Use RDD.mapPartitions to go over all the items in a partition with one Mapper
object. It will look something like this:
rdd.mapPartitions(iterator =>
val mapper = new myown.Mapper()
mapper.configure(conf)
val output = // {{create an OutputCollector that stores stuff in an
ArrayBuffer}}
for ((key, value) <- iterator) {
mapper.map(key, value, output, Reporter.NULL)
}
output
}
On Jun 5, 2014, at 8:12 AM, Wei Tan <[email protected]> wrote:
> Thanks Matei.
>
> Using your pointers I can import data frrom HDFS, what I want to do now is
> something like this in Spark:
>
> -----------------------
> import myown.mapper
>
> rdd.map (mapper.map)
> -----------------------
>
> The reason why I want this: myown.mapper is a java class I already developed.
> I used to run it in Hadoop. It is fairly complex and relies on a lot of
> utility java classes I wrote. Can I reuse the map function in java and port
> it into Spark?
>
> Best regards,
> Wei
>
>
> ---------------------------------
> Wei Tan, PhD
> Research Staff Member
> IBM T. J. Watson Research Center
> http://researcher.ibm.com/person/us-wtan
>
>
>
> From: Matei Zaharia <[email protected]>
> To: [email protected],
> Date: 06/04/2014 04:28 PM
> Subject: Re: reuse hadoop code in Spark
>
>
>
> Yes, you can write some glue in Spark to call these. Some functions to look
> at:
>
> - SparkContext.hadoopRDD lets you create an input RDD from an existing
> JobConf configured by Hadoop (including InputFormat, paths, etc)
> - RDD.mapPartitions lets you operate in all the values on one partition
> (block) at a time, similar to how Mappers in MapReduce work
> - PairRDDFunctions.reduceByKey and groupByKey can be used for aggregation.
> - RDD.pipe() can be used to call out to a script or binary, like Hadoop
> Streaming.
>
> A fair number of people have been running both Java and Hadoop Streaming apps
> like this.
>
> Matei
>
> On Jun 4, 2014, at 1:08 PM, Wei Tan <[email protected]> wrote:
>
> Hello,
>
> I am trying to use spark in such a scenario:
>
> I have code written in Hadoop and now I try to migrate to Spark. The mappers
> and reducers are fairly complex. So I wonder if I can reuse the map()
> functions I already wrote in Hadoop (Java), and use Spark to chain them,
> mixing the Java map() functions with Spark operators?
>
> Another related question, can I use binary as operators, like Hadoop
> streaming?
>
> Thanks!
> Wei
>
>
>