有时候我的查询不使用索引,但有时会这样做。你能解释一下为什么会这样吗?
这是表格结构。
MariaDB [crm]> desc vtiger_project;
+------------------------+---------------+------+-----+---------+-------+
| Field | Type | Null | Key | Default | Extra |
+------------------------+---------------+------+-----+---------+-------+
| projectid | int(11) | NO | PRI | 0 | |
| projectname | varchar(100) | YES | | NULL | |
| projecttype | varchar(50) | YES | | NULL | |
| siteaddress | varchar(500) | YES | | NULL | |
| state | varchar(100) | YES | | NULL | |
| district | varchar(100) | YES | | NULL | |
| city | varchar(100) | YES | | NULL | |
| pincode | varchar(100) | YES | | NULL | |
| phone | varchar(100) | YES | | NULL | |
| startdate | date | YES | MUL | NULL | |
| branch | varchar(100) | YES | | NULL | |
| customer | int(11) | YES | MUL | NULL | |
| dealer | int(11) | YES | | NULL | |
| contractor | int(11) | YES | MUL | NULL | |
| architect | int(11) | YES | MUL | NULL | |
| carpenter | int(11) | YES | MUL | NULL | |
| productcategory | varchar(100) | YES | | NULL | |
| brand_preferred | varchar(100) | YES | | NULL | |
| formal_spec_check | varchar(3) | YES | | NULL | |
| formal_spec_details | varchar(250) | YES | | NULL | |
| projectstatus | varchar(25) | YES | | NULL | |
| project_reason_loosing | varchar(100) | YES | | NULL | |
| reason_loosing_deatils | varchar(250) | YES | | NULL | |
| reason_winning_deatils | varchar(250) | YES | | NULL | |
| adjustment | decimal(25,8) | YES | | NULL | |
| exciseduty | decimal(25,3) | YES | | NULL | |
| total | decimal(25,8) | YES | | NULL | |
| subtotal | decimal(25,8) | YES | | NULL | |
| taxtype | varchar(25) | YES | | NULL | |
| discount_percent | decimal(25,3) | YES | | NULL | |
| discount_amount | decimal(25,8) | YES | | NULL | |
| s_h_amount | decimal(25,8) | YES | | NULL | |
| currency_id | int(19) | NO | | 1 | |
| conversion_rate | decimal(10,3) | NO | | 1.000 | |
| actual_sale | varchar(255) | YES | | NULL | |
| expected_sale_in_na | varchar(255) | YES | | NULL | |
| primary_decision_maker | varchar(100) | YES | | NULL | |
+------------------------+---------------+------+-----+---------+-------+
正如您在下面看到的那样输出第一个查询确实达到了索引,但第二个查询没有达到,但是我对该查询所做的唯一更改是 startdate 部分。 我做错了什么?
MariaDB [crm]> explain SELECT
-> COUNT(projectid)
-> FROM
-> vtiger_project
-> WHERE
-> 82582 IN (customer , contractor, architect, carpenter)
-> AND projectstatus NOT IN ('Supplied' , 'Closed As Complete', 'Closed As Lost')
-> AND actual_sale IS NULL
-> AND startdate > NOW()
-> ;
+------+-------------+----------------+-------+---------------+---------------+---------+------+------+------------------------------------+
| id | select_type | table | type | possible_keys | key | key_len | ref | rows | Extra |
+------+-------------+----------------+-------+---------------+---------------+---------+------+------+------------------------------------+
| 1 | SIMPLE | vtiger_project | range | startdate_idx | startdate_idx | 4 | NULL | 352 | Using index condition; Using where |
+------+-------------+----------------+-------+---------------+---------------+---------+------+------+------------------------------------+
1 row in set (0.00 sec)
MariaDB [crm]> explain SELECT
-> COUNT(projectid)
-> FROM
-> vtiger_project
-> WHERE
-> 82582 IN (customer , contractor, architect, carpenter)
-> AND projectstatus NOT IN ('Supplied' , 'Closed As Complete', 'Closed As Lost')
-> AND actual_sale IS NULL
-> AND startdate < NOW()
-> ;
+------+-------------+----------------+------+---------------+------+---------+------+-------+-------------+
| id | select_type | table | type | possible_keys | key | key_len | ref | rows | Extra |
+------+-------------+----------------+------+---------------+------+---------+------+-------+-------------+
| 1 | SIMPLE | vtiger_project | ALL | startdate_idx | NULL | NULL | NULL | 15779 | Using where |
+------+-------------+----------------+------+---------------+------+---------+------+-------+-------------+
&#13;
答案 0 :(得分:2)
答案在解释计划的行列中
在第一种情况下,您有352
在第二个15779
这表明在两个查询之间查询优化器的不同策略。
也可能是您正在不同的服务器中执行这两个查询,或者在解释
的情况下执行其他不同的条件答案 1 :(得分:0)
(sqlbot&#39;评论回答了这个问题。)
此复合索引可以进一步优化:
INDEX(actual_sale, startdate)
但是,这可能会很有效actual_sale
NULL
&#39;太频繁&#39;。否则,全表扫描会更好,但优化器可能无法解决这个问题。
您的许多列都被认为过大。这种过度行为可能会导致性能问题。