Question

创建实体化视图“mviews_price2”

id                 - Integer
price             - Integer
price2            - Integer
unit_price        - Integer
location_site_id - Integer  
leave_site_id    - Integer
web              - Boolean
category_id      - Integer

创建索引：

CREATE INDEX "location_site_id_idx" ON "mviews_price2" ("location_site_id")
CREATE INDEX "leave_site_id_idx" ON "mviews_price2" ("leave_site_id")
CREATE INDEX "web_idx" ON "mviews_price2" ("web")

获取查询：

SELECT DISTINCT "mviews_price2"."category_id" FROM "mviews_price2" WHERE ("mviews_price2"."location_site_id" = 1 OR "mviews_price2"."leave_site_id" = 1 OR "mviews_price2"."web" = true)

说明：

HashAggregate  (cost=42757.45..42757.82 rows=37 width=4) (actual time=596.252..596.257 rows=37 loops=1)'
Group Key: category_id'
->  Seq Scan on mviews_price2  (cost=0.00..39614.90 rows=1257021 width=4) (actual time=0.038..362.877 rows=1341021 loops=1)'
        Filter: ((location_site_id = 1) OR (leave_site_id = 1) OR web)'
        Rows Removed by Filter: 423439'
Planning time: 0.954 ms'
Execution time: 596.311 ms'

在视图中 - 1764460行

问题：

为什么不使用索引？
为什么查询速度很慢？

Answer 1

OR很难优化。尝试使用UNION编写查询：

SELECT "mviews_price2"."category_id"
FROM "mviews_price2"
WHERE "mviews_price2"."location_site_id" = 1
UNION  -- on purpose to remove duplicates
SELECT "mviews_price2"."category_id"
FROM "mviews_price2"
WHERE "mviews_price2"."leave_site_id" = 1
UNION  -- on purpose to remove duplicates
SELECT "mviews_price2"."category_id"
FROM "mviews_price2"
WHERE "mviews_price2"."web" = true;

每个子查询都应该能够利用其中一个索引。

注意：这仍然可能无法加快查询速度。 web上的索引 - 特别是 - 是一个只有2个值的字段（如果它可以接受NULL则为3）。除非web非常罕见，否则结果可能仍然是全表扫描。

物化视图中的Postgresql慢查询

1 个答案: