You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@drill.apache.org by GitBox <gi...@apache.org> on 2021/11/17 20:40:50 UTC
[GitHub] [drill] vdiravka commented on a change in pull request #2351: DRILL-1282: Move parquet to use v2 format as default
vdiravka commented on a change in pull request #2351:
URL: https://github.com/apache/drill/pull/2351#discussion_r751467675
##########
File path: exec/java-exec/src/test/java/org/apache/drill/exec/physical/impl/writer/TestParquetWriter.java
##########
@@ -963,6 +963,17 @@ private void compareParquetInt96Converters(String selection, String table) throw
}
}
+ @Test
+ public void testTPCHReadWriteFormatV2() throws Exception {
+ try {
+ alterSession(ExecConstants.PARQUET_WRITER_FORMAT_VERSION, "parquet_2_0");
Review comment:
Do we need enabling it, if the aim of the PR is enabling V2 by default?
##########
File path: exec/java-exec/src/main/java/org/apache/drill/exec/store/parquet/ParquetFormatPlugin.java
##########
@@ -78,6 +80,8 @@
public class ParquetFormatPlugin implements FormatPlugin {
+ public static final String[] PARQUET_VERSIONS = {"PARQUET_1_0", "PARQUET_2_0"};
Review comment:
I think we can simplify the values, the key already has `parquet` keyword - `PARQUET_VERSIONS`. Similar to `MetadataVersion.Constants` the array can be:
`{"1.0", "2.0"}`.
Or make it similar/corresponding to `ParquetProperties#WriterVersion` enum
##########
File path: exec/java-exec/src/main/java/org/apache/drill/exec/store/parquet/ParquetFormatConfig.java
##########
@@ -18,81 +18,95 @@
package org.apache.drill.exec.store.parquet;
import com.fasterxml.jackson.annotation.JsonCreator;
-import com.fasterxml.jackson.annotation.JsonIgnore;
import com.fasterxml.jackson.annotation.JsonInclude;
import com.fasterxml.jackson.annotation.JsonProperty;
-import java.util.Objects;
+import lombok.Builder;
+import lombok.EqualsAndHashCode;
+import lombok.Getter;
import org.apache.drill.common.PlanStringBuilder;
import org.apache.drill.common.logical.FormatPluginConfig;
import com.fasterxml.jackson.annotation.JsonTypeName;
+@EqualsAndHashCode
@JsonTypeName("parquet") @JsonInclude(JsonInclude.Include.NON_DEFAULT)
public class ParquetFormatConfig implements FormatPluginConfig {
- private final boolean autoCorrectCorruptDates;
- private final boolean enableStringsSignedMinMax;
-
- public ParquetFormatConfig() {
- this(true, false);
- }
-
- @JsonCreator
- public ParquetFormatConfig(@JsonProperty("autoCorrectCorruptDates") Boolean autoCorrectCorruptDates,
- @JsonProperty("enableStringsSignedMinMax") boolean enableStringsSignedMinMax) {
- this.autoCorrectCorruptDates = autoCorrectCorruptDates == null ? true : autoCorrectCorruptDates;
- this.enableStringsSignedMinMax = enableStringsSignedMinMax;
- }
+ @Getter private final boolean autoCorrectCorruptDates;
Review comment:
`enableStringsSignedMinMax` property has a good javadoc.
But `autoCorrectCorruptDates` has a good javadoc in `TestCorruptParquetDateCorrection`. Could you borrow some info from there or add a link to that javadoc?
##########
File path: exec/java-exec/src/main/java/org/apache/drill/exec/store/parquet/ParquetFormatPlugin.java
##########
@@ -141,40 +145,94 @@ public AbstractWriter getWriter(PhysicalOperator child, String location, List<St
}
public RecordWriter getRecordWriter(FragmentContext context, ParquetWriter writer) throws IOException, OutOfMemoryException {
- Map<String, String> options = new HashMap<>();
+ Map<String, String> writerOpts = new HashMap<>();
+ OptionManager contextOpts = context.getOptions();
- options.put("location", writer.getLocation());
+ writerOpts.put("location", writer.getLocation());
FragmentHandle handle = context.getHandle();
String fragmentId = String.format("%d_%d", handle.getMajorFragmentId(), handle.getMinorFragmentId());
- options.put("prefix", fragmentId);
-
- options.put(ExecConstants.PARQUET_BLOCK_SIZE, context.getOptions().getOption(ExecConstants.PARQUET_BLOCK_SIZE).num_val.toString());
- options.put(ExecConstants.PARQUET_WRITER_USE_SINGLE_FS_BLOCK,
- context.getOptions().getOption(ExecConstants.PARQUET_WRITER_USE_SINGLE_FS_BLOCK).bool_val.toString());
- options.put(ExecConstants.PARQUET_PAGE_SIZE, context.getOptions().getOption(ExecConstants.PARQUET_PAGE_SIZE).num_val.toString());
- options.put(ExecConstants.PARQUET_DICT_PAGE_SIZE, context.getOptions().getOption(ExecConstants.PARQUET_DICT_PAGE_SIZE).num_val.toString());
-
- options.put(ExecConstants.PARQUET_WRITER_COMPRESSION_TYPE,
- context.getOptions().getOption(ExecConstants.PARQUET_WRITER_COMPRESSION_TYPE).string_val);
-
- options.put(ExecConstants.PARQUET_WRITER_ENABLE_DICTIONARY_ENCODING,
- context.getOptions().getOption(ExecConstants.PARQUET_WRITER_ENABLE_DICTIONARY_ENCODING).bool_val.toString());
-
- options.put(ExecConstants.PARQUET_WRITER_LOGICAL_TYPE_FOR_DECIMALS,
- context.getOptions().getOption(ExecConstants.PARQUET_WRITER_LOGICAL_TYPE_FOR_DECIMALS).string_val);
-
- options.put(ExecConstants.PARQUET_WRITER_USE_PRIMITIVE_TYPES_FOR_DECIMALS,
- context.getOptions().getOption(ExecConstants.PARQUET_WRITER_USE_PRIMITIVE_TYPES_FOR_DECIMALS).bool_val.toString());
+ writerOpts.put("prefix", fragmentId);
+
+ // Many options which follow may be set as Drill config options or in the parquet format
+ // plugin config. If there is a Drill option set at session scope or narrower it takes precendence.
+ OptionValue.OptionScope minScope = OptionValue.OptionScope.SESSION;
+
+ writerOpts.put(ExecConstants.PARQUET_BLOCK_SIZE,
+ ObjectUtils.firstNonNull(
+ contextOpts.getOption(ExecConstants.PARQUET_BLOCK_SIZE).getValueMinScope(minScope),
+ config.getBlockSize(),
+ contextOpts.getInt(ExecConstants.PARQUET_BLOCK_SIZE)
+ ).toString()
+ );
+
+ writerOpts.put(ExecConstants.PARQUET_WRITER_USE_SINGLE_FS_BLOCK,
+ ObjectUtils.firstNonNull(
+ contextOpts.getOption(ExecConstants.PARQUET_WRITER_USE_SINGLE_FS_BLOCK).getValueMinScope(minScope),
+ config.getUseSingleFSBlock(),
+ contextOpts.getBoolean(ExecConstants.PARQUET_WRITER_USE_SINGLE_FS_BLOCK)
+ ).toString()
+ );
+
+ writerOpts.put(ExecConstants.PARQUET_PAGE_SIZE,
+ ObjectUtils.firstNonNull(
+ contextOpts.getOption(ExecConstants.PARQUET_PAGE_SIZE).getValueMinScope(minScope),
+ config.getPageSize(),
+ contextOpts.getInt(ExecConstants.PARQUET_PAGE_SIZE)
+ ).toString()
+ );
+
+ // "internal use" so not settable in format config
+ writerOpts.put(ExecConstants.PARQUET_DICT_PAGE_SIZE,
+ contextOpts.getOption(ExecConstants.PARQUET_DICT_PAGE_SIZE).num_val.toString()
+ );
+
+ // "internal use" so not settable in format config
+ writerOpts.put(ExecConstants.PARQUET_WRITER_ENABLE_DICTIONARY_ENCODING,
+ contextOpts.getOption(ExecConstants.PARQUET_WRITER_ENABLE_DICTIONARY_ENCODING).bool_val.toString()
+ );
+
+ writerOpts.put(ExecConstants.PARQUET_WRITER_COMPRESSION_TYPE,
+ ObjectUtils.firstNonNull(
+ contextOpts.getOption(ExecConstants.PARQUET_WRITER_COMPRESSION_TYPE).getValueMinScope(minScope),
+ config.getWriterCompressionType(),
+ contextOpts.getString(ExecConstants.PARQUET_WRITER_COMPRESSION_TYPE)
+ ).toString()
+ );
+
+ writerOpts.put(ExecConstants.PARQUET_WRITER_LOGICAL_TYPE_FOR_DECIMALS,
+ ObjectUtils.firstNonNull(
+ contextOpts.getOption(ExecConstants.PARQUET_WRITER_LOGICAL_TYPE_FOR_DECIMALS).getValueMinScope(minScope),
+ config.getWriterLogicalTypeForDecimals(),
+ contextOpts.getString(ExecConstants.PARQUET_WRITER_LOGICAL_TYPE_FOR_DECIMALS)
+ ).toString()
+ );
+
+ writerOpts.put(ExecConstants.PARQUET_WRITER_USE_PRIMITIVE_TYPES_FOR_DECIMALS,
+ ObjectUtils.firstNonNull(
+ contextOpts.getOption(ExecConstants.PARQUET_WRITER_USE_PRIMITIVE_TYPES_FOR_DECIMALS).getValueMinScope(minScope),
+ config.getWriterUsePrimitivesForDecimals(),
+ contextOpts.getBoolean(ExecConstants.PARQUET_WRITER_USE_PRIMITIVE_TYPES_FOR_DECIMALS)
+ ).toString()
+ );
+
+ writerOpts.put(ExecConstants.PARQUET_WRITER_FORMAT_VERSION,
+ ObjectUtils.firstNonNull(
+ contextOpts.getOption(ExecConstants.PARQUET_WRITER_FORMAT_VERSION).getValueMinScope(minScope),
+ config.getWriterFormatVersion(),
+ contextOpts.getString(ExecConstants.PARQUET_WRITER_FORMAT_VERSION)
+ ).toString()
+ );
RecordWriter recordWriter = new ParquetRecordWriter(context, writer);
- recordWriter.init(options);
+ recordWriter.init(writerOpts);
return recordWriter;
}
public WriterRecordBatch getWriterBatch(FragmentContext context, RecordBatch incoming, ParquetWriter writer)
throws ExecutionSetupException {
+ // getConfig().get
Review comment:
```suggestion
```
##########
File path: exec/java-exec/src/test/java/org/apache/drill/exec/store/parquet/TestParquetWriterConfig.java
##########
@@ -0,0 +1,67 @@
+/*
+ * 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;
+
+import org.apache.drill.exec.ExecConstants;
+import org.apache.drill.test.ClusterFixture;
+import org.apache.drill.test.ClusterFixtureBuilder;
+import org.apache.drill.test.ClusterTest;
+import org.junit.BeforeClass;
+import org.junit.Test;
+
+import java.nio.file.Paths;
+
+public class TestParquetWriterConfig extends ClusterTest {
Review comment:
I recommend introducing new test classes based on `JUnit5` framework. As an example you can check my recent update `TestDrillbitResilience` onto the JUnit5 (it is also extends `ClusterTest`)
##########
File path: exec/java-exec/src/main/java/org/apache/drill/exec/store/parquet/ParquetReaderConfig.java
##########
@@ -186,11 +187,13 @@ public ParquetReaderConfig build() {
readerConfig.enableTimeReadCounter = conf.getBoolean(ENABLE_TIME_READ_COUNTER, readerConfig.enableTimeReadCounter);
}
- // last assign values from session options, session options have higher priority than other configurations
+ // last assign values from session or query scoped options which have higher priority than other configurations
if (options != null) {
- String option = options.getOption(ExecConstants.PARQUET_READER_STRINGS_SIGNED_MIN_MAX_VALIDATOR);
- if (!option.isEmpty()) {
- readerConfig.enableStringsSignedMinMax = Boolean.valueOf(option);
+ String optVal = (String) options.getOption(
+ ExecConstants.PARQUET_READER_STRINGS_SIGNED_MIN_MAX
+ ).getValueMinScope(OptionValue.OptionScope.SESSION);
Review comment:
What about `OptionScope.QUERY` (according to above comment)?
##########
File path: exec/java-exec/src/main/java/org/apache/drill/exec/store/parquet/ParquetRecordWriter.java
##########
@@ -263,20 +269,29 @@ private void newSchema() throws IOException {
// We don't want this number to be too small either. Ideally, slightly bigger than the page size,
// but not bigger than the block buffer
int initialPageBufferSize = max(MINIMUM_BUFFER_SIZE, min(pageSize + pageSize / 10, initialBlockBufferSize));
+ ValuesWriterFactory valWriterFactory = writerVersion == WriterVersion.PARQUET_1_0
+ ? new DefaultV1ValuesWriterFactory()
+ : new DefaultV2ValuesWriterFactory();
+
ParquetProperties parquetProperties = ParquetProperties.builder()
.withPageSize(pageSize)
.withDictionaryEncoding(enableDictionary)
.withDictionaryPageSize(initialPageBufferSize)
- .withWriterVersion(writerVersion)
.withAllocator(new ParquetDirectByteBufferAllocator(oContext))
- .withValuesWriterFactory(new DefaultV1ValuesWriterFactory())
+ .withValuesWriterFactory(valWriterFactory)
+ .withWriterVersion(writerVersion)
.build();
+
// TODO: Replace ParquetColumnChunkPageWriteStore with ColumnChunkPageWriteStore from parquet library
// once DRILL-7906 (PARQUET-1006) will be resolved
pageStore = new ParquetColumnChunkPageWriteStore(codecFactory.getCompressor(codec), schema,
parquetProperties.getInitialSlabSize(), pageSize, parquetProperties.getAllocator(),
parquetProperties.getColumnIndexTruncateLength(), parquetProperties.getPageWriteChecksumEnabled());
- store = new ColumnWriteStoreV1(pageStore, parquetProperties);
+
+ store = writerVersion == WriterVersion.PARQUET_1_0
Review comment:
switch-case?
##########
File path: exec/java-exec/src/main/java/org/apache/drill/exec/store/parquet/ParquetFormatPlugin.java
##########
@@ -141,40 +145,94 @@ public AbstractWriter getWriter(PhysicalOperator child, String location, List<St
}
public RecordWriter getRecordWriter(FragmentContext context, ParquetWriter writer) throws IOException, OutOfMemoryException {
- Map<String, String> options = new HashMap<>();
+ Map<String, String> writerOpts = new HashMap<>();
+ OptionManager contextOpts = context.getOptions();
- options.put("location", writer.getLocation());
+ writerOpts.put("location", writer.getLocation());
FragmentHandle handle = context.getHandle();
String fragmentId = String.format("%d_%d", handle.getMajorFragmentId(), handle.getMinorFragmentId());
- options.put("prefix", fragmentId);
-
- options.put(ExecConstants.PARQUET_BLOCK_SIZE, context.getOptions().getOption(ExecConstants.PARQUET_BLOCK_SIZE).num_val.toString());
- options.put(ExecConstants.PARQUET_WRITER_USE_SINGLE_FS_BLOCK,
- context.getOptions().getOption(ExecConstants.PARQUET_WRITER_USE_SINGLE_FS_BLOCK).bool_val.toString());
- options.put(ExecConstants.PARQUET_PAGE_SIZE, context.getOptions().getOption(ExecConstants.PARQUET_PAGE_SIZE).num_val.toString());
- options.put(ExecConstants.PARQUET_DICT_PAGE_SIZE, context.getOptions().getOption(ExecConstants.PARQUET_DICT_PAGE_SIZE).num_val.toString());
-
- options.put(ExecConstants.PARQUET_WRITER_COMPRESSION_TYPE,
- context.getOptions().getOption(ExecConstants.PARQUET_WRITER_COMPRESSION_TYPE).string_val);
-
- options.put(ExecConstants.PARQUET_WRITER_ENABLE_DICTIONARY_ENCODING,
- context.getOptions().getOption(ExecConstants.PARQUET_WRITER_ENABLE_DICTIONARY_ENCODING).bool_val.toString());
-
- options.put(ExecConstants.PARQUET_WRITER_LOGICAL_TYPE_FOR_DECIMALS,
- context.getOptions().getOption(ExecConstants.PARQUET_WRITER_LOGICAL_TYPE_FOR_DECIMALS).string_val);
-
- options.put(ExecConstants.PARQUET_WRITER_USE_PRIMITIVE_TYPES_FOR_DECIMALS,
- context.getOptions().getOption(ExecConstants.PARQUET_WRITER_USE_PRIMITIVE_TYPES_FOR_DECIMALS).bool_val.toString());
+ writerOpts.put("prefix", fragmentId);
+
+ // Many options which follow may be set as Drill config options or in the parquet format
+ // plugin config. If there is a Drill option set at session scope or narrower it takes precendence.
+ OptionValue.OptionScope minScope = OptionValue.OptionScope.SESSION;
Review comment:
As far as I understand SESSION and QUERY scopes had precedence over plugin config earlier too. The aim of these changes for giving higher precedence for plugin config over SYSTEM options, right?
It is fine but could we manage plugin configs within a new OptionManager? And this manager can be in the managers fallback chain, so then the priority of the options will be resolved automatically for all options and plugins. Just consider it as an idea for enhancement (this PR or the new one)
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: dev-unsubscribe@drill.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org