[ 
https://issues.apache.org/jira/browse/FLINK-6232?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16057468#comment-16057468
 ] 

ASF GitHub Bot commented on FLINK-6232:
---------------------------------------

Github user fhueske commented on a diff in the pull request:

    https://github.com/apache/flink/pull/3715#discussion_r123205673
  
    --- Diff: 
flink-libraries/flink-table/src/main/scala/org/apache/flink/table/runtime/join/JoinUtil.scala
 ---
    @@ -0,0 +1,385 @@
    +/*
    + * Licensed to the Apache Software Foundation (ASF) under one
    + * or more contributor license agreements.  See the NOTICE file
    + * distributed with this work for additional information
    + * regarding copyright ownership.  The ASF licenses this file
    + * to you under the Apache License, Version 2.0 (the
    + * "License"); you may not use this file except in compliance
    + * with the License.  You may obtain a copy of the License at
    + *
    + *     http://www.apache.org/licenses/LICENSE-2.0
    + *
    + * Unless required by applicable law or agreed to in writing, software
    + * distributed under the License is distributed on an "AS IS" BASIS,
    + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
    + * See the License for the specific language governing permissions and
    + * limitations under the License.
    + */
    +package org.apache.flink.table.runtime.join
    +
    +import java.math.{BigDecimal => JBigDecimal}
    +import java.util
    +
    +import org.apache.calcite.plan.RelOptUtil
    +import org.apache.calcite.rel.`type`.RelDataType
    +import org.apache.calcite.rel.core.JoinRelType
    +import org.apache.calcite.rex._
    +import org.apache.calcite.sql.SqlKind
    +import org.apache.calcite.sql.fun.{SqlFloorFunction, SqlStdOperatorTable}
    +import org.apache.flink.api.common.functions.FlatJoinFunction
    +import org.apache.flink.api.common.typeinfo.TypeInformation
    +import org.apache.flink.table.api.{TableConfig, TableException}
    +import org.apache.flink.table.calcite.FlinkTypeFactory
    +import org.apache.flink.table.codegen.{CodeGenerator, ExpressionReducer}
    +import org.apache.flink.table.functions.TimeMaterializationSqlFunction
    +import org.apache.flink.table.plan.schema.{RowSchema, 
TimeIndicatorRelDataType}
    +import org.apache.flink.types.Row
    +
    +import scala.collection.JavaConversions._
    +
    +/**
    +  * An util class to help analyze and build join code .
    +  */
    +object JoinUtil {
    +
    +  /**
    +    * check if the join case is stream join stream
    +    *
    +    * @param  condition   other condtion include time-condition
    +    * @param  inputType   left and right connect stream type
    +    */
    +  private[flink] def isStreamStreamJoin(
    +      condition: RexNode,
    +      inputType: RelDataType) = {
    +
    +    def isExistTumble(expr: RexNode): Boolean = {
    +      expr match {
    +        case c: RexCall =>
    +          c.getOperator match {
    +            case _: SqlFloorFunction =>
    +              c.getOperands.map(analyzeSingleConditionTerm(_, 0, 
inputType)).exists(_.size > 0)
    +            case SqlStdOperatorTable.TUMBLE =>
    +              c.getOperands.map(analyzeSingleConditionTerm(_, 0, 
inputType)).exists(_.size > 0)
    +            case _ =>
    +              c.getOperands.map(isExistTumble(_)).exists(_ == true)
    +          }
    +        case _ => false
    +      }
    +    }
    +
    +    val isExistTimeIndicator = analyzeSingleConditionTerm(condition, 0, 
inputType).size > 0
    +    val isExistTumbleExpr = isExistTumble(condition)
    +
    +    !isExistTumbleExpr && isExistTimeIndicator
    +  }
    +
    +  /**
    +    * Analyze time-condtion to get time boundary for each stream and get 
the time type
    +    * and return remain condition.
    +    *
    +    * @param  condition   other condtion include time-condition
    +    * @param  leftLogicalFieldCnt left stream logical field num
    +    * @param  leftPhysicalFieldCnt left stream physical field num
    +    * @param  inputType   left and right connect stream type
    +    * @param  rexBuilder   util to build rexNode
    +    * @param  config      table environment config
    +    */
    +  private[flink] def analyzeTimeBoundary(
    +      condition: RexNode,
    +      leftLogicalFieldCnt: Int,
    +      leftPhysicalFieldCnt: Int,
    +      inputType: RelDataType,
    +      rexBuilder: RexBuilder,
    +      config: TableConfig): (RelDataType, Long, Long, Option[RexNode]) = {
    +
    +    // Converts the condition to conjunctive normal form (CNF)
    +    val cnfCondition = RexUtil.toCnf(rexBuilder, condition)
    +
    +    // split the condition into time indicator condition and other 
condition
    +    val (timeTerms, remainTerms) =
    +      splitJoinCondition(
    +        cnfCondition,
    +        leftLogicalFieldCnt,
    +        inputType
    +      )
    +
    +    if (timeTerms.size != 2) {
    +      throw new TableException("There only can and must have 2 time 
conditions.")
    +    }
    +
    +    // extract time offset from the time indicator conditon
    +    val streamTimeOffsets =
    +      timeTerms.map(x => extractTimeOffsetFromCondition(x._3, x._2, 
rexBuilder, config))
    +
    +    val (leftTableOffset, rightTableOffset) =
    +      streamTimeOffsets match {
    +        case Seq((x, true), (y, false)) => (x, y)
    +        case Seq((x, false), (y, true)) => (y, x)
    +        case _ =>
    +          throw new TableException("Both input need time boundary.")
    +      }
    +
    +    // compose the remain condition list into one condition
    +    val remainCondition =
    +      remainTerms match {
    +        case Seq() => None
    +        case _ =>
    +          // turn the logical field index to physical field index
    +          def transInputRef(expr: RexNode): RexNode = {
    +            expr match {
    +              case c: RexCall =>
    +                val newOps = c.operands.map(transInputRef(_))
    +                rexBuilder.makeCall(c.getType, c.getOperator, newOps)
    +              case i: RexInputRef if i.getIndex >= leftLogicalFieldCnt =>
    +                rexBuilder.makeInputRef(
    +                  i.getType,
    +                  i.getIndex - leftLogicalFieldCnt + leftPhysicalFieldCnt)
    +              case _ => expr
    +            }
    +          }
    +
    +          Some(remainTerms.map(transInputRef(_)).reduceLeft( (l, r) => {
    +            RelOptUtil.andJoinFilters(rexBuilder, l, r)
    +          }))
    +      }
    +
    +    (timeTerms.get(0)._1, leftTableOffset, rightTableOffset, 
remainCondition)
    +  }
    +
    +  /**
    +   * Split the join conditions into time condition and non-time condition
    +   */
    +  private def splitJoinCondition(
    +      cnfCondition: RexNode,
    +      leftFieldCount: Int,
    +      inputType: RelDataType): (Seq[(RelDataType, Boolean, RexNode)], 
Seq[RexNode]) = {
    +
    +    cnfCondition match {
    +      case c: RexCall if c.getKind == SqlKind.AND =>
    +        val timeIndicators =
    +          c.operands.map(splitJoinCondition(_, leftFieldCount, inputType))
    +        timeIndicators.reduceLeft { (l, r) =>
    +          (l._1 ++ r._1, l._2 ++ r._2)
    +        }
    +      case c: RexCall =>
    +        val timeIndicators = analyzeSingleConditionTerm(c, leftFieldCount, 
inputType)
    +        timeIndicators match {
    +          case Seq() =>
    +            (Seq(), Seq(c))
    +          case Seq(v1, v2) =>
    +            if (v1._1 != v2._1) {
    +              throw new TableException("The time indicators for each input 
should be the same.")
    +            }
    +            if (v1._2 == v2._2) {
    +              throw new TableException("Both input's time indicators is 
needed.")
    +            }
    +            (Seq((v1._1, v1._2, c)), Seq())
    +          case _ =>
    +            throw new TableException(
    +              "There only can and must have one time indicators for each 
input.")
    +        }
    +      case other =>
    +        val timeIndicators = analyzeSingleConditionTerm(other, 
leftFieldCount, inputType)
    +        timeIndicators match {
    +          case Seq() =>
    +            (Seq(), Seq(other))
    +          case _ =>
    +            throw new TableException("Time indicators can not be used in 
non time-condition.")
    +        }
    +    }
    +  }
    +
    +  /**
    +   * analysis if condition term has time indicator
    +   */
    +  def analyzeSingleConditionTerm(
    +      expression: RexNode,
    +      leftFieldCount: Int,
    +      inputType: RelDataType): Seq[(RelDataType, Boolean)] = {
    +
    +    expression match {
    +      case i: RexInputRef =>
    +        val idx = i.getIndex
    +        inputType.getFieldList.get(idx).getType match {
    +          case t: TimeIndicatorRelDataType if idx < leftFieldCount =>
    +            // left table time indicator
    +            Seq((t, true))
    +          case t: TimeIndicatorRelDataType =>
    +            // right table time indicator
    +            Seq((t, false))
    +          case _ => Seq()
    +        }
    +      case c: RexCall =>
    +        c.operands.map(analyzeSingleConditionTerm(_, leftFieldCount, 
inputType)).reduce(_++_)
    +      case _ => Seq()
    +    }
    +  }
    +
    +  /**
    +    * Extract time offset and determain which table the offset belong to
    +    */
    +  def extractTimeOffsetFromCondition(
    +      timeTerm: RexNode,
    +      isLeftExprBelongLeftTable: Boolean,
    +      rexBuilder: RexBuilder,
    +      config: TableConfig) = {
    +
    +    val timeCall: RexCall = timeTerm.asInstanceOf[RexCall]
    +    val leftLiteral =
    +      reduceTimeExpression(
    +        timeCall.operands.get(0),
    +        rexBuilder,
    +        config)
    +
    +    val rightLiteral =
    +      reduceTimeExpression(
    +        timeCall.operands.get(1),
    +        rexBuilder,
    +        config)
    +
    +    val (tmpTimeOffset: Long, isLeftTableTimeOffset: Boolean) =
    +      timeTerm.getKind match {
    +        // e.g a.proctime > b.proctime - 5 sec, we need to store stream a.
    +        // the left expr(a) belong to left table, so the offset belong to 
left table
    +        case kind @ (SqlKind.GREATER_THAN | SqlKind.GREATER_THAN_OR_EQUAL) 
=>
    +          (rightLiteral - leftLiteral, isLeftExprBelongLeftTable)
    +        // e.g a.proctime < b.proctime + 5 sec, we need to store stream b.
    +        case kind @ (SqlKind.LESS_THAN | SqlKind.LESS_THAN_OR_EQUAL) =>
    +          (leftLiteral - rightLiteral, !isLeftExprBelongLeftTable)
    +        case _ => 0
    +      }
    +
    +    val timeOffset =
    +      // only preceding offset need to store records
    +      if (tmpTimeOffset < 0)
    +        // determain the boudary value
    +        if (timeTerm.getKind == SqlKind.LESS_THAN || timeTerm.getKind == 
SqlKind.GREATER_THAN) {
    +          -tmpTimeOffset - 1
    +        } else {
    +          -tmpTimeOffset
    +        }
    +      else 0
    --- End diff --
    
    I think this might result in invalid bounds. For example the predicate 
`l.ptime > r.ptime - 10.secs AND l.ptime < r.ptime - 5.secs` would be changed 
to `l.ptime > r.ptime - 10.secs AND l.ptime <= r.ptime`.
    
    In principle bounds as in the example above are valid (join each event of 
`r` with all events from `l` that happened between 10 and 5 seconds before the 
right event). Changing the bound changes the semantics of the query. If we do 
not support this case in the initial version, we should throw an exception but 
not change the semantics.


> Support proctime inner equi-join between two streams in the SQL API
> -------------------------------------------------------------------
>
>                 Key: FLINK-6232
>                 URL: https://issues.apache.org/jira/browse/FLINK-6232
>             Project: Flink
>          Issue Type: Sub-task
>          Components: Table API & SQL
>            Reporter: hongyuhong
>            Assignee: hongyuhong
>
> The goal of this issue is to add support for inner equi-join on proc time 
> streams to the SQL interface.
> Queries similar to the following should be supported:
> {code}
> SELECT o.proctime, o.productId, o.orderId, s.proctime AS shipTime 
> FROM Orders AS o 
> JOIN Shipments AS s 
> ON o.orderId = s.orderId 
> AND o.proctime BETWEEN s.proctime AND s.proctime + INTERVAL '1' HOUR;
> {code}
> The following restrictions should initially apply:
> * The join hint only support inner join
> * The ON clause should include equi-join condition
> * The time-condition {{o.proctime BETWEEN s.proctime AND s.proctime + 
> INTERVAL '1' HOUR}} only can use proctime that is a system attribute, the 
> time condition only support bounded time range like {{o.proctime BETWEEN 
> s.proctime - INTERVAL '1' HOUR AND s.proctime + INTERVAL '1' HOUR}}, not 
> support unbounded like {{o.proctime > s.protime}},  and  should include both 
> two stream's proctime attribute, {{o.proctime between proctime() and 
> proctime() + 1}} should also not be supported.
> This issue includes:
> * Design of the DataStream operator to deal with stream join
> * Translation from Calcite's RelNode representation (LogicalJoin). 



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)

Reply via email to