我的查询是一个库存表。子查询连接所做的是获取每个库存 Assets 的工单总数。如果我使用设备类型、供应商、位置和房间的主要连接运行基本查询，它运行得很好。不到一秒返回结果。将它与子查询连接一起使用，需要 15 到 20 秒才能返回结果。
这是完整的查询:
SELECT `inventory`.inventory_id AS 'inventory_id',
`inventory`.media_tag AS 'media_tag',
`inventory`.asset_tag AS 'asset_tag',
`inventory`.idea_tag AS 'idea_tag',
`equipTypes`.equipment_type AS 'equipment_type',
`inventory`.equip_make AS 'equip_make',
`inventory`.equip_model AS 'equip_model',
`inventory`.equip_serial AS 'equip_serial',
`inventory`.sales_order AS 'sales_order',
`vendors`.vendor_name AS 'vendor_name',
`inventory`.purchase_order AS 'purchase_order',
`status`.status AS 'status',
`locations`.location_name AS 'location_name',
`rooms`.room_number AS 'room_number',
`inventory`.notes AS 'notes',
`inventory`.send_to AS 'send_to',
`inventory`.one_to_one AS 'one_to_one',
`enteredBy`.user_name AS 'user_name',
from_unixtime(`inventory`.enter_date, '%m/%d/%Y') AS 'enter_date',
from_unixtime(`inventory`.modified_date, '%m/%d/%Y') AS 'modified_date',
COALESCE(at.assets,0) AS assets
FROM mod_inventory_data AS `inventory`
LEFT JOIN mod_inventory_equip_types AS `equipTypes`
ON `equipTypes`.equip_type_id = `inventory`.equip_type_id
LEFT JOIN mod_vendors_main AS `vendors`
ON `vendors`.vendor_id = `inventory`.vendor_id
LEFT JOIN mod_inventory_status AS `status`
ON `status`.status_id = `inventory`.status_id
LEFT JOIN mod_locations_data AS `locations`
ON `locations`.location_id = `inventory`.location_id
LEFT JOIN mod_locations_rooms AS `rooms`
ON `rooms`.room_id = `inventory`.room_id
LEFT JOIN mod_users_data AS `enteredBy`
ON `enteredBy`.user_id = `inventory`.entered_by
LEFT JOIN
( SELECT asset_tag, count(*) AS assets
FROM mod_workorder_data
WHERE asset_tag IS NOT NULL
GROUP BY asset_tag ) AS at
ON at.asset_tag = inventory.asset_tag
ORDER BY inventory_id ASC LIMIT 0,20
MySQL EXPLAIN 数据在这里
+----+-------------+--------------------+--------+---------------+-----------+---------+-------------------------------------+-------+---------------------------------+
| id | select_type | table | type | possible_keys | key | key_len | ref | rows | Extra |
+----+-------------+--------------------+--------+---------------+-----------+---------+-------------------------------------+-------+---------------------------------+
| 1 | PRIMARY | inventory | ALL | NULL | NULL | NULL | NULL | 12612 | Using temporary; Using filesort |
| 1 | PRIMARY | equipTypes | eq_ref | PRIMARY | PRIMARY | 4 | spsd_woidbs.inventory.equip_type_id | 1 | |
| 1 | PRIMARY | vendors | eq_ref | PRIMARY | PRIMARY | 4 | spsd_woidbs.inventory.vendor_id | 1 | |
| 1 | PRIMARY | status | eq_ref | PRIMARY | PRIMARY | 4 | spsd_woidbs.inventory.status_id | 1 | |
| 1 | PRIMARY | locations | eq_ref | PRIMARY | PRIMARY | 4 | spsd_woidbs.inventory.location_id | 1 | |
| 1 | PRIMARY | rooms | eq_ref | PRIMARY | PRIMARY | 4 | spsd_woidbs.inventory.room_id | 1 | |
| 1 | PRIMARY | enteredBy | eq_ref | PRIMARY | PRIMARY | 4 | spsd_woidbs.inventory.entered_by | 1 | |
| 1 | PRIMARY | <derived2> | ALL | NULL | NULL | NULL | NULL | 4480 | |
| 2 | DERIVED | mod_workorder_data | range | asset_tag | asset_tag | 13 | NULL | 15897 | Using where; Using index |
+----+-------------+--------------------+--------+---------------+-----------+---------+-------------------------------------+-------+---------------------------------+
使用 MySql 查询分析我得到这个:
+--------------------------------+------------+
| Status | Time |
+--------------------------------+------------+
| starting | 0.000020 |
| checking query cache for query | 0.000263 |
| Opening tables | 0.000034 |
| System lock | 0.000013 |
| Table lock | 0.000079 |
| optimizing | 0.000011 |
| statistics | 0.000138 |
| preparing | 0.000019 |
| executing | 0.000010 |
| Sorting result | 0.000004 |
| Sending data | 0.015103 |
| init | 0.000094 |
| optimizing | 0.000009 |
| statistics | 0.000049 |
| preparing | 0.000022 |
| Creating tmp table | 0.000104 |
| executing | 0.000009 |
| Copying to tmp table | 15.410168 |
| Sorting result | 0.009488 |
| Sending data | 0.000215 |
| end | 0.000006 |
| removing tmp table | 0.001997 |
| end | 0.000018 |
| query end | 0.000005 |
| freeing items | 0.000112 |
| storing result in query cache | 0.000011 |
| removing tmp table | 0.000022 |
| closing tables | 0.000036 |
| logging slow query | 0.000005 |
| logging slow query | 0.000005 |
| cleaning up | 0.000013 |
+--------------------------------+------------+
这表明瓶颈正在复制到临时表，但我不确定如何加快速度。我可以配置服务器端的设置以加快速度吗？是否可以对现有查询进行更改以产生更快的相同结果？
在我看来，LEFT JOIN 子查询每次都会给出相同的结果数据矩阵，因此如果它必须对 list 列表中的每一行运行该查询，我可以理解为什么它会很慢。或者MySQL在运行时是否缓存子查询？我以为我在某处读到 MySQL 不缓存子查询，这是真的吗？
感谢任何帮助。
最佳答案
这是我所做的，看起来效果不错。我创建了一个名为 mod_workorder_counts 的表。该表有两个字段，唯一的 Assets 标签，以及 wo_count 和 INT(3) 字段。我正在用这个查询填充该表:
INSERT INTO mod_workorder_counts ( asset_tag, wo_count )
select s.asset_tag, ct
FROM
( SELECT t.asset_tag, count(*) as ct
FROM mod_workorder_data t
WHERE t.asset_tag IS NOT NULL
GROUP BY t.asset_tag
) as s
ON DUPLICATE KEY UPDATE mod_workorder_counts.wo_count = ct执行时间为 0.1580 秒，这可能会被认为稍微慢一点，但还不错。
现在，当我对原始查询运行此修改时:
SELECT `inventory`.inventory_id AS 'inventory_id',
`inventory`.media_tag AS 'media_tag',
`inventory`.asset_tag AS 'asset_tag',
`inventory`.idea_tag AS 'idea_tag',
`equipTypes`.equipment_type AS 'equipment_type',
`inventory`.equip_make AS 'equip_make',
`inventory`.equip_model AS 'equip_model',
`inventory`.equip_serial AS 'equip_serial',
`inventory`.sales_order AS 'sales_order',
`vendors`.vendor_name AS 'vendor_name',
`inventory`.purchase_order AS 'purchase_order',
`status`.status AS 'status',
`locations`.location_name AS 'location_name',
`rooms`.room_number AS 'room_number',
`inventory`.notes AS 'notes',
`inventory`.send_to AS 'send_to',
`inventory`.one_to_one AS 'one_to_one',
`enteredBy`.user_name AS 'user_name',
from_unixtime(`inventory`.enter_date, '%m/%d/%Y') AS 'enter_date',
from_unixtime(`inventory`.modified_date, '%m/%d/%Y') AS 'modified_date',
COALESCE(at.wo_count, 0) AS workorders
FROM mod_inventory_data AS `inventory`
LEFT JOIN mod_inventory_equip_types AS `equipTypes`
ON `equipTypes`.equip_type_id = `inventory`.equip_type_id
LEFT JOIN mod_vendors_main AS `vendors`
ON `vendors`.vendor_id = `inventory`.vendor_id
LEFT JOIN mod_inventory_status AS `status`
ON `status`.status_id = `inventory`.status_id
LEFT JOIN mod_locations_data AS `locations`
ON `locations`.location_id = `inventory`.location_id
LEFT JOIN mod_locations_rooms AS `rooms`
ON `rooms`.room_id = `inventory`.room_id
LEFT JOIN mod_users_data AS `enteredBy`
ON `enteredBy`.user_id = `inventory`.entered_by
LEFT JOIN mod_workorder_counts AS at
ON at.asset_tag = inventory.asset_tag
ORDER BY inventory_id ASC LIMIT 0,20它在 0.0051 秒内执行。这使得两个查询之间的总时间为 0.1631 秒，接近 1/10 秒，而原始子查询为 15+ 秒。
如果我只是包含字段“wo_count”而不使用 COALESCE，我会得到“mod_workorder_counts”表中未列出的任何 Assets 标签的 NULL 值。因此，对于任何 NULL 值，COALESCE 都会给我一个 0，这正是我想要的。
现在我将对其进行设置，以便在为 Assets 标签输入工作订单时，我将在那时对计数表更新进行 INSERT/UPDATE 查询，这样它就不会不必要地运行。
关于mysql - 我怎样才能加快我的查询。子查询太慢，我们在Stack Overflow上找到一个类似的问题： https://stackoverflow.com/questions/5683351/