[
https://issues.apache.org/jira/browse/DRILL-5356?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15954371#comment-15954371
]
ASF GitHub Bot commented on DRILL-5356:
---------------------------------------
Github user paul-rogers commented on a diff in the pull request:
https://github.com/apache/drill/pull/789#discussion_r109515637
--- Diff:
exec/java-exec/src/main/java/org/apache/drill/exec/store/parquet/columnreaders/ParquetSchema.java
---
@@ -0,0 +1,207 @@
+/*
+ * 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.columnreaders;
+
+import java.util.ArrayList;
+import java.util.Arrays;
+import java.util.Collection;
+import java.util.HashMap;
+import java.util.List;
+import java.util.Map;
+
+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.common.types.TypeProtos.DataMode;
+import org.apache.drill.exec.exception.SchemaChangeException;
+import org.apache.drill.exec.expr.TypeHelper;
+import org.apache.drill.exec.physical.impl.OutputMutator;
+import org.apache.drill.exec.record.MaterializedField;
+import org.apache.drill.exec.server.options.OptionManager;
+import org.apache.drill.exec.store.parquet.ParquetReaderUtility;
+import org.apache.drill.exec.vector.NullableIntVector;
+import org.apache.parquet.column.ColumnDescriptor;
+import org.apache.parquet.format.SchemaElement;
+import org.apache.parquet.hadoop.metadata.BlockMetaData;
+import org.apache.parquet.hadoop.metadata.ColumnChunkMetaData;
+import org.apache.parquet.hadoop.metadata.ParquetMetadata;
+
+import com.google.common.collect.Lists;
+
+/**
+ * Mapping from the schema of the Parquet file to that of the record reader
+ * to the schema that Drill and the Parquet reader uses.
+ */
+
+public class ParquetSchema {
+ private Collection<SchemaPath> selectedCols;
+ // This is a parallel list to the columns list above, it is used to
determine the subset of the project
+ // pushdown columns that do not appear in this file
+ private boolean[] columnsFound;
+ private ParquetMetadata footer;
+ private Map<String, SchemaElement> schemaElements;
+ private int columnsToScan;
+ private List<ColumnDescriptor> columns;
+ private List<ParquetColumnMetadata> columnMd = new ArrayList<>();
+ private int bitWidthAllFixedFields;
+ private boolean allFieldsFixedLength = true;
+ private long groupRecordCount;
+ private int recordsPerBatch;
+ private int rowGroupIndex;
+ private final OptionManager options;
+
+ public ParquetSchema(OptionManager options, int rowGroupIndex) {
+ selectedCols = null;
+ this.rowGroupIndex = rowGroupIndex;
+ this.options = options;
+ }
+
+ public ParquetSchema(OptionManager options, Collection<SchemaPath>
selectedCols) {
+ this.options = options;
+ this.selectedCols = selectedCols;
+ columnsFound = new boolean[selectedCols.size()];
+ }
+
+ public void buildSchema(ParquetMetadata footer, long batchSize) throws
Exception {
+ this.footer = footer;
+ columns = footer.getFileMetaData().getSchema().getColumns();
+ groupRecordCount = footer.getBlocks().get(rowGroupIndex).getRowCount();
+ loadParquetSchema();
+ computeFixedPart();
+// rowGroupOffset =
footer.getBlocks().get(rowGroupIndex).getColumns().get(0).getFirstDataPageOffset();
+
+ if (columnsToScan != 0 && allFieldsFixedLength) {
+ recordsPerBatch = (int) Math.min(Math.min(batchSize /
bitWidthAllFixedFields,
+ footer.getBlocks().get(0).getColumns().get(0).getValueCount()),
ParquetRecordReader.DEFAULT_RECORDS_TO_READ_IF_FIXED_WIDTH);
+ }
+ else {
+ recordsPerBatch =
ParquetRecordReader.DEFAULT_RECORDS_TO_READ_IF_VARIABLE_WIDTH;
+ }
+ }
+
+ private void loadParquetSchema() {
+ // TODO - figure out how to deal with this better once we add nested
reading, note also look where this map is used below
+ // store a map from column name to converted types if they are non-null
+ schemaElements =
ParquetReaderUtility.getColNameToSchemaElementMapping(footer);
+
+ // loop to add up the length of the fixed width columns and build the
schema
+ for (ColumnDescriptor column : columns) {
+ ParquetColumnMetadata colMd = new ParquetColumnMetadata(column);
+ colMd.resolveDrillType(schemaElements, options);
+ if (! fieldSelected(colMd.field)) {
+ continue;
+ }
+ columnMd.add(colMd);
--- End diff --
Fixed. And, taking a fresh look at the fields in this class, I found I
could get rid of a few that were either redundant, or used only locally in a
single method.
> Refactor Parquet Record Reader
> ------------------------------
>
> Key: DRILL-5356
> URL: https://issues.apache.org/jira/browse/DRILL-5356
> Project: Apache Drill
> Issue Type: Improvement
> Affects Versions: 1.10.0, 1.11.0
> Reporter: Paul Rogers
> Assignee: Paul Rogers
> Priority: Minor
> Fix For: 1.11.0
>
>
> The Parquet record reader class is a key part of Drill that has evolved over
> time to become somewhat hard to follow.
> A number of us are working on Parquet-related tasks and find we have to spend
> an uncomfortable amount of time trying to understand the code. In particular,
> this writer needs to figure out how to convince the reader to provide
> higher-density record batches.
> Rather than continue to decypher the complex code multiple times, this ticket
> requests to refactor the code to make it functionally identical, but
> structurally cleaner. The result will be faster time to value when working
> with this code.
> This is a lower-priority change and will be coordinated with others working
> on this code base. This ticket is only for the record reader class itself; it
> does not include the various readers and writers that Parquet uses since
> another project is actively modifying those classes.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
