SQL優化之外連接（一）

         前兩天接觸了一個項目，裏面的一個報表查詢功能，訪問數據庫很慢，嚴重時請求直接超時，寫的sql語句得有三四百行，真是頭疼，沒辦法，硬着頭皮，只能一段一段的分析sql，查詢sql慢的問題。解決思路就是使用EXPLAIN 命令來查看哪些表沒有用到索引。通過查找終於找到問題所在，現記錄如下：

1、問題所在：

join on 條件後面不能使用or連接，造成索引失效，會全表掃描。

sql語句實例：

SELECT
		sp.id,
		gcm.id AS gcmId,
		gcm.material_id
		FROM
		gt_supply_order sp
		LEFT JOIN gt_supply_order_material gm ON sp.id = gm.supply_order_id
		LEFT JOIN gt_contract gc ON sp.contract_id = gc.id
		left join gt_orderform_goods nm on nm.order_id=gc.id or nm.id=gm.order_form_goods_id
		LEFT JOIN gt_orderform_goods gcm ON gcm.id = nm.id
		OR gcm.main_contract_goods_id = nm.id
		AND gcm.deleted = 0
		WHERE
		sp.deleted = 0
		AND gm.deleted = 0
		AND sp.contract_id = #{jgMaterialSupplyVO.contractId} and sp.goods_class_name = #{jgMaterialSupplyVO.goodsName} and sp.supply_order_status_enum=4
		GROUP BY
		sp.id,gcm.id

2、解決方案：

上面查詢（join on or）可以用 union 替換（索引有效），缺點sql語句長（相比性能問題還是可取的）

中間有個小插曲：

union 操作會對結果去重且排序，union all 不會去重

優化後的SQL如下：

select  a.id,gcm.id AS gcmId,gcm.material_id from(
 SELECT
  sp.id,
    nm.id as nmid
  FROM
  gt_supply_order sp
  LEFT JOIN gt_supply_order_material gm ON sp.id = gm.supply_order_id
  LEFT JOIN gt_contract gc ON sp.contract_id = gc.id
  left join gt_orderform_goods nm on nm.order_id=gc.id  
 WHERE
  sp.deleted = 0
  AND gm.deleted = 0 AND nm.deleted=0
   AND sp.contract_id = 137957291194000 and sp.goods_class_name = '橋樑支座' and sp.supply_order_status_enum=4
 union
  SELECT
  sp.id,
  nm.id as nmid
  FROM
  gt_supply_order sp
  LEFT JOIN gt_supply_order_material gm ON sp.id = gm.supply_order_id
  LEFT JOIN gt_contract gc ON sp.contract_id = gc.id
  left join gt_orderform_goods nm on  nm.id=gm.order_form_goods_id
 WHERE
  sp.deleted = 0
  AND gm.deleted = 0 AND nm.deleted=0
   AND sp.contract_id = 137957291194000 and sp.goods_class_name = '橋樑支座' and sp.supply_order_status_enum=4
)a
 LEFT JOIN gt_orderform_goods gcm ON gcm.id = a.nmid 
where gcm.id IS NOT NULL
GROUP BY a.id,gcm.id
union
select  b.id,gcm.id AS gcmId,gcm.material_id from 
(
 SELECT
  sp.id,
    nm.id as nmid
  FROM
  gt_supply_order sp
  LEFT JOIN gt_supply_order_material gm ON sp.id = gm.supply_order_id
  LEFT JOIN gt_contract gc ON sp.contract_id = gc.id
  left join gt_orderform_goods nm on nm.order_id=gc.id  

 WHERE
  sp.deleted = 0
  AND gm.deleted = 0 AND nm.deleted=0
   AND sp.contract_id = 137957291194000 and sp.goods_class_name = '橋樑支座' and sp.supply_order_status_enum=4
 union
  SELECT
 sp.id,
  nm.id as nmid
  FROM
  gt_supply_order sp
  LEFT JOIN gt_supply_order_material gm ON sp.id = gm.supply_order_id
  LEFT JOIN gt_contract gc ON sp.contract_id = gc.id
  left join gt_orderform_goods nm on nm.id=gm.order_form_goods_id
 WHERE
  sp.deleted = 0
  AND gm.deleted = 0 AND nm.deleted=0
   AND sp.contract_id = 137957291194000 and sp.goods_class_name = '橋樑支座' and sp.supply_order_status_enum=4
)b
 LEFT JOIN gt_orderform_goods gcm ON  gcm.main_contract_goods_id = b.nmid 
where gcm.id IS NOT NULL
 GROUP BY
  b.id,gcm.id

 

3、總結：

（1）通過上面的優化，查詢速度由原來的77秒，提速到了0.107秒，提高700多倍，可見，SQL優化是多麼的重要呀。

（2）這種優化方式可能不是最優的（sql語句太長），如有高人還有其他優化方式，希望可以交流，謝謝。

（3）有興趣的小夥伴可以測試下，歡迎交流。