[ https://issues.apache.org/jira/browse/SPARK-6632?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Michael Armbrust resolved SPARK-6632. ------------------------------------- Resolution: Fixed Fix Version/s: 1.5.0 Starting with Spark 1.5 I believe all footer reading is delegated to a spark job. > Optimize the parquetSchema to metastore schema reconciliation, so that the > process is delegated to each map task itself > ----------------------------------------------------------------------------------------------------------------------- > > Key: SPARK-6632 > URL: https://issues.apache.org/jira/browse/SPARK-6632 > Project: Spark > Issue Type: Improvement > Components: SQL > Affects Versions: 1.3.0 > Reporter: Yash Datta > Fix For: 1.5.0 > > > Currently in ParquetRelation2, schema from all the part files is first > merged, and then reconciled with metastore schema. This approach does not > scale in case we have thousands of partitions for the table. We can take a > different approach where we can go ahead with the metastore schema, and > reconcile the names of the columns within each map task , using ReadSupport > hooks provided in parquet. -- This message was sent by Atlassian JIRA (v6.3.4#6332) --------------------------------------------------------------------- To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org For additional commands, e-mail: issues-h...@spark.apache.org