Question

I have the following query:

SELECT saturday_combinations.index, v.val AS  `row` , COUNT( * )  AS  `count` 
FROM saturday_combinations  
INNER JOIN (

SELECT ONE AS val
FROM saturday_combinations 
WHERE ONE IS NOT NULL 

UNION 
SELECT TWO AS val
FROM saturday_combinations
WHERE TWO IS NOT NULL 

UNION 
SELECT THREE AS val
FROM saturday_combinations
WHERE THREE IS NOT NULL 

UNION 
SELECT FOUR AS val
FROM saturday_combinations
WHERE FOUR IS NOT NULL 

UNION 
SELECT FIVE AS val
FROM saturday_combinations
WHERE FIVE IS NOT NULL


UNION 
SELECT SIX AS val
FROM saturday_combinations
WHERE SIX IS NOT NULL 

UNION 
SELECT SEVEN AS val
FROM saturday_combinations
WHERE SEVEN IS NOT NULL 

) v  ON v.val = saturday_combinations.ONE
  OR v.val = saturday_combinations.TWO
  OR v.val = saturday_combinations.THREE
  OR v.val = saturday_combinations.FOUR
  OR v.val = saturday_combinations.FIVE
  OR v.val = saturday_combinations.SIX
  OR v.val = saturday_combinations.SEVEN 
  GROUP BY v.val 

The purpose of the query is to provide a count of the different values contained in the columns ONE,TWO,THREE,FOUR,FIVE,SIX and SEVEN in the table saturday_combinations. However I want to put a desc limit 4 so that it only performs the count based on the last 4 rows (last four maximum indexes). But I am not getting it to work with the union. Adding order and limit at the very end only limits from the final select, rather than get the last 4 rows and calculate the distribution on them. Any tips?

The table schema is as follows:

index | ONE|TWO|THREE|FOUR|FIVE|SIX|SEVEN
 1       1   3   7     10    11  12  13
 2       3   4   5     30    31  22  23
 3       1   2   3      4     5   6   7
 4       1   2   3      4     5   6   7
 5       1   2   3      4     5   6   7
 6       1   2   3      4     5   6   7
 7       1   2   3      4     5   6   7
 8       1   2   3      4     5   6   7
 9       1   2   3      4     5   6   7
 10      1   2   3      4     5   6   7

Index is auto-increment and ONE-SEVEN has integer values.

There are about 3000 rows in the table and I want to count occurences for each value based on the last n rows.

Ideal result for the last n rows where n = last 3 rows should be
Numbers|Count
   1       3
   2       3
   3       3
   4       3
   5       3
   6       3
   7       3

If I increase n to include last 6 rows their count should increase. If I could last 10 rows the count should increase and other numbers should appear with their count.

Here is a link to a sample of the real table.
http://sqlfiddle.com/#!2/d035b

Answer 1

If answer to my comment is yes then, you could try the following. When you need to add limit, order by to union selects you need to wrap union queries with brackets ().

• SQLFIDDLE DEMO

Code:

(SELECT ONE AS val
FROM saturday_combinations 
WHERE ONE IS NOT NULL 
order by ONE desc limit 4)

UNION 
(SELECT TWO AS val
FROM saturday_combinations
WHERE TWO IS NOT NULL 
order by TWO desc limit 4)

UNION 
(SELECT THREE AS val
FROM saturday_combinations
WHERE THREE IS NOT NULL 
order by THREE desc limit 4)

If answer to my comment is no, then please clarify.

Here is the code based on your sample date:

select distinct x.one as uniqunumbers,
count(x.one) as counts
from(
sELECT DISTINCT 'one' 
AS col1, one FROM sat_comb
UNION ALL
SELECT DISTINCT 'two' 
AS col1, two FROM sat_comb
UNION ALL
SELECT DISTINCT 'three' 
AS col1, three FROM sat_comb
) as x
group by x.one;

UNIQUNUMBERS    COUNTS
1               1
3               2
4               1
5               1
7               1

EDIT as per OP has clarified and updated the question.

Quoted: “However I want to limit it so that it first takes the last n rows and then does the count on the values in those n rows. This means, if I have 3 columns with 3000 rows and 35 integers randomly appearing in these 3000 rows it should count how many times each integer appears.”

• SQLFIDDLE DEMO2

Query:

select x.one as uniqunumbers,
    count(x.one) as counts
    from(
    (sELECT DISTINCT 'one' 
    AS col1, one FROM sat_comb
     order by id desc limit 4)
    UNION ALL
    (SELECT DISTINCT 'two' 
    AS col1, two FROM sat_comb
     order by id desc limit 4)
    UNION ALL
    (SELECT DISTINCT 'three' 
    AS col1, three FROM sat_comb
     order by id desc limit 4)
      UNION ALL
    (SELECT DISTINCT 'four' 
    AS col1, four FROM sat_comb
     order by id desc limit 4)
      UNION ALL
    (SELECT DISTINCT 'five' 
    AS col1, five FROM sat_comb
     order by id desc limit 4)
      UNION ALL
    (SELECT DISTINCT 'six' 
    AS col1, six FROM sat_comb
     order by id desc limit 4)
      UNION ALL
    (SELECT DISTINCT 'seven' 
    AS col1, seven FROM sat_comb
     order by id desc limit 4)
    ) as x
    group by x.one;

Output:

UNIQUNUMBERS    COUNTS
2               4
3               3
4               3
5               4
6               4
8               3
9               4
20              3