如何以有效的方式消除带有列表的 mysql 表中的冗余

Question

Some time ago I made a fast MVP that now has become a more realistic project. Now I'm refactoring and improving it. I have a table like this

CREATE TABLE `records` (
  `id` int(11) NOT NULL,
  `type` int(11) NOT NULL,
  .....
  `ref` int(11) DEFAULT NULL,
  `enabled` tinyint(1) NOT NULL DEFAULT '1',
) ENGINE=MyISAM DEFAULT CHARSET=utf8 COLLATE=utf8_unicode_ci;

ALTER TABLE `mrecord`
  ADD PRIMARY KEY (`id`),
  ADD KEY `type` (`type`);
  ADD KEY `ref` (`ref`);

ref is a reference to a previous id or null if none and enabled lets me know if this item is the last version of the type. The point is that when an item type X is replaced for a new one, the old one will be disabled ( enabled = 0) and the new one will set ref to the old one id .

For example, we have this table with 3 types of items:

 --------------------------------------------
| ID | type | other columns | ref  | enabled |
|--------------------------------------------|
| 1  |   1  |               | null |    1    |
| 2  |   2  |               | null |    1    |
| 3  |   3  |               | null |    1    |
 --------------------------------------------

and now we add a new item version that replaces item type 2:

 --------------------------------------------
| ID | type | other columns | ref  | enabled |
|--------------------------------------------|
| 1  |   1  |               | null |    1    |
| 2  |   2  |               | null |    0    |
| 3  |   3  |               | null |    1    |
| 4  |   2  |               |  2   |    1    |
 --------------------------------------------

and if we update the brand new item we have:

 --------------------------------------------
| ID | type | other columns | ref  | enabled |
|--------------------------------------------|
| 1  |   1  |               | null |    1    |
| 2  |   2  |               | null |    0    |
| 3  |   3  |               | null |    1    |
| 4  |   2  |               |  2   |    0    |
| 5  |   2  |               |  4   |    1    |
 --------------------------------------------

What we have here are lists of item types with only last versions enabled.

But here enabled column is redundant because an enabled item is just an item that has not a new version.

So my problem is how to do a SQL query equivalent to:

SELECT * FROM `records` WHERE type='2' AND enabled='1'

without using enabled and in an efficient way (this query is <1ms).

Answer 1

You can use not exists :

select  r.*
from records r
where not exists (select 1
                  from records r2
                  where r2.ref = r.id
                 ) and
       r.type = 2;

However, the use of enabled makes the code clearer, in my opinion. An index on records(ref) is needed for performance.

If you assumes that the ids are always incrementing, you could also use the largest id .

Answer 2

For this query, you need an index on both columns that appear in the where clause:

create index myidx on records(type, enabled);

With the index, the database should be able to execute the query efficiently. You may also want to try inverting the columns order and see if it increases performance.