Oracle SQL

I’ve just noticed an interesting thing:

Assume, that we have a simple query with “MIN(ID)” that works through “Index full scan(MIN/MAX)”:

SQL> explain plan for
  2  select
  3     min(ID)      as x
  4  from tab1
  5  where ID is not null;

Explained.

SQL> select * from table(dbms_xplan.display);

PLAN_TABLE_OUTPUT
---------------------------------------------------------------------------------------
Plan hash value: 4170136576

---------------------------------------------------------------------------------------
| Id  | Operation                   | Name    | Rows  | Bytes | Cost (%CPU)| Time     |
---------------------------------------------------------------------------------------
|   0 | SELECT STATEMENT            |         |     1 |     4 |     3   (0)| 00:00:01 |
|   1 |  SORT AGGREGATE             |         |     1 |     4 |            |          |
|   2 |   FIRST ROW                 |         |     1 |     4 |     3   (0)| 00:00:01 |
|*  3 |    INDEX FULL SCAN (MIN/MAX)| IX_TAB1 |     1 |     4 |     3   (0)| 00:00:01 |
---------------------------------------------------------------------------------------

Predicate Information (identified by operation id):
---------------------------------------------------

   3 - filter("ID" IS NOT NULL)

Test tables

create table tab1(id, x, padding)
as 
   with gen as (select level n from dual connect by level<=1000)
   select g1.n, g2.n, rpad(rownum,10,'x')
   from gen g1,gen g2;
create index ix_tab1 on tab1(id, x);
exec dbms_stats.gather_table_stats('','TAB1');

[collapse]

But look what will happen if we add one more “MIN(ID)”:

SQL> explain plan for
  2  select
  3     min(ID)      as x
  4   , min(ID)+1000 as x1000
  5  from tab1
  6  where ID is not null;

Explained.

SQL> select * from table(dbms_xplan.display);

PLAN_TABLE_OUTPUT
---------------------------------------------------------------------------------
Plan hash value: 3397888171

---------------------------------------------------------------------------------
| Id  | Operation             | Name    | Rows  | Bytes | Cost (%CPU)| Time     |
---------------------------------------------------------------------------------
|   0 | SELECT STATEMENT      |         |     1 |     4 |  3075  (17)| 00:00:02 |
|   1 |  SORT AGGREGATE       |         |     1 |     4 |            |          |
|*  2 |   INDEX FAST FULL SCAN| IX_TAB1 |   999K|  3906K|  3075  (17)| 00:00:02 |
---------------------------------------------------------------------------------

Predicate Information (identified by operation id):
---------------------------------------------------

   2 - filter("ID" IS NOT NULL)

Continue reading→

A couple of days ago someone posted a question on the forum which at the first glance seemed old, boring, beaten up and down. Here it is:

There is a news feed. All news are divided into 10 categories (Politics, sport, auto, real estate, etc).
I need to get top 4 news sorted by time descending for each category with 1 query.
If you sort the results – you get 4 politics news, then 4 sport news etc.

But the task was to make it optimal, and the standard solution with usual TopN using row_number can not be called optimal in any way, especially in case of big tables, relatively small number of categories and uneven distribution or just overall low selectivity.

So after getting several more or less acceptable variants, and spotting the solution using PostgreSQL (I didn’t dig deep into it, as I saw recursion, min and predicate, and that was enough), I got a great variant.

But first things first:

1. Getting distinct values from the index Continue reading→