akashrn5 commented on a change in pull request #3908: URL: https://github.com/apache/carbondata/pull/3908#discussion_r486982098
########## File path: integration/spark/src/main/scala/org/apache/spark/util/PartitionCacheManager.scala ########## @@ -0,0 +1,181 @@ +/* + * 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.spark.util + +import java.net.URI +import java.util + +import scala.collection.JavaConverters._ +import scala.collection.mutable + +import org.apache.log4j.Logger +import org.apache.spark.sql.catalyst.catalog.{CatalogStorageFormat, CatalogTablePartition} + +import org.apache.carbondata.common.logging.LogServiceFactory +import org.apache.carbondata.core.cache.{Cache, Cacheable, CarbonLRUCache} +import org.apache.carbondata.core.constants.CarbonCommonConstants +import org.apache.carbondata.core.datastore.impl.FileFactory +import org.apache.carbondata.core.index.Segment +import org.apache.carbondata.core.metadata.{AbsoluteTableIdentifier, SegmentFileStore} +import org.apache.carbondata.core.statusmanager.{LoadMetadataDetails, SegmentStatusManager} +import org.apache.carbondata.core.util.path.CarbonTablePath + +object PartitionCacheManager extends Cache[PartitionCacheKey, + java.util.List[CatalogTablePartition]] { + + private val CACHE = new CarbonLRUCache( + CarbonCommonConstants.CARBON_PARTITION_MAX_DRIVER_LRU_CACHE_SIZE, + CarbonCommonConstants.CARBON_MAX_LRU_CACHE_SIZE_DEFAULT) + + val LOGGER: Logger = LogServiceFactory.getLogService(this.getClass.getName) + + def get(identifier: PartitionCacheKey): java.util.List[CatalogTablePartition] = { + LOGGER.info("Reading partition values from store") + // read the tableStatus file to get valid and invalid segments + val validInvalidSegments = new SegmentStatusManager(AbsoluteTableIdentifier.from( + identifier.tablePath, null, null, identifier.tableId)) + .getValidAndInvalidSegments + val cacheablePartitionSpecs = validInvalidSegments.getValidSegments.asScala.map { segment => + val segmentFileName = segment.getSegmentFileName + val segmentFilePath = FileFactory.getCarbonFile( + CarbonTablePath.getSegmentFilePath(identifier.tablePath, segmentFileName)) + // read the last modified time + val segmentFileModifiedTime = segmentFilePath.getLastModifiedTime + val existingCache = CACHE.get(identifier.tableId) + if (existingCache != null) { + val segmentCache = CACHE.get(identifier.tableId).asInstanceOf[CacheablePartitionSpec] + .partitionSpecs.get(segment.getSegmentNo) + segmentCache match { + case Some(c) => + // check if cache needs to be updated + if (segmentFileModifiedTime > c._2) { Review comment: As i can see from here, we are basically removing the cache if the segment file is updated. Actually same thing we do in `TableSegmentRefresher`, and use that for updating all cache, can move this piece of logic there also to unify for all? we use `TableSegmentRefresher` during query, but here we are doing in case of load, can you try if we can do that? If not this PR, may be you can give a try in another, as it will be unified and good. ########## File path: integration/spark/src/main/scala/org/apache/spark/util/PartitionCacheManager.scala ########## @@ -0,0 +1,181 @@ +/* + * 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.spark.util + +import java.net.URI +import java.util + +import scala.collection.JavaConverters._ +import scala.collection.mutable + +import org.apache.log4j.Logger +import org.apache.spark.sql.catalyst.catalog.{CatalogStorageFormat, CatalogTablePartition} + +import org.apache.carbondata.common.logging.LogServiceFactory +import org.apache.carbondata.core.cache.{Cache, Cacheable, CarbonLRUCache} +import org.apache.carbondata.core.constants.CarbonCommonConstants +import org.apache.carbondata.core.datastore.impl.FileFactory +import org.apache.carbondata.core.index.Segment +import org.apache.carbondata.core.metadata.{AbsoluteTableIdentifier, SegmentFileStore} +import org.apache.carbondata.core.statusmanager.{LoadMetadataDetails, SegmentStatusManager} +import org.apache.carbondata.core.util.path.CarbonTablePath + +object PartitionCacheManager extends Cache[PartitionCacheKey, + java.util.List[CatalogTablePartition]] { + + private val CACHE = new CarbonLRUCache( + CarbonCommonConstants.CARBON_PARTITION_MAX_DRIVER_LRU_CACHE_SIZE, + CarbonCommonConstants.CARBON_MAX_LRU_CACHE_SIZE_DEFAULT) + + val LOGGER: Logger = LogServiceFactory.getLogService(this.getClass.getName) + + def get(identifier: PartitionCacheKey): java.util.List[CatalogTablePartition] = { + LOGGER.info("Reading partition values from store") + // read the tableStatus file to get valid and invalid segments + val validInvalidSegments = new SegmentStatusManager(AbsoluteTableIdentifier.from( + identifier.tablePath, null, null, identifier.tableId)) + .getValidAndInvalidSegments + val cacheablePartitionSpecs = validInvalidSegments.getValidSegments.asScala.map { segment => + val segmentFileName = segment.getSegmentFileName + val segmentFilePath = FileFactory.getCarbonFile( + CarbonTablePath.getSegmentFilePath(identifier.tablePath, segmentFileName)) + // read the last modified time + val segmentFileModifiedTime = segmentFilePath.getLastModifiedTime + val existingCache = CACHE.get(identifier.tableId) + if (existingCache != null) { + val segmentCache = CACHE.get(identifier.tableId).asInstanceOf[CacheablePartitionSpec] + .partitionSpecs.get(segment.getSegmentNo) + segmentCache match { + case Some(c) => + // check if cache needs to be updated + if (segmentFileModifiedTime > c._2) { + (segment.getSegmentNo, (readPartition(identifier, + segmentFilePath.getAbsolutePath), segmentFileModifiedTime)) + } else { + (segment.getSegmentNo, c) + } + case None => + (segment.getSegmentNo, (readPartition(identifier, + segmentFilePath.getAbsolutePath), segmentFileModifiedTime)) + } + } else { + // read the partitions if not available in cache. + (segment.getSegmentNo, (readPartition(identifier, Review comment: As you mentioned first time it will take lot of time, but query time we will get benefit. You said matched partitions and since here for every load we are trying to cache, we wont be able to get any matched partitions right? It is something similar to our Indexes(MV, SI) etc where we load all segment at once in first time and from next time we do incremental, as @QiangCai mentioned. So here from subsequent loads, only that segment should be considered, which i think its being taken care. @ajantha-bhat are you trying to say to change to handle in query or load itself? Correct me, if i'm wrong. ########## File path: integration/spark/src/test/scala/org/apache/carbondata/spark/testsuite/standardpartition/StandardPartitionTableLoadingTestCase.scala ########## @@ -585,6 +587,30 @@ class StandardPartitionTableLoadingTestCase extends QueryTest with BeforeAndAfte } } + test("test partition caching") { + CarbonProperties.getInstance().addProperty("carbon.read.partition.hive.direct", "false") + sql("drop table if exists partition_cache") + sql("create table partition_cache(a string) partitioned by(b int) stored as carbondata") + sql("insert into partition_cache select 'k',1") + sql("select * from partition_cache where b = 1").collect() Review comment: can we add one assert to check the cache, after the load instead of query in one scenario? ########## File path: integration/spark/src/main/scala/org/apache/spark/util/PartitionCacheManager.scala ########## @@ -0,0 +1,181 @@ +/* + * 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.spark.util + +import java.net.URI +import java.util + +import scala.collection.JavaConverters._ +import scala.collection.mutable + +import org.apache.log4j.Logger +import org.apache.spark.sql.catalyst.catalog.{CatalogStorageFormat, CatalogTablePartition} + +import org.apache.carbondata.common.logging.LogServiceFactory +import org.apache.carbondata.core.cache.{Cache, Cacheable, CarbonLRUCache} +import org.apache.carbondata.core.constants.CarbonCommonConstants +import org.apache.carbondata.core.datastore.impl.FileFactory +import org.apache.carbondata.core.index.Segment +import org.apache.carbondata.core.metadata.{AbsoluteTableIdentifier, SegmentFileStore} +import org.apache.carbondata.core.statusmanager.{LoadMetadataDetails, SegmentStatusManager} +import org.apache.carbondata.core.util.path.CarbonTablePath + +object PartitionCacheManager extends Cache[PartitionCacheKey, + java.util.List[CatalogTablePartition]] { + + private val CACHE = new CarbonLRUCache( + CarbonCommonConstants.CARBON_PARTITION_MAX_DRIVER_LRU_CACHE_SIZE, + CarbonCommonConstants.CARBON_MAX_LRU_CACHE_SIZE_DEFAULT) + + val LOGGER: Logger = LogServiceFactory.getLogService(this.getClass.getName) + + def get(identifier: PartitionCacheKey): java.util.List[CatalogTablePartition] = { + LOGGER.info("Reading partition values from store") + // read the tableStatus file to get valid and invalid segments + val validInvalidSegments = new SegmentStatusManager(AbsoluteTableIdentifier.from( + identifier.tablePath, null, null, identifier.tableId)) + .getValidAndInvalidSegments + val cacheablePartitionSpecs = validInvalidSegments.getValidSegments.asScala.map { segment => + val segmentFileName = segment.getSegmentFileName + val segmentFilePath = FileFactory.getCarbonFile( + CarbonTablePath.getSegmentFilePath(identifier.tablePath, segmentFileName)) + // read the last modified time + val segmentFileModifiedTime = segmentFilePath.getLastModifiedTime + val existingCache = CACHE.get(identifier.tableId) + if (existingCache != null) { + val segmentCache = CACHE.get(identifier.tableId).asInstanceOf[CacheablePartitionSpec] + .partitionSpecs.get(segment.getSegmentNo) + segmentCache match { + case Some(c) => + // check if cache needs to be updated + if (segmentFileModifiedTime > c._2) { + (segment.getSegmentNo, (readPartition(identifier, + segmentFilePath.getAbsolutePath), segmentFileModifiedTime)) + } else { + (segment.getSegmentNo, c) + } + case None => + (segment.getSegmentNo, (readPartition(identifier, + segmentFilePath.getAbsolutePath), segmentFileModifiedTime)) + } + } else { + // read the partitions if not available in cache. + (segment.getSegmentNo, (readPartition(identifier, Review comment: and i also checked now, this caching happens in flow query too, so you can consider my reply specific to load. ########## File path: integration/spark/src/main/scala/org/apache/spark/util/PartitionCacheManager.scala ########## @@ -0,0 +1,181 @@ +/* + * 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.spark.util + +import java.net.URI +import java.util + +import scala.collection.JavaConverters._ +import scala.collection.mutable + +import org.apache.log4j.Logger +import org.apache.spark.sql.catalyst.catalog.{CatalogStorageFormat, CatalogTablePartition} + +import org.apache.carbondata.common.logging.LogServiceFactory +import org.apache.carbondata.core.cache.{Cache, Cacheable, CarbonLRUCache} +import org.apache.carbondata.core.constants.CarbonCommonConstants +import org.apache.carbondata.core.datastore.impl.FileFactory +import org.apache.carbondata.core.index.Segment +import org.apache.carbondata.core.metadata.{AbsoluteTableIdentifier, SegmentFileStore} +import org.apache.carbondata.core.statusmanager.{LoadMetadataDetails, SegmentStatusManager} +import org.apache.carbondata.core.util.path.CarbonTablePath + +object PartitionCacheManager extends Cache[PartitionCacheKey, + java.util.List[CatalogTablePartition]] { + + private val CACHE = new CarbonLRUCache( + CarbonCommonConstants.CARBON_PARTITION_MAX_DRIVER_LRU_CACHE_SIZE, + CarbonCommonConstants.CARBON_MAX_LRU_CACHE_SIZE_DEFAULT) + + val LOGGER: Logger = LogServiceFactory.getLogService(this.getClass.getName) + + def get(identifier: PartitionCacheKey): java.util.List[CatalogTablePartition] = { + LOGGER.info("Reading partition values from store") + // read the tableStatus file to get valid and invalid segments + val validInvalidSegments = new SegmentStatusManager(AbsoluteTableIdentifier.from( + identifier.tablePath, null, null, identifier.tableId)) + .getValidAndInvalidSegments + val cacheablePartitionSpecs = validInvalidSegments.getValidSegments.asScala.map { segment => + val segmentFileName = segment.getSegmentFileName + val segmentFilePath = FileFactory.getCarbonFile( + CarbonTablePath.getSegmentFilePath(identifier.tablePath, segmentFileName)) + // read the last modified time + val segmentFileModifiedTime = segmentFilePath.getLastModifiedTime + val existingCache = CACHE.get(identifier.tableId) + if (existingCache != null) { + val segmentCache = CACHE.get(identifier.tableId).asInstanceOf[CacheablePartitionSpec] + .partitionSpecs.get(segment.getSegmentNo) + segmentCache match { + case Some(c) => + // check if cache needs to be updated + if (segmentFileModifiedTime > c._2) { Review comment: yes, you are right, i didn't observe that module level. ---------------------------------------------------------------- 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: us...@infra.apache.org
