浅谈Slick（3）－ Slick201：从fp角度了解Slick

  我在上期讨论里已经成功的创建了一个简单的Slick项目，然后又尝试使用了一些最基本的功能。Slick是一个FRM（Functional Relational Mapper）,是为fp编程提供的scala SQL Query集成环境，可以让编程人员在scala编程语言里用函数式编程模式来实现对数据库操作的编程。在这篇讨论里我想以函数式思考模式来加深了解Slick。我对fp编程模式印象最深的就是类型匹配：从参数类型和返回结果类型来了解函数功能。所以上面我所指的函数式思考方式主要是从Slick函数的类型匹配角度来分析函数所起的作用和具体使用方式。

我们先了解一下建表过程：

import slick.driver.H2Driver.api._
object slick201 {
  //projection case classes 表列模版
  case class Coffee(
                     id: Option[Long]
                     ,name: String
                     ,sup_ID: Int
                     ,price: Double
                     ,grade: Grade
                     ,total: Int
                   )
  case class Supplier(
                       id: Option[Int]
                       ,name: String
                       ,address: String
                       ,website: Option[String]
                     )
  //自定义字段
  abstract class Grade(points: Int)
  object Grade {
    case object Premium extends Grade(2)
    case object Quality extends Grade(1)
    case object Bestbuy extends Grade(0)

    def fromInt(p: Int) = p match {
        case 2 => Premium
        case 1 => Quality
        case 0 => Bestbuy
    }
    def toInt(g: Grade) = g match {
      case Premium => 2
      case Quality => 1
      case Bestbuy => 0
    }
    implicit val customColumn: BaseColumnType[Grade] =
      MappedColumnType.base[Grade,Int](Grade.toInt, Grade.fromInt)
  }
  //schema 表行结构定义
  class Coffees(tag: Tag) extends Table[Coffee](tag, "COFFEES") {
    def id = column[Long]("COF_ID", O.AutoInc, O.PrimaryKey)
    def name = column[String]("COF_NAME")
    def price = column[Double]("COF_PRICE")
    def supID = column[Int]("COF_SUP")
    def grade = column[Grade]("COF_GRADE", O.Default(Grade.Bestbuy))
    def total = column[Int]("COF_TOTAL", O.Default(0))

    def * = (id.?,name,supID,price,grade,total) <> (Coffee.tupled, Coffee.unapply)

    def supplier = foreignKey("SUP_FK",supID,suppliers)(_.id,onUpdate = ForeignKeyAction.Restrict, onDelete = ForeignKeyAction.Cascade)
    def nameidx = index("NM_IX",name,unique = true)
  }
  val coffees = TableQuery[Coffees]

  class Suppliers(tag: Tag) extends Table[Supplier](tag, "SUPPLIERS") {
    def id = column[Int]("SUP_ID", O.PrimaryKey, O.AutoInc)
    def name = column[String]("SUP_NAME")
    def address = column[String]("SUP_ADDR", O.Default("-"))
    def website = column[Option[String]]("SUP_WEB")

    def * = (id.?, name, address, website) <> (Supplier.tupled, Supplier.unapply)
    def addidx = index("ADDR_IX",(name,address),unique = true)
  }
  val suppliers = TableQuery[Suppliers]

}

我尽量把经常会遇到的情况如：定义字段、建索引、默认值、自定义字段等都作了尝试。coffees和suppliers代表了最终的数据表Query，def * 定义了这个Query的默认返回结果字段。

所有的定义都是围绕着表行（Table Row）结构进行的，包括：表属性及操作（Table member methods）、字段（Column）、字段属性（ColumnOptions）。表行定义操作方法基本都在slick.lifted.AbstractTable里、表属性定义在slick.model命名空间里、而大部分的帮助支持函数都在slick.lifted命名空间的其它对象里。

表行的实际类型如下：

abstract class Table[T](_tableTag: Tag, _schemaName: Option[String], _tableName: String) extends AbstractTable[T](_tableTag, _schemaName, _tableName) { table => ...}
 
/** The profile-independent superclass of all table row objects.
  * @tparam T Row type for this table. Make sure it matches the type of your `*` projection. */
abstract class AbstractTable[T](val tableTag: Tag, val schemaName: Option[String], val tableName: String) extends Rep[T] {...}

如上所示，Table[T] extends AbstractTable[T]。现在所有表行定义操作函数应该在slick.profile.relationalTableComponent.Table里可以找得到。值得注意的是表行的最终类型是Rep[T]，T可能是case class或者Tuple，被升格（lift）到Rep[T]。所以大部分表行定义的支持函数都是在slick.lifted命名空间内的。

上面我们使用了模版对应表行定义方式，所有列都能和模版case class对应。那么在定义projection def * 时就需要使用<>函数：

def <>[R : ClassTag](f: (U => R), g: (R => Option[U])) = new MappedProjection[R, U](shape.toNode(value), MappedScalaType.Mapper(g.andThen(_.get).asInstanceOf[Any => Any], f.asInstanceOf[Any => Any], None), implicitly[ClassTag[R]])

f,g是两个case class <> Tuple转换函数。在上面的例子里我们提供的是tupled和unapply，效果就是这样的：

  Coffee.tupled
  //res2: ((Option[Long], String, Int, Double, Grade, Int)) => Coffee = <function1>
  Coffee.unapply _
  //res3: Coffee => Option[(Option[Long], String, Int, Double, Grade, Int)] = <function1>

res2 >>> 把tuple: (...)转成coffee，res2 >>> 把coffee转成Option[(...)]

TableQuery[T]继承了Query[T]：slick.lifted.Query.scala

/** Represents a database table. Profiles add extension methods to TableQuery
  * for operations that can be performed on tables but not on arbitrary
  * queries, e.g. getting the table DDL. */
class TableQuery[E <: AbstractTable[_]](cons: Tag => E) extends Query[E, E#TableElementType, Seq] {...}
...
sealed trait QueryBase[T] extends Rep[T]

/** An instance of Query represents a query or view, i.e. a computation of a
  * collection type (Rep[Seq[T]]). It is parameterized with both, the mixed
  * type (the type of values you see e.g. when you call map()) and the unpacked
  * type (the type of values that you get back when you run the query).
  *
  * Additional extension methods for queries containing a single column are
  * defined in [[slick.lifted.SingleColumnQueryExtensionMethods]].
  */
sealed abstract class Query[+E, U, C[_]] extends QueryBase[C[U]] { self =>...}

所有Query对象里提供的函数TableQuery类都可以调用。上面例子里coffees,suppliers实际是数据库表COFFEES,SUPPLIERS的Query实例，它们的默认字段集如：coffees.result是通过def * 定义的（除非用map或yield改变默认projection）。在slick.profile.RelationalProfile.TableQueryExtensionMethods里还有专门针对TableQuery类型的函数如schema等。

好了，来到了Query才算真正进入主题。Query可以说是Slick最核心的类型了。所有针对数据库的读写操作都是通过Query产生SQL语句发送到数据库实现的。Query是个函数式类型，即高阶类型Query[A]。A代表生成SQL语句的元素，通过转变A可以实现不同的SQL语句构建。不同功能的Query包括读取（retreive）、插入（insert）、更新（update）、删除（delete）都是通过Query变形（transformation）实现的。所有Query操作函数的款式：Query[A] => Query[B]，是典型的函数式编程方式，也是scala集合操作函数款式。我们先从数据读取Query开始，因为上面我们曾经提到过可以通过map来决定新的结果集结构（projection）：

val q1 = coffees.result
  q1.statements.head
  //res0: String = select "COF_ID", "COF_NAME", "COF_SUP", "COF_PRICE", "COF_GRADE", "COF_TOTAL" from "COFFEES"

  val q2 = coffees.map(r => (r.id, r.name)).result
  q2.statements.head
  //res1: String = select "COF_ID", "COF_NAME" from "COFFEES"

  val q3 = (for (c <- coffees) yield(c.id,c.name)).result
  q3.statements.head
  //res2: String = select "COF_ID", "COF_NAME" from "COFFEES"

因为map和flatMap的函数款式是：

map[A,B](Q[A])(A=>B]):Q[B], flatMap[A,B](Q[A])(A => Q[B]):Q[B]

所以不同的SQL语句基本上是通过Query[A] => Query[B]这种对高阶类型内嵌元素进行转变的函数式操作方式实现的。下面是一个带筛选条件的Query：

  val q = coffees.filter(_.price > 100.0).map(r => (r.id,r.name)).result
  q.statements.head
  //res3: String = select "COF_ID", "COF_NAME" from "COFFEES" where "COF_PRICE" > 100.0

  val q4 = coffees.filter(_.price > 100.0).take(4).map(_.name).result
  q4.statements.head
  //res4: String = select "COF_NAME" from "COFFEES" where "COF_PRICE" > 100.0 limit 4

  val q5 = coffees.sortBy(_.id.desc.nullsFirst).map(_.name).drop(3).result
  q5.statements.head
  //res5: String = select "COF_NAME" from "COFFEES" order by "COF_ID" desc nulls first limit -1 offset 3

再复杂一点的Query，比如说join两个表：

val q6 = for {
    (c,s) <- coffees join suppliers on (_.supID === _.id)
  } yield(c.id,c.name,s.name)
  q6.result.statements.head
  //res6: String = select x2."COF_ID", x2."COF_NAME", x3."SUP_NAME" from "COFFEES" x2, "SUPPLIERS" x3 where x2."COF_SUP" = x3."SUP_ID"
  
  val q7 = for {
    c <- coffees
    s <- suppliers.filter(c.supID === _.id)
  } yield(c.id,c.name,s.name)
  q7.result.statements.head
  //res7: String = select x2."COF_ID", x2."COF_NAME", x3."SUP_NAME" from "COFFEES" x2, "SUPPLIERS" x3 where x2."COF_SUP" = x3."SUP_ID"

还有汇总类型的Query:

  coffees.map(_.price).max.result.statements.head
  //res10: String = select max("COF_PRICE") from "COFFEES"
  coffees.map(_.total).sum.result.statements.head
  //res11: String = select sum("COF_TOTAL") from "COFFEES"
  coffees.length.result.statements.head
  //res12: String = select count(1) from "COFFEES"
  coffees.filter(_.price > 100.0).exists.result.statements.head
  //res13: String = select exists(select "COF_TOTAL", "COF_NAME", "COF_SUP", "COF_ID", "COF_PRICE", "COF_GRADE" from "COFFEES" where "COF_PRICE" > 100.0)

Query是个monad，它可以实现函数组合（functional composition）。如上所示：所有Query操作函数都是Query[A]=>Query[B]形式的。由于Query[A]里面的A类型是Rep[T]类型，是SQL语句组件类型。典型函数如flatMap的调用方式是：flatMap{a => MakeQuery(a ...)}，可以看到下一个Query的构成可能依赖a值，而a的类型是表行或列定义。所以Query的函数组合就是SQL语句的组合，最终结果是产生目标SQL语句。

Slick处理数据的方式是通过组合相应的SQL语句后发送给数据库去运算的，相关SQL语句的产生当然是通过Query来实现的：

  val qInsert = coffees += Coffee(Some(0),"American",101,56.0,Grade.Bestbuy,0)
  qInsert.statements.head
//res10: String = insert into "COFFEES" ("COF_NAME","COF_SUP","COF_PRICE","COF_GRADE","COF_TOTAL")  values (?,?,?,?,?)
  val qInsert2 = coffees.map{r => (r.name, r.supID, r.price)} += ("Columbia",101,102.0)
  qInsert2.statements.head
//res11: String = insert into "COFFEES" ("COF_NAME","COF_SUP","COF_PRICE")  values (?,?,?)
  val qInsert3 = (suppliers.map{r => (r.id,r.name)}).
    returning(suppliers.map(_.id)) += (101,"The Coffee Co.,")
  qInsert3.statements.head
//res12: String = insert into "SUPPLIERS" ("SUP_NAME")  values (?)

从qInsert3产生的SQL语句来看：jdbc返回数据后还必须由Slick进一步处理后才能返回用户要求的结果值。下面是一些其它更改数据的Query示范：

  val qDelete = coffees.filter(_.price === 0.0).delete
  qDelete.statements.head
  //res17: String = delete from "COFFEES" where "COFFEES"."COF_PRICE" = 0.0
  val qUpdate = for (c <- coffees if (c.name === "American")) yield c.price
  qUpdate.update(10.0).statements.head
  //res18: String = update "COFFEES" set "COF_PRICE" = ? where "COFFEES"."COF_NAME" = 'American'

update query必须通过for-comprehension的yield来确定更新字段。
Slick3.x最大的改进就是采用了functional I/O技术。具体做法就是引进DBIOAction类型，这是一个free monad。通过采用free monad的延迟运算模式来实现数据库操作动作的可组合性（composablility）及多线程运算（concurrency）。

DBIOAction类型款式如下：

sealed trait DBIOAction[+R, +S <: NoStream, -E <: Effect] extends Dumpable {
...}
package object dbio {
  /** Simplified type for a streaming [[DBIOAction]] without effect tracking */
  type StreamingDBIO[+R, +T] = DBIOAction[R, Streaming[T], Effect.All]

  /** Simplified type for a [[DBIOAction]] without streaming or effect tracking */
  type DBIO[+R] = DBIOAction[R, NoStream, Effect.All]
  val DBIO = DBIOAction
}

DBIO[+R]和StreamingDBIO[+R,+T]分别是固定类型参数S和E的类型别名，用它们来简化代码。所有的数据库操作函数包括result、insert、delete、update等都返回DBIOAction类型结果：

  def result: DriverAction[R, S, Effect.Read] = {...}
  def delete: DriverAction[Int, NoStream, Effect.Write] = {...}
  def update(value: T): DriverAction[Int, NoStream, Effect.Write] = {...}
  def += (value: U): DriverAction[SingleInsertResult, NoStream, Effect.Write] = {...}

上面的DriverAction是DBIOAction的子类。因为DBIOAction是个free monad，所以多个DBIOAction可以进行组合，而在过程中是不会立即产生DBIO副作用的。我们只能通过DBIOAction类型的运算器来对DBIOAction的组合进行运算才会正真进行数据库数据读写。DBIOAction运算函数款式如下：

/** Run an Action asynchronously and return the result as a Future. */
    final def run[R](a: DBIOAction[R, NoStream, Nothing]): Future[R] = runInternal(a, false)

run函数返回Future[R]，代表在异步线程运算完成后返回R类型值。一般来讲Query.result返回R类型为Seq[?]。

DBIOAction只是对数据库操作动作的描述，不是实际的读写，所以DBIOAction可以进行组合。所谓组合的意思实际上就是把几个动作连续起来。DBIOAction的函数组件除monad通用的map、flatMap、sequence等，还包括了andThen、zip等合并操作函数，andThen可以返回最后一个动作结果、zip在一个pair里返回两个动作的结果。因为DBIOAction是monad，所以for-comprehension应该是最灵活、最强大的组合方式了。我们来试试用上面Query产生的动作来进行一些组合示范：

  val initSupAction = suppliers.schema.create andThen qInsert3
  val createCoffeeAction = coffees.schema.create
  val insertCoffeeAction = qInsert zip qInsert2
  val initSupAndCoffee = for {
    _ <- initSupAction
    _ <- createCoffeeAction
    (i1,i2) <- insertCoffeeAction 
  } yield (i1,i2)

我们可以任意组合这些操作步骤，因为它们的返回结果类型都是DBIOAction[R]：一个free monad。大多数时间这些动作都是按照一定的流程顺序组合的。可能有些时候下一个动作需要依赖上一个动作产生的结果，这个时候用for-comprehension是最适合的了：

//先选出所有ESPRESSO开头的coffee名称，然后逐个删除
  val delESAction = (for {
    ns <- coffees.filter(_.name.startsWith("ESPRESSO")).map(_.name).result
    _ <- DBIO.seq(ns.map(n => coffees.filter(_.name === n).delete): _*)
  } yield ()).transactionally
  //delESAction: slick.dbio.DBIOAction[Unit,slick.dbio.NoStream,slick.dbio.Effect.Read ...

  //对一个品种价格升10%
  def raisePriceAction(i: Long, np: Double, pc: Double) =
    (for(c <- coffees if (c.id === i)) yield c.price).update(np * pc)
  //raisePriceAction: raisePriceAction[](val i: Long,val np: Double,val pc: Double) => slick.driver.H2Driver.DriverAction[Int,slick.dbio.NoStream,slick.dbio.Effect.Write]
  //对所有价格<100的coffee加价
  val updatePriceAction = (for {
    ips <- coffees.filter(_.price < 100.0).map(r => (r.id, r.price)).result
    _ <- DBIO.seq{ips.map { ip => raisePriceAction(ip._1, ip._2, 110.0)}: _* }
  } yield()).transactionally
  //updatePriceAction: slick.dbio.DBIOAction[Unit,slick.dbio.NoStream,slick.dbio.Effect.Read ...

另外，像monad的point：successful(R)可以把R升格成DBIOAction，failed(T)可以把T升格成DBIOAction[T]：

  DBIO.successful(Supplier(Some(102),"Coffee Company","",None))
  //res19: slick.dbio.DBIOAction[Supplier,slick.dbio.NoStream,slick.dbio.Effect] = SuccessAction(Supplier(Some(102),Coffee Company,,None))

  DBIO.failed(new Exception("oh my god..."))
  //res20: slick.dbio.DBIOAction[Nothing,slick.dbio.NoStream,slick.dbio.Effect] = FailureAction(java.lang.Exception: oh my god...)

DBIOAction还有比较完善的事后处理和异常处理机制：

//主要示范事后处理机制，不必理会功能的具体目的是否有任何意义
  qInsert.andFinally(qDelete)
  //res21: slick.dbio.DBIOAction[Int,slick.dbio.NoStream,slick.dbio.Effect.Write with slick.dbio.Effect.Write] = slick.dbio.SynchronousDatabaseAction$$anon$6@1d46b337

  updatePriceAction.cleanUp (
    { case Some(e) => initSupAction; DBIO.failed(new Exception("oh my..."))
      case _ => qInsert3
    }
      ,true
  )
  //res22: slick.dbio.DBIOAction[Unit,slick.dbio.NoStream,slick.dbio.Effect.Read ...

  raisePriceAction(101,10.0,110.0).asTry
  //res23: slick.dbio.DBIOAction[scala.util.Try[Int],slick.dbio.NoStream,slick.dbio.Effect.Write] = slick.dbio.SynchronousDatabaseAction$$anon$9@60304a44

从上面的这些示范例子我们认识到DBIOAction的函数组合就是数据库操作步骤组合、实际上就是程序的组合或者是功能组合：把一些简单的程序组合成功能更全面的程序，然后才运算这个组合而成的程序。DBIOAction的运算函数run的函数款式如下：

/** Run an Action asynchronously and return the result as a Future. */
    final def run[R](a: DBIOAction[R, NoStream, Nothing]): Future[R] = runInternal(a, false)

对DBIOAction进行运算后的结果是个Future类型，也是一个高阶类型，同样可以用map、flatMap、sequence、andThen等泛函组件进行函数组合。可以参考下面的这个示范：

  import slick.jdbc.meta.MTable
  import scala.concurrent.ExecutionContext.Implicits.global
  import scala.concurrent.duration.Duration
  import scala.concurrent.{Await, Future}
  import scala.util.{Success,Failure}

  val db = Database.forURL("jdbc:h2:mem:test1;DB_CLOSE_DELAY=-1", driver="org.h2.Driver")

  def recreateCoffeeTable: Future[Unit] = {
    db.run(MTable.getTables("Coffees")).flatMap {
      case tables if tables.isEmpty => db.run(coffees.schema.create).andThen {
        case Success(_) => println("coffee table created")
        case Failure(e) => println(s"failed to create! ${e.getMessage}")  
      }
      case _ => db.run((coffees.schema.drop andThen coffees.schema.create)).andThen {
        case Success(_) => println("coffee table recreated")
        case Failure(e) => println(s"failed to recreate! ${e.getMessage}")
      }   
    }
  }

好了，下面是这次讨论的示范代码：

import slick.driver.H2Driver.api._

object slick201 {
  //projection case classes 表列模版
  case class Coffee(
                     id: Option[Long]
                     ,name: String
                     ,sup_ID: Int
                     ,price: Double
                     ,grade: Grade
                     ,total: Int
                   )
  case class Supplier(
                       id: Option[Int]
                       ,name: String
                       ,address: String
                       ,website: Option[String]
                     )
  //自定义字段
  abstract class Grade(points: Int)
  object Grade {
    case object Premium extends Grade(2)
    case object Quality extends Grade(1)
    case object Bestbuy extends Grade(0)

    def fromInt(p: Int) = p match {
        case 2 => Premium
        case 1 => Quality
        case 0 => Bestbuy
    }
    def toInt(g: Grade) = g match {
      case Premium => 2
      case Quality => 1
      case Bestbuy => 0
    }
    implicit val customColumn: BaseColumnType[Grade] =
      MappedColumnType.base[Grade,Int](Grade.toInt, Grade.fromInt)
  }
  //schema 表行结构定义
  class Coffees(tag: Tag) extends Table[Coffee](tag, "COFFEES") {
    def id = column[Long]("COF_ID", O.AutoInc, O.PrimaryKey)
    def name = column[String]("COF_NAME")
    def price = column[Double]("COF_PRICE")
    def supID = column[Int]("COF_SUP")
    def grade = column[Grade]("COF_GRADE", O.Default(Grade.Bestbuy))
    def total = column[Int]("COF_TOTAL", O.Default(0))

    def * = (id.?,name,supID,price,grade,total) <> (Coffee.tupled, Coffee.unapply)

    def supplier = foreignKey("SUP_FK",supID,suppliers)(_.id,onUpdate = ForeignKeyAction.Restrict, onDelete = ForeignKeyAction.Cascade)
    def nameidx = index("NM_IX",name,unique = true)
  }
  val coffees = TableQuery[Coffees]

  class Suppliers(tag: Tag) extends Table[Supplier](tag, "SUPPLIERS") {
    def id = column[Int]("SUP_ID", O.PrimaryKey, O.AutoInc)
    def name = column[String]("SUP_NAME")
    def address = column[String]("SUP_ADDR", O.Default("-"))
    def website = column[Option[String]]("SUP_WEB")

    def * = (id.?, name, address, website) <> (Supplier.tupled, Supplier.unapply)
    def addidx = index("ADDR_IX",(name,address),unique = true)
  }
  val suppliers = TableQuery[Suppliers]

  class Bars(tag: Tag) extends Table[(Int,String)](tag,"BARS") {
    def id = column[Int]("BAR_ID",O.AutoInc,O.PrimaryKey)
    def name = column[String]("BAR_NAME")
    def * = (id, name)
  }
  val bars = TableQuery[Bars]

  Coffee.tupled
  //res2: ((Option[Long], String, Int, Double, Grade, Int)) => Coffee = <function1>
  Coffee.unapply _
  //res3: Coffee => Option[(Option[Long], String, Int, Double, Grade, Int)] = <function1>


  val q1 = coffees.result
  q1.statements.head
  //res0: String = select "COF_ID", "COF_NAME", "COF_SUP", "COF_PRICE", "COF_GRADE", "COF_TOTAL" from "COFFEES"
  
  val q2 = coffees.map(r => (r.id, r.name)).result
  q2.statements.head
  //res1: String = select "COF_ID", "COF_NAME" from "COFFEES"

  val q3 = (for (c <- coffees) yield(c.id,c.name)).result
  q3.statements.head
  //res2: String = select "COF_ID", "COF_NAME" from "COFFEES"


  val q = coffees.filter(_.price > 100.0).map(r => (r.id,r.name)).result
  q.statements.head
  //res3: String = select "COF_ID", "COF_NAME" from "COFFEES" where "COF_PRICE" > 100.0

  val q4 = coffees.filter(_.price > 100.0).take(4).map(_.name).result
  q4.statements.head
  //res4: String = select "COF_NAME" from "COFFEES" where "COF_PRICE" > 100.0 limit 4

  val q5 = coffees.sortBy(_.id.desc.nullsFirst).map(_.name).drop(3).result
  q5.statements.head
  //res5: String = select "COF_NAME" from "COFFEES" order by "COF_ID" desc nulls first limit -1 offset 3

  val q6 = for {
    (c,s) <- coffees join suppliers on (_.supID === _.id)
  } yield(c.id,c.name,s.name)
  q6.result.statements.head
  //res6: String = select x2."COF_ID", x2."COF_NAME", x3."SUP_NAME" from "COFFEES" x2, "SUPPLIERS" x3 where x2."COF_SUP" = x3."SUP_ID"

  val q7 = for {
    c <- coffees
    s <- suppliers.filter(c.supID === _.id)
  } yield(c.id,c.name,s.name)
  q7.result.statements.head
  //res7: String = select x2."COF_ID", x2."COF_NAME", x3."SUP_NAME" from "COFFEES" x2, "SUPPLIERS" x3 where x2."COF_SUP" = x3."SUP_ID"

  coffees.map(_.price).max.result.statements.head
  //res10: String = select max("COF_PRICE") from "COFFEES"
  coffees.map(_.total).sum.result.statements.head
  //res11: String = select sum("COF_TOTAL") from "COFFEES"
  coffees.length.result.statements.head
  //res12: String = select count(1) from "COFFEES"
  coffees.filter(_.price > 100.0).exists.result.statements.head
  //res13: String = select exists(select "COF_TOTAL", "COF_NAME", "COF_SUP", "COF_ID", "COF_PRICE", "COF_GRADE" from "COFFEES" where "COF_PRICE" > 100.0)
  val qInsert = coffees += Coffee(Some(0),"American",101,56.0,Grade.Bestbuy,0)
  qInsert.statements.head
  //res14: String = insert into "COFFEES" ("COF_NAME","COF_SUP","COF_PRICE","COF_GRADE","COF_TOTAL")  values (?,?,?,?,?)
  val qInsert2 = coffees.map{r => (r.name, r.supID, r.price)} += ("Columbia",101,102.0)
  qInsert2.statements.head
  //res15: String = insert into "COFFEES" ("COF_NAME","COF_SUP","COF_PRICE")  values (?,?,?)
  val qInsert3 = (suppliers.map{r => (r.id,r.name)}).
    returning(suppliers.map(_.id)) += (101,"The Coffee Co.,")
  qInsert3.statements.head
  //res16: String = insert into "SUPPLIERS" ("SUP_NAME")  values (?)

  val qDelete = coffees.filter(_.price === 0.0).delete
  qDelete.statements.head
  //res17: String = delete from "COFFEES" where "COFFEES"."COF_PRICE" = 0.0
  val qUpdate = for (c <- coffees if (c.name === "American")) yield c.price
  qUpdate.update(10.0).statements.head
  //res18: String = update "COFFEES" set "COF_PRICE" = ? where "COFFEES"."COF_NAME" = 'American'

  val initSupAction = suppliers.schema.create andThen qInsert3
  val createCoffeeAction = coffees.schema.create
  val insertCoffeeAction = qInsert zip qInsert2
  val initSupAndCoffee = for {
    _ <- initSupAction
    _ <- createCoffeeAction
    (i1,i2) <- insertCoffeeAction
  } yield (i1,i2)

  //先选出所有ESPRESSO开头的coffee名称，然后逐个删除
  val delESAction = (for {
    ns <- coffees.filter(_.name.startsWith("ESPRESSO")).map(_.name).result
    _ <- DBIO.seq(ns.map(n => coffees.filter(_.name === n).delete): _*)
  } yield ()).transactionally
  //delESAction: slick.dbio.DBIOAction[Unit,slick.dbio.NoStream,slick.dbio.Effect.Read with slick.dbio.Effect.Write with slick.dbio.Effect.Transactional] = CleanUpAction(AndThenAction(Vector(slick.driver.JdbcActionComponent$StartTransaction$@6e76c850, FlatMapAction(slick.driver.JdbcActionComponent$QueryActionExtensionMethodsImpl$$anon$1@2005bce5,<function1>,scala.concurrent.impl.ExecutionContextImpl@245036ad))),<function1>,true,slick.dbio.DBIOAction$sameThreadExecutionContext$@294c4c1d)

  //对一个品种价格升10%
  def raisePriceAction(i: Long, np: Double, pc: Double) =
    (for(c <- coffees if (c.id === i)) yield c.price).update(np * pc)
  //raisePriceAction: raisePriceAction[](val i: Long,val np: Double,val pc: Double) => slick.driver.H2Driver.DriverAction[Int,slick.dbio.NoStream,slick.dbio.Effect.Write]
  //对所有价格<100的coffee加价
  val updatePriceAction = (for {
    ips <- coffees.filter(_.price < 100.0).map(r => (r.id, r.price)).result
    _ <- DBIO.seq{ips.map { ip => raisePriceAction(ip._1, ip._2, 110.0)}: _* }
  } yield()).transactionally
  //updatePriceAction: slick.dbio.DBIOAction[Unit,slick.dbio.NoStream,slick.dbio.Effect.Read with slick.dbio.Effect.Write with slick.dbio.Effect.Transactional] = CleanUpAction(AndThenAction(Vector(slick.driver.JdbcActionComponent$StartTransaction$@6e76c850, FlatMapAction(slick.driver.JdbcActionComponent$QueryActionExtensionMethodsImpl$$anon$1@49c8a41f,<function1>,scala.concurrent.impl.ExecutionContextImpl@245036ad))),<function1>,true,slick.dbio.DBIOAction$sameThreadExecutionContext$@294c4c1d)

  DBIO.successful(Supplier(Some(102),"Coffee Company","",None))
  //res19: slick.dbio.DBIOAction[Supplier,slick.dbio.NoStream,slick.dbio.Effect] = SuccessAction(Supplier(Some(102),Coffee Company,,None))

  DBIO.failed(new Exception("oh my god..."))
  //res20: slick.dbio.DBIOAction[Nothing,slick.dbio.NoStream,slick.dbio.Effect] = FailureAction(java.lang.Exception: oh my god...)

  //示范事后处理机制，不必理会功能的具体目的
  qInsert.andFinally(qDelete)
  //res21: slick.dbio.DBIOAction[Int,slick.dbio.NoStream,slick.dbio.Effect.Write with slick.dbio.Effect.Write] = slick.dbio.SynchronousDatabaseAction$$anon$6@1d46b337

  updatePriceAction.cleanUp (
    { case Some(e) => initSupAction; DBIO.failed(new Exception("oh my..."))
      case _ => qInsert3
    }
      ,true
  )
  //res22: slick.dbio.DBIOAction[Unit,slick.dbio.NoStream,slick.dbio.Effect.Read with slick.dbio.Effect.Write with slick.dbio.Effect.Transactional with slick.dbio.Effect.Write] = CleanUpAction(CleanUpAction(AndThenAction(Vector(slick.driver.JdbcActionComponent$StartTransaction$@6e76c850, FlatMapAction(slick.driver.JdbcActionComponent$QueryActionExtensionMethodsImpl$$anon$1@1f7aad00,<function1>,scala.concurrent.impl.ExecutionContextImpl@245036ad))),<function1>,true,slick.dbio.DBIOAction$sameThreadExecutionContext$@294c4c1d),<function1>,true,scala.concurrent.impl.ExecutionContextImpl@245036ad)

  raisePriceAction(101,10.0,110.0).asTry
  //res23: slick.dbio.DBIOAction[scala.util.Try[Int],slick.dbio.NoStream,slick.dbio.Effect.Write] = slick.dbio.SynchronousDatabaseAction$$anon$9@60304a44


  import slick.jdbc.meta.MTable
  import scala.concurrent.ExecutionContext.Implicits.global
  import scala.concurrent.duration.Duration
  import scala.concurrent.{Await, Future}
  import scala.util.{Success,Failure}

  val db = Database.forURL("jdbc:h2:mem:test1;DB_CLOSE_DELAY=-1", driver="org.h2.Driver")

  def recreateCoffeeTable: Future[Unit] = {
    db.run(MTable.getTables("Coffees")).flatMap {
      case tables if tables.isEmpty => db.run(coffees.schema.create).andThen {
        case Success(_) => println("coffee table created")
        case Failure(e) => println(s"failed to create! ${e.getMessage}")
      }
      case _ => db.run((coffees.schema.drop andThen coffees.schema.create)).andThen {
        case Success(_) => println("coffee table recreated")
        case Failure(e) => println(s"failed to recreate! ${e.getMessage}")
      }
    }
  }

}