[GitHub] [spark] HeartSaVioR commented on a diff in pull request #38430: [SPARK-40957] Add in memory cache in HDFSMetadataLog

[GitBox <gi...@apache.org>]

HeartSaVioR commented on code in PR #38430:
URL: https://github.com/apache/spark/pull/38430#discussion_r1008968288


##########
sql/core/src/main/scala/org/apache/spark/sql/execution/streaming/HDFSMetadataLog.scala:
##########
@@ -64,6 +67,17 @@ class HDFSMetadataLog[T <: AnyRef : ClassTag](sparkSession: SparkSession, path:
     fileManager.mkdirs(metadataPath)
   }
 
+  protected val metadataCacheEnabled: Boolean
+  = sparkSession.sessionState.conf.getConf(SQLConf.STREAMING_METADATA_CACHE_ENABLED)
+
+  /**
+   * Cache the latest two batches. [[StreamExecution]] usually just accesses the latest two batches
+   * when committing offsets, this cache will save some file system operations.
+   */
+  protected[sql] val batchCache = Collections.synchronizedMap(new LinkedHashMap[Long, T](2) {

Review Comment:
   In the previous implementation we didn't use Guava cache. We don't want to add more coupling to Guava unless it is quite bothering to implement by our own.
   
   I provided the diff upon caching offset seq, see here https://github.com/apache/spark/pull/31495
   This is a generalization of previous cache from offset seq to HDFSMetadataLog, which I actually got requested in my PR as well. As of now, the cache is only effective to offset seq, but we plan to propose functionalities/features where caching commit log would be helpful.
   



-- 
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.

To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org

For queries about this service, please contact Infrastructure at:
users@infra.apache.org


---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org