IIUC, you can use mapPartitions transformation and pass a function f. The function is used to map a tuple of input iterator to an output iterator. Upon the input iterator, you can process multiple records at a time.
> 在 2019年5月6日,上午2:59,swastik mittal <smitt...@ncsu.edu> 写道: > > From my experience in spark, when working on hdfs data base, spark reads data > in form of records and does computation on every record as soon as it reads > it. I have multiple images as my data on hdfs, where each image is a record. > I want spark to read multiple records before doing any computation. Any idea > on how could I do this? > > > > -- > Sent from: http://apache-spark-user-list.1001560.n3.nabble.com/ > > --------------------------------------------------------------------- > To unsubscribe e-mail: user-unsubscr...@spark.apache.org >