如何在Slick中过滤部分索引的布尔列？ 变化1 变化2 版本3

如何解决如何在Slick中过滤部分索引的布尔列？ 变化1 变化2 版本3

我有一个用Slick查询的Postgres数据库。问题是我有一个被索引（部分索引）的BOOLEAN列（archived）。我通过EXPLAIN / ANALYZE发现，仅当查询采用WHERE archived IS FALSE形式时才使用索引，但是Slick似乎只输出WHERE archived = False或WHERE NOT archived ...我该如何修改我光滑的查询，或者，如果不可能的话，可以更改索引，使其与archived = false一起使用？这是更多信息：

变化1

光滑

val query = Messages
  .filterOpt(isArchived) { (row,archived) => 
    if (archived) { row.archived }
    else { !row.archived }
  }
  .sortBy(_.id.desc)

平滑输出

select "id","info","archived" from "messages" where not "archived" order by "id" desc

查询说明

Sort  (cost=37988.82..37988.90 rows=33 width=1033)
  Sort Key: id DESC
  ->  Seq Scan on messages  (cost=0.00..37987.99 rows=33 width=1033)
        Filter: (NOT archived)

变化2

光滑

val query = Messages
  .filterOpt(isArchived) { (row,archived) => 
    row.archived === archived
  }
  .sortBy(_.id.desc)

平滑输出

select "id","archived" from "messages" where "archived" = false order by "id" desc

查询说明（与说明1相同）

Sort  (cost=37988.82..37988.90 rows=33 width=1033)
  Sort Key: id DESC
  ->  Seq Scan on messages  (cost=0.00..37987.99 rows=33 width=1033)
        Filter: (NOT archived)

版本3

光滑

???

所需）平滑输出

select "id","archived" from "messages" where "archived" IS false order by "id" desc

（所需）查询说明

Sort  (cost=21.67..21.75 rows=33 width=1033)
  Sort Key: id DESC
  ->  Index Scan using messages_archived on messages  (cost=0.27..20.84 rows=33 width=1033)

解决方法

不是使用WHERE archived is false创建我们的部分索引，而是使用WHERE not archived或WHERE archived = false创建它。 Postgres似乎很聪明，可以相互转换其中的第二和第三，但不能与其他两个相互转换。

,

由于这个useful Medium Article，我找到了答案。它涉及扩展Slick PostgresProfile API

// ExtendedPostgresProfile.scala

import slick.ast.Library.SqlOperator
import slick.jdbc.PostgresProfile
import slick.jdbc.PostgresProfile.api._
import slick.lifted.BaseExtensionMethods
import scala.language.implicitConversions

/***
  * We need to Extend the PostgresProfile because the standard Slick filter on
  * `messages.archived` doesn't utilize the index on that column
  * (it uses = instead of IS). This explicitly creates an IS operator
  */
trait ExtendedPostgresProfile extends PostgresProfile {
  trait ExtApi extends API {
    implicit def extendedBoolColumn(c: Rep[Boolean]): BaseBooleanExtendedBooleanMethods[Boolean] =
      new BaseBooleanExtendedBooleanMethods[Boolean](c)
  }

  override val api: ExtApi = new ExtApi {}

  final class BaseBooleanExtendedBooleanMethods[P1](val c: Rep[P1]) extends BaseExtensionMethods[P1] {
    def is[P2,R](e: Rep[P2])(implicit om: o#arg[P1,P2]#to[P1,R]) =
      om.column(Operators.IS,n,e.toNode)
  }

  object Operators{
    val IS = new SqlOperator("IS")
  }
}

object ExtendedPostgresProfile extends ExtendedPostgresProfile

和实际的Slick代码：

// import slick.jdbc.PostgresProfile.api._ <- Removed
import ExtendedPostgresProfile.api._

val query = Messages
  .filterOpt(isArchived) { (row,archived) => 
    row.archived is archived
  }
  .sortBy(_.id.desc)

输出是

select "id","info","archived" from "messages" where "archived" IS false order by "id" desc