Question

How does the SQL Optimizer handle Select Distinct on a primary key? For eg

Employee Table – (EmpId int primary key, EmpName)

select distinct EmpId from Employee.

Would the optimizer reject the distinct keyword considering EmpId is already distinct?

Answer 1

Tried under Oracle 11g and they produce different results.

select USERID from TPM_USER

Produces:

SELECT STATEMENT    7.0 7   242479  667 2668    7                   ALL_ROWS                                            
TABLE ACCESS (FULL) 7.0 7   242479  667 2668    1   TPMDBO  TPM_USER    FULL    TABLE   ANALYZED    1                                       

And:

select distinct USERID from TPM_USER

Produces:

SELECT STATEMENT    8.0 7   24725738    667 2668    8                   ALL_ROWS                                            
HASH (UNIQUE)   8.0 7   24725738    667 2668    1           UNIQUE                                                  
TABLE ACCESS (FULL) 7.0 7   242479  667 2668    1   TPMDBO  TPM_USER    FULL    TABLE   ANALYZED    1                                       

USERID is the primary key on TPM_USER in case that wasn’t clear.

UPDATE:

I was somewhat irked that Oracle could be this stupid, so I tried the same queries on our production server which has about a thousand times more data. This time, the two query plans were identical (the index was used both times, and no full table scan was done). This leads me to believe the planner will take the table statistics into account when deciding which route to go.

Production (with or without DISTINCT):

SELECT STATEMENT    3.0 3   461492  3521    14084   3                   ALL_ROWS                                            
INDEX (FAST FULL SCAN)  3.0 3   461492  3521    14084   1   TPMDBO  TPM_USER_PK FAST FULL SCAN  INDEX (UNIQUE)  ANALYZED