You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@doris.apache.org by GitBox <gi...@apache.org> on 2020/06/04 13:23:50 UTC
[GitHub] [incubator-doris] wyb commented on a change in pull request #3729: (#3728) Using SparkDpp to complete some calculation in Spark Load
wyb commented on a change in pull request #3729:
URL: https://github.com/apache/incubator-doris/pull/3729#discussion_r435246339
##########
File path: fe/src/main/java/org/apache/doris/load/loadv2/dpp/ColumnParser.java
##########
@@ -0,0 +1,175 @@
+// 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.doris.load.loadv2.dpp;
+
+import org.apache.doris.common.UserException;
+import org.joda.time.DateTime;
+
+import java.io.Serializable;
+import java.util.Date;
+
+// Parser to validate value for different type
+public abstract class ColumnParser implements Serializable {
+ public static ColumnParser create(String columnType) throws UserException {
+ if (columnType.equalsIgnoreCase("TINYINT")) {
+ return new TinyIntParser();
+ } else if (columnType.equalsIgnoreCase("SMALLINT")) {
+ return new SmallIntParser();
+ } else if (columnType.equalsIgnoreCase("INT")) {
+ return new IntParser();
+ } else if (columnType.equalsIgnoreCase("BIGINT")) {
+ return new BigIntParser();
+ } else if (columnType.equalsIgnoreCase("FLOAT")) {
+ return new FloatParser();
+ } else if (columnType.equalsIgnoreCase("DOUBLE")) {
+ return new DoubleParser();
+ } else if (columnType.equalsIgnoreCase("BOOLEAN")) {
+ return new BooleanParser();
+ } else if (columnType.equalsIgnoreCase("DATE")) {
+ return new DateParser();
+ } else if (columnType.equalsIgnoreCase("DATETIME")) {
+ return new DatetimeParser();
+ } else if (columnType.equalsIgnoreCase("VARCHAR")
+ || columnType.equalsIgnoreCase("CHAR")
+ || columnType.equalsIgnoreCase("BITMAP")
+ || columnType.equalsIgnoreCase("HLL")) {
+ return new StringParser();
+ } else {
+ throw new UserException("unsupported type:" + columnType);
+ }
+ }
+
+ public abstract boolean parse(String value);
+}
+
+class TinyIntParser extends ColumnParser {
+ @Override
+ public boolean parse(String value) {
+ try {
+ Short parsed = Short.parseShort(value);
+ if (parsed > 127 || parsed < -128) {
+ return false;
+ }
+ } catch (NumberFormatException e) {
+ return false;
+ }
+ return true;
+ }
+}
+
+class SmallIntParser extends ColumnParser {
+ @Override
+ public boolean parse(String value) {
+ try {
+ Short.parseShort(value);
+ } catch (NumberFormatException e) {
+ return false;
+ }
+ return true;
+ }
+}
+
+class IntParser extends ColumnParser {
+ @Override
+ public boolean parse(String value) {
+ try {
+ Integer.parseInt(value);
+ } catch (NumberFormatException e) {
+ return false;
+ }
+ return true;
+ }
+}
+
+class BigIntParser extends ColumnParser {
+ @Override
+ public boolean parse(String value) {
+ try {
+ Integer.parseInt(value);
+ } catch (NumberFormatException e) {
+ return false;
+ }
+ return true;
+ }
+}
+
+class FloatParser extends ColumnParser {
+ @Override
+ public boolean parse(String value) {
+ try {
+ Float.parseFloat(value);
+ } catch (NumberFormatException e) {
+ return false;
+ }
+ return true;
+ }
+}
+
+class DoubleParser extends ColumnParser {
+ @Override
+ public boolean parse(String value) {
+ try {
+ Double.parseDouble(value);
+ } catch (NumberFormatException e) {
+ return false;
+ }
+ return true;
+ }
+}
+
+class BooleanParser extends ColumnParser {
+ @Override
+ public boolean parse(String value) {
+ if (value.equalsIgnoreCase("true")
+ || value.equalsIgnoreCase("false")) {
+ return true;
+ }
+ return false;
+ }
+}
+
+class DateParser extends ColumnParser {
+ @Override
+ public boolean parse(String value) {
+ try {
+ Date.parse(value);
+ } catch (IllegalArgumentException e) {
+ return false;
+ }
+ return true;
+ }
+}
+
+class DatetimeParser extends ColumnParser {
+ @Override
+ public boolean parse(String value) {
+ try {
+ DateTime.parse(value);
+ } catch (IllegalArgumentException e) {
+ return false;
+ }
+ return true;
+ }
+}
+
+class StringParser extends ColumnParser {
+ @Override
+ public boolean parse(String value) {
+ return true;
Review comment:
char and varchar type need to check byte length
##########
File path: fe/src/main/java/org/apache/doris/load/loadv2/dpp/SparkDpp.java
##########
@@ -0,0 +1,851 @@
+// 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.doris.load.loadv2.dpp;
+
+import com.google.common.base.Strings;
+import com.google.gson.Gson;
+import org.apache.commons.lang3.StringUtils;
+import org.apache.doris.common.UserException;
+import org.apache.doris.load.loadv2.etl.EtlJobConfig;
+import org.apache.hadoop.fs.FileStatus;
+import org.apache.hadoop.fs.Path;
+import org.apache.hadoop.conf.Configuration;
+import org.apache.parquet.column.ParquetProperties;
+import org.apache.parquet.hadoop.ParquetWriter;
+import org.apache.parquet.hadoop.metadata.CompressionCodecName;
+import org.apache.spark.Partitioner;
+import org.apache.spark.api.java.function.ForeachPartitionFunction;
+import org.apache.spark.api.java.function.PairFlatMapFunction;
+import org.apache.spark.api.java.function.PairFunction;
+import org.apache.spark.sql.catalyst.InternalRow;
+import org.apache.spark.sql.catalyst.encoders.ExpressionEncoder;
+import org.apache.spark.sql.catalyst.encoders.RowEncoder;
+import org.apache.spark.sql.execution.datasources.parquet.ParquetWriteSupport;
+import org.apache.spark.sql.functions;
+import org.apache.spark.sql.RowFactory;
+import org.apache.spark.sql.Row;
+import org.apache.spark.sql.Dataset;
+import org.apache.spark.sql.Column;
+import org.apache.spark.sql.SparkSession;
+import org.apache.spark.sql.types.DataTypes;
+import org.apache.spark.sql.types.StructField;
+import org.apache.spark.sql.types.StructType;
+import org.apache.spark.util.LongAccumulator;
+import org.apache.spark.api.java.JavaPairRDD;
+import org.apache.spark.api.java.JavaRDD;
+import org.apache.spark.TaskContext;
+
+import scala.Tuple2;
+import scala.collection.Seq;
+
+import java.io.IOException;
+import java.math.BigInteger;
+import java.net.URI;
+import java.net.URISyntaxException;
+import java.util.Map;
+import java.util.Set;
+import java.util.List;
+import java.util.ArrayList;
+import java.util.HashMap;
+import java.util.Iterator;
+import java.util.LinkedList;
+import java.util.HashSet;
+import java.util.Arrays;
+import java.util.stream.Collectors;
+
+import org.apache.hadoop.fs.FileSystem;
+import org.apache.hadoop.fs.FSDataOutputStream;
+import scala.collection.JavaConverters;
+import org.apache.log4j.LogManager;
+import org.apache.log4j.Logger;
+
+// This class is a Spark-based data preprocessing program,
+// which will make use of the distributed compute framework of spark to
+// do ETL job/sort/preaggregate jobs in spark job
+// to boost the process of large amount of data load.
+// the process steps are as following:
+// 1. load data
+// 1.1 load data from path/hive table
+// 1.2 do the etl process
+// 2. repartition data by using doris data model(partition and bucket)
+// 3. process aggregation if needed
+// 4. write data to parquet file
+public final class SparkDpp implements java.io.Serializable {
+ private static final Logger LOG = LogManager.getLogger(SparkDpp.class);
+
+ private static final String NULL_FLAG = "\\N";
+ private static final String DPP_RESULT_FILE = "dpp_result.json";
+ private static final String BITMAP_TYPE = "bitmap";
+ private SparkSession spark = null;
+ private EtlJobConfig etlJobConfig = null;
+ private LongAccumulator abnormalRowAcc = null;
+ private LongAccumulator unselectedRowAcc = null;
+ private LongAccumulator scannedRowsAcc = null;
+ private LongAccumulator fileNumberAcc = null;
+ private LongAccumulator fileSizeAcc = null;
+ private Map<String, Integer> bucketKeyMap = new HashMap<>();
+ // accumulator to collect invalid rows
+ private StringAccumulator invalidRows = new StringAccumulator();
+
+ public SparkDpp(SparkSession spark, EtlJobConfig etlJobConfig) {
+ this.spark = spark;
+ this.etlJobConfig = etlJobConfig;
+ }
+
+ public void init() {
+ abnormalRowAcc = spark.sparkContext().longAccumulator("abnormalRowAcc");
+ unselectedRowAcc = spark.sparkContext().longAccumulator("unselectedRowAcc");
+ scannedRowsAcc = spark.sparkContext().longAccumulator("scannedRowsAcc");
+ fileNumberAcc = spark.sparkContext().longAccumulator("fileNumberAcc");
+ fileSizeAcc = spark.sparkContext().longAccumulator("fileSizeAcc");
+ spark.sparkContext().register(invalidRows, "InvalidRowsAccumulator");
+ }
+
+ private Dataset<Row> processRDDAggAndRepartition(Dataset<Row> dataframe, EtlJobConfig.EtlIndex currentIndexMeta) throws UserException {
+ final boolean isDuplicateTable = !StringUtils.equalsIgnoreCase(currentIndexMeta.indexType, "AGGREGATE")
+ && !StringUtils.equalsIgnoreCase(currentIndexMeta.indexType, "UNIQUE");
+
+ // 1 make metadata for map/reduce
+ int keyLen = 0;
+ for (EtlJobConfig.EtlColumn etlColumn : currentIndexMeta.columns) {
+ keyLen = etlColumn.isKey ? keyLen + 1 : keyLen;
+ }
+
+ SparkRDDAggregator[] sparkRDDAggregators = new SparkRDDAggregator[currentIndexMeta.columns.size() - keyLen];
+
+ for (int i = 0 ; i < currentIndexMeta.columns.size(); i++) {
+ if (!currentIndexMeta.columns.get(i).isKey && !isDuplicateTable) {
+ sparkRDDAggregators[i - keyLen] = SparkRDDAggregator.buildAggregator(currentIndexMeta.columns.get(i));
+ }
+ }
+
+ PairFunction<Row, List<Object>, Object[]> encodePairFunction = isDuplicateTable ?
+ // add 1 to include bucketId
+ new EncodeDuplicateTableFunction(keyLen + 1, currentIndexMeta.columns.size() - keyLen)
+ : new EncodeAggregateTableFunction(sparkRDDAggregators, keyLen + 1);
+
+ // 2 convert dataframe to rdd
+ // TODO(wb) use rdd to avoid bitamp/hll serialize
+ JavaPairRDD<List<Object>, Object[]> currentRollupRDD = dataframe.toJavaRDD().mapToPair(encodePairFunction);
+
+ // 3 do aggregate
+ // TODO(wb) set the reduce concurrency by statistic instead of hard code 200
+ int aggregateConcurrency = 200;
+ JavaPairRDD<List<Object>, Object[]> reduceResultRDD = isDuplicateTable ? currentRollupRDD
+ : currentRollupRDD.reduceByKey(new AggregateReduceFunction(sparkRDDAggregators), aggregateConcurrency);
+
+ // 4 repartition
+ JavaRDD<Row> finalRDD = reduceResultRDD
+ .repartitionAndSortWithinPartitions(new BucketPartitioner(bucketKeyMap), new BucketComparator())
+ .map(record -> {
+ List<Object> keys = record._1;
+ Object[] values = record._2;
+ int size = keys.size() + values.length;
+ Object[] result = new Object[size];
+
+ for (int i = 0; i < keys.size(); i++) {
+ result[i] = keys.get(i);
+ }
+
+ for (int i = keys.size(); i < size; i++) {
+ int valueIdx = i - keys.size();
+ result[i] = isDuplicateTable ? values[valueIdx] : sparkRDDAggregators[valueIdx].finalize(values[valueIdx]);
+ }
+
+ return RowFactory.create(result);
+ });
+
+ // 4 convert to dataframe
+ StructType tableSchemaWithBucketId = DppUtils.createDstTableSchema(currentIndexMeta.columns, true, true);
+ dataframe = spark.createDataFrame(finalRDD, tableSchemaWithBucketId);
+ return dataframe;
+
+ }
+
+ // write data to parquet file by using writing the parquet scheme of spark.
+ private void writePartitionedAndSortedDataframeToParquet(Dataset<Row> dataframe,
+ String pathPattern,
+ long tableId,
+ EtlJobConfig.EtlIndex indexMeta) throws UserException {
+ StructType outputSchema = dataframe.schema();
+ StructType dstSchema = DataTypes.createStructType(
+ Arrays.asList(outputSchema.fields()).stream()
+ .filter(field -> !field.name().equalsIgnoreCase(DppUtils.BUCKET_ID))
+ .collect(Collectors.toList()));
+ ExpressionEncoder encoder = RowEncoder.apply(dstSchema);
+ dataframe.foreachPartition(new ForeachPartitionFunction<Row>() {
+ @Override
+ public void call(Iterator<Row> t) throws Exception {
+ // write the data to dst file
+ Configuration conf = new Configuration();
+ FileSystem fs = FileSystem.get(URI.create(etlJobConfig.outputPath), conf);
+ String lastBucketKey = null;
+ ParquetWriter<InternalRow> parquetWriter = null;
+ TaskContext taskContext = TaskContext.get();
+ long taskAttemptId = taskContext.taskAttemptId();
+ String dstPath = "";
+ String tmpPath = "";
+
+ while (t.hasNext()) {
+ Row row = t.next();
+ if (row.length() <= 1) {
+ LOG.warn("invalid row:" + row);
+ continue;
+ }
+
+
+ String curBucketKey = row.getString(0);
+ List<Object> columnObjects = new ArrayList<>();
+ for (int i = 1; i < row.length(); ++i) {
+ Object columnValue = row.get(i);
+ columnObjects.add(columnValue);
+ }
+ System.out.println();
+ Row rowWithoutBucketKey = RowFactory.create(columnObjects.toArray());
+ // if the bucket key is new, it will belong to a new tablet
+ if (lastBucketKey == null || !curBucketKey.equals(lastBucketKey)) {
+ if (parquetWriter != null) {
+ parquetWriter.close();
+ // rename tmpPath to path
+ try {
+ fs.rename(new Path(tmpPath), new Path(dstPath));
+ } catch (IOException ioe) {
+ LOG.warn("rename from tmpPath" + tmpPath + " to dstPath:" + dstPath + " failed. exception:" + ioe);
+ throw ioe;
+ }
+ }
+ // flush current writer and create a new writer
+ String[] bucketKey = curBucketKey.split("_");
+ if (bucketKey.length != 2) {
+ LOG.warn("invalid bucket key:" + curBucketKey);
+ continue;
+ }
+ int partitionId = Integer.parseInt(bucketKey[0]);
+ int bucketId = Integer.parseInt(bucketKey[1]);
+ dstPath = String.format(pathPattern, tableId, partitionId, indexMeta.indexId,
+ bucketId, indexMeta.schemaHash);
+ tmpPath = dstPath + "." + taskAttemptId;
+ conf.setBoolean("spark.sql.parquet.writeLegacyFormat", false);
+ conf.setBoolean("spark.sql.parquet.int64AsTimestampMillis", false);
+ conf.setBoolean("spark.sql.parquet.int96AsTimestamp", true);
+ conf.setBoolean("spark.sql.parquet.binaryAsString", false);
+ conf.set("spark.sql.parquet.outputTimestampType", "INT96");
+ ParquetWriteSupport.setSchema(dstSchema, conf);
+ ParquetWriteSupport parquetWriteSupport = new ParquetWriteSupport();
+ parquetWriter = new ParquetWriter<InternalRow>(new Path(tmpPath), parquetWriteSupport,
+ CompressionCodecName.SNAPPY, 256 * 1024 * 1024, 16 * 1024, 1024 * 1024,
+ true, false,
+ ParquetProperties.WriterVersion.PARQUET_1_0, conf);
+ if (parquetWriter != null) {
+ LOG.info("[HdfsOperate]>> initialize writer succeed! path:" + tmpPath);
+ }
+ lastBucketKey = curBucketKey;
+ }
+ InternalRow internalRow = encoder.toRow(rowWithoutBucketKey);
+ parquetWriter.write(internalRow);
+ }
+ if (parquetWriter != null) {
+ parquetWriter.close();
+ try {
+ fs.rename(new Path(tmpPath), new Path(dstPath));
+ } catch (IOException ioe) {
+ LOG.warn("rename from tmpPath" + tmpPath + " to dstPath:" + dstPath + " failed. exception:" + ioe);
+ throw ioe;
+ }
+ }
+ }
+ });
+ }
+
+ private void processRollupTree(RollupTreeNode rootNode,
+ Dataset<Row> rootDataframe,
+ long tableId, EtlJobConfig.EtlTable tableMeta,
+ EtlJobConfig.EtlIndex baseIndex) throws UserException {
+ Queue<RollupTreeNode> nodeQueue = new LinkedList<>();
+ nodeQueue.offer(rootNode);
+ int currentLevel = 0;
+ // level travel the tree
+ Map<Long, Dataset<Row>> parentDataframeMap = new HashMap<>();
+ parentDataframeMap.put(baseIndex.indexId, rootDataframe);
+ Map<Long, Dataset<Row>> childrenDataframeMap = new HashMap<>();
+ String pathPattern = etlJobConfig.outputPath + "/" + etlJobConfig.outputFilePattern;
+ while (!nodeQueue.isEmpty()) {
+ RollupTreeNode curNode = nodeQueue.poll();
+ LOG.info("start to process index:" + curNode.indexId);
+ if (curNode.children != null) {
+ for (RollupTreeNode child : curNode.children) {
+ nodeQueue.offer(child);
+ }
+ }
+ Dataset<Row> curDataFrame = null;
+ // column select for rollup
+ if (curNode.level != currentLevel) {
+ for (Dataset<Row> dataframe : parentDataframeMap.values()) {
+ dataframe.unpersist();
+ }
+ currentLevel = curNode.level;
+ parentDataframeMap.clear();
+ parentDataframeMap = childrenDataframeMap;
+ childrenDataframeMap = new HashMap<>();
+ }
+
+ long parentIndexId = baseIndex.indexId;
+ if (curNode.parent != null) {
+ parentIndexId = curNode.parent.indexId;
+ }
+
+ Dataset<Row> parentDataframe = parentDataframeMap.get(parentIndexId);
+ List<Column> columns = new ArrayList<>();
+ List<Column> keyColumns = new ArrayList<>();
+ Column bucketIdColumn = new Column(DppUtils.BUCKET_ID);
+ keyColumns.add(bucketIdColumn);
+ columns.add(bucketIdColumn);
+ for (String keyName : curNode.keyColumnNames) {
+ columns.add(new Column(keyName));
+ keyColumns.add(new Column(keyName));
+ }
+ for (String valueName : curNode.valueColumnNames) {
+ columns.add(new Column(valueName));
+ }
+ Seq<Column> columnSeq = JavaConverters.asScalaIteratorConverter(columns.iterator()).asScala().toSeq();
+ curDataFrame = parentDataframe.select(columnSeq);
+ // aggregate and repartition
+ curDataFrame = processRDDAggAndRepartition(curDataFrame, curNode.indexMeta);
+
+ childrenDataframeMap.put(curNode.indexId, curDataFrame);
+
+ if (curNode.children != null && curNode.children.size() > 1) {
+ // if the children number larger than 1, persist the dataframe for performance
+ curDataFrame.persist();
+ }
+ writePartitionedAndSortedDataframeToParquet(curDataFrame, pathPattern, tableId, curNode.indexMeta);
+ }
+ }
+
+ // repartition dataframe by partitionid_bucketid
+ // so data in the same bucket will be consecutive.
+ private Dataset<Row> repartitionDataframeByBucketId(SparkSession spark, Dataset<Row> dataframe,
+ EtlJobConfig.EtlPartitionInfo partitionInfo,
+ List<Integer> partitionKeyIndex,
+ List<Class> partitionKeySchema,
+ List<DorisRangePartitioner.PartitionRangeKey> partitionRangeKeys,
+ List<String> keyColumnNames,
+ List<String> valueColumnNames,
+ StructType dstTableSchema,
+ EtlJobConfig.EtlIndex baseIndex,
+ List<Long> validPartitionIds) throws UserException {
+ List<String> distributeColumns = partitionInfo.distributionColumnRefs;
+ Partitioner partitioner = new DorisRangePartitioner(partitionInfo, partitionKeyIndex, partitionRangeKeys);
+ Set<Integer> validPartitionIndex = new HashSet<>();
+ if (validPartitionIds == null) {
+ for (int i = 0; i < partitionInfo.partitions.size(); ++i) {
+ validPartitionIndex.add(i);
+ }
+ } else {
+ for (int i = 0; i < partitionInfo.partitions.size(); ++i) {
+ if (validPartitionIds.contains(partitionInfo.partitions.get(i).partitionId)) {
+ validPartitionIndex.add(i);
+ }
+ }
+ }
+ // use PairFlatMapFunction instead of PairMapFunction because the there will be
+ // 0 or 1 output row for 1 input row
+ JavaPairRDD<String, DppColumns> pairRDD = dataframe.javaRDD().flatMapToPair(
+ new PairFlatMapFunction<Row, String, DppColumns>() {
+ @Override
+ public Iterator<Tuple2<String, DppColumns>> call(Row row) {
+ List<Object> columns = new ArrayList<>();
+ List<Object> keyColumns = new ArrayList<>();
+ for (String columnName : keyColumnNames) {
+ Object columnObject = row.get(row.fieldIndex(columnName));
+ columns.add(columnObject);
+ keyColumns.add(columnObject);
+ }
+
+ for (String columnName : valueColumnNames) {
+ columns.add(row.get(row.fieldIndex(columnName)));
+ }
+ DppColumns dppColumns = new DppColumns(columns);
+ DppColumns key = new DppColumns(keyColumns);
+ List<Tuple2<String, DppColumns>> result = new ArrayList<>();
+ int pid = partitioner.getPartition(key);
+ if (!validPartitionIndex.contains(pid)) {
+ LOG.warn("invalid partition for row:" + row + ", pid:" + pid);
+ abnormalRowAcc.add(1);
+ LOG.info("abnormalRowAcc:" + abnormalRowAcc);
+ if (abnormalRowAcc.value() < 5) {
+ LOG.info("add row to invalidRows:" + row.toString());
+ invalidRows.add(row.toString());
+ LOG.info("invalid rows contents:" + invalidRows.value());
+ }
+ } else {
+ long hashValue = DppUtils.getHashValue(row, distributeColumns, dstTableSchema);
+ int bucketId = (int) ((hashValue & 0xffffffff) % partitionInfo.partitions.get(pid).bucketNum);
+ long partitionId = partitionInfo.partitions.get(pid).partitionId;
+ // bucketKey is partitionId_bucketId
+ String bucketKey = partitionId + "_" + bucketId;
+ Tuple2<String, DppColumns> newTuple = new Tuple2<String, DppColumns>(bucketKey, dppColumns);
+ result.add(newTuple);
+ }
+ return result.iterator();
+ }
+ });
+
+ JavaRDD<Row> resultRdd = pairRDD.map(record -> {
+ String bucketKey = record._1;
+ List<Object> row = new ArrayList<>();
+ // bucketKey as the first key
+ row.add(bucketKey);
+ row.addAll(record._2.columns);
+ return RowFactory.create(row.toArray());
+ }
+ );
+
+ // TODO(wb): using rdd instead of dataframe from here
+ StructType tableSchemaWithBucketId = DppUtils.createDstTableSchema(baseIndex.columns, true, false);
+ dataframe = spark.createDataFrame(resultRdd, tableSchemaWithBucketId);
+ // use bucket number as the parallel number
+ int reduceNum = 0;
+ for (EtlJobConfig.EtlPartition partition : partitionInfo.partitions) {
+ for (int i = 0; i < partition.bucketNum; i++) {
+ bucketKeyMap.put(partition.partitionId + "_" + i, reduceNum);
+ reduceNum++;
+ }
+ }
+
+ // print to system.out for easy to find log info
+ System.out.println("print bucket key map:" + bucketKeyMap.toString());
+
+ return dataframe;
+ }
+
+ // do the etl process
+ private Dataset<Row> convertSrcDataframeToDstDataframe(EtlJobConfig.EtlIndex baseIndex,
+ Dataset<Row> srcDataframe,
+ StructType dstTableSchema,
+ EtlJobConfig.EtlFileGroup fileGroup) throws UserException {
+ Dataset<Row> dataframe = srcDataframe;
+ StructType srcSchema = dataframe.schema();
+ Set<String> srcColumnNames = new HashSet<>();
+ for (StructField field : srcSchema.fields()) {
+ srcColumnNames.add(field.name());
+ }
+ Map<String, EtlJobConfig.EtlColumnMapping> columnMappings = fileGroup.columnMappings;
+ // 1. process simple columns
+ Set<String> mappingColumns = null;
+ if (columnMappings != null) {
+ mappingColumns = columnMappings.keySet();
+ }
+ List<String> dstColumnNames = new ArrayList<>();
+ for (StructField dstField : dstTableSchema.fields()) {
+ dstColumnNames.add(dstField.name());
+ EtlJobConfig.EtlColumn column = baseIndex.getColumn(dstField.name());
+ if (!srcColumnNames.contains(dstField.name())) {
+ if (mappingColumns != null && mappingColumns.contains(dstField.name())) {
+ // mapping columns will be processed in next step
+ continue;
+ }
+ if (column.defaultValue != null) {
+ if (column.defaultValue.equals(NULL_FLAG)) {
+ dataframe = dataframe.withColumn(dstField.name(), functions.lit(null));
+ } else {
+ dataframe = dataframe.withColumn(dstField.name(), functions.lit(column.defaultValue));
+ }
+ } else if (column.isAllowNull) {
+ dataframe = dataframe.withColumn(dstField.name(), functions.lit(null));
+ } else {
+ throw new UserException("Reason: no data for column:" + dstField.name());
+ }
+ }
+ if (column.columnType.equalsIgnoreCase("DATE")) {
+ dataframe = dataframe.withColumn(dstField.name(), dataframe.col(dstField.name()).cast("date"));
+ } else if (column.columnType.equalsIgnoreCase("BOOLEAN")) {
+ dataframe = dataframe.withColumn(dstField.name(),
+ functions.when(dataframe.col(dstField.name()).equalTo("true"), "1")
+ .otherwise("0"));
+ } else if (!column.columnType.equalsIgnoreCase(BITMAP_TYPE) && !dstField.dataType().equals(DataTypes.StringType)) {
+ dataframe = dataframe.withColumn(dstField.name(), dataframe.col(dstField.name()).cast(dstField.dataType()));
+ }
+ if (fileGroup.isNegative && !column.isKey) {
+ // negative load
+ // value will be convert te -1 * value
+ dataframe = dataframe.withColumn(dstField.name(), functions.expr("-1 *" + dstField.name()));
+ }
+ }
+ // 2. process the mapping columns
+ for (String mappingColumn : mappingColumns) {
+ String mappingDescription = columnMappings.get(mappingColumn).toDescription();
+ if (mappingDescription.toLowerCase().contains("hll_hash")) {
+ continue;
+ }
+ // here should cast data type to dst column type
+ dataframe = dataframe.withColumn(mappingColumn,
+ functions.expr(mappingDescription).cast(dstTableSchema.apply(mappingColumn).dataType()));
+ }
+ // projection and reorder the columns
+ dataframe.createOrReplaceTempView("src_table");
+ StringBuilder selectSqlBuilder = new StringBuilder();
+ selectSqlBuilder.append("select ");
+ for (String name : dstColumnNames) {
+ selectSqlBuilder.append(name + ",");
+ }
+ selectSqlBuilder.deleteCharAt(selectSqlBuilder.length() - 1);
+ selectSqlBuilder.append(" from src_table");
+ String selectSql = selectSqlBuilder.toString();
+ dataframe = spark.sql(selectSql);
+ return dataframe;
+ }
+
+ private Dataset<Row> loadDataFromPath(SparkSession spark,
+ EtlJobConfig.EtlFileGroup fileGroup,
+ String fileUrl,
+ EtlJobConfig.EtlIndex baseIndex,
+ List<EtlJobConfig.EtlColumn> columns) throws UserException {
+ List<String> columnValueFromPath = DppUtils.parseColumnsFromPath(fileUrl, fileGroup.columnsFromPath);
+ List<String> dataSrcColumns = fileGroup.fileFieldNames;
+ if (dataSrcColumns == null) {
+ // if there is no source columns info
+ // use base index columns as source columns
+ dataSrcColumns = new ArrayList<>();
+ for (EtlJobConfig.EtlColumn column : baseIndex.columns) {
+ dataSrcColumns.add(column.columnName);
+ }
+ }
+ List<String> dstTableNames = new ArrayList<>();
+ for (EtlJobConfig.EtlColumn column : baseIndex.columns) {
+ dstTableNames.add(column.columnName);
+ }
+ List<String> srcColumnsWithColumnsFromPath = new ArrayList<>();
+ srcColumnsWithColumnsFromPath.addAll(dataSrcColumns);
+ if (fileGroup.columnsFromPath != null) {
+ srcColumnsWithColumnsFromPath.addAll(fileGroup.columnsFromPath);
+ }
+ StructType srcSchema = createScrSchema(srcColumnsWithColumnsFromPath);
+ JavaRDD<String> sourceDataRdd = spark.read().textFile(fileUrl).toJavaRDD();
+ int columnSize = dataSrcColumns.size();
+ List<ColumnParser> parsers = new ArrayList<>();
+ for (EtlJobConfig.EtlColumn column : baseIndex.columns) {
+ parsers.add(ColumnParser.create(column.columnType));
+ }
+ // now we first support csv file
+ // TODO: support parquet file and orc file
+ JavaRDD<Row> rowRDD = sourceDataRdd.flatMap(
+ record -> {
+ scannedRowsAcc.add(1);
+ String[] attributes = record.split(fileGroup.columnSeparator);
+ List<Row> result = new ArrayList<>();
+ boolean validRow = true;
+ if (attributes.length != columnSize) {
+ LOG.warn("invalid src schema, data columns:"
+ + attributes.length + ", file group columns:"
+ + columnSize + ", row:" + record);
+ validRow = false;
+ } else {
+ for (int i = 0; i < attributes.length; ++i) {
+ if (attributes[i].equals(NULL_FLAG)) {
+ if (baseIndex.columns.get(i).isAllowNull) {
+ attributes[i] = null;
+ } else {
+ LOG.warn("colunm:" + i + " can not be null. row:" + record);
+ validRow = false;
+ break;
+ }
+ }
+ boolean isStrictMode = (boolean) etlJobConfig.properties.strictMode;
+ if (isStrictMode) {
+ StructField field = srcSchema.apply(i);
+ if (dstTableNames.contains(field.name())) {
+ String type = columns.get(i).columnType;
+ if (type.equalsIgnoreCase("CHAR")
+ || type.equalsIgnoreCase("VARCHAR")) {
+ continue;
+ }
+ ColumnParser parser = parsers.get(i);
+ boolean valid = parser.parse(attributes[i]);
+ if (!valid) {
+ validRow = false;
+ LOG.warn("invalid row:" + record
+ + ", attribute " + i + ": " + attributes[i] + " parsed failed");
+ break;
+ }
+ }
+ }
+ }
+ }
+ if (validRow) {
+ Row row = null;
+ if (fileGroup.columnsFromPath == null) {
+ row = RowFactory.create(attributes);
+ } else {
+ // process columns from path
+ // append columns from path to the tail
+ List<String> columnAttributes = new ArrayList<>();
+ columnAttributes.addAll(Arrays.asList(attributes));
+ columnAttributes.addAll(columnValueFromPath);
+ row = RowFactory.create(columnAttributes.toArray());
+ }
+ result.add(row);
+ } else {
+ abnormalRowAcc.add(1);
+ // at most add 5 rows to invalidRows
+ if (abnormalRowAcc.value() <= 5) {
+ invalidRows.add(record);
+ }
+ }
+ return result.iterator();
+ }
+ );
+
+ Dataset<Row> dataframe = spark.createDataFrame(rowRDD, srcSchema);
+ return dataframe;
+ }
+
+ private StructType createScrSchema(List<String> srcColumns) {
+ List<StructField> fields = new ArrayList<>();
+ for (String srcColumn : srcColumns) {
+ // user StringType to load source data
+ StructField field = DataTypes.createStructField(srcColumn, DataTypes.StringType, true);
+ fields.add(field);
+ }
+ StructType srcSchema = DataTypes.createStructType(fields);
+ return srcSchema;
+ }
+
+ // partition keys will be parsed into double from json
+ // so need to convert it to partition columns' type
+ private Object convertPartitionKey(Object srcValue, Class dstClass) throws UserException {
+ if (dstClass.equals(Float.class) || dstClass.equals(Double.class)) {
+ return null;
+ }
+ if (srcValue instanceof Double) {
+ if (dstClass.equals(Short.class)) {
+ return ((Double) srcValue).shortValue();
+ } else if (dstClass.equals(Integer.class)) {
+ return ((Double) srcValue).intValue();
+ } else if (dstClass.equals(Long.class)) {
+ return ((Double) srcValue).longValue();
+ } else if (dstClass.equals(BigInteger.class)) {
+ return new BigInteger(((Double) srcValue).toString());
+ } else {
+ // dst type is string
+ return srcValue.toString();
+ }
+ } else {
+ LOG.warn("unsupport partition key:" + srcValue);
+ throw new UserException("unsupport partition key:" + srcValue);
+ }
+ }
+
+ private List<DorisRangePartitioner.PartitionRangeKey> createPartitionRangeKeys(
+ EtlJobConfig.EtlPartitionInfo partitionInfo, List<Class> partitionKeySchema) throws UserException {
+ List<DorisRangePartitioner.PartitionRangeKey> partitionRangeKeys = new ArrayList<>();
+ for (EtlJobConfig.EtlPartition partition : partitionInfo.partitions) {
+ DorisRangePartitioner.PartitionRangeKey partitionRangeKey = new DorisRangePartitioner.PartitionRangeKey();
+ List<Object> startKeyColumns = new ArrayList<>();
+ for (int i = 0; i < partition.startKeys.size(); i++) {
+ Object value = partition.startKeys.get(i);
+ startKeyColumns.add(convertPartitionKey(value, partitionKeySchema.get(i)));
+ }
+ partitionRangeKey.startKeys = new DppColumns(startKeyColumns);
+ ;
Review comment:
Remove the semicolon
##########
File path: fe/src/main/java/org/apache/doris/load/loadv2/dpp/ColumnParser.java
##########
@@ -0,0 +1,175 @@
+// 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.doris.load.loadv2.dpp;
+
+import org.apache.doris.common.UserException;
+import org.joda.time.DateTime;
+
+import java.io.Serializable;
+import java.util.Date;
+
+// Parser to validate value for different type
+public abstract class ColumnParser implements Serializable {
+ public static ColumnParser create(String columnType) throws UserException {
+ if (columnType.equalsIgnoreCase("TINYINT")) {
+ return new TinyIntParser();
+ } else if (columnType.equalsIgnoreCase("SMALLINT")) {
+ return new SmallIntParser();
+ } else if (columnType.equalsIgnoreCase("INT")) {
+ return new IntParser();
+ } else if (columnType.equalsIgnoreCase("BIGINT")) {
+ return new BigIntParser();
+ } else if (columnType.equalsIgnoreCase("FLOAT")) {
+ return new FloatParser();
+ } else if (columnType.equalsIgnoreCase("DOUBLE")) {
+ return new DoubleParser();
+ } else if (columnType.equalsIgnoreCase("BOOLEAN")) {
+ return new BooleanParser();
+ } else if (columnType.equalsIgnoreCase("DATE")) {
+ return new DateParser();
+ } else if (columnType.equalsIgnoreCase("DATETIME")) {
+ return new DatetimeParser();
+ } else if (columnType.equalsIgnoreCase("VARCHAR")
+ || columnType.equalsIgnoreCase("CHAR")
+ || columnType.equalsIgnoreCase("BITMAP")
+ || columnType.equalsIgnoreCase("HLL")) {
+ return new StringParser();
+ } else {
+ throw new UserException("unsupported type:" + columnType);
Review comment:
Decimal?
##########
File path: fe/src/main/java/org/apache/doris/load/loadv2/dpp/SparkDpp.java
##########
@@ -0,0 +1,851 @@
+// 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.doris.load.loadv2.dpp;
+
+import com.google.common.base.Strings;
+import com.google.gson.Gson;
+import org.apache.commons.lang3.StringUtils;
+import org.apache.doris.common.UserException;
+import org.apache.doris.load.loadv2.etl.EtlJobConfig;
+import org.apache.hadoop.fs.FileStatus;
+import org.apache.hadoop.fs.Path;
+import org.apache.hadoop.conf.Configuration;
+import org.apache.parquet.column.ParquetProperties;
+import org.apache.parquet.hadoop.ParquetWriter;
+import org.apache.parquet.hadoop.metadata.CompressionCodecName;
+import org.apache.spark.Partitioner;
+import org.apache.spark.api.java.function.ForeachPartitionFunction;
+import org.apache.spark.api.java.function.PairFlatMapFunction;
+import org.apache.spark.api.java.function.PairFunction;
+import org.apache.spark.sql.catalyst.InternalRow;
+import org.apache.spark.sql.catalyst.encoders.ExpressionEncoder;
+import org.apache.spark.sql.catalyst.encoders.RowEncoder;
+import org.apache.spark.sql.execution.datasources.parquet.ParquetWriteSupport;
+import org.apache.spark.sql.functions;
+import org.apache.spark.sql.RowFactory;
+import org.apache.spark.sql.Row;
+import org.apache.spark.sql.Dataset;
+import org.apache.spark.sql.Column;
+import org.apache.spark.sql.SparkSession;
+import org.apache.spark.sql.types.DataTypes;
+import org.apache.spark.sql.types.StructField;
+import org.apache.spark.sql.types.StructType;
+import org.apache.spark.util.LongAccumulator;
+import org.apache.spark.api.java.JavaPairRDD;
+import org.apache.spark.api.java.JavaRDD;
+import org.apache.spark.TaskContext;
+
+import scala.Tuple2;
+import scala.collection.Seq;
+
+import java.io.IOException;
+import java.math.BigInteger;
+import java.net.URI;
+import java.net.URISyntaxException;
+import java.util.Map;
+import java.util.Set;
+import java.util.List;
+import java.util.ArrayList;
+import java.util.HashMap;
+import java.util.Iterator;
+import java.util.LinkedList;
+import java.util.HashSet;
+import java.util.Arrays;
+import java.util.stream.Collectors;
+
+import org.apache.hadoop.fs.FileSystem;
+import org.apache.hadoop.fs.FSDataOutputStream;
+import scala.collection.JavaConverters;
+import org.apache.log4j.LogManager;
+import org.apache.log4j.Logger;
+
+// This class is a Spark-based data preprocessing program,
+// which will make use of the distributed compute framework of spark to
+// do ETL job/sort/preaggregate jobs in spark job
+// to boost the process of large amount of data load.
+// the process steps are as following:
+// 1. load data
+// 1.1 load data from path/hive table
+// 1.2 do the etl process
+// 2. repartition data by using doris data model(partition and bucket)
+// 3. process aggregation if needed
+// 4. write data to parquet file
+public final class SparkDpp implements java.io.Serializable {
+ private static final Logger LOG = LogManager.getLogger(SparkDpp.class);
+
+ private static final String NULL_FLAG = "\\N";
+ private static final String DPP_RESULT_FILE = "dpp_result.json";
+ private static final String BITMAP_TYPE = "bitmap";
+ private SparkSession spark = null;
+ private EtlJobConfig etlJobConfig = null;
+ private LongAccumulator abnormalRowAcc = null;
+ private LongAccumulator unselectedRowAcc = null;
+ private LongAccumulator scannedRowsAcc = null;
+ private LongAccumulator fileNumberAcc = null;
+ private LongAccumulator fileSizeAcc = null;
+ private Map<String, Integer> bucketKeyMap = new HashMap<>();
+ // accumulator to collect invalid rows
+ private StringAccumulator invalidRows = new StringAccumulator();
+
+ public SparkDpp(SparkSession spark, EtlJobConfig etlJobConfig) {
+ this.spark = spark;
+ this.etlJobConfig = etlJobConfig;
+ }
+
+ public void init() {
+ abnormalRowAcc = spark.sparkContext().longAccumulator("abnormalRowAcc");
+ unselectedRowAcc = spark.sparkContext().longAccumulator("unselectedRowAcc");
+ scannedRowsAcc = spark.sparkContext().longAccumulator("scannedRowsAcc");
+ fileNumberAcc = spark.sparkContext().longAccumulator("fileNumberAcc");
+ fileSizeAcc = spark.sparkContext().longAccumulator("fileSizeAcc");
+ spark.sparkContext().register(invalidRows, "InvalidRowsAccumulator");
+ }
+
+ private Dataset<Row> processRDDAggAndRepartition(Dataset<Row> dataframe, EtlJobConfig.EtlIndex currentIndexMeta) throws UserException {
+ final boolean isDuplicateTable = !StringUtils.equalsIgnoreCase(currentIndexMeta.indexType, "AGGREGATE")
+ && !StringUtils.equalsIgnoreCase(currentIndexMeta.indexType, "UNIQUE");
+
+ // 1 make metadata for map/reduce
+ int keyLen = 0;
+ for (EtlJobConfig.EtlColumn etlColumn : currentIndexMeta.columns) {
+ keyLen = etlColumn.isKey ? keyLen + 1 : keyLen;
+ }
+
+ SparkRDDAggregator[] sparkRDDAggregators = new SparkRDDAggregator[currentIndexMeta.columns.size() - keyLen];
+
+ for (int i = 0 ; i < currentIndexMeta.columns.size(); i++) {
+ if (!currentIndexMeta.columns.get(i).isKey && !isDuplicateTable) {
+ sparkRDDAggregators[i - keyLen] = SparkRDDAggregator.buildAggregator(currentIndexMeta.columns.get(i));
+ }
+ }
+
+ PairFunction<Row, List<Object>, Object[]> encodePairFunction = isDuplicateTable ?
+ // add 1 to include bucketId
+ new EncodeDuplicateTableFunction(keyLen + 1, currentIndexMeta.columns.size() - keyLen)
+ : new EncodeAggregateTableFunction(sparkRDDAggregators, keyLen + 1);
+
+ // 2 convert dataframe to rdd
+ // TODO(wb) use rdd to avoid bitamp/hll serialize
+ JavaPairRDD<List<Object>, Object[]> currentRollupRDD = dataframe.toJavaRDD().mapToPair(encodePairFunction);
+
+ // 3 do aggregate
+ // TODO(wb) set the reduce concurrency by statistic instead of hard code 200
+ int aggregateConcurrency = 200;
+ JavaPairRDD<List<Object>, Object[]> reduceResultRDD = isDuplicateTable ? currentRollupRDD
+ : currentRollupRDD.reduceByKey(new AggregateReduceFunction(sparkRDDAggregators), aggregateConcurrency);
+
+ // 4 repartition
+ JavaRDD<Row> finalRDD = reduceResultRDD
+ .repartitionAndSortWithinPartitions(new BucketPartitioner(bucketKeyMap), new BucketComparator())
+ .map(record -> {
+ List<Object> keys = record._1;
+ Object[] values = record._2;
+ int size = keys.size() + values.length;
+ Object[] result = new Object[size];
+
+ for (int i = 0; i < keys.size(); i++) {
+ result[i] = keys.get(i);
+ }
+
+ for (int i = keys.size(); i < size; i++) {
+ int valueIdx = i - keys.size();
+ result[i] = isDuplicateTable ? values[valueIdx] : sparkRDDAggregators[valueIdx].finalize(values[valueIdx]);
+ }
+
+ return RowFactory.create(result);
+ });
+
+ // 4 convert to dataframe
+ StructType tableSchemaWithBucketId = DppUtils.createDstTableSchema(currentIndexMeta.columns, true, true);
+ dataframe = spark.createDataFrame(finalRDD, tableSchemaWithBucketId);
+ return dataframe;
+
+ }
+
+ // write data to parquet file by using writing the parquet scheme of spark.
+ private void writePartitionedAndSortedDataframeToParquet(Dataset<Row> dataframe,
+ String pathPattern,
+ long tableId,
+ EtlJobConfig.EtlIndex indexMeta) throws UserException {
+ StructType outputSchema = dataframe.schema();
+ StructType dstSchema = DataTypes.createStructType(
+ Arrays.asList(outputSchema.fields()).stream()
+ .filter(field -> !field.name().equalsIgnoreCase(DppUtils.BUCKET_ID))
+ .collect(Collectors.toList()));
+ ExpressionEncoder encoder = RowEncoder.apply(dstSchema);
+ dataframe.foreachPartition(new ForeachPartitionFunction<Row>() {
+ @Override
+ public void call(Iterator<Row> t) throws Exception {
+ // write the data to dst file
+ Configuration conf = new Configuration();
+ FileSystem fs = FileSystem.get(URI.create(etlJobConfig.outputPath), conf);
+ String lastBucketKey = null;
+ ParquetWriter<InternalRow> parquetWriter = null;
+ TaskContext taskContext = TaskContext.get();
+ long taskAttemptId = taskContext.taskAttemptId();
+ String dstPath = "";
+ String tmpPath = "";
+
+ while (t.hasNext()) {
+ Row row = t.next();
+ if (row.length() <= 1) {
+ LOG.warn("invalid row:" + row);
+ continue;
+ }
+
+
+ String curBucketKey = row.getString(0);
+ List<Object> columnObjects = new ArrayList<>();
+ for (int i = 1; i < row.length(); ++i) {
+ Object columnValue = row.get(i);
+ columnObjects.add(columnValue);
+ }
+ System.out.println();
+ Row rowWithoutBucketKey = RowFactory.create(columnObjects.toArray());
+ // if the bucket key is new, it will belong to a new tablet
+ if (lastBucketKey == null || !curBucketKey.equals(lastBucketKey)) {
+ if (parquetWriter != null) {
+ parquetWriter.close();
+ // rename tmpPath to path
+ try {
+ fs.rename(new Path(tmpPath), new Path(dstPath));
+ } catch (IOException ioe) {
+ LOG.warn("rename from tmpPath" + tmpPath + " to dstPath:" + dstPath + " failed. exception:" + ioe);
+ throw ioe;
+ }
+ }
+ // flush current writer and create a new writer
+ String[] bucketKey = curBucketKey.split("_");
+ if (bucketKey.length != 2) {
+ LOG.warn("invalid bucket key:" + curBucketKey);
+ continue;
+ }
+ int partitionId = Integer.parseInt(bucketKey[0]);
+ int bucketId = Integer.parseInt(bucketKey[1]);
+ dstPath = String.format(pathPattern, tableId, partitionId, indexMeta.indexId,
+ bucketId, indexMeta.schemaHash);
+ tmpPath = dstPath + "." + taskAttemptId;
+ conf.setBoolean("spark.sql.parquet.writeLegacyFormat", false);
+ conf.setBoolean("spark.sql.parquet.int64AsTimestampMillis", false);
+ conf.setBoolean("spark.sql.parquet.int96AsTimestamp", true);
+ conf.setBoolean("spark.sql.parquet.binaryAsString", false);
+ conf.set("spark.sql.parquet.outputTimestampType", "INT96");
+ ParquetWriteSupport.setSchema(dstSchema, conf);
+ ParquetWriteSupport parquetWriteSupport = new ParquetWriteSupport();
+ parquetWriter = new ParquetWriter<InternalRow>(new Path(tmpPath), parquetWriteSupport,
+ CompressionCodecName.SNAPPY, 256 * 1024 * 1024, 16 * 1024, 1024 * 1024,
+ true, false,
+ ParquetProperties.WriterVersion.PARQUET_1_0, conf);
+ if (parquetWriter != null) {
+ LOG.info("[HdfsOperate]>> initialize writer succeed! path:" + tmpPath);
+ }
+ lastBucketKey = curBucketKey;
+ }
+ InternalRow internalRow = encoder.toRow(rowWithoutBucketKey);
+ parquetWriter.write(internalRow);
+ }
+ if (parquetWriter != null) {
+ parquetWriter.close();
+ try {
+ fs.rename(new Path(tmpPath), new Path(dstPath));
+ } catch (IOException ioe) {
+ LOG.warn("rename from tmpPath" + tmpPath + " to dstPath:" + dstPath + " failed. exception:" + ioe);
+ throw ioe;
+ }
+ }
+ }
+ });
+ }
+
+ private void processRollupTree(RollupTreeNode rootNode,
+ Dataset<Row> rootDataframe,
+ long tableId, EtlJobConfig.EtlTable tableMeta,
+ EtlJobConfig.EtlIndex baseIndex) throws UserException {
+ Queue<RollupTreeNode> nodeQueue = new LinkedList<>();
+ nodeQueue.offer(rootNode);
+ int currentLevel = 0;
+ // level travel the tree
+ Map<Long, Dataset<Row>> parentDataframeMap = new HashMap<>();
+ parentDataframeMap.put(baseIndex.indexId, rootDataframe);
+ Map<Long, Dataset<Row>> childrenDataframeMap = new HashMap<>();
+ String pathPattern = etlJobConfig.outputPath + "/" + etlJobConfig.outputFilePattern;
+ while (!nodeQueue.isEmpty()) {
+ RollupTreeNode curNode = nodeQueue.poll();
+ LOG.info("start to process index:" + curNode.indexId);
+ if (curNode.children != null) {
+ for (RollupTreeNode child : curNode.children) {
+ nodeQueue.offer(child);
+ }
+ }
+ Dataset<Row> curDataFrame = null;
+ // column select for rollup
+ if (curNode.level != currentLevel) {
+ for (Dataset<Row> dataframe : parentDataframeMap.values()) {
+ dataframe.unpersist();
+ }
+ currentLevel = curNode.level;
+ parentDataframeMap.clear();
+ parentDataframeMap = childrenDataframeMap;
+ childrenDataframeMap = new HashMap<>();
+ }
+
+ long parentIndexId = baseIndex.indexId;
+ if (curNode.parent != null) {
+ parentIndexId = curNode.parent.indexId;
+ }
+
+ Dataset<Row> parentDataframe = parentDataframeMap.get(parentIndexId);
+ List<Column> columns = new ArrayList<>();
+ List<Column> keyColumns = new ArrayList<>();
+ Column bucketIdColumn = new Column(DppUtils.BUCKET_ID);
+ keyColumns.add(bucketIdColumn);
+ columns.add(bucketIdColumn);
+ for (String keyName : curNode.keyColumnNames) {
+ columns.add(new Column(keyName));
+ keyColumns.add(new Column(keyName));
+ }
+ for (String valueName : curNode.valueColumnNames) {
+ columns.add(new Column(valueName));
+ }
+ Seq<Column> columnSeq = JavaConverters.asScalaIteratorConverter(columns.iterator()).asScala().toSeq();
+ curDataFrame = parentDataframe.select(columnSeq);
+ // aggregate and repartition
+ curDataFrame = processRDDAggAndRepartition(curDataFrame, curNode.indexMeta);
+
+ childrenDataframeMap.put(curNode.indexId, curDataFrame);
+
+ if (curNode.children != null && curNode.children.size() > 1) {
+ // if the children number larger than 1, persist the dataframe for performance
+ curDataFrame.persist();
+ }
+ writePartitionedAndSortedDataframeToParquet(curDataFrame, pathPattern, tableId, curNode.indexMeta);
+ }
+ }
+
+ // repartition dataframe by partitionid_bucketid
+ // so data in the same bucket will be consecutive.
+ private Dataset<Row> repartitionDataframeByBucketId(SparkSession spark, Dataset<Row> dataframe,
+ EtlJobConfig.EtlPartitionInfo partitionInfo,
+ List<Integer> partitionKeyIndex,
+ List<Class> partitionKeySchema,
+ List<DorisRangePartitioner.PartitionRangeKey> partitionRangeKeys,
+ List<String> keyColumnNames,
+ List<String> valueColumnNames,
+ StructType dstTableSchema,
+ EtlJobConfig.EtlIndex baseIndex,
+ List<Long> validPartitionIds) throws UserException {
+ List<String> distributeColumns = partitionInfo.distributionColumnRefs;
+ Partitioner partitioner = new DorisRangePartitioner(partitionInfo, partitionKeyIndex, partitionRangeKeys);
+ Set<Integer> validPartitionIndex = new HashSet<>();
+ if (validPartitionIds == null) {
+ for (int i = 0; i < partitionInfo.partitions.size(); ++i) {
+ validPartitionIndex.add(i);
+ }
+ } else {
+ for (int i = 0; i < partitionInfo.partitions.size(); ++i) {
+ if (validPartitionIds.contains(partitionInfo.partitions.get(i).partitionId)) {
+ validPartitionIndex.add(i);
+ }
+ }
+ }
+ // use PairFlatMapFunction instead of PairMapFunction because the there will be
+ // 0 or 1 output row for 1 input row
+ JavaPairRDD<String, DppColumns> pairRDD = dataframe.javaRDD().flatMapToPair(
+ new PairFlatMapFunction<Row, String, DppColumns>() {
+ @Override
+ public Iterator<Tuple2<String, DppColumns>> call(Row row) {
+ List<Object> columns = new ArrayList<>();
+ List<Object> keyColumns = new ArrayList<>();
+ for (String columnName : keyColumnNames) {
+ Object columnObject = row.get(row.fieldIndex(columnName));
+ columns.add(columnObject);
+ keyColumns.add(columnObject);
+ }
+
+ for (String columnName : valueColumnNames) {
+ columns.add(row.get(row.fieldIndex(columnName)));
+ }
+ DppColumns dppColumns = new DppColumns(columns);
+ DppColumns key = new DppColumns(keyColumns);
+ List<Tuple2<String, DppColumns>> result = new ArrayList<>();
+ int pid = partitioner.getPartition(key);
+ if (!validPartitionIndex.contains(pid)) {
+ LOG.warn("invalid partition for row:" + row + ", pid:" + pid);
+ abnormalRowAcc.add(1);
+ LOG.info("abnormalRowAcc:" + abnormalRowAcc);
+ if (abnormalRowAcc.value() < 5) {
+ LOG.info("add row to invalidRows:" + row.toString());
+ invalidRows.add(row.toString());
+ LOG.info("invalid rows contents:" + invalidRows.value());
+ }
+ } else {
+ long hashValue = DppUtils.getHashValue(row, distributeColumns, dstTableSchema);
+ int bucketId = (int) ((hashValue & 0xffffffff) % partitionInfo.partitions.get(pid).bucketNum);
+ long partitionId = partitionInfo.partitions.get(pid).partitionId;
+ // bucketKey is partitionId_bucketId
+ String bucketKey = partitionId + "_" + bucketId;
+ Tuple2<String, DppColumns> newTuple = new Tuple2<String, DppColumns>(bucketKey, dppColumns);
+ result.add(newTuple);
+ }
+ return result.iterator();
+ }
+ });
+
+ JavaRDD<Row> resultRdd = pairRDD.map(record -> {
+ String bucketKey = record._1;
+ List<Object> row = new ArrayList<>();
+ // bucketKey as the first key
+ row.add(bucketKey);
+ row.addAll(record._2.columns);
+ return RowFactory.create(row.toArray());
+ }
+ );
+
+ // TODO(wb): using rdd instead of dataframe from here
+ StructType tableSchemaWithBucketId = DppUtils.createDstTableSchema(baseIndex.columns, true, false);
+ dataframe = spark.createDataFrame(resultRdd, tableSchemaWithBucketId);
+ // use bucket number as the parallel number
+ int reduceNum = 0;
+ for (EtlJobConfig.EtlPartition partition : partitionInfo.partitions) {
+ for (int i = 0; i < partition.bucketNum; i++) {
+ bucketKeyMap.put(partition.partitionId + "_" + i, reduceNum);
+ reduceNum++;
+ }
+ }
+
+ // print to system.out for easy to find log info
+ System.out.println("print bucket key map:" + bucketKeyMap.toString());
+
+ return dataframe;
+ }
+
+ // do the etl process
+ private Dataset<Row> convertSrcDataframeToDstDataframe(EtlJobConfig.EtlIndex baseIndex,
+ Dataset<Row> srcDataframe,
+ StructType dstTableSchema,
+ EtlJobConfig.EtlFileGroup fileGroup) throws UserException {
+ Dataset<Row> dataframe = srcDataframe;
+ StructType srcSchema = dataframe.schema();
+ Set<String> srcColumnNames = new HashSet<>();
+ for (StructField field : srcSchema.fields()) {
+ srcColumnNames.add(field.name());
+ }
+ Map<String, EtlJobConfig.EtlColumnMapping> columnMappings = fileGroup.columnMappings;
+ // 1. process simple columns
+ Set<String> mappingColumns = null;
+ if (columnMappings != null) {
+ mappingColumns = columnMappings.keySet();
+ }
+ List<String> dstColumnNames = new ArrayList<>();
+ for (StructField dstField : dstTableSchema.fields()) {
+ dstColumnNames.add(dstField.name());
+ EtlJobConfig.EtlColumn column = baseIndex.getColumn(dstField.name());
+ if (!srcColumnNames.contains(dstField.name())) {
+ if (mappingColumns != null && mappingColumns.contains(dstField.name())) {
+ // mapping columns will be processed in next step
+ continue;
+ }
+ if (column.defaultValue != null) {
+ if (column.defaultValue.equals(NULL_FLAG)) {
+ dataframe = dataframe.withColumn(dstField.name(), functions.lit(null));
+ } else {
+ dataframe = dataframe.withColumn(dstField.name(), functions.lit(column.defaultValue));
+ }
+ } else if (column.isAllowNull) {
+ dataframe = dataframe.withColumn(dstField.name(), functions.lit(null));
+ } else {
+ throw new UserException("Reason: no data for column:" + dstField.name());
+ }
+ }
+ if (column.columnType.equalsIgnoreCase("DATE")) {
+ dataframe = dataframe.withColumn(dstField.name(), dataframe.col(dstField.name()).cast("date"));
+ } else if (column.columnType.equalsIgnoreCase("BOOLEAN")) {
+ dataframe = dataframe.withColumn(dstField.name(),
+ functions.when(dataframe.col(dstField.name()).equalTo("true"), "1")
+ .otherwise("0"));
+ } else if (!column.columnType.equalsIgnoreCase(BITMAP_TYPE) && !dstField.dataType().equals(DataTypes.StringType)) {
+ dataframe = dataframe.withColumn(dstField.name(), dataframe.col(dstField.name()).cast(dstField.dataType()));
+ }
+ if (fileGroup.isNegative && !column.isKey) {
+ // negative load
+ // value will be convert te -1 * value
+ dataframe = dataframe.withColumn(dstField.name(), functions.expr("-1 *" + dstField.name()));
+ }
+ }
+ // 2. process the mapping columns
+ for (String mappingColumn : mappingColumns) {
+ String mappingDescription = columnMappings.get(mappingColumn).toDescription();
+ if (mappingDescription.toLowerCase().contains("hll_hash")) {
+ continue;
+ }
+ // here should cast data type to dst column type
+ dataframe = dataframe.withColumn(mappingColumn,
+ functions.expr(mappingDescription).cast(dstTableSchema.apply(mappingColumn).dataType()));
+ }
+ // projection and reorder the columns
+ dataframe.createOrReplaceTempView("src_table");
+ StringBuilder selectSqlBuilder = new StringBuilder();
+ selectSqlBuilder.append("select ");
+ for (String name : dstColumnNames) {
+ selectSqlBuilder.append(name + ",");
+ }
+ selectSqlBuilder.deleteCharAt(selectSqlBuilder.length() - 1);
+ selectSqlBuilder.append(" from src_table");
+ String selectSql = selectSqlBuilder.toString();
+ dataframe = spark.sql(selectSql);
+ return dataframe;
+ }
+
+ private Dataset<Row> loadDataFromPath(SparkSession spark,
+ EtlJobConfig.EtlFileGroup fileGroup,
+ String fileUrl,
+ EtlJobConfig.EtlIndex baseIndex,
+ List<EtlJobConfig.EtlColumn> columns) throws UserException {
+ List<String> columnValueFromPath = DppUtils.parseColumnsFromPath(fileUrl, fileGroup.columnsFromPath);
+ List<String> dataSrcColumns = fileGroup.fileFieldNames;
+ if (dataSrcColumns == null) {
+ // if there is no source columns info
+ // use base index columns as source columns
+ dataSrcColumns = new ArrayList<>();
+ for (EtlJobConfig.EtlColumn column : baseIndex.columns) {
+ dataSrcColumns.add(column.columnName);
+ }
+ }
+ List<String> dstTableNames = new ArrayList<>();
+ for (EtlJobConfig.EtlColumn column : baseIndex.columns) {
+ dstTableNames.add(column.columnName);
+ }
+ List<String> srcColumnsWithColumnsFromPath = new ArrayList<>();
+ srcColumnsWithColumnsFromPath.addAll(dataSrcColumns);
+ if (fileGroup.columnsFromPath != null) {
+ srcColumnsWithColumnsFromPath.addAll(fileGroup.columnsFromPath);
+ }
+ StructType srcSchema = createScrSchema(srcColumnsWithColumnsFromPath);
+ JavaRDD<String> sourceDataRdd = spark.read().textFile(fileUrl).toJavaRDD();
+ int columnSize = dataSrcColumns.size();
+ List<ColumnParser> parsers = new ArrayList<>();
+ for (EtlJobConfig.EtlColumn column : baseIndex.columns) {
+ parsers.add(ColumnParser.create(column.columnType));
+ }
+ // now we first support csv file
+ // TODO: support parquet file and orc file
+ JavaRDD<Row> rowRDD = sourceDataRdd.flatMap(
+ record -> {
+ scannedRowsAcc.add(1);
+ String[] attributes = record.split(fileGroup.columnSeparator);
+ List<Row> result = new ArrayList<>();
+ boolean validRow = true;
+ if (attributes.length != columnSize) {
+ LOG.warn("invalid src schema, data columns:"
+ + attributes.length + ", file group columns:"
+ + columnSize + ", row:" + record);
+ validRow = false;
+ } else {
+ for (int i = 0; i < attributes.length; ++i) {
+ if (attributes[i].equals(NULL_FLAG)) {
+ if (baseIndex.columns.get(i).isAllowNull) {
+ attributes[i] = null;
+ } else {
+ LOG.warn("colunm:" + i + " can not be null. row:" + record);
+ validRow = false;
+ break;
+ }
+ }
+ boolean isStrictMode = (boolean) etlJobConfig.properties.strictMode;
+ if (isStrictMode) {
+ StructField field = srcSchema.apply(i);
+ if (dstTableNames.contains(field.name())) {
+ String type = columns.get(i).columnType;
+ if (type.equalsIgnoreCase("CHAR")
+ || type.equalsIgnoreCase("VARCHAR")) {
+ continue;
+ }
+ ColumnParser parser = parsers.get(i);
+ boolean valid = parser.parse(attributes[i]);
+ if (!valid) {
+ validRow = false;
+ LOG.warn("invalid row:" + record
+ + ", attribute " + i + ": " + attributes[i] + " parsed failed");
+ break;
+ }
+ }
+ }
+ }
+ }
+ if (validRow) {
+ Row row = null;
+ if (fileGroup.columnsFromPath == null) {
+ row = RowFactory.create(attributes);
+ } else {
+ // process columns from path
+ // append columns from path to the tail
+ List<String> columnAttributes = new ArrayList<>();
+ columnAttributes.addAll(Arrays.asList(attributes));
+ columnAttributes.addAll(columnValueFromPath);
+ row = RowFactory.create(columnAttributes.toArray());
+ }
+ result.add(row);
+ } else {
+ abnormalRowAcc.add(1);
+ // at most add 5 rows to invalidRows
+ if (abnormalRowAcc.value() <= 5) {
+ invalidRows.add(record);
+ }
+ }
+ return result.iterator();
+ }
+ );
+
+ Dataset<Row> dataframe = spark.createDataFrame(rowRDD, srcSchema);
+ return dataframe;
+ }
+
+ private StructType createScrSchema(List<String> srcColumns) {
+ List<StructField> fields = new ArrayList<>();
+ for (String srcColumn : srcColumns) {
+ // user StringType to load source data
+ StructField field = DataTypes.createStructField(srcColumn, DataTypes.StringType, true);
+ fields.add(field);
+ }
+ StructType srcSchema = DataTypes.createStructType(fields);
+ return srcSchema;
+ }
+
+ // partition keys will be parsed into double from json
+ // so need to convert it to partition columns' type
+ private Object convertPartitionKey(Object srcValue, Class dstClass) throws UserException {
+ if (dstClass.equals(Float.class) || dstClass.equals(Double.class)) {
+ return null;
+ }
+ if (srcValue instanceof Double) {
+ if (dstClass.equals(Short.class)) {
+ return ((Double) srcValue).shortValue();
+ } else if (dstClass.equals(Integer.class)) {
+ return ((Double) srcValue).intValue();
+ } else if (dstClass.equals(Long.class)) {
+ return ((Double) srcValue).longValue();
+ } else if (dstClass.equals(BigInteger.class)) {
+ return new BigInteger(((Double) srcValue).toString());
+ } else {
+ // dst type is string
+ return srcValue.toString();
+ }
+ } else {
+ LOG.warn("unsupport partition key:" + srcValue);
+ throw new UserException("unsupport partition key:" + srcValue);
+ }
+ }
+
+ private List<DorisRangePartitioner.PartitionRangeKey> createPartitionRangeKeys(
+ EtlJobConfig.EtlPartitionInfo partitionInfo, List<Class> partitionKeySchema) throws UserException {
+ List<DorisRangePartitioner.PartitionRangeKey> partitionRangeKeys = new ArrayList<>();
+ for (EtlJobConfig.EtlPartition partition : partitionInfo.partitions) {
+ DorisRangePartitioner.PartitionRangeKey partitionRangeKey = new DorisRangePartitioner.PartitionRangeKey();
+ List<Object> startKeyColumns = new ArrayList<>();
+ for (int i = 0; i < partition.startKeys.size(); i++) {
+ Object value = partition.startKeys.get(i);
+ startKeyColumns.add(convertPartitionKey(value, partitionKeySchema.get(i)));
+ }
+ partitionRangeKey.startKeys = new DppColumns(startKeyColumns);
+ ;
+ if (!partition.isMaxPartition) {
+ partitionRangeKey.isMaxPartition = false;
+ List<Object> endKeyColumns = new ArrayList<>();
+ for (int i = 0; i < partition.endKeys.size(); i++) {
+ Object value = partition.endKeys.get(i);
+ endKeyColumns.add(convertPartitionKey(value, partitionKeySchema.get(i)));
+ }
+ partitionRangeKey.endKeys = new DppColumns(endKeyColumns);
+ } else {
+ partitionRangeKey.isMaxPartition = false;
+ }
+ partitionRangeKeys.add(partitionRangeKey);
+ }
+ return partitionRangeKeys;
+ }
+
+ private Dataset<Row> loadDataFromFilePaths(SparkSession spark,
+ EtlJobConfig.EtlIndex baseIndex,
+ List<String> filePaths,
+ EtlJobConfig.EtlFileGroup fileGroup,
+ StructType dstTableSchema)
+ throws UserException, IOException, URISyntaxException {
+ Dataset<Row> fileGroupDataframe = null;
+ for (String filePath : filePaths) {
+ fileNumberAcc.add(1);
+ try {
+ Configuration conf = new Configuration();
+ URI uri = new URI(filePath);
+ FileSystem fs = FileSystem.get(uri, conf);
+ FileStatus fileStatus = fs.getFileStatus(new Path(filePath));
+ fileSizeAcc.add(fileStatus.getLen());
+ } catch (Exception e) {
+ LOG.warn("parse path failed:" + filePath);
+ throw e;
+ }
+ if (fileGroup.columnSeparator == null) {
+ LOG.warn("invalid null column separator!");
+ throw new UserException("Reason: invalid null column separator!");
+ }
+ Dataset<Row> dataframe = null;
+
+ dataframe = loadDataFromPath(spark, fileGroup, filePath, baseIndex, baseIndex.columns);
+ dataframe = convertSrcDataframeToDstDataframe(baseIndex, dataframe, dstTableSchema, fileGroup);
+ if (fileGroupDataframe == null) {
+ fileGroupDataframe = dataframe;
+ } else {
+ fileGroupDataframe.union(dataframe);
+ }
+ }
+ return fileGroupDataframe;
+ }
+
+ private Dataset<Row> loadDataFromHiveTable(SparkSession spark,
+ String hiveTableName,
+ EtlJobConfig.EtlIndex baseIndex,
+ EtlJobConfig.EtlFileGroup fileGroup,
+ StructType dstTableSchema) throws UserException {
+ Dataset<Row> dataframe = spark.sql("select * from " + hiveTableName);
+ dataframe = convertSrcDataframeToDstDataframe(baseIndex, dataframe, dstTableSchema, fileGroup);
+ return dataframe;
+ }
+
+ private DppResult process() throws Exception {
+ DppResult dppResult = new DppResult();
+ try {
+ for (Map.Entry<Long, EtlJobConfig.EtlTable> entry : etlJobConfig.tables.entrySet()) {
+ Long tableId = entry.getKey();
+ EtlJobConfig.EtlTable etlTable = entry.getValue();
+
+ // get the base index meta
+ EtlJobConfig.EtlIndex baseIndex = null;
+ for (EtlJobConfig.EtlIndex indexMeta : etlTable.indexes) {
+ if (indexMeta.isBaseIndex) {
+ baseIndex = indexMeta;
+ break;
+ }
+ }
+
+ // get key column names and value column names seperately
+ List<String> keyColumnNames = new ArrayList<>();
+ List<String> valueColumnNames = new ArrayList<>();
+ for (EtlJobConfig.EtlColumn etlColumn : baseIndex.columns) {
+ if (etlColumn.isKey) {
+ keyColumnNames.add(etlColumn.columnName);
+ } else {
+ valueColumnNames.add(etlColumn.columnName);
+ }
+ }
+
+ EtlJobConfig.EtlPartitionInfo partitionInfo = etlTable.partitionInfo;
+ List<Integer> partitionKeyIndex = new ArrayList<Integer>();
+ List<Class> partitionKeySchema = new ArrayList<>();
+ for (String key : partitionInfo.partitionColumnRefs) {
+ for (int i = 0; i < baseIndex.columns.size(); ++i) {
+ EtlJobConfig.EtlColumn column = baseIndex.columns.get(i);
+ if (column.columnName.equals(key)) {
+ partitionKeyIndex.add(i);
+ partitionKeySchema.add(DppUtils.getClassFromColumn(column));
+ break;
+ }
+ }
+ }
+ List<DorisRangePartitioner.PartitionRangeKey> partitionRangeKeys = createPartitionRangeKeys(partitionInfo, partitionKeySchema);
+ StructType dstTableSchema = DppUtils.createDstTableSchema(baseIndex.columns, false, false);
+ RollupTreeBuilder rollupTreeParser = new MinimumCoverageRollupTreeBuilder();
+ RollupTreeNode rootNode = rollupTreeParser.build(etlTable);
+ LOG.info("Start to process rollup tree:" + rootNode);
+
+ Dataset<Row> tableDataframe = null;
+ for (EtlJobConfig.EtlFileGroup fileGroup : etlTable.fileGroups) {
+ List<String> filePaths = fileGroup.filePaths;
+ Dataset<Row> fileGroupDataframe = null;
+ if (Strings.isNullOrEmpty(fileGroup.hiveTableName)) {
+ fileGroupDataframe = loadDataFromFilePaths(spark, baseIndex, filePaths, fileGroup, dstTableSchema);
+ } else {
+ String taskId = etlJobConfig.outputPath.substring(etlJobConfig.outputPath.lastIndexOf("/") + 1);
+ String dorisIntermediateHiveTable = String.format(EtlJobConfig.DORIS_INTERMEDIATE_HIVE_TABLE_NAME,
+ tableId, taskId);
+ dorisIntermediateHiveTable = "kylin2x_test.doris_intermediate_hive_table_31105_66130"; // 3kw
Review comment:
Remove the temporary code
##########
File path: fe/src/main/java/org/apache/doris/load/loadv2/dpp/DppUtils.java
##########
@@ -0,0 +1,252 @@
+// 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.doris.load.loadv2.dpp;
+
+import org.apache.doris.common.UserException;
+import org.apache.doris.load.loadv2.etl.EtlJobConfig;
+
+import org.apache.spark.sql.types.DataType;
+import org.apache.spark.sql.types.DataTypes;
+import org.apache.spark.sql.types.StructField;
+import org.apache.spark.sql.types.StructType;
+import org.apache.spark.sql.types.DecimalType;
+import org.apache.spark.sql.Row;
+
+import com.google.common.collect.Lists;
+
+import java.math.BigDecimal;
+import java.nio.ByteBuffer;
+import java.nio.ByteOrder;
+import java.util.ArrayList;
+import java.util.Collections;
+import java.util.Date;
+import java.util.List;
+import java.util.zip.CRC32;
+
+public class DppUtils {
+ public static final String BUCKET_ID = "__bucketId__";
+ public static Class getClassFromDataType(DataType dataType) {
+ if (dataType == null) {
+ return null;
+ }
+ if (dataType.equals(DataTypes.BooleanType)) {
+ return Boolean.class;
+ } else if (dataType.equals(DataTypes.ShortType)) {
+ return Short.class;
+ } else if (dataType.equals(DataTypes.IntegerType)) {
+ return Integer.class;
+ } else if (dataType.equals(DataTypes.LongType)) {
+ return Long.class;
+ } else if (dataType.equals(DataTypes.FloatType)) {
+ return Float.class;
+ } else if (dataType.equals(DataTypes.DoubleType)) {
+ return Double.class;
+ } else if (dataType.equals(DataTypes.DateType)) {
+ return Date.class;
+ } else if (dataType.equals(DataTypes.StringType)) {
+ return String.class;
+ } else if (dataType instanceof DecimalType) {
+ DecimalType decimalType = (DecimalType)dataType;
+ return BigDecimal.valueOf(decimalType.precision(), decimalType.scale()).getClass();
+ } else if (dataType.equals(DataTypes.TimestampType)) {
+ return Long.class;
+ }
+ return null;
+ }
+
+ public static Class getClassFromColumn(EtlJobConfig.EtlColumn column) throws UserException {
+ switch (column.columnType) {
+ case "BOOLEAN":
+ return Boolean.class;
+ case "TINYINT":
+ case "SMALLINT":
+ return Short.class;
+ case "INT":
+ return Integer.class;
+ case "DATETIME":
+ case "BIGINT":
+ return Long.class;
+ case "LARGEINT":
+ throw new UserException("LARGEINT is not supported now");
+ case "FLOAT":
+ return Float.class;
+ case "DOUBLE":
+ return Double.class;
+ case "DATE":
+ return Date.class;
+ case "HLL":
+ case "CHAR":
+ case "VARCHAR":
+ case "BITMAP":
+ case "OBJECT":
+ return String.class;
+ case "DECIMALV2":
+ return BigDecimal.valueOf(column.precision, column.scale).getClass();
+ default:
+ return String.class;
+ }
+ }
+
+ public static DataType getDataTypeFromColumn(EtlJobConfig.EtlColumn column, boolean regardDistinctColumnAsBinary) {
+ DataType dataType = DataTypes.StringType;
+ switch (column.columnType) {
+ case "BOOLEAN":
+ dataType = DataTypes.StringType;
+ break;
+ case "TINYINT":
+ case "SMALLINT":
+ dataType = DataTypes.ShortType;
+ break;
+ case "INT":
+ dataType = DataTypes.IntegerType;
+ break;
+ case "DATETIME":
+ dataType = DataTypes.TimestampType;
+ break;
+ case "BIGINT":
+ dataType = DataTypes.LongType;
+ break;
+ case "LARGEINT":
+ dataType = DataTypes.StringType;
+ break;
+ case "FLOAT":
+ dataType = DataTypes.FloatType;
+ break;
+ case "DOUBLE":
+ dataType = DataTypes.DoubleType;
+ break;
+ case "DATE":
+ dataType = DataTypes.DateType;
+ break;
+ case "CHAR":
+ case "VARCHAR":
+ case "OBJECT":
+ dataType = DataTypes.StringType;
+ break;
+ case "HLL":
+ case "BITMAP":
+ dataType = regardDistinctColumnAsBinary ? DataTypes.BinaryType : DataTypes.StringType;
+ break;
+ case "DECIMALV2":
+ dataType = DecimalType.apply(column.precision, column.scale);
+ break;
+ default:
+ throw new RuntimeException("Reason: invalid column type:" + column);
+ }
+ return dataType;
+ }
+
+ public static ByteBuffer getHashValue(Object o, DataType type) {
+ ByteBuffer buffer = ByteBuffer.allocate(8);
+ buffer.order(ByteOrder.LITTLE_ENDIAN);
+ if (o == null) {
+ buffer.putInt(0);
+ return buffer;
+ }
+ if (type.equals(DataTypes.ShortType)) {
+ buffer.putShort((Short)o);
Review comment:
Don't use short value for getting tinyint type hash value, this will cause the tinyint data partitioned into wrong bucket.
##########
File path: fe/src/main/java/org/apache/doris/load/loadv2/dpp/SparkDpp.java
##########
@@ -0,0 +1,851 @@
+// 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.doris.load.loadv2.dpp;
+
+import com.google.common.base.Strings;
+import com.google.gson.Gson;
+import org.apache.commons.lang3.StringUtils;
+import org.apache.doris.common.UserException;
+import org.apache.doris.load.loadv2.etl.EtlJobConfig;
+import org.apache.hadoop.fs.FileStatus;
+import org.apache.hadoop.fs.Path;
+import org.apache.hadoop.conf.Configuration;
+import org.apache.parquet.column.ParquetProperties;
+import org.apache.parquet.hadoop.ParquetWriter;
+import org.apache.parquet.hadoop.metadata.CompressionCodecName;
+import org.apache.spark.Partitioner;
+import org.apache.spark.api.java.function.ForeachPartitionFunction;
+import org.apache.spark.api.java.function.PairFlatMapFunction;
+import org.apache.spark.api.java.function.PairFunction;
+import org.apache.spark.sql.catalyst.InternalRow;
+import org.apache.spark.sql.catalyst.encoders.ExpressionEncoder;
+import org.apache.spark.sql.catalyst.encoders.RowEncoder;
+import org.apache.spark.sql.execution.datasources.parquet.ParquetWriteSupport;
+import org.apache.spark.sql.functions;
+import org.apache.spark.sql.RowFactory;
+import org.apache.spark.sql.Row;
+import org.apache.spark.sql.Dataset;
+import org.apache.spark.sql.Column;
+import org.apache.spark.sql.SparkSession;
+import org.apache.spark.sql.types.DataTypes;
+import org.apache.spark.sql.types.StructField;
+import org.apache.spark.sql.types.StructType;
+import org.apache.spark.util.LongAccumulator;
+import org.apache.spark.api.java.JavaPairRDD;
+import org.apache.spark.api.java.JavaRDD;
+import org.apache.spark.TaskContext;
+
+import scala.Tuple2;
+import scala.collection.Seq;
+
+import java.io.IOException;
+import java.math.BigInteger;
+import java.net.URI;
+import java.net.URISyntaxException;
+import java.util.Map;
+import java.util.Set;
+import java.util.List;
+import java.util.ArrayList;
+import java.util.HashMap;
+import java.util.Iterator;
+import java.util.LinkedList;
+import java.util.HashSet;
+import java.util.Arrays;
+import java.util.stream.Collectors;
+
+import org.apache.hadoop.fs.FileSystem;
+import org.apache.hadoop.fs.FSDataOutputStream;
+import scala.collection.JavaConverters;
+import org.apache.log4j.LogManager;
+import org.apache.log4j.Logger;
+
+// This class is a Spark-based data preprocessing program,
+// which will make use of the distributed compute framework of spark to
+// do ETL job/sort/preaggregate jobs in spark job
+// to boost the process of large amount of data load.
+// the process steps are as following:
+// 1. load data
+// 1.1 load data from path/hive table
+// 1.2 do the etl process
+// 2. repartition data by using doris data model(partition and bucket)
+// 3. process aggregation if needed
+// 4. write data to parquet file
+public final class SparkDpp implements java.io.Serializable {
+ private static final Logger LOG = LogManager.getLogger(SparkDpp.class);
+
+ private static final String NULL_FLAG = "\\N";
+ private static final String DPP_RESULT_FILE = "dpp_result.json";
+ private static final String BITMAP_TYPE = "bitmap";
+ private SparkSession spark = null;
+ private EtlJobConfig etlJobConfig = null;
+ private LongAccumulator abnormalRowAcc = null;
+ private LongAccumulator unselectedRowAcc = null;
+ private LongAccumulator scannedRowsAcc = null;
+ private LongAccumulator fileNumberAcc = null;
+ private LongAccumulator fileSizeAcc = null;
+ private Map<String, Integer> bucketKeyMap = new HashMap<>();
+ // accumulator to collect invalid rows
+ private StringAccumulator invalidRows = new StringAccumulator();
+
+ public SparkDpp(SparkSession spark, EtlJobConfig etlJobConfig) {
+ this.spark = spark;
+ this.etlJobConfig = etlJobConfig;
+ }
+
+ public void init() {
+ abnormalRowAcc = spark.sparkContext().longAccumulator("abnormalRowAcc");
+ unselectedRowAcc = spark.sparkContext().longAccumulator("unselectedRowAcc");
+ scannedRowsAcc = spark.sparkContext().longAccumulator("scannedRowsAcc");
+ fileNumberAcc = spark.sparkContext().longAccumulator("fileNumberAcc");
+ fileSizeAcc = spark.sparkContext().longAccumulator("fileSizeAcc");
+ spark.sparkContext().register(invalidRows, "InvalidRowsAccumulator");
+ }
+
+ private Dataset<Row> processRDDAggAndRepartition(Dataset<Row> dataframe, EtlJobConfig.EtlIndex currentIndexMeta) throws UserException {
+ final boolean isDuplicateTable = !StringUtils.equalsIgnoreCase(currentIndexMeta.indexType, "AGGREGATE")
+ && !StringUtils.equalsIgnoreCase(currentIndexMeta.indexType, "UNIQUE");
+
+ // 1 make metadata for map/reduce
+ int keyLen = 0;
+ for (EtlJobConfig.EtlColumn etlColumn : currentIndexMeta.columns) {
+ keyLen = etlColumn.isKey ? keyLen + 1 : keyLen;
+ }
+
+ SparkRDDAggregator[] sparkRDDAggregators = new SparkRDDAggregator[currentIndexMeta.columns.size() - keyLen];
+
+ for (int i = 0 ; i < currentIndexMeta.columns.size(); i++) {
+ if (!currentIndexMeta.columns.get(i).isKey && !isDuplicateTable) {
+ sparkRDDAggregators[i - keyLen] = SparkRDDAggregator.buildAggregator(currentIndexMeta.columns.get(i));
+ }
+ }
+
+ PairFunction<Row, List<Object>, Object[]> encodePairFunction = isDuplicateTable ?
+ // add 1 to include bucketId
+ new EncodeDuplicateTableFunction(keyLen + 1, currentIndexMeta.columns.size() - keyLen)
+ : new EncodeAggregateTableFunction(sparkRDDAggregators, keyLen + 1);
+
+ // 2 convert dataframe to rdd
+ // TODO(wb) use rdd to avoid bitamp/hll serialize
+ JavaPairRDD<List<Object>, Object[]> currentRollupRDD = dataframe.toJavaRDD().mapToPair(encodePairFunction);
+
+ // 3 do aggregate
+ // TODO(wb) set the reduce concurrency by statistic instead of hard code 200
+ int aggregateConcurrency = 200;
+ JavaPairRDD<List<Object>, Object[]> reduceResultRDD = isDuplicateTable ? currentRollupRDD
+ : currentRollupRDD.reduceByKey(new AggregateReduceFunction(sparkRDDAggregators), aggregateConcurrency);
+
+ // 4 repartition
+ JavaRDD<Row> finalRDD = reduceResultRDD
+ .repartitionAndSortWithinPartitions(new BucketPartitioner(bucketKeyMap), new BucketComparator())
+ .map(record -> {
+ List<Object> keys = record._1;
+ Object[] values = record._2;
+ int size = keys.size() + values.length;
+ Object[] result = new Object[size];
+
+ for (int i = 0; i < keys.size(); i++) {
+ result[i] = keys.get(i);
+ }
+
+ for (int i = keys.size(); i < size; i++) {
+ int valueIdx = i - keys.size();
+ result[i] = isDuplicateTable ? values[valueIdx] : sparkRDDAggregators[valueIdx].finalize(values[valueIdx]);
+ }
+
+ return RowFactory.create(result);
+ });
+
+ // 4 convert to dataframe
+ StructType tableSchemaWithBucketId = DppUtils.createDstTableSchema(currentIndexMeta.columns, true, true);
+ dataframe = spark.createDataFrame(finalRDD, tableSchemaWithBucketId);
+ return dataframe;
+
+ }
+
+ // write data to parquet file by using writing the parquet scheme of spark.
+ private void writePartitionedAndSortedDataframeToParquet(Dataset<Row> dataframe,
+ String pathPattern,
+ long tableId,
+ EtlJobConfig.EtlIndex indexMeta) throws UserException {
+ StructType outputSchema = dataframe.schema();
+ StructType dstSchema = DataTypes.createStructType(
+ Arrays.asList(outputSchema.fields()).stream()
+ .filter(field -> !field.name().equalsIgnoreCase(DppUtils.BUCKET_ID))
+ .collect(Collectors.toList()));
+ ExpressionEncoder encoder = RowEncoder.apply(dstSchema);
+ dataframe.foreachPartition(new ForeachPartitionFunction<Row>() {
+ @Override
+ public void call(Iterator<Row> t) throws Exception {
+ // write the data to dst file
+ Configuration conf = new Configuration();
+ FileSystem fs = FileSystem.get(URI.create(etlJobConfig.outputPath), conf);
+ String lastBucketKey = null;
+ ParquetWriter<InternalRow> parquetWriter = null;
+ TaskContext taskContext = TaskContext.get();
+ long taskAttemptId = taskContext.taskAttemptId();
+ String dstPath = "";
+ String tmpPath = "";
+
+ while (t.hasNext()) {
+ Row row = t.next();
+ if (row.length() <= 1) {
+ LOG.warn("invalid row:" + row);
+ continue;
+ }
+
+
+ String curBucketKey = row.getString(0);
+ List<Object> columnObjects = new ArrayList<>();
+ for (int i = 1; i < row.length(); ++i) {
+ Object columnValue = row.get(i);
+ columnObjects.add(columnValue);
+ }
+ System.out.println();
+ Row rowWithoutBucketKey = RowFactory.create(columnObjects.toArray());
+ // if the bucket key is new, it will belong to a new tablet
+ if (lastBucketKey == null || !curBucketKey.equals(lastBucketKey)) {
+ if (parquetWriter != null) {
+ parquetWriter.close();
+ // rename tmpPath to path
+ try {
+ fs.rename(new Path(tmpPath), new Path(dstPath));
+ } catch (IOException ioe) {
+ LOG.warn("rename from tmpPath" + tmpPath + " to dstPath:" + dstPath + " failed. exception:" + ioe);
+ throw ioe;
+ }
+ }
+ // flush current writer and create a new writer
+ String[] bucketKey = curBucketKey.split("_");
+ if (bucketKey.length != 2) {
+ LOG.warn("invalid bucket key:" + curBucketKey);
+ continue;
+ }
+ int partitionId = Integer.parseInt(bucketKey[0]);
+ int bucketId = Integer.parseInt(bucketKey[1]);
+ dstPath = String.format(pathPattern, tableId, partitionId, indexMeta.indexId,
+ bucketId, indexMeta.schemaHash);
+ tmpPath = dstPath + "." + taskAttemptId;
+ conf.setBoolean("spark.sql.parquet.writeLegacyFormat", false);
+ conf.setBoolean("spark.sql.parquet.int64AsTimestampMillis", false);
+ conf.setBoolean("spark.sql.parquet.int96AsTimestamp", true);
+ conf.setBoolean("spark.sql.parquet.binaryAsString", false);
+ conf.set("spark.sql.parquet.outputTimestampType", "INT96");
+ ParquetWriteSupport.setSchema(dstSchema, conf);
+ ParquetWriteSupport parquetWriteSupport = new ParquetWriteSupport();
+ parquetWriter = new ParquetWriter<InternalRow>(new Path(tmpPath), parquetWriteSupport,
+ CompressionCodecName.SNAPPY, 256 * 1024 * 1024, 16 * 1024, 1024 * 1024,
+ true, false,
+ ParquetProperties.WriterVersion.PARQUET_1_0, conf);
+ if (parquetWriter != null) {
+ LOG.info("[HdfsOperate]>> initialize writer succeed! path:" + tmpPath);
+ }
+ lastBucketKey = curBucketKey;
+ }
+ InternalRow internalRow = encoder.toRow(rowWithoutBucketKey);
+ parquetWriter.write(internalRow);
+ }
+ if (parquetWriter != null) {
+ parquetWriter.close();
+ try {
+ fs.rename(new Path(tmpPath), new Path(dstPath));
+ } catch (IOException ioe) {
+ LOG.warn("rename from tmpPath" + tmpPath + " to dstPath:" + dstPath + " failed. exception:" + ioe);
+ throw ioe;
+ }
+ }
+ }
+ });
+ }
+
+ private void processRollupTree(RollupTreeNode rootNode,
+ Dataset<Row> rootDataframe,
+ long tableId, EtlJobConfig.EtlTable tableMeta,
+ EtlJobConfig.EtlIndex baseIndex) throws UserException {
+ Queue<RollupTreeNode> nodeQueue = new LinkedList<>();
+ nodeQueue.offer(rootNode);
+ int currentLevel = 0;
+ // level travel the tree
+ Map<Long, Dataset<Row>> parentDataframeMap = new HashMap<>();
+ parentDataframeMap.put(baseIndex.indexId, rootDataframe);
+ Map<Long, Dataset<Row>> childrenDataframeMap = new HashMap<>();
+ String pathPattern = etlJobConfig.outputPath + "/" + etlJobConfig.outputFilePattern;
+ while (!nodeQueue.isEmpty()) {
+ RollupTreeNode curNode = nodeQueue.poll();
+ LOG.info("start to process index:" + curNode.indexId);
+ if (curNode.children != null) {
+ for (RollupTreeNode child : curNode.children) {
+ nodeQueue.offer(child);
+ }
+ }
+ Dataset<Row> curDataFrame = null;
+ // column select for rollup
+ if (curNode.level != currentLevel) {
+ for (Dataset<Row> dataframe : parentDataframeMap.values()) {
+ dataframe.unpersist();
+ }
+ currentLevel = curNode.level;
+ parentDataframeMap.clear();
+ parentDataframeMap = childrenDataframeMap;
+ childrenDataframeMap = new HashMap<>();
+ }
+
+ long parentIndexId = baseIndex.indexId;
+ if (curNode.parent != null) {
+ parentIndexId = curNode.parent.indexId;
+ }
+
+ Dataset<Row> parentDataframe = parentDataframeMap.get(parentIndexId);
+ List<Column> columns = new ArrayList<>();
+ List<Column> keyColumns = new ArrayList<>();
+ Column bucketIdColumn = new Column(DppUtils.BUCKET_ID);
+ keyColumns.add(bucketIdColumn);
+ columns.add(bucketIdColumn);
+ for (String keyName : curNode.keyColumnNames) {
+ columns.add(new Column(keyName));
+ keyColumns.add(new Column(keyName));
+ }
+ for (String valueName : curNode.valueColumnNames) {
+ columns.add(new Column(valueName));
+ }
+ Seq<Column> columnSeq = JavaConverters.asScalaIteratorConverter(columns.iterator()).asScala().toSeq();
+ curDataFrame = parentDataframe.select(columnSeq);
+ // aggregate and repartition
+ curDataFrame = processRDDAggAndRepartition(curDataFrame, curNode.indexMeta);
+
+ childrenDataframeMap.put(curNode.indexId, curDataFrame);
+
+ if (curNode.children != null && curNode.children.size() > 1) {
+ // if the children number larger than 1, persist the dataframe for performance
+ curDataFrame.persist();
+ }
+ writePartitionedAndSortedDataframeToParquet(curDataFrame, pathPattern, tableId, curNode.indexMeta);
+ }
+ }
+
+ // repartition dataframe by partitionid_bucketid
+ // so data in the same bucket will be consecutive.
+ private Dataset<Row> repartitionDataframeByBucketId(SparkSession spark, Dataset<Row> dataframe,
+ EtlJobConfig.EtlPartitionInfo partitionInfo,
+ List<Integer> partitionKeyIndex,
+ List<Class> partitionKeySchema,
+ List<DorisRangePartitioner.PartitionRangeKey> partitionRangeKeys,
+ List<String> keyColumnNames,
+ List<String> valueColumnNames,
+ StructType dstTableSchema,
+ EtlJobConfig.EtlIndex baseIndex,
+ List<Long> validPartitionIds) throws UserException {
+ List<String> distributeColumns = partitionInfo.distributionColumnRefs;
+ Partitioner partitioner = new DorisRangePartitioner(partitionInfo, partitionKeyIndex, partitionRangeKeys);
+ Set<Integer> validPartitionIndex = new HashSet<>();
+ if (validPartitionIds == null) {
+ for (int i = 0; i < partitionInfo.partitions.size(); ++i) {
+ validPartitionIndex.add(i);
+ }
+ } else {
+ for (int i = 0; i < partitionInfo.partitions.size(); ++i) {
+ if (validPartitionIds.contains(partitionInfo.partitions.get(i).partitionId)) {
+ validPartitionIndex.add(i);
+ }
+ }
+ }
+ // use PairFlatMapFunction instead of PairMapFunction because the there will be
+ // 0 or 1 output row for 1 input row
+ JavaPairRDD<String, DppColumns> pairRDD = dataframe.javaRDD().flatMapToPair(
+ new PairFlatMapFunction<Row, String, DppColumns>() {
+ @Override
+ public Iterator<Tuple2<String, DppColumns>> call(Row row) {
+ List<Object> columns = new ArrayList<>();
+ List<Object> keyColumns = new ArrayList<>();
+ for (String columnName : keyColumnNames) {
+ Object columnObject = row.get(row.fieldIndex(columnName));
+ columns.add(columnObject);
+ keyColumns.add(columnObject);
+ }
+
+ for (String columnName : valueColumnNames) {
+ columns.add(row.get(row.fieldIndex(columnName)));
+ }
+ DppColumns dppColumns = new DppColumns(columns);
+ DppColumns key = new DppColumns(keyColumns);
+ List<Tuple2<String, DppColumns>> result = new ArrayList<>();
+ int pid = partitioner.getPartition(key);
+ if (!validPartitionIndex.contains(pid)) {
+ LOG.warn("invalid partition for row:" + row + ", pid:" + pid);
+ abnormalRowAcc.add(1);
+ LOG.info("abnormalRowAcc:" + abnormalRowAcc);
+ if (abnormalRowAcc.value() < 5) {
+ LOG.info("add row to invalidRows:" + row.toString());
+ invalidRows.add(row.toString());
+ LOG.info("invalid rows contents:" + invalidRows.value());
+ }
+ } else {
+ long hashValue = DppUtils.getHashValue(row, distributeColumns, dstTableSchema);
+ int bucketId = (int) ((hashValue & 0xffffffff) % partitionInfo.partitions.get(pid).bucketNum);
+ long partitionId = partitionInfo.partitions.get(pid).partitionId;
+ // bucketKey is partitionId_bucketId
+ String bucketKey = partitionId + "_" + bucketId;
+ Tuple2<String, DppColumns> newTuple = new Tuple2<String, DppColumns>(bucketKey, dppColumns);
+ result.add(newTuple);
+ }
+ return result.iterator();
+ }
+ });
+
+ JavaRDD<Row> resultRdd = pairRDD.map(record -> {
+ String bucketKey = record._1;
+ List<Object> row = new ArrayList<>();
+ // bucketKey as the first key
+ row.add(bucketKey);
+ row.addAll(record._2.columns);
+ return RowFactory.create(row.toArray());
+ }
+ );
+
+ // TODO(wb): using rdd instead of dataframe from here
+ StructType tableSchemaWithBucketId = DppUtils.createDstTableSchema(baseIndex.columns, true, false);
+ dataframe = spark.createDataFrame(resultRdd, tableSchemaWithBucketId);
+ // use bucket number as the parallel number
+ int reduceNum = 0;
+ for (EtlJobConfig.EtlPartition partition : partitionInfo.partitions) {
+ for (int i = 0; i < partition.bucketNum; i++) {
+ bucketKeyMap.put(partition.partitionId + "_" + i, reduceNum);
+ reduceNum++;
+ }
+ }
+
+ // print to system.out for easy to find log info
+ System.out.println("print bucket key map:" + bucketKeyMap.toString());
+
+ return dataframe;
+ }
+
+ // do the etl process
+ private Dataset<Row> convertSrcDataframeToDstDataframe(EtlJobConfig.EtlIndex baseIndex,
+ Dataset<Row> srcDataframe,
+ StructType dstTableSchema,
+ EtlJobConfig.EtlFileGroup fileGroup) throws UserException {
+ Dataset<Row> dataframe = srcDataframe;
+ StructType srcSchema = dataframe.schema();
+ Set<String> srcColumnNames = new HashSet<>();
+ for (StructField field : srcSchema.fields()) {
+ srcColumnNames.add(field.name());
+ }
+ Map<String, EtlJobConfig.EtlColumnMapping> columnMappings = fileGroup.columnMappings;
+ // 1. process simple columns
+ Set<String> mappingColumns = null;
+ if (columnMappings != null) {
+ mappingColumns = columnMappings.keySet();
+ }
+ List<String> dstColumnNames = new ArrayList<>();
+ for (StructField dstField : dstTableSchema.fields()) {
+ dstColumnNames.add(dstField.name());
+ EtlJobConfig.EtlColumn column = baseIndex.getColumn(dstField.name());
+ if (!srcColumnNames.contains(dstField.name())) {
+ if (mappingColumns != null && mappingColumns.contains(dstField.name())) {
+ // mapping columns will be processed in next step
+ continue;
+ }
+ if (column.defaultValue != null) {
+ if (column.defaultValue.equals(NULL_FLAG)) {
+ dataframe = dataframe.withColumn(dstField.name(), functions.lit(null));
+ } else {
+ dataframe = dataframe.withColumn(dstField.name(), functions.lit(column.defaultValue));
+ }
+ } else if (column.isAllowNull) {
+ dataframe = dataframe.withColumn(dstField.name(), functions.lit(null));
+ } else {
+ throw new UserException("Reason: no data for column:" + dstField.name());
+ }
+ }
+ if (column.columnType.equalsIgnoreCase("DATE")) {
+ dataframe = dataframe.withColumn(dstField.name(), dataframe.col(dstField.name()).cast("date"));
+ } else if (column.columnType.equalsIgnoreCase("BOOLEAN")) {
+ dataframe = dataframe.withColumn(dstField.name(),
+ functions.when(dataframe.col(dstField.name()).equalTo("true"), "1")
+ .otherwise("0"));
+ } else if (!column.columnType.equalsIgnoreCase(BITMAP_TYPE) && !dstField.dataType().equals(DataTypes.StringType)) {
+ dataframe = dataframe.withColumn(dstField.name(), dataframe.col(dstField.name()).cast(dstField.dataType()));
+ }
+ if (fileGroup.isNegative && !column.isKey) {
+ // negative load
+ // value will be convert te -1 * value
+ dataframe = dataframe.withColumn(dstField.name(), functions.expr("-1 *" + dstField.name()));
+ }
+ }
+ // 2. process the mapping columns
+ for (String mappingColumn : mappingColumns) {
+ String mappingDescription = columnMappings.get(mappingColumn).toDescription();
+ if (mappingDescription.toLowerCase().contains("hll_hash")) {
+ continue;
+ }
+ // here should cast data type to dst column type
+ dataframe = dataframe.withColumn(mappingColumn,
+ functions.expr(mappingDescription).cast(dstTableSchema.apply(mappingColumn).dataType()));
+ }
+ // projection and reorder the columns
+ dataframe.createOrReplaceTempView("src_table");
+ StringBuilder selectSqlBuilder = new StringBuilder();
+ selectSqlBuilder.append("select ");
+ for (String name : dstColumnNames) {
+ selectSqlBuilder.append(name + ",");
+ }
+ selectSqlBuilder.deleteCharAt(selectSqlBuilder.length() - 1);
+ selectSqlBuilder.append(" from src_table");
+ String selectSql = selectSqlBuilder.toString();
+ dataframe = spark.sql(selectSql);
+ return dataframe;
+ }
+
+ private Dataset<Row> loadDataFromPath(SparkSession spark,
+ EtlJobConfig.EtlFileGroup fileGroup,
+ String fileUrl,
+ EtlJobConfig.EtlIndex baseIndex,
+ List<EtlJobConfig.EtlColumn> columns) throws UserException {
+ List<String> columnValueFromPath = DppUtils.parseColumnsFromPath(fileUrl, fileGroup.columnsFromPath);
+ List<String> dataSrcColumns = fileGroup.fileFieldNames;
+ if (dataSrcColumns == null) {
+ // if there is no source columns info
+ // use base index columns as source columns
+ dataSrcColumns = new ArrayList<>();
+ for (EtlJobConfig.EtlColumn column : baseIndex.columns) {
+ dataSrcColumns.add(column.columnName);
+ }
+ }
+ List<String> dstTableNames = new ArrayList<>();
+ for (EtlJobConfig.EtlColumn column : baseIndex.columns) {
+ dstTableNames.add(column.columnName);
+ }
+ List<String> srcColumnsWithColumnsFromPath = new ArrayList<>();
+ srcColumnsWithColumnsFromPath.addAll(dataSrcColumns);
+ if (fileGroup.columnsFromPath != null) {
+ srcColumnsWithColumnsFromPath.addAll(fileGroup.columnsFromPath);
+ }
+ StructType srcSchema = createScrSchema(srcColumnsWithColumnsFromPath);
+ JavaRDD<String> sourceDataRdd = spark.read().textFile(fileUrl).toJavaRDD();
+ int columnSize = dataSrcColumns.size();
+ List<ColumnParser> parsers = new ArrayList<>();
+ for (EtlJobConfig.EtlColumn column : baseIndex.columns) {
+ parsers.add(ColumnParser.create(column.columnType));
+ }
+ // now we first support csv file
+ // TODO: support parquet file and orc file
+ JavaRDD<Row> rowRDD = sourceDataRdd.flatMap(
+ record -> {
+ scannedRowsAcc.add(1);
+ String[] attributes = record.split(fileGroup.columnSeparator);
+ List<Row> result = new ArrayList<>();
+ boolean validRow = true;
+ if (attributes.length != columnSize) {
+ LOG.warn("invalid src schema, data columns:"
+ + attributes.length + ", file group columns:"
+ + columnSize + ", row:" + record);
+ validRow = false;
+ } else {
+ for (int i = 0; i < attributes.length; ++i) {
+ if (attributes[i].equals(NULL_FLAG)) {
+ if (baseIndex.columns.get(i).isAllowNull) {
+ attributes[i] = null;
+ } else {
+ LOG.warn("colunm:" + i + " can not be null. row:" + record);
+ validRow = false;
+ break;
+ }
+ }
+ boolean isStrictMode = (boolean) etlJobConfig.properties.strictMode;
+ if (isStrictMode) {
+ StructField field = srcSchema.apply(i);
+ if (dstTableNames.contains(field.name())) {
+ String type = columns.get(i).columnType;
+ if (type.equalsIgnoreCase("CHAR")
+ || type.equalsIgnoreCase("VARCHAR")) {
+ continue;
+ }
+ ColumnParser parser = parsers.get(i);
+ boolean valid = parser.parse(attributes[i]);
+ if (!valid) {
+ validRow = false;
+ LOG.warn("invalid row:" + record
+ + ", attribute " + i + ": " + attributes[i] + " parsed failed");
+ break;
+ }
+ }
+ }
+ }
+ }
+ if (validRow) {
+ Row row = null;
+ if (fileGroup.columnsFromPath == null) {
+ row = RowFactory.create(attributes);
+ } else {
+ // process columns from path
+ // append columns from path to the tail
+ List<String> columnAttributes = new ArrayList<>();
+ columnAttributes.addAll(Arrays.asList(attributes));
+ columnAttributes.addAll(columnValueFromPath);
+ row = RowFactory.create(columnAttributes.toArray());
+ }
+ result.add(row);
+ } else {
+ abnormalRowAcc.add(1);
+ // at most add 5 rows to invalidRows
+ if (abnormalRowAcc.value() <= 5) {
+ invalidRows.add(record);
+ }
+ }
+ return result.iterator();
+ }
+ );
+
+ Dataset<Row> dataframe = spark.createDataFrame(rowRDD, srcSchema);
+ return dataframe;
+ }
+
+ private StructType createScrSchema(List<String> srcColumns) {
+ List<StructField> fields = new ArrayList<>();
+ for (String srcColumn : srcColumns) {
+ // user StringType to load source data
+ StructField field = DataTypes.createStructField(srcColumn, DataTypes.StringType, true);
+ fields.add(field);
+ }
+ StructType srcSchema = DataTypes.createStructType(fields);
+ return srcSchema;
+ }
+
+ // partition keys will be parsed into double from json
+ // so need to convert it to partition columns' type
+ private Object convertPartitionKey(Object srcValue, Class dstClass) throws UserException {
+ if (dstClass.equals(Float.class) || dstClass.equals(Double.class)) {
+ return null;
+ }
+ if (srcValue instanceof Double) {
+ if (dstClass.equals(Short.class)) {
+ return ((Double) srcValue).shortValue();
+ } else if (dstClass.equals(Integer.class)) {
+ return ((Double) srcValue).intValue();
+ } else if (dstClass.equals(Long.class)) {
+ return ((Double) srcValue).longValue();
+ } else if (dstClass.equals(BigInteger.class)) {
+ return new BigInteger(((Double) srcValue).toString());
+ } else {
+ // dst type is string
+ return srcValue.toString();
+ }
+ } else {
+ LOG.warn("unsupport partition key:" + srcValue);
+ throw new UserException("unsupport partition key:" + srcValue);
+ }
+ }
+
+ private List<DorisRangePartitioner.PartitionRangeKey> createPartitionRangeKeys(
+ EtlJobConfig.EtlPartitionInfo partitionInfo, List<Class> partitionKeySchema) throws UserException {
+ List<DorisRangePartitioner.PartitionRangeKey> partitionRangeKeys = new ArrayList<>();
+ for (EtlJobConfig.EtlPartition partition : partitionInfo.partitions) {
+ DorisRangePartitioner.PartitionRangeKey partitionRangeKey = new DorisRangePartitioner.PartitionRangeKey();
+ List<Object> startKeyColumns = new ArrayList<>();
+ for (int i = 0; i < partition.startKeys.size(); i++) {
+ Object value = partition.startKeys.get(i);
+ startKeyColumns.add(convertPartitionKey(value, partitionKeySchema.get(i)));
+ }
+ partitionRangeKey.startKeys = new DppColumns(startKeyColumns);
+ ;
+ if (!partition.isMaxPartition) {
+ partitionRangeKey.isMaxPartition = false;
+ List<Object> endKeyColumns = new ArrayList<>();
+ for (int i = 0; i < partition.endKeys.size(); i++) {
+ Object value = partition.endKeys.get(i);
+ endKeyColumns.add(convertPartitionKey(value, partitionKeySchema.get(i)));
+ }
+ partitionRangeKey.endKeys = new DppColumns(endKeyColumns);
+ } else {
+ partitionRangeKey.isMaxPartition = false;
Review comment:
isMaxPartition = true;
xuyang already fixed, maybe the code is not up to date
----------------------------------------------------------------
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.
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscribe@doris.apache.org
For additional commands, e-mail: commits-help@doris.apache.org