diff options
Diffstat (limited to 'scalding/src/main/scala/sandcrawler/HBaseStatusCountJob.scala')
-rw-r--r-- | scalding/src/main/scala/sandcrawler/HBaseStatusCountJob.scala | 15 |
1 files changed, 6 insertions, 9 deletions
diff --git a/scalding/src/main/scala/sandcrawler/HBaseStatusCountJob.scala b/scalding/src/main/scala/sandcrawler/HBaseStatusCountJob.scala index f88f9db..fd0b4e2 100644 --- a/scalding/src/main/scala/sandcrawler/HBaseStatusCountJob.scala +++ b/scalding/src/main/scala/sandcrawler/HBaseStatusCountJob.scala @@ -15,21 +15,18 @@ import parallelai.spyglass.hbase.HBaseSource class HBaseStatusCountJob(args: Args) extends JobBase(args) with HBasePipeConversions { - val colSpec = "grobid0:status_code" - val output = args("output") - HBaseBuilder.parseColSpec(colSpec) - val Col: String = colSpec.split(":")(1) - - val source : TypedPipe[Long] = HBaseCountJob.getHBaseSource( + val source = HBaseCountJob.getHBaseSource( args("hbase-table"), args("zookeeper-hosts"), - colSpec) + "grobid0:status_code") + + val statusPipe : TypedPipe[Long] = source .read .toTypedPipe[(ImmutableBytesWritable,ImmutableBytesWritable)]('key, 'status_code) .map { case (key, raw_code) => Bytes.toLong(raw_code.copyBytes()) } - source.groupBy { identity } + statusPipe.groupBy { identity } .size .debug - .write(TypedTsv[(Long,Long)](output)) + .write(TypedTsv[(Long,Long)](args("output"))) } |