drill-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From paul-rogers <...@git.apache.org>
Subject [GitHub] drill pull request #729: Drill 1328: Support table statistics for Parquet
Date Fri, 24 Feb 2017 03:53:43 GMT
Github user paul-rogers commented on a diff in the pull request:

    https://github.com/apache/drill/pull/729#discussion_r102869874
  
    --- Diff: exec/java-exec/src/main/java/org/apache/drill/exec/physical/impl/statistics/AvgWidthMergedStatistic.java
---
    @@ -0,0 +1,135 @@
    +/*
    + * 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
    + * <p/>
    + * http://www.apache.org/licenses/LICENSE-2.0
    + * <p/>
    + * 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.drill.exec.physical.impl.statistics;
    +
    +import org.apache.drill.common.types.TypeProtos;
    +import org.apache.drill.exec.expr.holders.NullableFloat8Holder;
    +import org.apache.drill.exec.expr.holders.ValueHolder;
    +import org.apache.drill.exec.vector.NullableFloat8Vector;
    +import org.apache.drill.exec.vector.ValueVector;
    +import org.apache.drill.exec.vector.complex.MapVector;
    +
    +import java.util.HashMap;
    +import java.util.List;
    +import java.util.Map;
    +
    +public class AvgWidthMergedStatistic extends AbstractMergedStatistic {
    +
    +  private String name;
    +  private String inputName;
    +  private boolean configureComplete = false;
    +  private boolean mergeComplete = false;
    +  private Map<String, ValueHolder> sumHolder;
    +  MergedStatistic types, nonNullStatCounts, statCounts;
    +
    +  public AvgWidthMergedStatistic (String name, String inputName) {
    +    this.name = name;
    +    this.inputName = inputName;
    +    this.sumHolder = new HashMap<>();
    +    types = nonNullStatCounts = statCounts = null;
    +  }
    +
    +  @Override
    +  public String getName() {
    +    return name;
    +  }
    +
    +  @Override
    +  public String getInput() {
    +    return inputName;
    +  }
    +
    +  @Override
    +  public void merge(ValueVector input) {
    +    // Check the input is a Map Vector
    +    assert (input.getField().getType().getMinorType() == TypeProtos.MinorType.MAP);
    +    MapVector inputMap = (MapVector) input;
    +    for (ValueVector vv : inputMap) {
    +      String colName = vv.getField().getLastName();
    +      NullableFloat8Holder colSumHolder;
    +      if (sumHolder.get(colName) != null) {
    +        colSumHolder = (NullableFloat8Holder) sumHolder.get(colName);
    +      } else {
    +        colSumHolder = new NullableFloat8Holder();
    +        sumHolder.put(colName, colSumHolder);
    +      }
    +      Object val = vv.getAccessor().getObject(0);
    --- End diff --
    
    This call is quite expensive: it materializes the value as a Java object, which you will
unbox later. Since the only way you can get here is via creating a `NullableFloat8Holder`,
we can take a step back.
    
    You really don't need a holder; they are meant for code generation. Heck, they don't even
do anything useful here.
    
    What you want is the value vector itself, of the proper type.
    
    ```
      for (ValueVector vv : inputMap) {
        NullableFloat8Vector fv = (NullableFloat8Vector) vv;
        NullableFloat8Vector.Accessor accessor = fv.getAccessor();
        if (! accessor.isNull()) {
          colSumHolder.value += accessor.get(0);
          colSumHolder.isSet = 1;
        }
      }
    ```
    
    Note that you also don't need a holder for the sum. A simple double should be fine.
    
    Can the input values actually be null? Under what situation? If they can, then a simple
`boolean isSet;` would handle the case of propagating null values.
    
    Using holder introduces overhead, causes more heap thrashing, and really does not buy
you anything here.


---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastructure@apache.org or file a JIRA ticket
with INFRA.
---

Mime
View raw message