MySQL选择包含字符串的行+特定的int列比较

Question

I have found numerous questions related to mine but still can't solve this issue.

In my table there are 3 columns filled with integer values and 3 columns with string values. I have several rows.

Table structure example:

  INT_1 | INT_2 | INT_3 | VALUE1 | VALUE2 | VALUE3
   33   |  25   |   10  | "nice"| "hello"| "goodbye"
  ---------------------------------------------------
   10   |  15   |   28  | "dice"| "hay"  | "bird"

I have a string that I use to select the rows based on the VALUE columns. The way I want to select it is with inclusion which means if the string is "llo" I should get the row where at least one of the values (VALUE,VALUE2,VALUE3) contains "llo" (would select the row with "hello" in VALUE2, for example).

However if two different rows have VALUE columns that contain the string (like in the example if the string is "ice") I want to retrieve the row where the INT column associated to that VALUE is higher. In the example since the string was compared to VALUE1 I should compare INT_1 of the upper column with INT_1 of the lower column and retrieve the row where INT_1 is higher. (INT_1 -> VALUE1, INT_2 -> VALUE2, INT_3 -> VALUE3).

Well not much but I could figure this myself:

 SELECT * FROM my_table WHERE VALUE1 = "+string+" OR VALUE2= "+string+" OR VALUE3= "+string+"";

• I am not sure how should I include "LIKE" to check for containing string when I have the values like "+string+".
• I don't know how to compare the specific INT column with the specific VALUE column when I have more than one row where VALUE contains the string.

Answer 1

First normalize your table using UNION ALL . That means every row have to be splitted into three. One for each group ( INT_1 VALUE1 , INT_2 VALUE2 , INT_3 VALUE3 ). Since you don't have an explicit primary key, you need to include all columns to identify the source row.

select t.*, 1 as position, INT_1 as i, VALUE1 as v from my_table t
union all
select t.*, 2 as position, INT_2 as i, VALUE2 as v from my_table t
union all
select t.*, 3 as position, INT_3 as i, VALUE3 as v from my_table t

Result:

| INT_1 | INT_2 | INT_3 | VALUE1 | VALUE2 |  VALUE3 | position |  i |       v |
|-------|-------|-------|--------|--------|---------|----------|----|---------|
|    33 |    25 |    10 |   nice |  hello | goodbye |        1 | 33 |    nice |
|    10 |    15 |    28 |   dice |    hay |    bird |        1 | 10 |    dice |
|    33 |    25 |    10 |   nice |  hello | goodbye |        2 | 25 |   hello |
|    10 |    15 |    28 |   dice |    hay |    bird |        2 | 15 |     hay |
|    33 |    25 |    10 |   nice |  hello | goodbye |        3 | 10 | goodbye |
|    10 |    15 |    28 |   dice |    hay |    bird |        3 | 28 |    bird |

http://sqlfiddle.com/#!9/9086d5/1

Now put it in a subquery and search for your string in the v column using WHERE v LIKE '%ice%' .

select *
from (
    select t.*, 1 as position, INT_1 as i, VALUE1 as v from my_table t
    union all
    select t.*, 2 as position, INT_2 as i, VALUE2 as v from my_table t
    union all
    select t.*, 3 as position, INT_3 as i, VALUE3 as v from my_table t
) n
where v like '%ice%'

Result:

| INT_1 | INT_2 | INT_3 | VALUE1 | VALUE2 |  VALUE3 | position |  i |    v |
|-------|-------|-------|--------|--------|---------|----------|----|------|
|    33 |    25 |    10 |   nice |  hello | goodbye |        1 | 33 | nice |
|    10 |    15 |    28 |   dice |    hay |    bird |        1 | 10 | dice |

http://sqlfiddle.com/#!9/9086d5/4

Last step - Pick the row with the highest value in i using ORDER BY i DESC LIMIT 1 :

select `INT_1`, `INT_2`, `INT_3`, `VALUE1`, `VALUE2`, `VALUE3`
from (
    select t.*, 1 as position, INT_1 as i, VALUE1 as v from my_table t
    union all
    select t.*, 2 as position, INT_2 as i, VALUE2 as v from my_table t
    union all
    select t.*, 3 as position, INT_3 as i, VALUE3 as v from my_table t
) n
where v like '%ice%'
order by i desc
limit 1

Result:

| INT_1 | INT_2 | INT_3 | VALUE1 | VALUE2 |  VALUE3 |
|-------|-------|-------|--------|--------|---------|
|    33 |    25 |    10 |   nice |  hello | goodbye |

http://sqlfiddle.com/#!9/9086d5/5

The query can be shorter if you use a HAVING clause instead of WHERE , so you don't need to use a subquery. But then you get two columns ( i and v ), that you might not need. On the other hand, they might be the only columns you need.

select t.*, INT_1 as i, VALUE1 as v from my_table t union all
select t.*, INT_2 as i, VALUE2 as v from my_table t union all
select t.*, INT_3 as i, VALUE3 as v from my_table t
having v like '%ice%'
order by i desc
limit 1

And one more modification which might improve the performance a little bit:

select t.*, INT_1 as i from my_table t where VALUE1 like '%ice%' union all
select t.*, INT_2 as i from my_table t where VALUE2 like '%ice%' union all
select t.*, INT_3 as i from my_table t where VALUE3 like '%ice%'
order by i desc
limit 1

Answer 2

This is a horrible data structure. But here is one way to do this?

SELECT t.*
FROM my_table t
WHERE VALUE1 LIKE '%string%' OR VALUE2 LIKE '%string%' OR VALUE3 LIKE '%string%'
ORDER BY greatest( (case when VALUE1 LIKE '%string%' then int_1 else -1 end),
                   (case when VALUE1 LIKE '%string%' then int_2 else -1 end),
                   (case when VALUE1 LIKE '%string%' then int_3 else -1 end) ) desc
LIMIT 1;