Scalaz（37）－ Free ：实践－DB Transaction free style
我一直在不断的提示大家：FP就是Monadic Programming，是一种特殊的编程风格。在我们熟悉的数据库编程领域能不能实现FP风格呢？我们先设计一些示范例子来分析一下惯用的数据库编程过程：
import scalaz._
import Scalaz._
import scala.language.higherKinds
import scala.language.implicitConversions
import com.jolbox.bonecp.BoneCP
import com.jolbox.bonecp.BoneCPConfig
import java.sql.Connection
import java.sql.ResultSet object freedbtxns {
def getTutorId(courseId: Int, conn: Connection): Int = {
val sqlString = "select TUTOR from COURSES where ID=" + courseId
conn.createStatement().executeQuery(sqlString).getInt("ID")
}
def getTutorPay(courseId: Int, conn: Connection): Double = {
val sqlString = "select PAYAMT from COURSES where ID=" + courseId
conn.createStatement().executeQuery(sqlString).getDouble("PAYAMT")
}
def getStudentFee(courseId: Int, conn: Connection): Double = {
val sqlString = "select FEEAMT from COURSES where ID=" + courseId
conn.createStatement().executeQuery(sqlString).getDouble("FEEAMT")
}
def updateTutorPay(tutorId: Int, plusAmt: Double, conn: Connection): Unit = {
val sqlString = "update TUTORS set PAYABLE = PAYABLE+"+plusAmt.toString + " where ID=" + tutorId
conn.createStatement().executeUpdate(sqlString)
}
def updateStudentFee(studentId: Int, plusAmt: Double, conn: Connection): Unit = {
val sqlString = "update STUDENTS set DUEAMT = DUEAMT+"+plusAmt.toString + " where ID=" + studentId
conn.createStatement().executeUpdate(sqlString)
}
def findEmptySeat(courseId: Int, conn: Connection): Int = {
val sqlString = "select ID from SEATS where OCCUPIED='T' AND ID=" + courseId
conn.createStatement().executeQuery(sqlString).getInt("ID")
}
def updateSeatsStatus(seatId: Int, taken: Boolean, conn: Connection): Unit = {
val sqlString = "update SEATS set OCCUPIED ='"+taken.toString.toUpperCase.head + "' where ID=" + seatId
conn.createStatement().executeUpdate(sqlString)
}
我这里模拟了一个培训学校内的一些业务。上面设计的是一些基本函数，可以分别对学员、导师、座位进行查询和更新。如果我们需要把更新工作放入事务处理（transaction）内的话我们可以这样做：
def updateStudent(studentId: Int, courseId: Int): Unit = {
val config = new BoneCPConfig()
val bonecp = new BoneCP(config)
val conn = bonecp.getConnection()
conn.setReadOnly(false)
conn.setAutoCommit(false)
conn.rollback()
try {
val fee = getStudentFee(courseId, conn)
updateStudentFee(studentId,fee, conn)
conn.commit()
} catch {
case (e:Exception) => conn.rollback()
} finally {
conn.close()
}
}
def updateStudentAndSeat(studentId: Int, courseId: Int): Unit = {
val config = new BoneCPConfig()
val bonecp = new BoneCP(config)
val conn = bonecp.getConnection()
conn.setReadOnly(false)
conn.setAutoCommit(false)
conn.rollback()
try {
val fee = getStudentFee(courseId, conn)
updateStudentFee(studentId,fee, conn)
val seatId = findEmptySeat(courseId, conn)
updateSeatsStatus(seatId, true, conn)
conn.commit()
} catch {
case (e:Exception) => conn.rollback()
} finally {
conn.close()
}
}
马上可以发现在我们对这些函数在事务处理内进行组合使用时我们必须重新对事务处理进行设置，无法实现真正意义上的函数组合。如果我们认可FP风格的话，这里起码有两项弊处：一是源代码增加了大量的铺垫（boilerplate code）,重复事务处理设置、二是每个更新函数都会产生副作用，换句话说就是这里那里都会有副作用影响，很难控制，这样就增加了程序的复杂程度，造成代码分析的困难。
我们希望达到的目标：
/*
def updateStudentAndSeat(studentId: Int): program {
// findEmptySeat
// updateStudentFee
// updateSeatStatus
} def runDBTxn(prg: program) {
//conn＝ getConnection
//try
// run(pre)
//commit
//catch
//rollback
}
runDBTxn(updateStudent)
runDBTxn(updateStudentAndSeat)
runDBTxn(updateSeatStatus)
*/
我们只在一个地方设置和运行事务处理。我们希望能把不同的program传入runDBTxn去运算。这不就是Free Monad的编程、运算关注分离模式嘛。那我们就试着用Free Monad来提供数据库事务处理支持。按上篇讨论的设计流程我们先设计ADT：
case class SqlOp[A](run: Connection => A)
模拟sql指令很简单，两种情况：query或者update。两者都可以用函数run表示：传入Connection,返回结果A，A有可能是Unit。要成为Free Monad就必须先获取SqlOp的Functor实例：
case class SqlOp[A](run: Connection => A)
implicit val sqlOpFunctor = new Functor[SqlOp] {
def map[A,B](sa: SqlOp[A])(f: A => B): SqlOp[B] =
SqlOp{ (conn: Connection) => f(sa.run(conn)) }
}
基本功能的sql操作函数及升格Free：
type Sql[A] = Free[SqlOp,A]
def getTutorId(courseId: Int): Sql[Int] =
Free.liftF(SqlOp{
(conn: Connection) => {
val sqlString = "select TUTOR from COURSES where ID=" + courseId
conn.createStatement().executeQuery(sqlString).getInt("ID")
}
}) def getTutorPay(courseId: Int): Sql[Double] =
Free.liftF(SqlOp{
(conn: Connection) => {
val sqlString = "select PAYAMT from COURSES where ID=" + courseId
conn.createStatement().executeQuery(sqlString).getDouble("PAYAMT")
}
})
def getStudentFee(courseId: Int): Sql[Double] =
Free.liftF(SqlOp{
(conn: Connection) => {
val sqlString = "select FEEAMT from COURSES where ID=" + courseId
conn.createStatement().executeQuery(sqlString).getDouble("FEEAMT")
}
})
def updateTutorPay(tutorId: Int, plusAmt: Double): Sql[Unit] =
Free.liftF(SqlOp{
(conn: Connection) => {
val sqlString = "update TUTORS set PAYABLE = PAYABLE+"+plusAmt.toString + " where ID=" + tutorId
conn.createStatement().executeUpdate(sqlString)
}
})
def updateStudentFee(studentId: Int, plusAmt: Double): Sql[Unit] =
Free.liftF(SqlOp{
(conn: Connection) => {
val sqlString = "update STUDENTS set DUEAMT = DUEAMT+"+plusAmt.toString + " where ID=" + studentId
conn.createStatement().executeUpdate(sqlString)
}
})
def findEmptySeat(courseId: Int): Sql[Int] =
Free.liftF(SqlOp{
(conn: Connection) => {
val sqlString = "select ID from SEATS where OCCUPIED='T' AND ID=" + courseId
conn.createStatement().executeQuery(sqlString).getInt("ID")
}
})
def updateSeatsStatus(seatId: Int, taken: Boolean): Sql[Unit] =
Free.liftF(SqlOp{
(conn: Connection) => {
val sqlString = "update SEATS set OCCUPIED ='"+taken.toString.toUpperCase.head + "' where ID=" + seatId
conn.createStatement().executeUpdate(sqlString)
}
})
我们现在可以用这些升格成Free的函数来建设AST示范例子：
def takeSeat(courseId: Int): Sql[Unit] = for {
emptySeat <- findEmptySeat(courseId)
_ <- updateSeatsStatus(emptySeat, true)
} yield()
def addCourse(studentId: Int, courseId: Int): Sql[Unit] = for {
fee <- getStudentFee(courseId)
pay <- getTutorPay(courseId)
tutorId <- getTutorId(courseId)
_ <- updateStudentFee(studentId, fee)
_ <- updateTutorPay(tutorId, pay)
_ <- takeSeat(courseId)
} yield()
addCourse对基本函数进行了组合，又调用了已经组合过一次的takeSeat，证明AST可以实现高度的函数组合。
下面示范实现相关的Interpreter：
def runTransactionImpl[A](conn: Connection, ast: Sql[A]): A =
ast.resume.fold ({
case x: SqlOp[Sql[A]] => runTransactionImpl(conn, x.run(conn))
},
(a: A) => a
)
我们需要一个通用的事务处理方法：
def runTransaction[A](ast: Sql[A]): Exception \/ A = {
val config = new BoneCPConfig()
val bonecp = new BoneCP(config)
val conn = bonecp.getConnection()
conn.setReadOnly(false)
conn.setAutoCommit(false)
conn.rollback()
try {
val result: A = runTransactionImpl(conn, ast)
result.right[Exception]
} catch {
case e: Exception => e.left[A]
} finally {
conn.close
}
}
这样，我们可以在一个地方使用事务处理来运算任何事先设计的AST。
我们可以用不同的方法来实现Interpreter。下面就是用Free.foldMap来运算AST的示范。由于我们需要注入Connection，所以采用了Sql to State的自然转换（natural transformation）：
type SqlState[A] = State[Connection, A]
object SqlToState extends (SqlOp ~> SqlState) {
def apply[A](sa: SqlOp[A]): SqlState[A] = sa match {
case SqlOp(f) => State {
conn => (conn,f(conn))
}
}
}
def runTransactionImplState[A](conn: Connection, ast: Sql[A]) =
ast.foldMap(SqlToState).run(conn)
下面是这个用Free来实现FP风格数据库事务处理的完整示范代码：
import scalaz._
import Scalaz._
import scala.language.higherKinds
import scala.language.implicitConversions
import com.jolbox.bonecp.BoneCP
import com.jolbox.bonecp.BoneCPConfig
import java.sql.Connection
import java.sql.ResultSet object freedbtxns { case class SqlOp[A](run: Connection => A)
implicit val sqlOpFunctor = new Functor[SqlOp] {
def map[A,B](sa: SqlOp[A])(f: A => B): SqlOp[B] =
SqlOp{ (conn: Connection) => f(sa.run(conn)) }
}
type Sql[A] = Free[SqlOp,A]
def getTutorId(courseId: Int): Sql[Int] =
Free.liftF(SqlOp{
(conn: Connection) => {
val sqlString = "select TUTOR from COURSES where ID=" + courseId
conn.createStatement().executeQuery(sqlString).getInt("ID")
}
}) def getTutorPay(courseId: Int): Sql[Double] =
Free.liftF(SqlOp{
(conn: Connection) => {
val sqlString = "select PAYAMT from COURSES where ID=" + courseId
conn.createStatement().executeQuery(sqlString).getDouble("PAYAMT")
}
})
def getStudentFee(courseId: Int): Sql[Double] =
Free.liftF(SqlOp{
(conn: Connection) => {
val sqlString = "select FEEAMT from COURSES where ID=" + courseId
conn.createStatement().executeQuery(sqlString).getDouble("FEEAMT")
}
})
def updateTutorPay(tutorId: Int, plusAmt: Double): Sql[Unit] =
Free.liftF(SqlOp{
(conn: Connection) => {
val sqlString = "update TUTORS set PAYABLE = PAYABLE+"+plusAmt.toString + " where ID=" + tutorId
conn.createStatement().executeUpdate(sqlString)
}
})
def updateStudentFee(studentId: Int, plusAmt: Double): Sql[Unit] =
Free.liftF(SqlOp{
(conn: Connection) => {
val sqlString = "update STUDENTS set DUEAMT = DUEAMT+"+plusAmt.toString + " where ID=" + studentId
conn.createStatement().executeUpdate(sqlString)
}
})
def findEmptySeat(courseId: Int): Sql[Int] =
Free.liftF(SqlOp{
(conn: Connection) => {
val sqlString = "select ID from SEATS where OCCUPIED='T' AND ID=" + courseId
conn.createStatement().executeQuery(sqlString).getInt("ID")
}
})
def updateSeatsStatus(seatId: Int, taken: Boolean): Sql[Unit] =
Free.liftF(SqlOp{
(conn: Connection) => {
val sqlString = "update SEATS set OCCUPIED ='"+taken.toString.toUpperCase.head + "' where ID=" + seatId
conn.createStatement().executeUpdate(sqlString)
}
}) def takeSeat(courseId: Int): Sql[Unit] = for {
emptySeat <- findEmptySeat(courseId)
_ <- updateSeatsStatus(emptySeat, true)
} yield()
def addCourse(studentId: Int, courseId: Int): Sql[Unit] = for {
fee <- getStudentFee(courseId)
pay <- getTutorPay(courseId)
tutorId <- getTutorId(courseId)
_ <- updateStudentFee(studentId, fee)
_ <- updateTutorPay(tutorId, pay)
_ <- takeSeat(courseId)
} yield() def runTransactionImpl[A](conn: Connection, ast: Sql[A]): A =
ast.resume.fold ({
case x: SqlOp[Sql[A]] => runTransactionImpl(conn, x.run(conn))
},
(a: A) => a
)
def runTransaction[A](ast: Sql[A]): Exception \/ A = {
val config = new BoneCPConfig()
val bonecp = new BoneCP(config)
val conn = bonecp.getConnection()
conn.setReadOnly(false)
conn.setAutoCommit(false)
conn.rollback()
try {
val result: A = runTransactionImpl(conn, ast)
result.right[Exception]
} catch {
case e: Exception => e.left[A]
} finally {
conn.close
}
}
}
Scalaz（37）－ Free ：实践－DB Transaction free style的更多相关文章
- 在 laravel 的 DB::transaction 中，为外部变量赋值
例如,我想在 laravel 的事务中,对某个外部变量赋值,然后在后续的逻辑中判断该变量的属性 $user = null; // init DB::transaction(function() use ...
- Django深入----django.db.transaction
django 的事务: transaction.py atomic---原子性 def atomic(using=None, savepoint=True): # Bare decorator: @a ...
- laravel transaction : laravel 的事务是不支持eloquent的， 要用DB::的方式
数据库事务处理# 你可以使用 transaction 方法,去执行一组数据库事务处理的操作: DB::transaction(function() { DB::table('users')->u ...
- js-新兴的API,最佳实践,离线应用于客户端存储
离线应用于客户端存储: 1.离线检测:online以及offline事件,都是在window对象上触发 navigator.online为true的时候是表示设备能够上网 2.使用一个描述文件(man ...
- [PWA] 13. New db and object store
Create a db: import idb from 'idb'; var dbPromise = idb.open('test-db', 2, function (upgradeDb) { sw ...
- File already exists: filesystem '/path/file', transaction svn常见错误解决方法
前言 多人任务基本都会用到SVN,于是提交的时候如果不先更新在提交或者操作顺序不对,会经常出现错误,其中File already exists: filesystem这个就是个常见问题,上网找了半天没 ...
- yii源码三 -- db
<AR> CActiveRecord:path:/framework/db/ar/CActiveRecord.phpoverview:is the base class for class ...
- The transaction associated with this command is not the connection's active transaction
The fix is fairly simple: if you want a Dapper query to participate in a connection, explicitly deno ...
- PhoneGap下Web SQL实践
HTML5里的Web SQL数据库,内置了SQLite数据库, 对数据库的操作使用executeSql执行增删改查 1. 创建数据库 function creatDatabase(){ db = op ...
随机推荐
- 用css来写一些简单的图形
在写网页的过程中,有时我们需要用到一些简单的图片但是手头又没有合适的,我们其实可以自己来写,下面我就简单的介绍几个例子: 1.上三角 Triangle Up #triangle-up { width: ...
- 批处理集锦——(5)使用dir查找文件
eg dir /a-d /s /b *.bat -d表示不显示.bat结尾的文件夹
- salesforce 零基础学习（三十三）通过REST方式访问外部数据以及JAVA通过rest方式访问salesforce
本篇参考Trail教程: https://developer.salesforce.com/trailhead/force_com_dev_intermediate/apex_integration_ ...
- java实现栈与队列
一.栈 栈是一种特殊的线性表.其特殊性在于限定插入和删除数据元素的操作只能在线性表的一端进行.(先进后出) 访问权限:栈限制了访问权限,只可以访问尾节点,也就是最后添加的元素 即栈顶的元素 /** * ...
- nginx上部署python web
nginx上部署python web http://uwsgi-docs.readthedocs.io/en/latest/tutorials/Django_and_nginx.html
- C++中“类”相关知识点汇总
一:类中默认的成员函数 一个空的class在C++编译器处理过后就不再为空,编译器会自动地为我们声明一些member function,如果你写 class Empty{}; 就相当于: class ...
- knockoutjs中使用mapping插件绑定数据列表
使用KO绑定数据列表示例: 1.先申请V,T,T2三个辅助方法,方便调试.声明viewModel和加载数据时的映射条件mapping 2.先使用ko.mapping.fromJS()将原来的 ...
- Unity 游戏存档 PlayerPrefs类的用法
unity3d提供了一个用于本地持久化保存与读取的类--PlayerPrefs.工作原理非常简单,以键值对的形式将数据保存在文件中,然后程序可以根据这个名称取出上次保存的数值. PlayerPr ...
- js断点调试心得
虽然网上已经有多的数不清的调试教程了,但仍然没有发现哪篇文章写的通俗易懂,索性自己尝试写写自己的一些使用习惯或者说是心得,希望对那些还不是很懂得使用断点调试的孩子有一些帮助(大神请无视~). 1.断点 ...
- JAVA 设计模式 策略模式
用途 Title 它定义了算法家族,分别封装起来,让它们之间可以互相替换,此模式让算法的变化,不会影响到使用算法的客户. 策略模式是一种行为型模式. 结构