT * NOT IN ([col1], [col2], ...).

Now, I can see in both queries:

• a list of undesired columns;
• an IN clause;
• a three characters clause - FOR and NOT;

It means that you need to write a bit more in my approach as you need a fake aggregation (you don't need to give it an alias as I did, just write MAX(1)) and the PIVOT clause... It's too much for me if I spend two seconds to write these keywords and also, I have no problem to remember how to write this query - it is as easy to remember as the probable "standard". I can't really see a writer's cramp here.