sankarh commented on a change in pull request #1109:
URL: https://github.com/apache/hive/pull/1109#discussion_r452816142
##########
File path:
standalone-metastore/metastore-server/src/main/java/org/apache/hadoop/hive/metastore/cache/SharedCache.java
##########
@@ -261,44 +283,57 @@ public int getObjectSize(Class<?> clazz, Object obj) {
private Map<String, String> parameters;
private byte[] sdHash;
private int otherSize;
- private int tableColStatsCacheSize;
- private int partitionCacheSize;
- private int partitionColStatsCacheSize;
- private int aggrColStatsCacheSize;
+
+ // Arrays to hold the size/updated bit of cached objects.
+ // These arrays are to be referenced using MemberName enum only.
+ private int[] memberObjectsSize = new int[MemberName.values().length];
+ private AtomicBoolean[] memberCacheUpdated = new
AtomicBoolean[MemberName.values().length];
private ReentrantReadWriteLock tableLock = new
ReentrantReadWriteLock(true);
// For caching column stats for an unpartitioned table
// Key is column name and the value is the col stat object
private Map<String, ColumnStatisticsObj> tableColStatsCache = new
ConcurrentHashMap<String, ColumnStatisticsObj>();
- private AtomicBoolean isTableColStatsCacheDirty = new AtomicBoolean(false);
// For caching partition objects
// Ket is partition values and the value is a wrapper around the partition
object
private Map<String, PartitionWrapper> partitionCache = new
ConcurrentHashMap<String, PartitionWrapper>();
- private AtomicBoolean isPartitionCacheDirty = new AtomicBoolean(false);
// For caching column stats for a partitioned table
// Key is aggregate of partition values, column name and the value is the
col stat object
private Map<String, ColumnStatisticsObj> partitionColStatsCache =
new ConcurrentHashMap<String, ColumnStatisticsObj>();
- private AtomicBoolean isPartitionColStatsCacheDirty = new
AtomicBoolean(false);
// For caching aggregate column stats for all and all minus default
partition
// Key is column name and the value is a list of 2 col stat objects
// (all partitions and all but default)
private Map<String, List<ColumnStatisticsObj>> aggrColStatsCache =
new ConcurrentHashMap<String, List<ColumnStatisticsObj>>();
- private AtomicBoolean isAggrPartitionColStatsCacheDirty = new
AtomicBoolean(false);
+
+ private Map<String, SQLPrimaryKey> primaryKeyCache = new
ConcurrentHashMap<>();
+
+ private Map<String, SQLForeignKey> foreignKeyCache = new
ConcurrentHashMap<>();
+
+ private Map<String, SQLNotNullConstraint> notNullConstraintCache = new
ConcurrentHashMap<>();
+
+ private Map<String, SQLUniqueConstraint> uniqueConstraintCache = new
ConcurrentHashMap<>();
TableWrapper(Table t, byte[] sdHash, String location, Map<String, String>
parameters) {
this.t = t;
this.sdHash = sdHash;
this.location = location;
this.parameters = parameters;
- this.tableColStatsCacheSize = 0;
- this.partitionCacheSize = 0;
- this.partitionColStatsCacheSize = 0;
- this.aggrColStatsCacheSize = 0;
+ for(MemberName mn : MemberName.values()) {
+ this.memberObjectsSize[mn.getValue()] = 0;
Review comment:
In second thought, I think ordinal is better as we freshly load cache
entries during HMS startup. So, the ordering doesn't matter. However, setting
values can be a problem if someone pass incorrect value or remove an element
without updating other values.
----------------------------------------------------------------
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
---------------------------------------------------------------------
To unsubscribe, e-mail: gitbox-unsubscr...@hive.apache.org
For additional commands, e-mail: gitbox-h...@hive.apache.org
