鉴于我有标记的文章，一篇文章可以有n个标签。 目前，大约有250,000个标签条目都指向后面 在他们的归属文章上。
现在我想找到符合特定条件的文章中的所有标签。 我提出了两种不同的方法。两者都有缺点而且很慢。 可能有人可以指出我正确的方向，如何加快甚至来 提供更好的解决方案。
键（ind，rindex）是varchar（255），遗憾的是无法更改
取7.5 - subselect在50ms内返回60条记录
SELECT count(*) AS tagscount, tags.value FROM tags
WHERE tags.`rindex` IN
(
SELECT article.ind
FROM article
INNER JOIN struktur ON (struktur.ind = article.struktur)
WHERE article.date = '2011-12-21'
)
AND tags.`rtable` = 'article'
GROUP BY tags.value ORDER BY tagscount DESC LIMIT 20
需要60ms
SELECT count(*) AS tagscount, tags.value FROM tags
INNER JOIN article ON (article.ind = tags.rindex AND tags.rtable = 'article')
LEFT JOIN structure ON (article.structure = structure.ind)
WHERE article.date = '2011-12-21'
GROUP BY tags.value ORDER BY tagscount DESC LIMIT 20
当我将article.date = '2011-12-21'更改为article.date >= '2009-12-21'时
查询＃1
查询＃2
如果您需要更多信息，我将很乐意提供
mysql> SHOW COLUMNS FROM tags;
+---------+--------------+------+-----+---------+-------+
| Field | Type | Null | Key | Default | Extra |
+---------+--------------+------+-----+---------+-------+
| ind | varchar(255) | NO | PRI | | |
| rtable | varchar(255) | NO | MUL | | |
| rindex | varchar(255) | NO | MUL | | |
| value | varchar(40) | YES | MUL | NULL | |
+---------+--------------+------+-----+---------+-------+
mysql> SHOW indexes FROM tags
+-------+------------+---------------------+--------------+-------------+-----------+-------------+----------+--------+------+------------+---------+
| Table | Non_unique | Key_name | Seq_in_index | Column_name | Collation | Cardinality | Sub_part | Packed | Null | Index_type | Comment |
+-------+------------+---------------------+--------------+-------------+-----------+-------------+----------+--------+------+------------+---------+
| tags | 0 | tags_ind | 1 | ind | A | 275834 | NULL | NULL | | BTREE | |
| tags | 1 | tags_tag | 1 | tag | A | 27583 | NULL | NULL | YES | BTREE | |
| tags | 1 | tags_rindex | 1 | rindex | A | 55166 | NULL | NULL | | BTREE | |
| tags | 1 | tags_rindex_tabelle | 1 | tabelle | A | 4 | 30 | NULL | | BTREE | |
| tags | 1 | tags_rindex_tabelle | 2 | rindex | A | 55166 | 50 | NULL | | BTREE | |
+-------+------------+---------------------+--------------+-------------+-----------+-------------+----------+--------+------+------------+---------+
mysql> SHOW COLUMNS FROM structure;
+------------------------+--------------+------+-----+---------+-------+
| Field | Type | Null | Key | Default | Extra |
+------------------------+--------------+------+-----+---------+-------+
| ind | varchar(255) | NO | PRI | | |
+------------------------+--------------+------+-----+---------+-------+
mysql> SHOW COLUMNS FROM artikel;
+--------------------+--------------+------+-----+------------+-------+
| Field | Type | Null | Key | Default | Extra |
+--------------------+--------------+------+-----+------------+-------+
| ind | varchar(255) | NO | PRI | | |
| date | date | NO | MUL | 0000-00-00 | |
+--------------------+--------------+------+-----+------------+-------+
mysql> explain #1
+----+--------------------+----------+--------+-------------------------------------------------------------------------------------+---------------------+---------+---------------------+--------+----------------------------------------------+
| id | select_type | table | type | possible_keys | key | key_len | ref | rows | Extra |
+----+--------------------+----------+--------+-------------------------------------------------------------------------------------+---------------------+---------+---------------------+--------+----------------------------------------------+
| 1 | PRIMARY | tags | ref | tags_rindex_tabelle | tags_rindex_tabelle | 32 | const | 177175 | Using where; Using temporary; Using filesort |
| 2 | DEPENDENT SUBQUERY | artikel | eq_ref | artikel_ind,zeitraum_start_i,freigabe_i,korrektur_i,struktur_i,artikel_start_slot_i | artikel_ind | 257 | func | 1 | Using where |
| 2 | DEPENDENT SUBQUERY | struktur | eq_ref | struktur_ind,struktur_host | struktur_ind | 257 | ec.artikel.struktur | 1 | Using where |
+----+--------------------+----------+--------+-------------------------------------------------------------------------------------+---------------------+---------+---------------------+--------+----------------------------------------------+
mysql> explain #2
+----+-------------+----------+--------+-------------------------------------------------------------------------------------+---------------------+---------+---------------------+--------+----------------------------------------------+
| id | select_type | table | type | possible_keys | key | key_len | ref | rows | Extra |
+----+-------------+----------+--------+-------------------------------------------------------------------------------------+---------------------+---------+---------------------+--------+----------------------------------------------+
| 1 | SIMPLE | tags | ref | tags_rindex,tags_rindex_tabelle | tags_rindex_tabelle | 32 | const | 177175 | Using where; Using temporary; Using filesort |
| 1 | SIMPLE | artikel | eq_ref | artikel_ind,zeitraum_start_i,freigabe_i,korrektur_i,struktur_i,artikel_start_slot_i | artikel_ind | 257 | ec.tags.rindex | 1 | Using where |
| 1 | SIMPLE | struktur | eq_ref | struktur_ind,struktur_host | struktur_ind | 257 | ec.artikel.struktur | 1 | Using where |
+----+-------------+----------+--------+-------------------------------------------------------------------------------------+---------------------+---------+---------------------+--------+----------------------------------------------+
答案 0 :(得分：1)
我认为artikel.ind不限于以与artikel.date相同的顺序提升词汇顺序。如果是，那么显而易见的解决方案是在对应于日期范围的rindex中添加限制。
实际上，它似乎正在使用一个合适的计划。
不改变数据类型的最佳选择是创建在(artikel.date, tags.value, artikel.ind)上编制索引的物化视图，然后查询。