org.apache.calcite.rel.RelFieldCollation.getFieldIndex()方法的使用及代码示例

x33g5p2x  于2022-01-29 转载在 其他  
字(11.4k)|赞(0)|评价(0)|浏览(74)

本文整理了Java中org.apache.calcite.rel.RelFieldCollation.getFieldIndex方法的一些代码示例，展示了RelFieldCollation.getFieldIndex的具体用法。这些代码示例主要来源于Github/Stackoverflow/Maven等平台，是从一些精选项目中提取出来的代码，具有较强的参考意义，能在一定程度帮忙到你。RelFieldCollation.getFieldIndex方法的具体详情如下：
包路径：org.apache.calcite.rel.RelFieldCollation
类名称：RelFieldCollation
方法名：getFieldIndex

RelFieldCollation.getFieldIndex介绍

暂无

代码示例

代码示例来源：origin: apache/hive

/** Returns references to fields for a given collation. */
public ImmutableList<RexNode> fields(RelCollation collation) {
 final ImmutableList.Builder<RexNode> nodes = ImmutableList.builder();
 for (RelFieldCollation fieldCollation : collation.getFieldCollations()) {
  RexNode node = field(fieldCollation.getFieldIndex());
  switch (fieldCollation.direction) {
  case DESCENDING:
   node = desc(node);
  }
  switch (fieldCollation.nullDirection) {
  case FIRST:
   node = nullsFirst(node);
   break;
  case LAST:
   node = nullsLast(node);
   break;
  }
  nodes.add(node);
 }
 return nodes.build();
}

代码示例来源：origin: apache/hive

public RelNode align(Aggregate rel, List<RelFieldCollation> collations) {
 // 1) We extract the group by positions that are part of the collations and
 // sort them so they respect it
 LinkedHashSet<Integer> aggregateColumnsOrder = new LinkedHashSet<>();
 ImmutableList.Builder<RelFieldCollation> propagateCollations = ImmutableList.builder();
 if (rel.getGroupType() == Group.SIMPLE && !collations.isEmpty()) {
  for (RelFieldCollation c : collations) {
   if (c.getFieldIndex() < rel.getGroupCount()) {
    // Group column found
    if (aggregateColumnsOrder.add(c.getFieldIndex())) {
     propagateCollations.add(c.copy(rel.getGroupSet().nth(c.getFieldIndex())));
    }
   }
  }
 }
 for (int i = 0; i < rel.getGroupCount(); i++) {
  if (!aggregateColumnsOrder.contains(i)) {
   // Not included in the input collations, but can be propagated as this Aggregate
   // will enforce it
   propagateCollations.add(new RelFieldCollation(rel.getGroupSet().nth(i)));
  }
 }
 // 2) We propagate
 final RelNode child = dispatchAlign(rel.getInput(), propagateCollations.build());
 // 3) We annotate the Aggregate operator with this info
 final HiveAggregate newAggregate = (HiveAggregate) rel.copy(rel.getTraitSet(),
     ImmutableList.of(child));
 newAggregate.setAggregateColumnsOrder(aggregateColumnsOrder);
 return newAggregate;
}

代码示例来源：origin: apache/hive

public RelNode align(Project rel, List<RelFieldCollation> collations) {
 // 1) We extract the collations indices
 boolean containsWindowing = false;
 for (RexNode childExp : rel.getChildExps()) {
  if (childExp instanceof RexOver) {
   // TODO: support propagation for partitioning/ordering in windowing
   containsWindowing = true;
   break;
  }
 }
 ImmutableList.Builder<RelFieldCollation> propagateCollations = ImmutableList.builder();
 if (!containsWindowing) {
  for (RelFieldCollation c : collations) {
   RexNode rexNode = rel.getChildExps().get(c.getFieldIndex());
   if (rexNode instanceof RexInputRef) {
    int newIdx = ((RexInputRef) rexNode).getIndex();
    propagateCollations.add(c.copy((newIdx)));
   }
  }
 }
 // 2) We propagate
 final RelNode child = dispatchAlign(rel.getInput(), propagateCollations.build());
 // 3) Return new Project
 return rel.copy(rel.getTraitSet(), ImmutableList.of(child));
}

代码示例来源：origin: apache/hive

RexNode equals = idxToConjuncts.get(c.getFieldIndex());
if (equals != null) {
 conjuncts.add(equals);
 idxToConjuncts.remove(c.getFieldIndex());
 idxToConjuncts.remove(refToRef.get(c.getFieldIndex()));
 if (c.getFieldIndex() < nLeftColumns) {
  propagateCollationsLeft.add(c.copy(c.getFieldIndex()));
  propagateCollationsRight.add(c.copy(refToRef.get(c.getFieldIndex()) - nLeftColumns));
 } else {
  propagateCollationsLeft.add(c.copy(refToRef.get(c.getFieldIndex())));
  propagateCollationsRight.add(c.copy(c.getFieldIndex() - nLeftColumns));

代码示例来源：origin: apache/drill

RexNode equals = idxToConjuncts.get(c.getFieldIndex());
if (equals != null) {
 conjuncts.add(equals);
 idxToConjuncts.remove(c.getFieldIndex());
 idxToConjuncts.remove(refToRef.get(c.getFieldIndex()));
 if (c.getFieldIndex() < nLeftColumns) {
  propagateCollationsLeft.add(c.copy(c.getFieldIndex()));
  propagateCollationsRight.add(c.copy(refToRef.get(c.getFieldIndex()) - nLeftColumns));
 } else {
  propagateCollationsLeft.add(c.copy(refToRef.get(c.getFieldIndex())));
  propagateCollationsRight.add(c.copy(c.getFieldIndex() - nLeftColumns));

代码示例来源：origin: apache/drill

public RelNode align(Aggregate rel, List<RelFieldCollation> collations) {
 // 1) We extract the group by positions that are part of the collations and
 // sort them so they respect it
 LinkedHashSet<Integer> aggregateColumnsOrder = new LinkedHashSet<>();
 ImmutableList.Builder<RelFieldCollation> propagateCollations = ImmutableList.builder();
 if (!rel.indicator && !collations.isEmpty()) {
  for (RelFieldCollation c : collations) {
   if (c.getFieldIndex() < rel.getGroupCount()) {
    // Group column found
    if (aggregateColumnsOrder.add(c.getFieldIndex())) {
     propagateCollations.add(c.copy(rel.getGroupSet().nth(c.getFieldIndex())));
    }
   }
  }
 }
 for (int i = 0; i < rel.getGroupCount(); i++) {
  if (!aggregateColumnsOrder.contains(i)) {
   // Not included in the input collations, but can be propagated as this Aggregate
   // will enforce it
   propagateCollations.add(new RelFieldCollation(rel.getGroupSet().nth(i)));
  }
 }
 // 2) We propagate
 final RelNode child = dispatchAlign(rel.getInput(), propagateCollations.build());
 // 3) We annotate the Aggregate operator with this info
 final HiveAggregate newAggregate = (HiveAggregate) rel.copy(rel.getTraitSet(),
     ImmutableList.of(child));
 newAggregate.setAggregateColumnsOrder(aggregateColumnsOrder);
 return newAggregate;
}

代码示例来源：origin: apache/hive

Set<Integer> needed = new HashSet<>();
for (RelFieldCollation fc : sort.getCollation().getFieldCollations()) {
 needed.add(fc.getFieldIndex());
 final RexNode node = project.getProjects().get(map.getTarget(fc.getFieldIndex()));
 if (node.isA(SqlKind.CAST)) {
 fieldCollations.add(new RelFieldCollation(m.get(fc.getFieldIndex()), fc.direction,
   fc.nullDirection));

代码示例来源：origin: apache/hive

StringBuilder nullOrder = new StringBuilder();
for (RelFieldCollation sortInfo : sortRel.getCollation().getFieldCollations()) {
 int sortColumnPos = sortInfo.getFieldIndex();
 ColumnInfo columnInfo = new ColumnInfo(inputOp.getSchema().getSignature()
   .get(sortColumnPos));

代码示例来源：origin: apache/hive

List<RelFieldCollation> fieldCollations = new ArrayList<>();
for (RelFieldCollation fc : sort.getCollation().getFieldCollations()) {
 final int target = mapping.getTargetOpt(fc.getFieldIndex());
 if (target < 0) {

代码示例来源：origin: apache/drill

StringBuilder nullOrder = new StringBuilder();
for (RelFieldCollation sortInfo : sortRel.getCollation().getFieldCollations()) {
 int sortColumnPos = sortInfo.getFieldIndex();
 ColumnInfo columnInfo = new ColumnInfo(inputOp.getSchema().getSignature()
   .get(sortColumnPos));

代码示例来源：origin: apache/kylin

@Override
public void implementRewrite(RewriteImplementor implementor) {
  implementor.visitChild(this, getInput());
  // No need to rewrite "order by" applied on non-olap context.
  // Occurs in sub-query like "select ... from (...) inner join (...) order by ..."
  if (this.context.realization == null)
    return;
  for (RelFieldCollation fieldCollation : this.collation.getFieldCollations()) {
    int index = fieldCollation.getFieldIndex();
    SQLDigest.OrderEnum order = getOrderEnum(fieldCollation.getDirection());
    OLAPRel olapChild = (OLAPRel) this.getInput();
    TblColRef orderCol = olapChild.getColumnRowType().getAllColumns().get(index);
    this.context.addSort(orderCol, order);
    this.context.storageContext.markSort();
  }
  this.rowType = this.deriveRowType();
  this.columnRowType = buildColumnRowType();
}

代码示例来源：origin: apache/drill

List<RelFieldCollation> fieldCollations = new ArrayList<>();
for (RelFieldCollation fc : sort.getCollation().getFieldCollations()) {
 final int target = mapping.getTargetOpt(fc.getFieldIndex());
 if (target < 0) {

代码示例来源：origin: apache/drill

/** Returns references to fields for a given collation. */
public ImmutableList<RexNode> fields(RelCollation collation) {
 final ImmutableList.Builder<RexNode> nodes = ImmutableList.builder();
 for (RelFieldCollation fieldCollation : collation.getFieldCollations()) {
  RexNode node = field(fieldCollation.getFieldIndex());
  switch (fieldCollation.direction) {
   case DESCENDING:
    node = desc(node);
  }
  switch (fieldCollation.nullDirection) {
   case FIRST:
    node = nullsFirst(node);
    break;
   case LAST:
    node = nullsLast(node);
    break;
  }
  nodes.add(node);
 }
 return nodes.build();
}

代码示例来源：origin: apache/drill

public RelNode align(Project rel, List<RelFieldCollation> collations) {
 // 1) We extract the collations indices
 boolean containsWindowing = false;
 for (RexNode childExp : rel.getChildExps()) {
  if (childExp instanceof RexOver) {
   // TODO: support propagation for partitioning/ordering in windowing
   containsWindowing = true;
   break;
  }
 }
 ImmutableList.Builder<RelFieldCollation> propagateCollations = ImmutableList.builder();
 if (!containsWindowing) {
  for (RelFieldCollation c : collations) {
   RexNode rexNode = rel.getChildExps().get(c.getFieldIndex());
   if (rexNode instanceof RexInputRef) {
    int newIdx = ((RexInputRef) rexNode).getIndex();
    propagateCollations.add(c.copy((newIdx)));
   }
  }
 }
 // 2) We propagate
 final RelNode child = dispatchAlign(rel.getInput(), propagateCollations.build());
 // 3) Return new Project
 return rel.copy(rel.getTraitSet(), ImmutableList.of(child));
}

代码示例来源：origin: apache/hive

obExpr = obRefToCallMap.get(c.getFieldIndex());
ColumnInfo cI = schema.get(c.getFieldIndex());

代码示例来源：origin: apache/drill

obExpr = obRefToCallMap.get(c.getFieldIndex());
ColumnInfo cI = schema.get(c.getFieldIndex());

代码示例来源：origin: apache/hive

for (RelFieldCollation relFieldCollation
  : sortLimit.getCollation().getFieldCollations()) {
 if (relFieldCollation.getFieldIndex()
   >= join.getLeft().getRowType().getFieldCount()) {
  return false;
for (RelFieldCollation relFieldCollation
  : sortLimit.getCollation().getFieldCollations()) {
 if (relFieldCollation.getFieldIndex()
   < join.getLeft().getRowType().getFieldCount()) {
  return false;

代码示例来源：origin: apache/drill

for (RelFieldCollation relFieldCollation
  : sortLimit.getCollation().getFieldCollations()) {
 if (relFieldCollation.getFieldIndex()
   >= join.getLeft().getRowType().getFieldCount()) {
  return false;
for (RelFieldCollation relFieldCollation
  : sortLimit.getCollation().getFieldCollations()) {
 if (relFieldCollation.getFieldIndex()
   < join.getLeft().getRowType().getFieldCount()) {
  return false;

代码示例来源：origin: apache/drill

public void onMatch(RelOptRuleCall call) {
 final HiveSortLimit sort = call.rel(0);
 final HiveProject project = call.rel(1);
 // Determine mapping between project input and output fields. If sort
 // relies on non-trivial expressions, we can't push.
 final Mappings.TargetMapping map =
   RelOptUtil.permutation(
     project.getProjects(), project.getInput().getRowType());
 for (RelFieldCollation fc : sort.getCollation().getFieldCollations()) {
  if (map.getTargetOpt(fc.getFieldIndex()) < 0) {
   return;
  }
 }
 // Create new collation
 final RelCollation newCollation =
   RelCollationTraitDef.INSTANCE.canonize(
     RexUtil.apply(map, sort.getCollation()));
 // New operators
 final HiveSortLimit newSort = sort.copy(sort.getTraitSet().replace(newCollation),
     project.getInput(), newCollation, sort.offset, sort.fetch);
 final RelNode newProject = project.copy(sort.getTraitSet(),
     ImmutableList.<RelNode>of(newSort));
 call.transformTo(newProject);
}

代码示例来源：origin: apache/drill

public void onMatch(RelOptRuleCall call) {
 final HiveProject project = call.rel(0);
 final HiveSortLimit sort = call.rel(1);
 // Determine mapping between project input and output fields. If sort
 // relies on non-trivial expressions, we can't push.
 final Mappings.TargetMapping map =
   RelOptUtil.permutation(
     project.getProjects(), project.getInput().getRowType()).inverse();
 for (RelFieldCollation fc : sort.getCollation().getFieldCollations()) {
  if (map.getTarget(fc.getFieldIndex()) < 0) {
   return;
  }
 }
 // Create new collation
 final RelCollation newCollation =
   RelCollationTraitDef.INSTANCE.canonize(
     RexUtil.apply(map, sort.getCollation()));
 // New operators
 final RelNode newProject = project.copy(sort.getInput().getTraitSet(),
     ImmutableList.<RelNode>of(sort.getInput()));
 final HiveSortLimit newSort = sort.copy(newProject.getTraitSet(),
     newProject, newCollation, sort.offset, sort.fetch);
 call.transformTo(newSort);
}

内容来源于网络，如有侵权，请联系作者删除！

RelFieldCollation

相关文章

    查看更多
    微信公众号

    热门标签

    更多
    JavaquerypythonNode开发语言requestUtil数据库Table后端算法LoggerMessageElementParser

    最新文章

    更多
    • 蜀ICP备13028337号-1 大数据知识库 https://www.saoniuhuo.com © All rights reserved
    • 本站内容来源互联网,如果侵犯您的权益请联系我们删除, 联系方式：448109455@qq.com