PICE（2）：JDBCStreaming - gRPC-JDBC Service

在一个akka-cluster环境里，从数据调用的角度上，JDBC数据库与集群中其它节点是脱离的。这是因为JDBC数据库不是分布式的，不具备节点位置透明化特性。所以，JDBC数据库服务器必须通过服务方式来向外提供数据操。在这种场景里服务端是JDBC服务，其它节点，包括其它的JDBC数据库节点都是这个JDBC服务的调用客户端。因为我们已经明确选择了在akka-cluster集群环境里实施gRPC服务模式，通过akka-stream的流控制方式实现数据库操作的程序控制，所以在本次讨论里我们将示范说明gRPC-JDBC-Streaming的具体实现和使用方式。

在上次的讨论里我们已经示范了最简单的JDBC-Streaming Unary request/response模式：从客户端向JDBC-Service发送一个JDBCQuery、JDBC服务端运行JDBCQuery后向客户端返回一个数据流DataRows。jdbc.proto文件里用IDL定义的数据和服务类型如下：

message JDBCDataRow {
 string year = ;
 string state = ;
 string county = ;
 string value = ;
}

message JDBCQuery {
  string dbName = ;
  string statement = ;
  bytes parameters = ;
  google.protobuf.Int32Value fetchSize= ;
  google.protobuf.BoolValue autoCommit = ;
  google.protobuf.Int32Value queryTimeout = ;
}

service JDBCServices {
  rpc runQuery(JDBCQuery) returns (stream JDBCDataRow) {}
}

以上数据类型JDBCDataRow和JDBCQuery分别对应JDBC-Streaming工具的流元素结构和JDBCQueryContext，如下：

  val toRow = (rs: WrappedResultSet) => JDBCDataRow(
    year = rs.string("REPORTYEAR"),
    state = rs.string("STATENAME"),
    county = rs.string("COUNTYNAME"),
    value = rs.string("VALUE")
  )

   val ctx = JDBCQueryContext[JDBCDataRow](
     dbName = Symbol(q.dbName),
     statement = q.statement,
     parameters = params，
     fetchSize = q.fetchSize.getOrElse(),
     autoCommit = q.autoCommit.getOrElse(false),
     queryTimeout = q.queryTimeout
   )
   jdbcAkkaStream(ctx, toRow)

用scalaPB编译后自动产生服务端和客户端框架代码（boilerplate-code）。我们需要实现具体的JDBC服务：

class JDBCStreamingServices(implicit ec: ExecutionContextExecutor) extends JdbcGrpcAkkaStream.JDBCServices {
  val logger = Logger.getLogger(classOf[JDBCStreamingServices].getName)

  val toRow = (rs: WrappedResultSet) => JDBCDataRow(
    year = rs.string("REPORTYEAR"),
    state = rs.string("STATENAME"),
    county = rs.string("COUNTYNAME"),
    value = rs.string("VALUE")
  )
  override def runQuery: Flow[JDBCQuery, JDBCDataRow, NotUsed] = {
    logger.info("**** runQuery called on service side ***")
    Flow[JDBCQuery]
      .flatMapConcat { q =>
        //unpack JDBCQuery and construct the context
        val params: Seq[Any] = unmarshal[Seq[Any]](q.parameters)
        logger.info(s"**** query parameters: ${params} ****")
        val ctx = JDBCQueryContext[JDBCDataRow](
          dbName = Symbol(q.dbName),
          statement = q.statement,
          parameters = params,
          fetchSize = q.fetchSize.getOrElse(),
          autoCommit = q.autoCommit.getOrElse(false),
          queryTimeout = q.queryTimeout
        )
        jdbcAkkaStream(ctx, toRow)
      }
  }
}

下面是客户端调用服务示范：

  val query = JDBCQuery (
    dbName = "h2",
    statement = "select * from AQMRPT where STATENAME = ? and VALUE = ?",
    parameters = marshal(Seq("Arizona", ))
  )
  def queryRows: Source[JDBCDataRow,NotUsed] = {
    logger.info(s"running queryRows ...")
    Source
      .single(query)
      .via(stub.runQuery)
  }

这个程序的运行方式如下：

object QueryRows extends App {
  implicit val system = ActorSystem("QueryRows")
  implicit val mat = ActorMaterializer.create(system)

  val client = new JDBCStreamClient("localhost", )

  client.queryRows.runForeach(println)

  scala.io.StdIn.readLine()
  mat.shutdown()
  system.terminate()
}

那么如果从客户端发出一串的JDBCQuery又如何呢？这也是所谓的BiDi-Streaming模式，在jdbc.proto的服务描述如下：

service JDBCServices {
  rpc runQuery(JDBCQuery) returns (stream JDBCDataRow) {}
  rpc batQuery(stream JDBCQuery) returns (stream JDBCDataRow) {}
}

我们看到batQuery的入参是一个stream。自动产生的服务函数batQuery款式是这样的：

  override def runQuery: Flow[JDBCQuery, JDBCDataRow, NotUsed] = { ... }
  override def batQuery: Flow[JDBCQuery, JDBCDataRow, NotUsed] = runQuery

runQuery和batQuery的函数款式是一样的。这就说明服务端提供的服务模式是一样的。在我们这个例子里它们都是对每个收到的JDBCQuery发还相关的数据流。实际上这两项服务的区别在客户方。下面是scalaPB产生的源代码：

   override def runQuery: Flow[grpc.jdbc.services.JDBCQuery, grpc.jdbc.services.JDBCDataRow, NotUsed] =
      Flow.fromGraph(
        new GrpcGraphStage[grpc.jdbc.services.JDBCQuery, grpc.jdbc.services.JDBCDataRow]({ outputObserver =>
          new StreamObserver[grpc.jdbc.services.JDBCQuery] {
            override def onError(t: Throwable): Unit = ()
            override def onCompleted(): Unit = ()
            override def onNext(request: grpc.jdbc.services.JDBCQuery): Unit =
              ClientCalls.asyncServerStreamingCall(
                channel.newCall(METHOD_RUN_QUERY, options),
                request,
                outputObserver
              )
          }
        })
      )
 ...
      override def batQuery: Flow[grpc.jdbc.services.JDBCQuery, grpc.jdbc.services.JDBCDataRow, NotUsed] =
        Flow.fromGraph(new GrpcGraphStage[grpc.jdbc.services.JDBCQuery, grpc.jdbc.services.JDBCDataRow](outputObserver =>
          ClientCalls.asyncBidiStreamingCall(
            channel.newCall(METHOD_BAT_QUERY, options),
            outputObserver
          )
        ))

所以在客户端我们调用batQuery：

  def batQueryRows: Source[JDBCDataRow,NotUsed] = {
    logger.info(s"running batQueryRows ...")
    Source
      .fromIterator(() => List(query,query2,query3).toIterator)
      .via(stub.batQuery)
  }

JDBC操作除Query之外还应该具备数据更新部分，包括Schema DDL和database-updates。JDBC-update是通过JDBCContext来传递更新要求的：

case class JDBCContext(
                        dbName: Symbol,
                        statements: Seq[String] = Nil,
                        parameters: Seq[Seq[Any]] = Nil,
                        fetchSize: Int = ,
                        queryTimeout: Option[Int] = None,
                        queryTags: Seq[String] = Nil,
                        sqlType: JDBCContext.SQLTYPE = JDBCContext.SQL_UPDATE,
                        batch: Boolean = false,
                        returnGeneratedKey: Seq[Option[Any]] = Nil,
                        // no return: None, return by index: Some(1), by name: Some("id")
                        preAction: Option[PreparedStatement => Unit] = None,
                        postAction: Option[PreparedStatement => Unit] = None) {... }

这个class对应的protobuf message定义如下：

message JDBCResult {
  bytes result = ;
}

message JDBCUpdate {
  string dbName = ;
  repeated string statements = ;
  bytes parameters = ;
  google.protobuf.Int32Value fetchSize= ;
  google.protobuf.Int32Value queryTimeout = ;
  int32 sqlType = ;
  google.protobuf.Int32Value batch = ;
  bytes returnGeneratedKey = ;
}

service JDBCServices {
  rpc runQuery(JDBCQuery) returns (stream JDBCDataRow) {}
  rpc batQuery(stream JDBCQuery) returns (stream JDBCDataRow) {}
  rpc runDDL(JDBCUpdate) returns (JDBCResult) {}
}

服务函数runDDL返回消息类型JDBCResult： 包嵌一个Seq[Any]类型的返回值。下面是JDBCContext的protobuf message打包、还原使用方法示范，在服务端把JDBCUpdate拆解构建JDBCContext后调用jdbcExecuteDDL：

 override def runDDL: Flow[JDBCUpdate, JDBCResult, NotUsed] = {
    logger.info("**** runDDL called on service side ***")
    Flow[JDBCUpdate]
      .flatMapConcat { context =>
        //unpack JDBCUpdate and construct the context

        val ctx = JDBCContext(
          dbName = Symbol(context.dbName),
          statements = context.statements,
          sqlType = JDBCContext.SQL_EXEDDL,
          queryTimeout = context.queryTimeout
        )

        logger.info(s"**** JDBCContext => ${ctx} ***")

        Source
          .fromFuture(jdbcExecuteDDL(ctx))
          .map { r => JDBCResult(marshal(r)) }

      }
  }

jdbcExecuteDDL返回Future[String]，如下：

  def jdbcExecuteDDL(ctx: JDBCContext)(implicit ec: ExecutionContextExecutor): Future[String] = {
    if (ctx.sqlType != SQL_EXEDDL) {
      Future.failed(new IllegalStateException("JDBCContex setting error: sqlType must be 'SQL_EXEDDL'!"))
    }
    else {
      Future {
        NamedDB(ctx.dbName) localTx { implicit session =>
          ctx.statements.foreach { stm =>
            val ddl = new SQLExecution(statement = stm, parameters = Nil)(
              before = WrappedResultSet => {})(
              after = WrappedResultSet => {})

            ddl.apply()
          }
          "SQL_EXEDDL executed succesfully."
        }
      }
    }
  }

我们可以用Source.fromFuture(jdbcExecuteDDL(cox))来构建一个akka-stream Source。 在客户端构建一个JDBCUpdate结构传给服务端进行运算：

  val dropSQL: String ="""
      drop table members
    """

  val createSQL: String ="""
    create table members (
      id serial not null primary key,
      name varchar() not null,
      description varchar(),
      birthday date,
      created_at timestamp not null,
      picture blob
    )"""
  val ctx = JDBCUpdate (
    dbName = "h2",
    sqlType = JDBCContext.SQL_EXEDDL,
    statements = Seq(dropSQL,createSQL)
  )

  def createTbl: Source[JDBCResult,NotUsed] = {
    logger.info(s"running createTbl ...")
    Source
      .single(ctx)
      .via(stub.runDDL)
  }

注意：statements = Seq(dropSQL,createSQL)包含了两个独立的SQL运算。

下面我们示范一下从客户端传送一个数据流（stream MemberRow），由服务端插入数据库操作。DDL数据类型和服务函数定义如下：

message JDBCDate {
  int32 yyyy = ;
  int32 mm   = ;
  int32 dd   = ;
}

message JDBCTime {
  int32 hh   = ;
  int32 mm   = ;
  int32 ss   = ;
  int32 nnn  = ;
}

message JDBCDateTime {
   JDBCDate date = ;
   JDBCTime time = ;
}
message MemberRow {
  string name = ;
  JDBCDate birthday = ;
  string description = ;
  JDBCDateTime created_at = ;
  bytes picture = ;
}

service JDBCServices {
  rpc runQuery(JDBCQuery) returns (stream JDBCDataRow) {}
  rpc batQuery(stream JDBCQuery) returns (stream JDBCDataRow) {}
  rpc runDDL(JDBCUpdate) returns (JDBCResult) {}
  rpc insertRows(stream MemberRow) returns(JDBCResult) {}
}

insertRows服务函数的实现如下：

 override def insertRows: Flow[MemberRow, JDBCResult, NotUsed] = {
    logger.info("**** insertRows called on service side ***")
    val insertSQL = """
      insert into members(
        name,
        birthday,
        description,
        created_at
      ) values (?, ?, ?, ?)
    """
    Flow[MemberRow]
      .flatMapConcat { row =>
        val ctx = JDBCContext('h2)
          .setUpdateCommand(true,insertSQL,
             row.name,
             jdbcSetDate(row.birthday.get.yyyy,row.birthday.get.mm,row.birthday.get.dd),
             row.description,
             jdbcSetNow
          )

        logger.info(s"**** JDBCContext => ${ctx} ***")

        Source
          .fromFuture(jdbcTxUpdates[Vector](ctx))
          .map { r => JDBCResult(marshal(r)) }
      }
  }

同样，这个jdbcTxUpdates返回结果是Future类型。具体实现在附件的JDBCEngine.scala中。

客户端构建一个MemberRow流，然后经过stub.insertRows发送给服务端：

  val p1 = MemberRow( "Peter Chan",Some(JDBCDate(,,)),"new member1",None,_root_.com.google.protobuf.ByteString.EMPTY)
  val p2 = MemberRow( "Alanda Wong",Some(JDBCDate(,,)),"new member2",None,_root_.com.google.protobuf.ByteString.EMPTY)
  val p3 = MemberRow( "Kate Zhang",Some(JDBCDate(,,)),"new member3",None,_root_.com.google.protobuf.ByteString.EMPTY)
  val p4 = MemberRow( "Tiger Chan",Some(JDBCDate(,,)),"new member4",None,_root_.com.google.protobuf.ByteString.EMPTY)

  def insertRows: Source[JDBCResult,NotUsed] = {
    logger.info(s"running insertRows ...")
    Source
      .fromIterator(() => List(p1,p2,p3,p4).toIterator)
      .via(stub.insertRows)
  }

最后，我们再示范jdbcBatchUpdate函数的使用。我们从服务端读取MemberRow再传回服务端进行更新操作。DDL如下：

message MemberRow {
  string name = ;
  JDBCDate birthday = ;
  string description = ;
  JDBCDateTime created_at = ;
  bytes picture = ;
}

service JDBCServices {
  rpc runQuery(JDBCQuery) returns (stream JDBCDataRow) {}
  rpc batQuery(stream JDBCQuery) returns (stream JDBCDataRow) {}
  rpc runDDL(JDBCUpdate) returns (JDBCResult) {}
  rpc insertRows(stream MemberRow) returns(JDBCResult) {}
  rpc updateRows(stream MemberRow) returns(JDBCResult) {}
  rpc getMembers(JDBCQuery) returns (stream MemberRow) {}
}

服务端函数定义如下：

 val toMemberRow = (rs: WrappedResultSet) => MemberRow(
    name = rs.string("name"),
    description = rs.string("description"),
    birthday = None,
    createdAt = None,
    picture = _root_.com.google.protobuf.ByteString.EMPTY
  )
  override def getMembers: Flow[JDBCQuery, MemberRow, NotUsed] = {
    logger.info("**** getMembers called on service side ***")
    Flow[JDBCQuery]
      .flatMapConcat { q =>
        //unpack JDBCQuery and construct the context
        var params: Seq[Any] =  Nil
        if (q.parameters != _root_.com.google.protobuf.ByteString.EMPTY)
           params = unmarshal[Seq[Any]](q.parameters)
        logger.info(s"**** query parameters: ${params} ****")
        val ctx = JDBCQueryContext[MemberRow](
          dbName = Symbol(q.dbName),
          statement = q.statement,
          parameters = params,
          fetchSize = q.fetchSize.getOrElse(),
          autoCommit = q.autoCommit.getOrElse(false),
          queryTimeout = q.queryTimeout
        )
        jdbcAkkaStream(ctx, toMemberRow)
      }
  }
override def updateRows: Flow[MemberRow, JDBCResult, NotUsed] = {
    logger.info("**** updateRows called on service side ***")
    val updateSQL = "update members set description = ?, created_at = ? where name = ?"

    Flow[MemberRow]
      .flatMapConcat { row =>
        val ctx = JDBCContext('h2)
            .setBatchCommand(updateSQL)
            .appendBatchParameters(
              row.name + " updated.",
              jdbcSetNow,
              row.name
            ).setBatchReturnGeneratedKeyOption(true)

        logger.info(s"**** JDBCContext => ${ctx} ***")

        Source
          .fromFuture(jdbcBatchUpdate[Vector](ctx))
          .map { r => JDBCResult(marshal(r)) }
      }
  }

jdbcBatchUpdate函数的源代码在附件JDBCEngine.scala中。客户端代码如下：

  val queryMember = JDBCQuery (
    dbName = "h2",
    statement = "select * from members"
  )
  def updateRows: Source[JDBCResult,NotUsed] = {
    logger.info(s"running updateRows ...")
    Source
      .single(queryMember)
      .via(stub.getMembers)
      .via(stub.updateRows)
  }

下面的例子示范了如何利用JDBCActionStream来批量处理数据。服务端的源代码如下：

  val params: JDBCDataRow => Seq[Any] = row => {
    Seq((row.value.toInt * ), row.state, row.county, row.year) }
  val sql = "update AQMRPT set total = ? where statename = ? and countyname = ? and reportyear = ?"

  val jdbcActionStream = JDBCActionStream('h2,sql ,params)
    .setParallelism().setProcessOrder(false)
  val jdbcActionFlow = jdbcActionStream.performOnRow

  override def updateBat: Flow[JDBCDataRow, JDBCDataRow, NotUsed] = {
    logger.info("**** updateBat called on service side ***")
    Flow[JDBCDataRow]
         .via(jdbcActionFlow)
  }

jdbcActionFlow是一个Flow[R,R,_]，所以我们直接用via把它连接到上一个Flow。下面是JDBCActionStream的定义代码：

  case class JDBCActionStream[R](dbName: Symbol, parallelism: Int = , processInOrder: Boolean = true,
                                 statement: String, prepareParams: R => Seq[Any]) {
    jas =>
    def setDBName(db: Symbol): JDBCActionStream[R] = jas.copy(dbName = db)
    def setParallelism(parLevel: Int): JDBCActionStream[R] = jas.copy(parallelism = parLevel)
    def setProcessOrder(ordered: Boolean): JDBCActionStream[R] = jas.copy(processInOrder = ordered)

    private def perform(r: R)(implicit ec: ExecutionContextExecutor) = {
      import scala.concurrent._
      val params = prepareParams(r)
      Future {
        NamedDB(dbName) autoCommit { session =>
          session.execute(statement, params: _*)
        }
        r
      }
    }
    def performOnRow(implicit ec: ExecutionContextExecutor): Flow[R, R, NotUsed] =
      if (processInOrder)
        Flow[R].mapAsync(parallelism)(perform)
      else
        Flow[R].mapAsyncUnordered(parallelism)(perform)

  }
  object JDBCActionStream {
    def apply[R](_dbName: Symbol, _statement: String, params: R => Seq[Any]): JDBCActionStream[R] =
      new JDBCActionStream[R](dbName = _dbName, statement=_statement, prepareParams = params)
  }

函数performOnRow是个passthrough处理过程，使用了mapAsync来支持多线程运算。客户端调用方式如下：

  def updateBatches: Source[JDBCDataRow,NotUsed] = {
    logger.info(s"running updateBatches ...")
      Source
        .fromIterator(() => List(query,query2,query3).toIterator)
        .via(stub.batQuery)
        .via(stub.updateBat)

  }

下面是本次示范的完整源代码：

jdbc.proto

syntax = "proto3";

import "google/protobuf/wrappers.proto";
import "google/protobuf/any.proto";
import "scalapb/scalapb.proto";

package grpc.jdbc.services;

option (scalapb.options) = {
  // use a custom Scala package name
  // package_name: "io.ontherocks.introgrpc.demo"

  // don't append file name to package
  flat_package: true

  // generate one Scala file for all messages (services still get their own file)
  single_file: true

  // add imports to generated file
  // useful when extending traits or using custom types
  // import: "io.ontherocks.hellogrpc.RockingMessage"

  // code to put at the top of generated file
  // works only with `single_file: true`
  //preamble: "sealed trait SomeSealedTrait"
};

/*
 * Demoes various customization options provided by ScalaPBs.
 */

message JDBCDataRow {
 string year = ;
 string state = ;
 string county = ;
 string value = ;
}

message JDBCQuery {
  string dbName = ;
  string statement = ;
  bytes parameters = ;
  google.protobuf.Int32Value fetchSize= ;
  google.protobuf.BoolValue autoCommit = ;
  google.protobuf.Int32Value queryTimeout = ;
}

message JDBCResult {
  bytes result = ;
}

message JDBCUpdate {
  string dbName = ;
  repeated string statements = ;
  bytes parameters = ;
  google.protobuf.Int32Value fetchSize= ;
  google.protobuf.Int32Value queryTimeout = ;
  int32 sqlType = ;
  google.protobuf.Int32Value batch = ;
  bytes returnGeneratedKey = ;
}

message JDBCDate {
  int32 yyyy = ;
  int32 mm   = ;
  int32 dd   = ;
}

message JDBCTime {
  int32 hh   = ;
  int32 mm   = ;
  int32 ss   = ;
  int32 nnn  = ;
}

message JDBCDateTime {
   JDBCDate date = ;
   JDBCTime time = ;
}

message MemberRow {
  string name = ;
  JDBCDate birthday = ;
  string description = ;
  JDBCDateTime created_at = ;
  bytes picture = ;
}

service JDBCServices {
  rpc runQuery(JDBCQuery) returns (stream JDBCDataRow) {}
  rpc batQuery(stream JDBCQuery) returns (stream JDBCDataRow) {}
  rpc runDDL(JDBCUpdate) returns (JDBCResult) {}
  rpc insertRows(stream MemberRow) returns(JDBCResult) {}
  rpc updateRows(stream MemberRow) returns(JDBCResult) {}
  rpc getMembers(JDBCQuery) returns (stream MemberRow) {}
}

JDBCEngine.scala

package sdp.jdbc.engine
import java.sql.PreparedStatement
import scala.collection.generic.CanBuildFrom
import akka.stream.scaladsl._
import scalikejdbc._
import scalikejdbc.streams._
import akka.NotUsed
import akka.stream._
import java.time._
import scala.concurrent.duration._
import scala.concurrent._
import sdp.jdbc.FileStreaming._

import scalikejdbc.TxBoundary.Try._

import scala.concurrent.ExecutionContextExecutor
import java.io.InputStream

object JDBCContext {
  type SQLTYPE = Int
  val SQL_EXEDDL=
  val SQL_UPDATE =
  val RETURN_GENERATED_KEYVALUE = true
  val RETURN_UPDATED_COUNT = false

}

case class JDBCQueryContext[M](
                                dbName: Symbol,
                                statement: String,
                                parameters: Seq[Any] = Nil,
                                fetchSize: Int = ,
                                autoCommit: Boolean = false,
                                queryTimeout: Option[Int] = None)

case class JDBCContext(
                        dbName: Symbol,
                        statements: Seq[String] = Nil,
                        parameters: Seq[Seq[Any]] = Nil,
                        fetchSize: Int = ,
                        queryTimeout: Option[Int] = None,
                        queryTags: Seq[String] = Nil,
                        sqlType: JDBCContext.SQLTYPE = JDBCContext.SQL_UPDATE,
                        batch: Boolean = false,
                        returnGeneratedKey: Seq[Option[Any]] = Nil,
                        // no return: None, return by index: Some(1), by name: Some("id")
                        preAction: Option[PreparedStatement => Unit] = None,
                        postAction: Option[PreparedStatement => Unit] = None) {

  ctx =>

  //helper functions

  def appendTag(tag: String): JDBCContext = ctx.copy(queryTags = ctx.queryTags :+ tag)

  def appendTags(tags: Seq[String]): JDBCContext = ctx.copy(queryTags = ctx.queryTags ++ tags)

  def setFetchSize(size: Int): JDBCContext = ctx.copy(fetchSize = size)

  def setQueryTimeout(time: Option[Int]): JDBCContext = ctx.copy(queryTimeout = time)

  def setPreAction(action: Option[PreparedStatement => Unit]): JDBCContext = {
    if (ctx.sqlType == JDBCContext.SQL_UPDATE &&
      !ctx.batch && ctx.statements.size == )
      ctx.copy(preAction = action)
    else
      throw new IllegalStateException("JDBCContex setting error: preAction not supported!")
  }

  def setPostAction(action: Option[PreparedStatement => Unit]): JDBCContext = {
    if (ctx.sqlType == JDBCContext.SQL_UPDATE &&
      !ctx.batch && ctx.statements.size == )
      ctx.copy(postAction = action)
    else
      throw new IllegalStateException("JDBCContex setting error: preAction not supported!")
  }

  def appendDDLCommand(_statement: String, _parameters: Any*): JDBCContext = {
    if (ctx.sqlType == JDBCContext.SQL_EXEDDL) {
      ctx.copy(
        statements = ctx.statements ++ Seq(_statement),
        parameters = ctx.parameters ++ Seq(Seq(_parameters))
      )
    } else
      throw new IllegalStateException("JDBCContex setting error: option not supported!")
  }

  def appendUpdateCommand(_returnGeneratedKey: Boolean, _statement: String,_parameters: Any*): JDBCContext = {
    if (ctx.sqlType == JDBCContext.SQL_UPDATE && !ctx.batch) {
      ctx.copy(
        statements = ctx.statements ++ Seq(_statement),
        parameters = ctx.parameters ++ Seq(_parameters),
        returnGeneratedKey = ctx.returnGeneratedKey ++ (if (_returnGeneratedKey) Seq(Some()) else Seq(None))
      )
    } else
      throw new IllegalStateException("JDBCContex setting error: option not supported!")
  }

  def appendBatchParameters(_parameters: Any*): JDBCContext = {
    if (ctx.sqlType != JDBCContext.SQL_UPDATE || !ctx.batch)
      throw new IllegalStateException("JDBCContex setting error: batch parameters only supported for SQL_UPDATE and batch = true!")

    var matchParams = true
    if (ctx.parameters != Nil)
      if (ctx.parameters.head.size != _parameters.size)
        matchParams = false
    if (matchParams) {
      ctx.copy(
        parameters = ctx.parameters ++ Seq(_parameters)
      )
    } else
      throw new IllegalStateException("JDBCContex setting error: batch command parameters not match!")
  }

  def setBatchReturnGeneratedKeyOption(returnKey: Boolean): JDBCContext = {
    if (ctx.sqlType != JDBCContext.SQL_UPDATE || !ctx.batch)
      throw new IllegalStateException("JDBCContex setting error: only supported in batch update commands!")
    ctx.copy(
      returnGeneratedKey = if (returnKey) Seq(Some()) else Nil
    )
  }

  def setDDLCommand(_statement: String, _parameters: Any*): JDBCContext = {
    ctx.copy(
      statements = Seq(_statement),
      parameters = Seq(_parameters),
      sqlType = JDBCContext.SQL_EXEDDL,
      batch = false
    )
  }

  def setUpdateCommand(_returnGeneratedKey: Boolean, _statement: String,_parameters: Any*): JDBCContext = {
    ctx.copy(
      statements = Seq(_statement),
      parameters = Seq(_parameters),
      returnGeneratedKey = if (_returnGeneratedKey) Seq(Some()) else Seq(None),
      sqlType = JDBCContext.SQL_UPDATE,
      batch = false
    )
  }
  def setBatchCommand(_statement: String): JDBCContext = {
    ctx.copy (
      statements = Seq(_statement),
      sqlType = JDBCContext.SQL_UPDATE,
      batch = true
    )
  }

}

object JDBCEngine {

  import JDBCContext._

  type JDBCDate = LocalDate
  type JDBCDateTime = LocalDateTime
  type JDBCTime = LocalTime

  def jdbcSetDate(yyyy: Int, mm: Int, dd: Int) = LocalDate.of(yyyy,mm,dd)
  def jdbcSetTime(hh: Int, mm: Int, ss: Int, nn: Int) = LocalTime.of(hh,mm,ss,nn)
  def jdbcSetDateTime(date: JDBCDate, time: JDBCTime) =  LocalDateTime.of(date,time)
  def jdbcSetNow = LocalDateTime.now()

  type JDBCBlob = InputStream

  def fileToJDBCBlob(fileName: String, timeOut: FiniteDuration =  seconds)(
    implicit mat: Materializer) = FileToInputStream(fileName,timeOut)

  def jdbcBlobToFile(blob: JDBCBlob, fileName: String)(
    implicit mat: Materializer) =  InputStreamToFile(blob,fileName)

  private def noExtractor(message: String): WrappedResultSet => Nothing = { (rs: WrappedResultSet) =>
    throw new IllegalStateException(message)
  }

  def jdbcAkkaStream[A](ctx: JDBCQueryContext[A],extractor: WrappedResultSet => A)
                       (implicit ec: ExecutionContextExecutor): Source[A,NotUsed] = {
      val publisher: DatabasePublisher[A] = NamedDB(ctx.dbName) readOnlyStream {
      val rawSql = new SQLToCollectionImpl[A, NoExtractor](ctx.statement, ctx.parameters)(noExtractor(""))
      ctx.queryTimeout.foreach(rawSql.queryTimeout(_))
      val sql: SQL[A, HasExtractor] = rawSql.map(extractor)

      sql.iterator
        .withDBSessionForceAdjuster(session => {
          session.connection.setAutoCommit(ctx.autoCommit)
          session.fetchSize(ctx.fetchSize)
        })
    }
    Source.fromPublisher[A](publisher)
  }

  def jdbcQueryResult[C[_] <: TraversableOnce[_], A](ctx: JDBCQueryContext[A],
                                                     extractor: WrappedResultSet => A)(
                                                      implicit cbf: CanBuildFrom[Nothing, A, C[A]]): C[A] = {

    val rawSql = new SQLToCollectionImpl[A, NoExtractor](ctx.statement, ctx.parameters)(noExtractor(""))
    ctx.queryTimeout.foreach(rawSql.queryTimeout(_))
    rawSql.fetchSize(ctx.fetchSize)
    implicit val session = NamedAutoSession(ctx.dbName)
    val sql: SQL[A, HasExtractor] = rawSql.map(extractor)
    sql.collection.apply[C]()

  }

  def jdbcExecuteDDL(ctx: JDBCContext)(implicit ec: ExecutionContextExecutor): Future[String] = {
    if (ctx.sqlType != SQL_EXEDDL) {
      Future.failed(new IllegalStateException("JDBCContex setting error: sqlType must be 'SQL_EXEDDL'!"))
    }
    else {
      Future {
        NamedDB(ctx.dbName) localTx { implicit session =>
          ctx.statements.foreach { stm =>
            val ddl = new SQLExecution(statement = stm, parameters = Nil)(
              before = WrappedResultSet => {})(
              after = WrappedResultSet => {})

            ddl.apply()
          }
          "SQL_EXEDDL executed succesfully."
        }
      }
    }
  }

  def jdbcBatchUpdate[C[_] <: TraversableOnce[_]](ctx: JDBCContext)(
    implicit ec: ExecutionContextExecutor,
             cbf: CanBuildFrom[Nothing, Long, C[Long]]): Future[C[Long]] = {
    if (ctx.statements == Nil)
     Future.failed ( new IllegalStateException("JDBCContex setting error: statements empty!"))
    if (ctx.sqlType != SQL_UPDATE) {
      Future.failed(new IllegalStateException("JDBCContex setting error: sqlType must be 'SQL_UPDATE'!"))
    }
    else {
      if (ctx.batch) {
        if (noReturnKey(ctx)) {
          val usql = SQL(ctx.statements.head)
            .tags(ctx.queryTags: _*)
            .batch(ctx.parameters: _*)
          Future {
            NamedDB(ctx.dbName) localTx { implicit session =>
              ctx.queryTimeout.foreach(session.queryTimeout(_))
              usql.apply[Seq]()
              Seq.empty[Long].to[C]
            }
          }
        } else {
          val usql = new SQLBatchWithGeneratedKey(ctx.statements.head, ctx.parameters, ctx.queryTags)(None)
          Future {
            NamedDB(ctx.dbName) localTx { implicit session =>
              ctx.queryTimeout.foreach(session.queryTimeout(_))
              usql.apply[C]()
            }
          }
        }

      } else {
        Future.failed(new IllegalStateException("JDBCContex setting error: must set batch = true !"))
      }
    }
  }
  private def singleTxUpdateWithReturnKey[C[_] <: TraversableOnce[_]](ctx: JDBCContext)(
          implicit ec: ExecutionContextExecutor,
                    cbf: CanBuildFrom[Nothing, Long, C[Long]]): Future[C[Long]] = {
    val Some(key) :: xs = ctx.returnGeneratedKey
    val params: Seq[Any] = ctx.parameters match {
      case Nil => Nil
      case p@_ => p.head
    }
    val usql = new SQLUpdateWithGeneratedKey(ctx.statements.head, params, ctx.queryTags)(key)
    Future {
      NamedDB(ctx.dbName) localTx { implicit session =>
        session.fetchSize(ctx.fetchSize)
        ctx.queryTimeout.foreach(session.queryTimeout(_))
        val result = usql.apply()
        Seq(result).to[C]
      }
    }
  }

  private def singleTxUpdateNoReturnKey[C[_] <: TraversableOnce[_]](ctx: JDBCContext)(
          implicit ec: ExecutionContextExecutor,
                   cbf: CanBuildFrom[Nothing, Long, C[Long]]): Future[C[Long]] = {
    val params: Seq[Any] = ctx.parameters match {
      case Nil => Nil
      case p@_ => p.head
    }
    val before = ctx.preAction match {
      case None => pstm: PreparedStatement => {}
      case Some(f) => f
    }
    val after = ctx.postAction match {
      case None => pstm: PreparedStatement => {}
      case Some(f) => f
    }
    val usql = new SQLUpdate(ctx.statements.head,params,ctx.queryTags)(before)(after)
    Future {
      NamedDB(ctx.dbName) localTx {implicit session =>
        session.fetchSize(ctx.fetchSize)
        ctx.queryTimeout.foreach(session.queryTimeout(_))
        val result = usql.apply()
        Seq(result.toLong).to[C]
      }
    }

  }

  private def singleTxUpdate[C[_] <: TraversableOnce[_]](ctx: JDBCContext)(
    implicit ec: ExecutionContextExecutor,
             cbf: CanBuildFrom[Nothing, Long, C[Long]]): Future[C[Long]] = {
    if (noReturnKey(ctx))
      singleTxUpdateNoReturnKey(ctx)
    else
      singleTxUpdateWithReturnKey(ctx)
  }

  private def noReturnKey(ctx: JDBCContext): Boolean = {
    if (ctx.returnGeneratedKey != Nil) {
      val k :: xs = ctx.returnGeneratedKey
      k match {
        case None => true
        case Some(k) => false
      }
    } else true
  }

  def noActon: PreparedStatement=>Unit = pstm => {}

  def multiTxUpdates[C[_] <: TraversableOnce[_]](ctx: JDBCContext)(
    implicit ec: ExecutionContextExecutor,
             cbf: CanBuildFrom[Nothing, Long, C[Long]]): Future[C[Long]] = {
    Future {
      NamedDB(ctx.dbName) localTx { implicit session =>
        session.fetchSize(ctx.fetchSize)
        ctx.queryTimeout.foreach(session.queryTimeout(_))
        val keys: Seq[Option[Any]] = ctx.returnGeneratedKey match {
          case Nil => Seq.fill(ctx.statements.size)(None)
          case k@_ => k
        }
        val sqlcmd = ctx.statements zip ctx.parameters zip keys
        val results = sqlcmd.map { case ((stm, param), key) =>
          key match {
            case None =>
              new SQLUpdate(stm, param, Nil)(noActon)(noActon).apply().toLong
            case Some(k) =>
              new SQLUpdateWithGeneratedKey(stm, param, Nil)(k).apply().toLong
          }
        }
        results.to[C]
      }
    }
  }

  def jdbcTxUpdates[C[_] <: TraversableOnce[_]](ctx: JDBCContext)(
    implicit ec: ExecutionContextExecutor,
             cbf: CanBuildFrom[Nothing, Long, C[Long]]): Future[C[Long]] = {
    if (ctx.statements == Nil)
      Future.failed( new IllegalStateException("JDBCContex setting error: statements empty!"))
    if (ctx.sqlType != SQL_UPDATE) {
      Future.failed(new IllegalStateException("JDBCContex setting error: sqlType must be 'SQL_UPDATE'!"))
    }
    else {
      if (!ctx.batch) {
        if (ctx.statements.size == )
          singleTxUpdate(ctx)
        else
          multiTxUpdates(ctx)
      } else
        Future.failed(new IllegalStateException("JDBCContex setting error: must set batch = false !"))

    }
  }

  case class JDBCActionStream[R](dbName: Symbol, parallelism: Int = , processInOrder: Boolean = true,
                                 statement: String, prepareParams: R => Seq[Any]) {
    jas =>
    def setDBName(db: Symbol): JDBCActionStream[R] = jas.copy(dbName = db)
    def setParallelism(parLevel: Int): JDBCActionStream[R] = jas.copy(parallelism = parLevel)
    def setProcessOrder(ordered: Boolean): JDBCActionStream[R] = jas.copy(processInOrder = ordered)

    private def perform(r: R)(implicit ec: ExecutionContextExecutor) = {
      import scala.concurrent._
      val params = prepareParams(r)
      Future {
        NamedDB(dbName) autoCommit { session =>
          session.execute(statement, params: _*)
        }
        r
      }
     // Future.successful(r)
    }
    def performOnRow(implicit ec: ExecutionContextExecutor): Flow[R, R, NotUsed] =
      if (processInOrder)
        Flow[R].mapAsync(parallelism)(perform)
      else
        Flow[R].mapAsyncUnordered(parallelism)(perform)

  }
  object JDBCActionStream {
    def apply[R](_dbName: Symbol, _statement: String, params: R => Seq[Any]): JDBCActionStream[R] =
      new JDBCActionStream[R](dbName = _dbName, statement=_statement, prepareParams = params)
  }
}

JDBCService.scala

package demo.grpc.jdbc.services

import akka.NotUsed
import akka.stream.scaladsl.{Source,Flow}
import grpc.jdbc.services._
import java.util.logging.Logger
import protobuf.bytes.Converter._
import sdp.jdbc.engine._
import JDBCEngine._
import scalikejdbc.WrappedResultSet
import scala.concurrent.ExecutionContextExecutor

class JDBCStreamingServices(implicit ec: ExecutionContextExecutor) extends JdbcGrpcAkkaStream.JDBCServices {
  val logger = Logger.getLogger(classOf[JDBCStreamingServices].getName)

  val toRow = (rs: WrappedResultSet) => JDBCDataRow(
    year = rs.string("REPORTYEAR"),
    state = rs.string("STATENAME"),
    county = rs.string("COUNTYNAME"),
    value = rs.string("VALUE")
  )
  override def runQuery: Flow[JDBCQuery, JDBCDataRow, NotUsed] = {
    logger.info("**** runQuery called on service side ***")
    Flow[JDBCQuery]
      .flatMapConcat { q =>
        //unpack JDBCQuery and construct the context
        var params: Seq[Any] =  Nil
        if (q.parameters != _root_.com.google.protobuf.ByteString.EMPTY)
          params = unmarshal[Seq[Any]](q.parameters)
        logger.info(s"**** query parameters: ${params} ****")
        val ctx = JDBCQueryContext[JDBCDataRow](
          dbName = Symbol(q.dbName),
          statement = q.statement,
          parameters = params,
          fetchSize = q.fetchSize.getOrElse(),
          autoCommit = q.autoCommit.getOrElse(false),
          queryTimeout = q.queryTimeout
        )
        jdbcAkkaStream(ctx, toRow)
      }
  }

  override def batQuery: Flow[JDBCQuery, JDBCDataRow, NotUsed] = runQuery

  override def runDDL: Flow[JDBCUpdate, JDBCResult, NotUsed] = {
    logger.info("**** runDDL called on service side ***")
    Flow[JDBCUpdate]
      .flatMapConcat { context =>
        //unpack JDBCUpdate and construct the context

        val ctx = JDBCContext(
          dbName = Symbol(context.dbName),
          statements = context.statements,
          sqlType = JDBCContext.SQL_EXEDDL,
          queryTimeout = context.queryTimeout
        )

        logger.info(s"**** JDBCContext => ${ctx} ***")

        Source
          .fromFuture(jdbcExecuteDDL(ctx))
          .map { r => JDBCResult(marshal(r)) }

      }

  }

  override def insertRows: Flow[MemberRow, JDBCResult, NotUsed] = {
    logger.info("**** insertRows called on service side ***")
    val insertSQL = """
      insert into members(
        name,
        birthday,
        description,
        created_at
      ) values (?, ?, ?, ?)
    """
    Flow[MemberRow]
      .flatMapConcat { row =>
        val ctx = JDBCContext('h2)
          .setUpdateCommand(true,insertSQL,
             row.name,
             jdbcSetDate(row.birthday.get.yyyy,row.birthday.get.mm,row.birthday.get.dd),
             row.description,
             jdbcSetNow
          )

        logger.info(s"**** JDBCContext => ${ctx} ***")

        Source
          .fromFuture(jdbcTxUpdates[Vector](ctx))
          .map { r => JDBCResult(marshal(r)) }
      }
  }

  override def updateRows: Flow[MemberRow, JDBCResult, NotUsed] = {
    logger.info("**** updateRows called on service side ***")
    val updateSQL = "update members set description = ?, created_at = ? where name = ?"

    Flow[MemberRow]
      .flatMapConcat { row =>
        val ctx = JDBCContext('h2)
            .setBatchCommand(updateSQL)
            .appendBatchParameters(
              row.name + " updated.",
              jdbcSetNow,
              row.name
            ).setBatchReturnGeneratedKeyOption(true)

        logger.info(s"**** JDBCContext => ${ctx} ***")

        Source
          .fromFuture(jdbcBatchUpdate[Vector](ctx))
          .map { r => JDBCResult(marshal(r)) }
      }
  }

  val toMemberRow = (rs: WrappedResultSet) => MemberRow(
    name = rs.string("name"),
    description = rs.string("description"),
    birthday = None,
    createdAt = None,
    picture = _root_.com.google.protobuf.ByteString.EMPTY
  )
  override def getMembers: Flow[JDBCQuery, MemberRow, NotUsed] = {
    logger.info("**** getMembers called on service side ***")
    Flow[JDBCQuery]
      .flatMapConcat { q =>
        //unpack JDBCQuery and construct the context
        var params: Seq[Any] =  Nil
        if (q.parameters != _root_.com.google.protobuf.ByteString.EMPTY)
           params = unmarshal[Seq[Any]](q.parameters)
        logger.info(s"**** query parameters: ${params} ****")
        val ctx = JDBCQueryContext[MemberRow](
          dbName = Symbol(q.dbName),
          statement = q.statement,
          parameters = params,
          fetchSize = q.fetchSize.getOrElse(),
          autoCommit = q.autoCommit.getOrElse(false),
          queryTimeout = q.queryTimeout
        )
        jdbcAkkaStream(ctx, toMemberRow)
      }

  }
}

JDBCServer.scala

package demo.grpc.jdbc.server

import java.util.logging.Logger

import akka.actor.ActorSystem
import akka.stream.ActorMaterializer
import io.grpc.Server
import demo.grpc.jdbc.services._
import io.grpc.ServerBuilder
import grpc.jdbc.services._

class gRPCServer(server: Server) {

  val logger: Logger = Logger.getLogger(classOf[gRPCServer].getName)

  def start(): Unit = {
    server.start()
    logger.info(s"Server started, listening on ${server.getPort}")
    sys.addShutdownHook {
      // Use stderr here since the logger may has been reset by its JVM shutdown hook.
      System.err.println("*** shutting down gRPC server since JVM is shutting down")
      stop()
      System.err.println("*** server shut down")
    }
    ()
  }

  def stop(): Unit = {
    server.shutdown()
  }

  /**
    * Await termination on the main thread since the grpc library uses daemon threads.
    */
  def blockUntilShutdown(): Unit = {
    server.awaitTermination()
  }
}

object JDBCServer extends App {
  import sdp.jdbc.config._

  implicit val system = ActorSystem("JDBCServer")
  implicit val mat = ActorMaterializer.create(system)
  implicit val ec = system.dispatcher

  ConfigDBsWithEnv("dev").setup('h2)
  ConfigDBsWithEnv("dev").loadGlobalSettings()

  val server = new gRPCServer(
    ServerBuilder
      .forPort()
      .addService(
        JdbcGrpcAkkaStream.bindService(
          new JDBCStreamingServices
        )
      ).build()
  )
  server.start()
  //  server.blockUntilShutdown()
  scala.io.StdIn.readLine()
  ConfigDBsWithEnv("dev").close('h2)
  mat.shutdown()
  system.terminate()
}

JDBCClient.scala

package demo.grpc.jdbc.client
import grpc.jdbc.services._
import java.util.logging.Logger

import protobuf.bytes.Converter._
import akka.stream.scaladsl._
import akka.NotUsed
import akka.actor.ActorSystem
import akka.stream.{ActorMaterializer, ThrottleMode}
import io.grpc._
import sdp.jdbc.engine._

class JDBCStreamClient(host: String, port: Int) {
  val logger: Logger = Logger.getLogger(classOf[JDBCStreamClient].getName)

  val channel = ManagedChannelBuilder
    .forAddress(host,port)
    .usePlaintext(true)
    .build()

  val stub = JdbcGrpcAkkaStream.stub(channel)
  val query = JDBCQuery (
    dbName = "h2",
    statement = "select * from AQMRPT where STATENAME = ? and VALUE = ?",
    parameters = marshal(Seq("Arizona", ))
  )
  val query2 = JDBCQuery (
    dbName = "h2",
    statement = "select * from AQMRPT where STATENAME = ? and VALUE = ?",
    parameters = marshal(Seq("Colorado", ))
  )
  val query3= JDBCQuery (
    dbName = "h2",
    statement = "select * from AQMRPT where STATENAME = ? and VALUE = ?",
    parameters = marshal(Seq("Arkansas", ))
  )

  def queryRows: Source[JDBCDataRow,NotUsed] = {
    logger.info(s"running queryRows ...")
    Source
        .single(query)
        .via(stub.runQuery)
  }

  def batQueryRows: Source[JDBCDataRow,NotUsed] = {
    logger.info(s"running batQueryRows ...")
    Source
      .fromIterator(() => List(query,query2,query3).toIterator)
      .via(stub.batQuery)
  }

  val dropSQL: String ="""
      drop table members
    """

  val createSQL: String ="""
    create table members (
      id serial not null primary key,
      name varchar() not null,
      description varchar(),
      birthday date,
      created_at timestamp not null,
      picture blob
    )"""
  val ctx = JDBCUpdate (
    dbName = "h2",
    sqlType = JDBCContext.SQL_EXEDDL,
    statements = Seq(dropSQL,createSQL)
  )

  def createTbl: Source[JDBCResult,NotUsed] = {
    logger.info(s"running createTbl ...")
    Source
      .single(ctx)
      .via(stub.runDDL)
  }

  val p1 = MemberRow( "Peter Chan",Some(JDBCDate(,,)),"new member1",None,_root_.com.google.protobuf.ByteString.EMPTY)
  val p2 = MemberRow( "Alanda Wong",Some(JDBCDate(,,)),"new member2",None,_root_.com.google.protobuf.ByteString.EMPTY)
  val p3 = MemberRow( "Kate Zhang",Some(JDBCDate(,,)),"new member3",None,_root_.com.google.protobuf.ByteString.EMPTY)
  val p4 = MemberRow( "Tiger Chan",Some(JDBCDate(,,)),"new member4",None,_root_.com.google.protobuf.ByteString.EMPTY)

  def insertRows: Source[JDBCResult,NotUsed] = {
    logger.info(s"running insertRows ...")
    Source
      .fromIterator(() => List(p1,p2,p3,p4).toIterator)
      .via(stub.insertRows)
  }

  val queryMember = JDBCQuery (
    dbName = "h2",
    statement = "select * from members"
  )
  def updateRows: Source[JDBCResult,NotUsed] = {
    logger.info(s"running updateRows ...")
    Source
      .single(queryMember)
      .via(stub.getMembers)
      .via(stub.updateRows)
  }
  def updateBatches: Source[JDBCDataRow,NotUsed] = {
    logger.info(s"running updateBatches ...")
      Source
        .fromIterator(() => List(query,query2,query3).toIterator)
        .via(stub.batQuery)
        .via(stub.updateBat)

  }

}

object TestConversion extends App {

  val orgval: Seq[Option[Any]] = Seq(Some(),Some("id"),None,Some())
  println(s"original value: ${orgval}")

  val marval = marshal(orgval)

  println(s"marshal value: ${marval}")

  val unmval = unmarshal[Seq[Option[Any]]](marval)

  println(s"marshal value: ${unmval}")

  val m1 = MemberRow(name = "Peter")
  val m2 = m1.update(
    _.birthday.yyyy := ,
    _.birthday.mm := ,
    _.birthday.dd := ,
    _.description := "a new member"
  )

}

object QueryRows extends App {
  implicit val system = ActorSystem("QueryRows")
  implicit val mat = ActorMaterializer.create(system)

  val client = new JDBCStreamClient("localhost", )

  client.queryRows.runForeach { r => println(r) }

  scala.io.StdIn.readLine()
  mat.shutdown()
  system.terminate()
}

object BatQueryRows extends App {
  implicit val system = ActorSystem("BatQueryRows")
  implicit val mat = ActorMaterializer.create(system)

  val client = new JDBCStreamClient("localhost", )

  client.batQueryRows.runForeach(println)

  scala.io.StdIn.readLine()
  mat.shutdown()
  system.terminate()
}

object RunDDL extends App {
  implicit val system = ActorSystem("RunDDL")
  implicit val mat = ActorMaterializer.create(system)

  val client = new JDBCStreamClient("localhost", )

  client.createTbl.runForeach{r => println(unmarshal[Seq[Any]](r.result))}

  scala.io.StdIn.readLine()
  mat.shutdown()
  system.terminate()

}

object InsertRows extends App {
  implicit val system = ActorSystem("InsertRows")
  implicit val mat = ActorMaterializer.create(system)

  val client = new JDBCStreamClient("localhost", )

  client.insertRows.runForeach { r => println(unmarshal[Vector[Long]](r.result)) }

  scala.io.StdIn.readLine()
  mat.shutdown()
  system.terminate()
}

object UpdateRows extends App {
  implicit val system = ActorSystem("UpdateRows")
  implicit val mat = ActorMaterializer.create(system)

  val client = new JDBCStreamClient("localhost", )

  client.updateRows.runForeach{ r => println(unmarshal[Vector[Long]](r.result)) }

  scala.io.StdIn.readLine()
  mat.shutdown()
  system.terminate()
}

object BatUpdates extends App {
  implicit val system = ActorSystem("BatUpdates")
  implicit val mat = ActorMaterializer.create(system)

  val client = new JDBCStreamClient("localhost", )

  client.updateBatches.runForeach(println)

  scala.io.StdIn.readLine()
  mat.shutdown()
  system.terminate()
}

ByteConverter.scala

package protobuf.bytes
import java.io.{ByteArrayInputStream,ByteArrayOutputStream,ObjectInputStream,ObjectOutputStream}
import com.google.protobuf.ByteString
object Converter {

  def marshal(value: Any): ByteString = {
    val stream: ByteArrayOutputStream = new ByteArrayOutputStream()
    val oos = new ObjectOutputStream(stream)
    oos.writeObject(value)
    oos.close()
    ByteString.copyFrom(stream.toByteArray())
  }

  def unmarshal[A](bytes: ByteString): A = {
    val ois = new ObjectInputStream(new ByteArrayInputStream(bytes.toByteArray))
    val value = ois.readObject()
    ois.close()
    value.asInstanceOf[A]
  }

}

其它部分的源代码和系统设置可以从上次的讨论稿中获取。