SQL获取具有相同列A但不同B的记录

Question

Suppose we have the following table called meals :

| meal  | stars |
-----------------
| steak |   1   |
| steak |   2   |
| fish  |   4   |
| fish  |   4   |
| salad |   5   |

How can I get records with the same meal, but different stars? I need the records whose only have different stars.

Result for the table above should be as follows:

| meal  | stars |
-----------------
| steak |   1   |
| steak |   2   |

I've tried the following query:

SELECT DISTINCT t1.*
FROM meals t1
INNER JOIN meals t2 ON t1.meal = t2.meal
AND t1.stars <> t2.stars;

But it consumes too much time and some noticeable amount of memory.

The actual size of my table is:

SELECT pg_size_pretty(pg_relation_size('table_name')); 
 pg_size_pretty 
----------------
 2295 MB

So I need to come up with something else and I am asking for your help!

Answer 1

SELECT  a.*
FROM    meals a
        INNER JOIN
        (
            SELECT  meal
            FROM    meals
            GROUP   BY meal
            HAVING  COUNT(DISTINCT stars) > 1
        ) b ON a.meal = b.meal

• SQLFiddle Demo

OUTPUT

╔═══════╦═══════╗
║ MEAL  ║ STARS ║
╠═══════╬═══════╣
║ steak ║     1 ║
║ steak ║     2 ║
╚═══════╩═══════╝

Answer 2

 SELECT meal,stars FROM meals
 GROUP BY meal,stars
 HAVING count(*)=1 and meal in (
    SELECT meal FROM meals
    GROUP BY meal
    HAVING count(*)>1 )

Answer 3

This should be the fastest way:

SELECT  *
FROM    meals m
WHERE   EXISTS (SELECT FROM meals WHERE meal = m.meal AND stars <> m.stars);

"Get all rows, where at least one other meal exists, with the same name but different stars."

Assuming NULL values do not occur.

SQL Fiddle.

Answer 4

A simple and faster way to achieve the same results

SELECT  
  m.*
FROM    
  meals m
  JOIN meals m2 ON ( m2.meal = m.meal AND m2.stars<>m.stars)