[ https://issues.apache.org/jira/browse/SPARK-11519?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Sean Owen resolved SPARK-11519. ------------------------------- Resolution: Not A Problem Not quite sure what you're reporting, but it is expected that objects returned from InputFormats are reused. Callers have to clone the objects if storing them. > Spark MemoryStore with hadoop SequenceFile cache the values is same record. > --------------------------------------------------------------------------- > > Key: SPARK-11519 > URL: https://issues.apache.org/jira/browse/SPARK-11519 > Project: Spark > Issue Type: Bug > Affects Versions: 1.1.0 > Environment: jdk.1.7.0, spark1.1.0, hadoop2.3.0 > Reporter: xukaiqiang > > use spark create newAPIHadoopFile which is SequenceFile format, when use > spark memory cache, the cache save the same java object . > read hadoop file with SequenceFileRecordReader save as NewHadoopRDD. the kv > values as : > [1, com.data.analysis.domain.RecordObject@54cdb594] > [2, com.data.analysis.domain.RecordObject@54cdb594] > [3, com.data.analysis.domain.RecordObject@54cdb594] > although the value is the same java object , but i am sure the context is not > same . > jsut use spark memory cache, the MemoryStore vector save all records, but > the value is the last vlaue from newHadoopRDD. -- 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