• Home
• com.nvidia
• rapids-4-spark_2.12
• 24.08.1
• source code
• GpuParquetReader.java

×

Project download

Many resources are needed to download a project. Please understand that we have to compensate our server costs. Thank you in advance.
Project price only 1 $

You can buy this project and download/modify it how often you want.

com.nvidia.spark.rapids.iceberg.parquet.GpuParquetReader Maven / Gradle / Ivy

/*
 * Copyright (c) 2022-2024, NVIDIA CORPORATION.
 *
 * Licensed 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 com.nvidia.spark.rapids.iceberg.parquet;

import java.io.IOException;
import java.io.UncheckedIOException;
import java.util.List;
import java.util.Map;
import java.util.Objects;
import java.util.Set;

import scala.collection.Seq;

import com.nvidia.spark.rapids.DateTimeRebaseCorrected$;
import com.nvidia.spark.rapids.GpuMetric;
import com.nvidia.spark.rapids.GpuParquetUtils;
import com.nvidia.spark.rapids.ParquetPartitionReader;
import com.nvidia.spark.rapids.PartitionReaderWithBytesRead;
import com.nvidia.spark.rapids.iceberg.data.GpuDeleteFilter;
import com.nvidia.spark.rapids.iceberg.spark.SparkSchemaUtil;
import com.nvidia.spark.rapids.iceberg.spark.source.GpuIcebergReader;
import org.apache.hadoop.conf.Configuration;
import org.apache.hadoop.fs.Path;
import org.apache.iceberg.MetadataColumns;
import org.apache.iceberg.Schema;
import org.apache.iceberg.expressions.Expression;
import org.apache.iceberg.io.CloseableGroup;
import org.apache.iceberg.io.CloseableIterable;
import org.apache.iceberg.io.InputFile;
import org.apache.iceberg.mapping.NameMapping;
import org.apache.iceberg.relocated.com.google.common.base.Preconditions;
import org.apache.iceberg.relocated.com.google.common.collect.ImmutableList;
import org.apache.iceberg.relocated.com.google.common.collect.Lists;
import org.apache.iceberg.relocated.com.google.common.collect.Maps;
import org.apache.iceberg.relocated.com.google.common.collect.Sets;
import org.apache.iceberg.types.Types;
import org.apache.parquet.ParquetReadOptions;
import org.apache.parquet.hadoop.ParquetFileReader;
import org.apache.parquet.hadoop.metadata.BlockMetaData;
import org.apache.parquet.schema.GroupType;
import org.apache.parquet.schema.MessageType;
import org.apache.parquet.schema.PrimitiveType;
import org.apache.parquet.schema.Type;
import org.apache.parquet.schema.Types.MessageTypeBuilder;

import org.apache.spark.sql.execution.datasources.PartitionedFile;
import org.apache.spark.sql.types.ArrayType;
import org.apache.spark.sql.types.DataType;
import org.apache.spark.sql.types.DecimalType$;
import org.apache.spark.sql.types.DoubleType$;
import org.apache.spark.sql.types.FloatType$;
import org.apache.spark.sql.types.IntegerType$;
import org.apache.spark.sql.types.LongType$;
import org.apache.spark.sql.types.MapType;
import org.apache.spark.sql.types.Metadata;
import org.apache.spark.sql.types.StructField;
import org.apache.spark.sql.types.StructType;
import org.apache.spark.sql.vectorized.ColumnarBatch;

/** GPU version of Apache Iceberg's ParquetReader class */
public class GpuParquetReader extends CloseableGroup implements CloseableIterable {
  private final InputFile input;
  private final Schema expectedSchema;
  private final ParquetReadOptions options;
  private final Expression filter;
  private final boolean caseSensitive;
  private final NameMapping nameMapping;
  private final Map idToConstant;
  private final GpuDeleteFilter deleteFilter;
  private final PartitionedFile partFile;
  private final Configuration conf;
  private final int maxBatchSizeRows;
  private final long maxBatchSizeBytes;
  private final long targetBatchSizeBytes;
  private final boolean useChunkedReader;
  private final long maxChunkedReaderMemoryUsageSizeBytes;
  private final scala.Option debugDumpPrefix;
  private final boolean debugDumpAlways;
  private final scala.collection.immutable.Map metrics;

  public GpuParquetReader(
      InputFile input, Schema expectedSchema, ParquetReadOptions options,
      NameMapping nameMapping, Expression filter, boolean caseSensitive,
      Map idToConstant, GpuDeleteFilter deleteFilter,
      PartitionedFile partFile, Configuration conf, int maxBatchSizeRows,
      long maxBatchSizeBytes, long targetBatchSizeBytes, boolean useChunkedReader,
      long maxChunkedReaderMemoryUsageSizeBytes,
      scala.Option debugDumpPrefix, boolean debugDumpAlways,
      scala.collection.immutable.Map metrics) {
    this.input = input;
    this.expectedSchema = expectedSchema;
    this.options = options;
    this.nameMapping = nameMapping;
    this.filter = filter;
    this.caseSensitive = caseSensitive;
    this.idToConstant = idToConstant;
    this.deleteFilter = deleteFilter;
    this.partFile = partFile;
    this.conf = conf;
    this.maxBatchSizeRows = maxBatchSizeRows;
    this.maxBatchSizeBytes = maxBatchSizeBytes;
    this.targetBatchSizeBytes = targetBatchSizeBytes;
    this.useChunkedReader = useChunkedReader;
    this.maxChunkedReaderMemoryUsageSizeBytes = maxChunkedReaderMemoryUsageSizeBytes;
    this.debugDumpPrefix = debugDumpPrefix;
    this.debugDumpAlways = debugDumpAlways;
    this.metrics = metrics;
  }

  @Override
  public org.apache.iceberg.io.CloseableIterator iterator() {
    try (ParquetFileReader reader = newReader(input, options)) {
      MessageType fileSchema = reader.getFileMetaData().getSchema();
      List filteredRowGroups = filterRowGroups(reader, nameMapping,
          expectedSchema, filter, caseSensitive);

      ReorderColumns reorder = ParquetSchemaUtil.hasIds(fileSchema) ? new ReorderColumns(idToConstant)
          : new ReorderColumnsFallback(idToConstant);
      MessageType fileReadSchema = (MessageType) TypeWithSchemaVisitor.visit(
          expectedSchema.asStruct(), fileSchema, reorder);

      Seq clippedBlocks = GpuParquetUtils.clipBlocksToSchema(
          fileReadSchema, filteredRowGroups, caseSensitive);
      StructType partReaderSparkSchema = (StructType) TypeWithSchemaVisitor.visit(
          expectedSchema.asStruct(), fileReadSchema, new SparkSchemaConverter());

      // reuse Parquet scan code to read the raw data from the file
      ParquetPartitionReader parquetPartReader = new ParquetPartitionReader(conf, partFile,
          new Path(input.location()), clippedBlocks, fileReadSchema, caseSensitive,
          partReaderSparkSchema, debugDumpPrefix, debugDumpAlways,
          maxBatchSizeRows, maxBatchSizeBytes, targetBatchSizeBytes, useChunkedReader,
          maxChunkedReaderMemoryUsageSizeBytes,
          metrics,
          DateTimeRebaseCorrected$.MODULE$, // dateRebaseMode
          DateTimeRebaseCorrected$.MODULE$, // timestampRebaseMode
          true, // hasInt96Timestamps
          false // useFieldId
      );
      PartitionReaderWithBytesRead partReader = new PartitionReaderWithBytesRead(parquetPartReader);

      Map updatedConstants = addNullsForMissingFields(idToConstant, reorder.getMissingFields());
      return new GpuIcebergReader(expectedSchema, partReader, deleteFilter, updatedConstants);
    } catch (IOException e) {
      throw new UncheckedIOException("Failed to create/close reader for file: " + input, e);
    }
  }

  public static List filterRowGroups(ParquetFileReader reader,
      NameMapping nameMapping, Schema expectedSchema, Expression filter, boolean caseSensitive) {
    MessageType fileSchema = reader.getFileMetaData().getSchema();

    MessageType typeWithIds;
    if (ParquetSchemaUtil.hasIds(fileSchema)) {
      typeWithIds = fileSchema;
    } else if (nameMapping != null) {
      typeWithIds = ParquetSchemaUtil.applyNameMapping(fileSchema, nameMapping);
    } else {
      typeWithIds = ParquetSchemaUtil.addFallbackIds(fileSchema);
    }

    List rowGroups = reader.getRowGroups();
    List filteredRowGroups = Lists.newArrayListWithCapacity(rowGroups.size());

    if (expectedSchema.findField(MetadataColumns.ROW_POSITION.fieldId()) != null) {
      throw new UnsupportedOperationException("row position meta column not implemented");
    }

    ParquetMetricsRowGroupFilter statsFilter = null;
    ParquetDictionaryRowGroupFilter dictFilter = null;
    if (filter != null) {
      statsFilter = new ParquetMetricsRowGroupFilter(expectedSchema, filter, caseSensitive);
      dictFilter = new ParquetDictionaryRowGroupFilter(expectedSchema, filter, caseSensitive);
    }

    for (BlockMetaData rowGroup : rowGroups) {
      boolean shouldRead = filter == null || (
          statsFilter.shouldRead(typeWithIds, rowGroup) &&
              dictFilter.shouldRead(typeWithIds, rowGroup, reader.getDictionaryReader(rowGroup)));
      if (shouldRead) {
        filteredRowGroups.add(rowGroup);
      }
    }
    return filteredRowGroups;
  }

  public static ParquetFileReader newReader(InputFile file, ParquetReadOptions options) {
    try {
      return ParquetFileReader.open(ParquetIO.file(file), options);
    } catch (IOException e) {
      throw new UncheckedIOException("Failed to open Parquet file: " + file.location(), e);
    }
  }

  public static Map addNullsForMissingFields(Map idToConstant,
                                                          Set missingFields) {
    if (missingFields.isEmpty()) {
      return idToConstant;
    }
    Map updated = Maps.newHashMap(idToConstant);
    for (Integer field : missingFields) {
      updated.put(field, null);
    }
    return updated;
  }

  /** Generate the Spark schema corresponding to a Parquet schema and expected Iceberg schema */
  public static class SparkSchemaConverter extends TypeWithSchemaVisitor {
    @Override
    public DataType message(Types.StructType iStruct, MessageType message, List fields) {
      return struct(iStruct, message, fields);
    }

    @Override
    public DataType struct(Types.StructType iStruct, GroupType struct, List fieldTypes) {
      List parquetFields = struct.getFields();
      List fields = Lists.newArrayListWithExpectedSize(fieldTypes.size());

      for (int i = 0; i < parquetFields.size(); i++) {
        Type parquetField = parquetFields.get(i);

        Preconditions.checkArgument(
            !parquetField.isRepetition(Type.Repetition.REPEATED),
            "Fields cannot have repetition REPEATED: %s", parquetField);

        boolean isNullable = parquetField.isRepetition(Type.Repetition.OPTIONAL);
        StructField field = new StructField(parquetField.getName(), fieldTypes.get(i),
            isNullable, Metadata.empty());
        fields.add(field);
      }

      return new StructType(fields.toArray(new StructField[0]));
    }

    @Override
    public DataType list(Types.ListType iList, GroupType array, DataType elementType) {
      GroupType repeated = array.getType(0).asGroupType();
      Type element = repeated.getType(0);

      Preconditions.checkArgument(
          !element.isRepetition(Type.Repetition.REPEATED),
          "Elements cannot have repetition REPEATED: %s", element);

      boolean isNullable = element.isRepetition(Type.Repetition.OPTIONAL);
      return new ArrayType(elementType, isNullable);
    }

    @Override
    public DataType map(Types.MapType iMap, GroupType map, DataType keyType, DataType valueType) {
      GroupType keyValue = map.getType(0).asGroupType();
      Type value = keyValue.getType(1);

      Preconditions.checkArgument(
          !value.isRepetition(Type.Repetition.REPEATED),
          "Values cannot have repetition REPEATED: %s", value);

      boolean isValueNullable = value.isRepetition(Type.Repetition.OPTIONAL);
      return new MapType(keyType, valueType, isValueNullable);
    }

    @Override
    public DataType primitive(org.apache.iceberg.types.Type.PrimitiveType iPrimitive, PrimitiveType primitiveType) {
      // If up-casts are needed, load as the pre-cast Spark type, and this will be up-cast in GpuIcebergReader.
      switch (iPrimitive.typeId()) {
        case LONG:
          if (primitiveType.getPrimitiveTypeName().equals(PrimitiveType.PrimitiveTypeName.INT32)) {
            return IntegerType$.MODULE$;
          }
          return LongType$.MODULE$;
        case DOUBLE:
          if (primitiveType.getPrimitiveTypeName().equals(PrimitiveType.PrimitiveTypeName.FLOAT)) {
            return FloatType$.MODULE$;
          }
          return DoubleType$.MODULE$;
        case DECIMAL:
          @SuppressWarnings("deprecation")
          org.apache.parquet.schema.DecimalMetadata metadata = primitiveType.getDecimalMetadata();
          return DecimalType$.MODULE$.apply(metadata.getPrecision(), metadata.getScale());
        default:
          return SparkSchemaUtil.convert(iPrimitive);
      }
    }
  }

  public static class ReorderColumns extends TypeWithSchemaVisitor {
    private final Map idToConstant;
    private final Set missingFields = Sets.newHashSet();

    public ReorderColumns(Map idToConstant) {
      this.idToConstant = idToConstant;
    }

    public Set getMissingFields() {
      return missingFields;
    }

    @Override
    public Type message(Types.StructType expected, MessageType message, List fields) {
      MessageTypeBuilder builder = org.apache.parquet.schema.Types.buildMessage();
      List newFields = filterAndReorder(expected, fields);
      for (Type type : newFields) {
        builder.addField(type);
      }
      return builder.named(message.getName());
    }

    @Override
    public Type struct(Types.StructType expected, GroupType struct, List fields) {
      List newFields = filterAndReorder(expected, fields);
      return struct.withNewFields(newFields);
    }

    @Override
    public Type list(Types.ListType expectedList, GroupType list, Type element) {
      if (expectedList != null) {
        boolean hasConstant = expectedList.fields().stream()
          .anyMatch(f -> idToConstant.containsKey(f.fieldId()));
        if (hasConstant) {
          throw new UnsupportedOperationException("constant column in list");
        }
      }
      GroupType repeated = list.getType(0).asGroupType();
      Type originalElement = repeated.getType(0);
      if (Objects.equals(element, originalElement)) {
        return list;
      }
      return list.withNewFields(repeated.withNewFields(element));
    }

    @Override
    public Type map(Types.MapType expectedMap, GroupType map, Type key, Type value) {
      if (expectedMap != null) {
        boolean hasConstant = expectedMap.fields().stream()
          .anyMatch(f -> idToConstant.containsKey(f.fieldId()));
        if (hasConstant) {
          throw new UnsupportedOperationException("constant column in map");
        }
      }
      GroupType repeated = map.getFields().get(0).asGroupType();
      Type originalKey = repeated.getType(0);
      Type originalValue = repeated.getType(0);
      if (Objects.equals(key, originalKey) && Objects.equals(value, originalValue)) {
        return map;
      }
      return map.withNewFields(repeated.withNewFields(key, value));
    }

    @Override
    public Type primitive(org.apache.iceberg.types.Type.PrimitiveType expected, PrimitiveType primitive) {
      return primitive;
    }

    /** Returns true if a column with the specified ID should be ignored when loading the file data */
    private boolean shouldIgnoreFileColumn(int id) {
      return idToConstant.containsKey(id) ||
          id == MetadataColumns.ROW_POSITION.fieldId() &&
          id == MetadataColumns.IS_DELETED.fieldId();
    }

    private List filterAndReorder(Types.StructType expected, List fields) {
      // match the expected struct's order
      Map typesById = Maps.newHashMap();
      for (Type fieldType : fields) {
        if (fieldType.getId() != null) {
          int id = fieldType.getId().intValue();
          typesById.put(id, fieldType);
        }
      }

      List expectedFields = expected != null ?
          expected.fields() : ImmutableList.of();
      List reorderedFields = Lists.newArrayListWithCapacity(expectedFields.size());
      for (Types.NestedField field : expectedFields) {
        int id = field.fieldId();
        if (!shouldIgnoreFileColumn(id)) {
          Type newField = typesById.get(id);
          if (newField != null) {
            reorderedFields.add(newField);
          } else {
            missingFields.add(id);
          }
        }
      }

      return reorderedFields;
    }
  }

  public static class ReorderColumnsFallback extends ReorderColumns {
    public ReorderColumnsFallback(Map idToConstant) {
      super(idToConstant);
    }

    @Override
    public Type message(Types.StructType expected, MessageType message, List fields) {
      // the top level matches by ID, but the remaining IDs are missing
      return super.struct(expected, message, fields);
    }

    @Override
    public Type struct(Types.StructType ignored, GroupType struct, List fields) {
      // the expected struct is ignored because nested fields are never found when the IDs are missing
      return struct;
    }
  }
}