lanyuanxiaoyao/hudi
1
0
Fork 0
Code Activity

[HUDI-3985] Refactor DLASyncTool to support read hoodie table as spark datasource table (#5532)

Browse Source
This commit is contained in:
huberylee
2022-05-20 22:25:32 +08:00
committed by GitHub
parent c7576f7613
commit 85b146d3d5
26 changed files with 1281 additions and 974 deletions
Download Patch File Download Diff File Expand all files Collapse all files

84
hudi-sync/hudi-hive-sync/src/main/java/org/apache/hudi/hive/HiveSyncTool.java
View File

@@ -27,9 +27,8 @@ import org.apache.hudi.common.util.StringUtils;
import org.apache.hudi.exception.HoodieException;
import org.apache.hudi.exception.InvalidTableException;
import org.apache.hudi.hadoop.utils.HoodieInputFormatUtils;
import org.apache.hudi.hive.util.ConfigUtils;
import org.apache.hudi.sync.common.util.ConfigUtils;
import org.apache.hudi.hive.util.HiveSchemaUtil;
import org.apache.hudi.hive.util.Parquet2SparkSchemaUtils;
import org.apache.hudi.sync.common.AbstractSyncHoodieClient.PartitionEvent;
import org.apache.hudi.sync.common.AbstractSyncHoodieClient.PartitionEvent.PartitionEventType;
import org.apache.hudi.sync.common.AbstractSyncTool;
@@ -43,20 +42,13 @@ import org.apache.hadoop.hive.conf.HiveConf;
import org.apache.hadoop.hive.metastore.api.FieldSchema;
import org.apache.log4j.LogManager;
import org.apache.log4j.Logger;
import org.apache.parquet.schema.GroupType;
import org.apache.parquet.schema.MessageType;
import org.apache.parquet.schema.PrimitiveType;
import org.apache.parquet.schema.Type;
import java.util.ArrayList;
import java.util.HashMap;
import java.util.List;
import java.util.Map;
import java.util.stream.Collectors;
import static org.apache.parquet.schema.OriginalType.UTF8;
import static org.apache.parquet.schema.PrimitiveType.PrimitiveTypeName.BINARY;
/**
* Tool to sync a hoodie HDFS table with a hive metastore table. Either use it as a api
* HiveSyncTool.syncHoodieTable(HiveSyncConfig) or as a command line java -cp hoodie-hive-sync.jar HiveSyncTool [args]
@@ -248,8 +240,9 @@ public class HiveSyncTool extends AbstractSyncTool implements AutoCloseable {
Map<String, String> tableProperties = ConfigUtils.toMap(hiveSyncConfig.tableProperties);
Map<String, String> serdeProperties = ConfigUtils.toMap(hiveSyncConfig.serdeProperties);
if (hiveSyncConfig.syncAsSparkDataSourceTable) {
Map<String, String> sparkTableProperties = getSparkTableProperties(hiveSyncConfig.sparkSchemaLengthThreshold, schema);
Map<String, String> sparkSerdeProperties = getSparkSerdeProperties(readAsOptimized);
Map<String, String> sparkTableProperties = getSparkTableProperties(hiveSyncConfig.partitionFields,
hiveSyncConfig.sparkVersion, hiveSyncConfig.sparkSchemaLengthThreshold, schema);
Map<String, String> sparkSerdeProperties = getSparkSerdeProperties(readAsOptimized, hiveSyncConfig.basePath);
tableProperties.putAll(sparkTableProperties);
serdeProperties.putAll(sparkSerdeProperties);
}
@@ -309,75 +302,6 @@ public class HiveSyncTool extends AbstractSyncTool implements AutoCloseable {
return schemaChanged;
}
/**
* Get Spark Sql related table properties. This is used for spark datasource table.
* @param schema The schema to write to the table.
* @return A new parameters added the spark's table properties.
*/
private Map<String, String> getSparkTableProperties(int schemaLengthThreshold, MessageType schema) {
// Convert the schema and partition info used by spark sql to hive table properties.
// The following code refers to the spark code in
// https://github.com/apache/spark/blob/master/sql/hive/src/main/scala/org/apache/spark/sql/hive/HiveExternalCatalog.scala
GroupType originGroupType = schema.asGroupType();
List<String> partitionNames = hiveSyncConfig.partitionFields;
List<Type> partitionCols = new ArrayList<>();
List<Type> dataCols = new ArrayList<>();
Map<String, Type> column2Field = new HashMap<>();
for (Type field : originGroupType.getFields()) {
column2Field.put(field.getName(), field);
}
// Get partition columns and data columns.
for (String partitionName : partitionNames) {
// Default the unknown partition fields to be String.
// Keep the same logical with HiveSchemaUtil#getPartitionKeyType.
partitionCols.add(column2Field.getOrDefault(partitionName,
new PrimitiveType(Type.Repetition.REQUIRED, BINARY, partitionName, UTF8)));
}
for (Type field : originGroupType.getFields()) {
if (!partitionNames.contains(field.getName())) {
dataCols.add(field);
}
}
List<Type> reOrderedFields = new ArrayList<>();
reOrderedFields.addAll(dataCols);
reOrderedFields.addAll(partitionCols);
GroupType reOrderedType = new GroupType(originGroupType.getRepetition(), originGroupType.getName(), reOrderedFields);
Map<String, String> sparkProperties = new HashMap<>();
sparkProperties.put("spark.sql.sources.provider", "hudi");
if (!StringUtils.isNullOrEmpty(hiveSyncConfig.sparkVersion)) {
sparkProperties.put("spark.sql.create.version", hiveSyncConfig.sparkVersion);
}
// Split the schema string to multi-parts according the schemaLengthThreshold size.
String schemaString = Parquet2SparkSchemaUtils.convertToSparkSchemaJson(reOrderedType);
int numSchemaPart = (schemaString.length() + schemaLengthThreshold - 1) / schemaLengthThreshold;
sparkProperties.put("spark.sql.sources.schema.numParts", String.valueOf(numSchemaPart));
// Add each part of schema string to sparkProperties
for (int i = 0; i < numSchemaPart; i++) {
int start = i * schemaLengthThreshold;
int end = Math.min(start + schemaLengthThreshold, schemaString.length());
sparkProperties.put("spark.sql.sources.schema.part." + i, schemaString.substring(start, end));
}
// Add partition columns
if (!partitionNames.isEmpty()) {
sparkProperties.put("spark.sql.sources.schema.numPartCols", String.valueOf(partitionNames.size()));
for (int i = 0; i < partitionNames.size(); i++) {
sparkProperties.put("spark.sql.sources.schema.partCol." + i, partitionNames.get(i));
}
}
return sparkProperties;
}
private Map<String, String> getSparkSerdeProperties(boolean readAsOptimized) {
Map<String, String> sparkSerdeProperties = new HashMap<>();
sparkSerdeProperties.put("path", hiveSyncConfig.basePath);
sparkSerdeProperties.put(ConfigUtils.IS_QUERY_AS_RO_TABLE, String.valueOf(readAsOptimized));
return sparkSerdeProperties;
}
/**
* Syncs the list of storage partitions passed in (checks if the partition is in hive, if not adds it or if the
* partition path does not match, it updates the partition path).

78
hudi-sync/hudi-hive-sync/src/main/java/org/apache/hudi/hive/util/ConfigUtils.java
View File

@@ -1,78 +0,0 @@
/*
* 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.hudi.hive.util;
import java.util.HashMap;
import java.util.Map;
import org.apache.hudi.common.util.StringUtils;
public class ConfigUtils {
/**
* Config stored in hive serde properties to tell query engine (spark/flink) to
* read the table as a read-optimized table when this config is true.
*/
public static final String IS_QUERY_AS_RO_TABLE = "hoodie.query.as.ro.table";
/**
* Convert the key-value config to a map.The format of the config
* is a key-value pair just like "k1=v1\nk2=v2\nk3=v3".
* @param keyValueConfig
* @return
*/
public static Map<String, String> toMap(String keyValueConfig) {
if (StringUtils.isNullOrEmpty(keyValueConfig)) {
return new HashMap<>();
}
String[] keyvalues = keyValueConfig.split("\n");
Map<String, String> tableProperties = new HashMap<>();
for (String keyValue : keyvalues) {
String[] keyValueArray = keyValue.split("=");
if (keyValueArray.length == 1 || keyValueArray.length == 2) {
String key = keyValueArray[0].trim();
String value = keyValueArray.length == 2 ? keyValueArray[1].trim() : "";
tableProperties.put(key, value);
} else {
throw new IllegalArgumentException("Bad key-value config: " + keyValue + ", must be the"
+ " format 'key = value'");
}
}
return tableProperties;
}
/**
* Convert map config to key-value string.The format of the config
* is a key-value pair just like "k1=v1\nk2=v2\nk3=v3".
* @param config
* @return
*/
public static String configToString(Map<String, String> config) {
if (config == null) {
return null;
}
StringBuilder sb = new StringBuilder();
for (Map.Entry<String, String> entry : config.entrySet()) {
if (sb.length() > 0) {
sb.append("\n");
}
sb.append(entry.getKey()).append("=").append(entry.getValue());
}
return sb.toString();
}
}

171
hudi-sync/hudi-hive-sync/src/main/java/org/apache/hudi/hive/util/Parquet2SparkSchemaUtils.java
View File

@@ -1,171 +0,0 @@
/*
* 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.hudi.hive.util;
import org.apache.hudi.common.util.ValidationUtils;
import org.apache.parquet.schema.GroupType;
import org.apache.parquet.schema.OriginalType;
import org.apache.parquet.schema.PrimitiveType;
import org.apache.parquet.schema.Type;
import static org.apache.parquet.schema.Type.Repetition.OPTIONAL;
/**
* Convert the parquet schema to spark schema' json string.
* This code is refer to org.apache.spark.sql.execution.datasources.parquet.ParquetToSparkSchemaConverter
* in spark project.
*/
public class Parquet2SparkSchemaUtils {
public static String convertToSparkSchemaJson(GroupType parquetSchema) {
String fieldsJsonString = parquetSchema.getFields().stream().map(field -> {
switch (field.getRepetition()) {
case OPTIONAL:
return "{\"name\":\"" + field.getName() + "\",\"type\":" + convertFieldType(field)
+ ",\"nullable\":true,\"metadata\":{}}";
case REQUIRED:
return "{\"name\":\"" + field.getName() + "\",\"type\":" + convertFieldType(field)
+ ",\"nullable\":false,\"metadata\":{}}";
case REPEATED:
String arrayType = arrayType(field, false);
return "{\"name\":\"" + field.getName() + "\",\"type\":" + arrayType
+ ",\"nullable\":false,\"metadata\":{}}";
default:
throw new UnsupportedOperationException("Unsupport convert " + field + " to spark sql type");
}
}).reduce((a, b) -> a + "," + b).orElse("");
return "{\"type\":\"struct\",\"fields\":[" + fieldsJsonString + "]}";
}
private static String convertFieldType(Type field) {
if (field instanceof PrimitiveType) {
return "\"" + convertPrimitiveType((PrimitiveType) field) + "\"";
} else {
assert field instanceof GroupType;
return convertGroupField((GroupType) field);
}
}
private static String convertPrimitiveType(PrimitiveType field) {
PrimitiveType.PrimitiveTypeName typeName = field.getPrimitiveTypeName();
OriginalType originalType = field.getOriginalType();
switch (typeName) {
case BOOLEAN: return "boolean";
case FLOAT: return "float";
case DOUBLE: return "double";
case INT32:
if (originalType == null) {
return "integer";
}
switch (originalType) {
case INT_8: return "byte";
case INT_16: return "short";
case INT_32: return "integer";
case DATE: return "date";
case DECIMAL:
return "decimal(" + field.getDecimalMetadata().getPrecision() + ","
+ field.getDecimalMetadata().getScale() + ")";
default: throw new UnsupportedOperationException("Unsupport convert " + typeName + " to spark sql type");
}
case INT64:
if (originalType == null) {
return "long";
}
switch (originalType) {
case INT_64: return "long";
case DECIMAL:
return "decimal(" + field.getDecimalMetadata().getPrecision() + ","
+ field.getDecimalMetadata().getScale() + ")";
case TIMESTAMP_MICROS:
case TIMESTAMP_MILLIS:
return "timestamp";
default:
throw new UnsupportedOperationException("Unsupport convert " + typeName + " to spark sql type");
}
case INT96: return "timestamp";
case BINARY:
if (originalType == null) {
return "binary";
}
switch (originalType) {
case UTF8:
case ENUM:
case JSON:
return "string";
case BSON: return "binary";
case DECIMAL:
return "decimal(" + field.getDecimalMetadata().getPrecision() + ","
+ field.getDecimalMetadata().getScale() + ")";
default:
throw new UnsupportedOperationException("Unsupport convert " + typeName + " to spark sql type");
}
case FIXED_LEN_BYTE_ARRAY:
switch (originalType) {
case DECIMAL:
return "decimal(" + field.getDecimalMetadata().getPrecision() + ","
+ field.getDecimalMetadata().getScale() + ")";
default:
throw new UnsupportedOperationException("Unsupport convert " + typeName + " to spark sql type");
}
default:
throw new UnsupportedOperationException("Unsupport convert " + typeName + " to spark sql type");
}
}
private static String convertGroupField(GroupType field) {
if (field.getOriginalType() == null) {
return convertToSparkSchemaJson(field);
}
switch (field.getOriginalType()) {
case LIST:
ValidationUtils.checkArgument(field.getFieldCount() == 1, "Illegal List type: " + field);
Type repeatedType = field.getType(0);
if (isElementType(repeatedType, field.getName())) {
return arrayType(repeatedType, false);
} else {
Type elementType = repeatedType.asGroupType().getType(0);
boolean optional = elementType.isRepetition(OPTIONAL);
return arrayType(elementType, optional);
}
case MAP:
case MAP_KEY_VALUE:
GroupType keyValueType = field.getType(0).asGroupType();
Type keyType = keyValueType.getType(0);
Type valueType = keyValueType.getType(1);
boolean valueOptional = valueType.isRepetition(OPTIONAL);
return "{\"type\":\"map\", \"keyType\":" + convertFieldType(keyType)
+ ",\"valueType\":" + convertFieldType(valueType)
+ ",\"valueContainsNull\":" + valueOptional + "}";
default:
throw new UnsupportedOperationException("Unsupport convert " + field + " to spark sql type");
}
}
private static String arrayType(Type elementType, boolean containsNull) {
return "{\"type\":\"array\", \"elementType\":" + convertFieldType(elementType) + ",\"containsNull\":" + containsNull + "}";
}
private static boolean isElementType(Type repeatedType, String parentName) {
return repeatedType.isPrimitive() || repeatedType.asGroupType().getFieldCount() > 1
|| repeatedType.getName().equals("array") || repeatedType.getName().equals(parentName + "_tuple");
}
}

2
hudi-sync/hudi-hive-sync/src/test/java/org/apache/hudi/hive/TestHiveSyncTool.java
View File

@@ -28,7 +28,7 @@ import org.apache.hudi.common.util.Option;
import org.apache.hudi.common.util.StringUtils;
import org.apache.hudi.common.util.collection.ImmutablePair;
import org.apache.hudi.hive.testutils.HiveTestUtil;
import org.apache.hudi.hive.util.ConfigUtils;
import org.apache.hudi.sync.common.util.ConfigUtils;
import org.apache.hudi.sync.common.AbstractSyncHoodieClient.PartitionEvent;
import org.apache.hudi.sync.common.AbstractSyncHoodieClient.PartitionEvent.PartitionEventType;

2
hudi-sync/hudi-hive-sync/src/test/java/org/apache/hudi/hive/TestParquet2SparkSchemaUtils.java
View File

@@ -18,7 +18,7 @@
package org.apache.hudi.hive;
import org.apache.hudi.hive.util.Parquet2SparkSchemaUtils;
import org.apache.hudi.sync.common.util.Parquet2SparkSchemaUtils;
import org.apache.spark.sql.execution.SparkSqlParser;
import org.apache.spark.sql.execution.datasources.parquet.SparkToParquetSchemaConverter;
import org.apache.spark.sql.internal.SQLConf;