core/src/main/java/org/apache/carbondata/core/indexstore/BlockletIndexStore.java

/*
 * 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.carbondata.core.indexstore;

import java.io.IOException;
import java.util.ArrayList;
import java.util.HashMap;
import java.util.HashSet;
import java.util.List;
import java.util.Map;
import java.util.Set;
import java.util.concurrent.ConcurrentHashMap;

import org.apache.carbondata.common.logging.LogServiceFactory;
import org.apache.carbondata.core.cache.Cache;
import org.apache.carbondata.core.cache.CarbonLRUCache;
import org.apache.carbondata.core.constants.CarbonCommonConstants;
import org.apache.carbondata.core.datastore.block.SegmentPropertiesAndSchemaHolder;
import org.apache.carbondata.core.index.dev.Index;
import org.apache.carbondata.core.indexstore.blockletindex.BlockIndex;
import org.apache.carbondata.core.indexstore.blockletindex.BlockletIndexFactory;
import org.apache.carbondata.core.indexstore.blockletindex.BlockletIndexModel;
import org.apache.carbondata.core.indexstore.blockletindex.SegmentIndexFileStore;
import org.apache.carbondata.core.metadata.blocklet.DataFileFooter;
import org.apache.carbondata.core.metadata.schema.table.CarbonTable;
import org.apache.carbondata.core.util.BlockletIndexUtil;
import org.apache.carbondata.core.util.CarbonUtil;

import org.apache.hadoop.conf.Configuration;
import org.apache.log4j.Logger;

/**
 * Class to handle loading, unloading,clearing,storing of the table
 * blocks
 */
public class BlockletIndexStore
    implements Cache<TableBlockIndexUniqueIdentifierWrapper, BlockletIndexWrapper> {
  private static final Logger LOGGER =
      LogServiceFactory.getLogService(BlockletIndexStore.class.getName());
  /**
   * CarbonLRU cache
   */
  protected CarbonLRUCache lruCache;

  /**
   * map of block info to lock object map, while loading the btree this will be filled
   * and removed after loading the tree for that particular block info, this will be useful
   * while loading the tree concurrently so only block level lock will be applied another
   * block can be loaded concurrently
   */
  private Map<String, Object> segmentLockMap;

  /**
   * constructor to initialize the SegmentTaskIndexStore
   *
   * @param lruCache
   */
  public BlockletIndexStore(CarbonLRUCache lruCache) {
    this.lruCache = lruCache;
    segmentLockMap = new ConcurrentHashMap<String, Object>();
  }

  @Override
  public BlockletIndexWrapper get(TableBlockIndexUniqueIdentifierWrapper identifierWrapper) {
    return get(identifierWrapper, null);
  }

  public BlockletIndexWrapper get(TableBlockIndexUniqueIdentifierWrapper identifierWrapper,
      Map<String, Map<String, BlockMetaInfo>> segInfoCache) {
    TableBlockIndexUniqueIdentifier identifier =
        identifierWrapper.getTableBlockIndexUniqueIdentifier();
    String lruCacheKey = identifier.getUniqueTableSegmentIdentifier();
    BlockletIndexWrapper blockletIndexWrapper =
        (BlockletIndexWrapper) lruCache.get(lruCacheKey);
    List<BlockIndex> indexes = new ArrayList<>();
    if (blockletIndexWrapper == null) {
      try {
        SegmentIndexFileStore indexFileStore =
            new SegmentIndexFileStore(identifierWrapper.getConfiguration());
        Set<String> filesRead = new HashSet<>();
        String segmentFilePath = identifier.getIndexFilePath();
        if (segInfoCache == null) {
          segInfoCache = new HashMap<>();
        }
        Map<String, BlockMetaInfo> carbonDataFileBlockMetaInfoMapping =
            segInfoCache.get(segmentFilePath);
        if (carbonDataFileBlockMetaInfoMapping == null) {
          carbonDataFileBlockMetaInfoMapping =
              BlockletIndexUtil.createCarbonDataFileBlockMetaInfoMapping(segmentFilePath,
                  identifierWrapper.getConfiguration());
          segInfoCache.put(segmentFilePath, carbonDataFileBlockMetaInfoMapping);
        }
        // if the identifier is not a merge file we can directly load the indexes
        if (identifier.getMergeIndexFileName() == null) {
          List<DataFileFooter> indexInfos = new ArrayList<>();
          Map<String, BlockMetaInfo> blockMetaInfoMap = BlockletIndexUtil
              .getBlockMetaInfoMap(identifierWrapper, indexFileStore, filesRead,
                  carbonDataFileBlockMetaInfoMapping, indexInfos);
          BlockIndex blockIndex =
              loadAndGetIndex(identifier, indexFileStore, blockMetaInfoMap,
                  identifierWrapper.getCarbonTable(),
                  identifierWrapper.isAddToUnsafe(),
                  identifierWrapper.getConfiguration(),
                  identifierWrapper.isSerializeDmStore(),
                  indexInfos);
          indexes.add(blockIndex);
          blockletIndexWrapper =
              new BlockletIndexWrapper(identifier.getSegmentId(), indexes);
        } else {
          // if the identifier is a merge file then collect the index files and load the indexes
          List<TableBlockIndexUniqueIdentifier> tableBlockIndexUniqueIdentifiers =
              BlockletIndexUtil.getIndexFileIdentifiersFromMergeFile(identifier, indexFileStore);
          for (TableBlockIndexUniqueIdentifier blockIndexUniqueIdentifier :
              tableBlockIndexUniqueIdentifiers) {
            List<DataFileFooter> indexInfos = new ArrayList<>();
            Map<String, BlockMetaInfo> blockMetaInfoMap = BlockletIndexUtil.getBlockMetaInfoMap(
                new TableBlockIndexUniqueIdentifierWrapper(blockIndexUniqueIdentifier,
                    identifierWrapper.getCarbonTable()), indexFileStore, filesRead,
                carbonDataFileBlockMetaInfoMapping, indexInfos);
            if (!blockMetaInfoMap.isEmpty()) {
              BlockIndex blockIndex =
                  loadAndGetIndex(blockIndexUniqueIdentifier, indexFileStore, blockMetaInfoMap,
                      identifierWrapper.getCarbonTable(),
                      identifierWrapper.isAddToUnsafe(),
                      identifierWrapper.getConfiguration(),
                      identifierWrapper.isSerializeDmStore(),
                      indexInfos);
              indexes.add(blockIndex);
            }
          }
          blockletIndexWrapper =
              new BlockletIndexWrapper(identifier.getSegmentId(), indexes);
        }
        if (identifierWrapper.isAddTableBlockToUnsafeAndLRUCache()) {
          long expiration_time = CarbonUtil.getExpiration_time(identifierWrapper.getCarbonTable());
          lruCache.put(identifier.getUniqueTableSegmentIdentifier(), blockletIndexWrapper,
                  blockletIndexWrapper.getMemorySize(), expiration_time);
        }
      } catch (Throwable e) {
        // clear all the memory used by indexes loaded
        for (Index index : indexes) {
          index.clear();
        }
        LOGGER.error("memory exception when loading index: " + e.getMessage(), e);
        throw new RuntimeException(e);
      }
    }
    return blockletIndexWrapper;
  }

  @Override
  public List<BlockletIndexWrapper> getAll(
      List<TableBlockIndexUniqueIdentifierWrapper> tableSegmentUniqueIdentifiers)
      throws IOException {
    Map<String, Map<String, BlockMetaInfo>> segInfoCache =
        new HashMap<String, Map<String, BlockMetaInfo>>();

    List<BlockletIndexWrapper> blockletIndexWrappers =
        new ArrayList<>(tableSegmentUniqueIdentifiers.size());
    List<TableBlockIndexUniqueIdentifierWrapper> missedIdentifiersWrapper = new ArrayList<>();
    BlockletIndexWrapper blockletIndexWrapper = null;
    // Get the indexes for each index file from cache.
    try {
      for (TableBlockIndexUniqueIdentifierWrapper
               identifierWrapper : tableSegmentUniqueIdentifiers) {
        BlockletIndexWrapper indexWrapper = getIfPresent(identifierWrapper);
        if (indexWrapper != null) {
          blockletIndexWrappers.add(indexWrapper);
        } else {
          missedIdentifiersWrapper.add(identifierWrapper);
        }
      }
      if (missedIdentifiersWrapper.size() > 0) {
        for (TableBlockIndexUniqueIdentifierWrapper identifierWrapper : missedIdentifiersWrapper) {
          blockletIndexWrapper = get(identifierWrapper, segInfoCache);
          blockletIndexWrappers.add(blockletIndexWrapper);
        }
      }
    } catch (Throwable e) {
      if (null != blockletIndexWrapper) {
        List<BlockIndex> indexes = blockletIndexWrapper.getIndexes();
        for (Index index : indexes) {
          index.clear();
        }
      }
      throw new IOException("Problem in loading segment blocks: " + e.getMessage(), e);
    }

    return blockletIndexWrappers;
  }

  /**
   * returns the SegmentTaskIndexWrapper
   *
   * @param tableSegmentUniqueIdentifierWrapper
   * @return
   */
  @Override
  public BlockletIndexWrapper getIfPresent(
      TableBlockIndexUniqueIdentifierWrapper tableSegmentUniqueIdentifierWrapper) {
    return (BlockletIndexWrapper) lruCache.get(
        tableSegmentUniqueIdentifierWrapper.getTableBlockIndexUniqueIdentifier()
            .getUniqueTableSegmentIdentifier());
  }

  /**
   * method invalidate the segment cache for segment
   *
   * @param tableSegmentUniqueIdentifierWrapper
   */
  @Override
  public void invalidate(
      TableBlockIndexUniqueIdentifierWrapper tableSegmentUniqueIdentifierWrapper) {
    BlockletIndexWrapper blockletIndexWrapper =
        getIfPresent(tableSegmentUniqueIdentifierWrapper);
    if (null != blockletIndexWrapper) {
      // clear the segmentProperties cache
      List<BlockIndex> indexes = blockletIndexWrapper.getIndexes();
      if (null != indexes && !indexes.isEmpty()) {
        String segmentId =
            tableSegmentUniqueIdentifierWrapper.getTableBlockIndexUniqueIdentifier().getSegmentId();
        // as segmentId will be same for all the indexes and segmentProperties cache is
        // maintained at segment level so it need to be called only once for clearing
        SegmentPropertiesAndSchemaHolder.getInstance()
            .invalidate(segmentId, indexes.get(0).getSegmentPropertiesWrapper(),
                tableSegmentUniqueIdentifierWrapper.isAddTableBlockToUnsafeAndLRUCache());
      }
    }
    lruCache.remove(tableSegmentUniqueIdentifierWrapper.getTableBlockIndexUniqueIdentifier()
        .getUniqueTableSegmentIdentifier());
  }

  @Override
  public void put(TableBlockIndexUniqueIdentifierWrapper tableBlockIndexUniqueIdentifierWrapper,
      BlockletIndexWrapper wrapper) throws IOException {
    // As index will use unsafe memory, it is not recommended to overwrite an existing entry
    // as in that case clearing unsafe memory need to be taken card. If at all index entry
    // in the cache need to be overwritten then use the invalidate interface
    // and then use the put interface
    if (null == getIfPresent(tableBlockIndexUniqueIdentifierWrapper)) {
      List<BlockIndex> indexes = wrapper.getIndexes();
      try {
        for (BlockIndex blockIndex : indexes) {
          blockIndex.convertToUnsafeDMStore();
        }
        // get cacheExpirationTime for table from tableProperties
        long expirationTime =
            CarbonUtil.getExpiration_time(tableBlockIndexUniqueIdentifierWrapper.getCarbonTable());
        // Locking is not required here because in LRU cache map add method is synchronized to add
        // only one entry at a time and if a key already exists it will not overwrite the entry
        lruCache.put(tableBlockIndexUniqueIdentifierWrapper.getTableBlockIndexUniqueIdentifier()
            .getUniqueTableSegmentIdentifier(), wrapper, wrapper.getMemorySize(), expirationTime);
      } catch (Throwable e) {
        // clear all the memory acquired by index in case of any failure
        for (Index blockletIndex : indexes) {
          blockletIndex.clear();
        }
        throw new IOException("Problem in adding index to cache.", e);
      }
    }
  }

  /**
   * Below method will be used to load the segment of segments
   * One segment may have multiple task , so  table segment will be loaded
   * based on task id and will return the map of taskId to table segment
   * map
   *
   * @return map of taks id to segment mapping
   * @throws IOException
   */
  private BlockIndex loadAndGetIndex(TableBlockIndexUniqueIdentifier identifier,
      SegmentIndexFileStore indexFileStore, Map<String, BlockMetaInfo> blockMetaInfoMap,
      CarbonTable carbonTable, boolean addTableBlockToUnsafe, Configuration configuration,
      boolean serializeDmStore, List<DataFileFooter> indexInfos) throws IOException {
    String uniqueTableSegmentIdentifier =
        identifier.getUniqueTableSegmentIdentifier();
    Object lock = segmentLockMap.get(uniqueTableSegmentIdentifier);
    if (lock == null) {
      lock = addAndGetSegmentLock(uniqueTableSegmentIdentifier);
    }
    BlockIndex blockIndex;
    synchronized (lock) {
      blockIndex = (BlockIndex) BlockletIndexFactory.createIndex(carbonTable);
      final BlockletIndexModel blockletIndexModel = new BlockletIndexModel(carbonTable,
          identifier.getIndexFilePath() + CarbonCommonConstants.FILE_SEPARATOR + identifier
              .getIndexFileName(), indexFileStore.getFileData(identifier.getIndexFileName()),
          blockMetaInfoMap, identifier.getSegmentId(), addTableBlockToUnsafe, configuration,
          serializeDmStore);
      blockletIndexModel.setIndexInfos(indexInfos);
      blockIndex.init(blockletIndexModel);
    }
    return blockIndex;
  }

  /**
   * Below method will be used to get the segment level lock object
   *
   * @param uniqueIdentifier
   * @return lock object
   */
  private synchronized Object addAndGetSegmentLock(String uniqueIdentifier) {
    // get the segment lock object if it is present then return
    // otherwise add the new lock and return
    Object segmentLockObject = segmentLockMap.get(uniqueIdentifier);
    if (null == segmentLockObject) {
      segmentLockObject = new Object();
      segmentLockMap.put(uniqueIdentifier, segmentLockObject);
    }
    return segmentLockObject;
  }

  /**
   * The method clears the access count of table segments
   *
   * @param tableSegmentUniqueIdentifiersWrapper
   */
  @Override
  public void clearAccessCount(
      List<TableBlockIndexUniqueIdentifierWrapper> tableSegmentUniqueIdentifiersWrapper) {
    for (TableBlockIndexUniqueIdentifierWrapper
             identifierWrapper : tableSegmentUniqueIdentifiersWrapper) {
      BlockIndex cacheable = (BlockIndex) lruCache.get(
          identifierWrapper.getTableBlockIndexUniqueIdentifier().getUniqueTableSegmentIdentifier());
      cacheable.clear();
    }
  }
}