【发布时间】:2014-03-07 16:30:04
【问题描述】:
我有一个 PostgreSQL 字符串数组作为表中的一列。我使用 GIN 方法创建了一个索引。但是任何查询都不会使用索引(相反,它们使用过滤器对整个表进行顺序扫描)。我错过了什么?
这是我的迁移:
class CreateDocuments < ActiveRecord::Migration
def up
create_table :documents do |t|
t.string :title
t.string :tags, array: true, default: []
t.timestamps
end
add_index :documents, :tags, using: 'gin'
(1..100000).each do |i|
tags = []
tags << 'even' if (i % 2) == 0
tags << 'odd' if (i % 2) == 1
tags << 'divisible by 3' if (i % 3) == 0
tags << 'divisible by 4' if (i % 4) == 0
tags << 'divisible by 5' if (i % 5) == 0
Document.create(
title: i,
tags: tags
)
end
end
def down
drop_table :documents
end
end
这是我的查询,结果行数。
Document.where("'divisible by 5' = ANY (tags)").explain
Document Load (249.8ms) SELECT "documents".* FROM "documents" WHERE ('divisible by 5' = ANY (tags))
D, [2014-03-07T17:09:49.689709 #41937] DEBUG -- : Document Load (249.8ms) SELECT "documents".* FROM "documents" WHERE ('divisible by 5' = ANY (tags))
=> EXPLAIN for: SELECT "documents".* FROM "documents" WHERE ('divisible by 5' = ANY (tags))
QUERY PLAN
-----------------------------------------------------------------
Seq Scan on documents (cost=0.00..3500.00 rows=20057 width=69)
Filter: ('divisible by 5'::text = ANY ((tags)::text[]))
(2 rows)
Document.where("'divisible by 5' = ANY (tags)").length
Document Load (258.0ms) SELECT "documents".* FROM "documents" WHERE ('divisible by 5' = ANY (tags))
D, [2014-03-07T17:09:55.536517 #41937] DEBUG -- : Document Load (258.0ms) SELECT "documents".* FROM "documents" WHERE ('divisible by 5' = ANY (tags))
=> 20000
【问题讨论】:
-
表
products有多少行? -
该列的基数是多少?报告的行数有多少唯一记录?
-
@IgorRomanchenko 该表刚刚超过 100,000 行
-
@Mihai 有超过 100,000 个唯一行。有问题的列有超过 18,000 个唯一值
-
可能是 ANY 关键字使优化器选择顺序扫描作为最佳选项。该查询返回多少行?
标签: sql ruby-on-rails postgresql indexing query-performance