Github user jinfengni commented on a diff in the pull request:
https://github.com/apache/drill/pull/637#discussion_r86465657
--- Diff:
exec/java-exec/src/main/java/org/apache/drill/exec/store/parquet/stat/ParquetMetaStatCollector.java
---
@@ -0,0 +1,146 @@
+/**
+ * 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.drill.exec.store.parquet.stat;
+
+import org.apache.drill.common.expression.SchemaPath;
+import org.apache.drill.common.types.TypeProtos;
+import org.apache.drill.common.types.Types;
+import org.apache.drill.exec.store.parquet.Metadata;
+import org.apache.drill.exec.store.parquet.ParquetGroupScan;
+import org.apache.parquet.column.statistics.BinaryStatistics;
+import org.apache.parquet.column.statistics.DoubleStatistics;
+import org.apache.parquet.column.statistics.FloatStatistics;
+import org.apache.parquet.column.statistics.IntStatistics;
+import org.apache.parquet.column.statistics.LongStatistics;
+import org.apache.parquet.column.statistics.Statistics;
+import org.apache.parquet.schema.OriginalType;
+import org.apache.parquet.schema.PrimitiveType;
+import org.joda.time.DateTimeConstants;
+
+import java.util.HashMap;
+import java.util.List;
+import java.util.Map;
+import java.util.Set;
+
+public class ParquetMetaStatCollector implements ColumnStatCollector{
+ private final Metadata.ParquetTableMetadataBase parquetTableMetadata;
+ private final List<? extends Metadata.ColumnMetadata>
columnMetadataList;
+ final Map<String, String> implicitColValues;
+
+ public ParquetMetaStatCollector(Metadata.ParquetTableMetadataBase
parquetTableMetadata, List<? extends Metadata.ColumnMetadata>
columnMetadataList, Map<String, String> implicitColValues) {
+ this.parquetTableMetadata = parquetTableMetadata;
+ this.columnMetadataList = columnMetadataList;
+ this.implicitColValues = implicitColValues;
+ }
+
+ @Override
+ public Map<SchemaPath, ColumnStatistics> collectColStat(Set<SchemaPath>
fields) {
+ // map from column to ColumnMetadata
+ final Map<SchemaPath, Metadata.ColumnMetadata> columnMetadataMap = new
HashMap<>();
+
+ // map from column name to column statistics.
+ final Map<SchemaPath, ColumnStatistics> statMap = new HashMap<>();
+
+ for (final Metadata.ColumnMetadata columnMetadata :
columnMetadataList) {
+ SchemaPath schemaPath =
SchemaPath.getCompoundPath(columnMetadata.getName());
+ columnMetadataMap.put(schemaPath, columnMetadata);
+ }
+
+ for (final SchemaPath schemaPath : fields) {
+ final PrimitiveType.PrimitiveTypeName primitiveType;
+ final OriginalType originalType;
+
+ final Metadata.ColumnMetadata columnMetadata =
columnMetadataMap.get(schemaPath);
+
+ if (columnMetadata != null) {
+ final Object min = columnMetadata.getMinValue();
+ final Object max = columnMetadata.getMaxValue();
+ final Long numNull = columnMetadata.getNulls();
+
+ primitiveType =
this.parquetTableMetadata.getPrimitiveType(columnMetadata.getName());
+ originalType =
this.parquetTableMetadata.getOriginalType(columnMetadata.getName());
+ final Integer repetitionLevel =
this.parquetTableMetadata.getRepetitionLevel(columnMetadata.getName());
+
+ statMap.put(schemaPath, getStat(min, max, numNull, primitiveType,
originalType, repetitionLevel));
+ } else {
+ final String columnName = schemaPath.getRootSegment().getPath();
+ if (implicitColValues.containsKey(columnName)) {
--- End diff --
Without knowledge of implicit columns, expression materialization will
treat dir0 in dir0 = 1995 as NULLEXPRESSION, and could not differentiate from a
regular non-exist column. A condition on a regular non-exist column will
always lead to canDrop = true.
That's the main reason we have to pass in the list of implicit columns when
do expression materialization.
In addition to implicit column name, we also pass in implicit column
values, and wrap them in Statistics instance with min and max having same
value. That is done to expand the possibility of pruning.
For example, if we have regCol = 5 or dir0 = 1995. If regCol is not a
partition column, we would not do any partition pruning in the current
partition pruning logical. Pass the implicit column values may allow us to
prune some row groups using condition regCol = 5 or dir0 = 1995.
I added explanations in comment.
---
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 [email protected] or file a JIRA ticket
with INFRA.
---
