aboutsummaryrefslogtreecommitdiffstats
path: root/scalding/src/main
diff options
context:
space:
mode:
authorEllen Spertus <ellen.spertus@gmail.com>2018-06-06 09:40:04 -0700
committerEllen Spertus <ellen.spertus@gmail.com>2018-06-06 09:40:04 -0700
commitcc54dea9c77f8386bafd71977b2be1b3b21f0aae (patch)
tree6f61a189276ea99c4a15dde55969155ba619809f /scalding/src/main
parente746171526f578789c55d7aa690cc8381e139439 (diff)
downloadsandcrawler-cc54dea9c77f8386bafd71977b2be1b3b21f0aae.tar.gz
sandcrawler-cc54dea9c77f8386bafd71977b2be1b3b21f0aae.zip
Added job and test for counting mime types.
Diffstat (limited to 'scalding/src/main')
-rw-r--r--scalding/src/main/scala/sandcrawler/HBaseMimeCountJob.scala28
1 files changed, 28 insertions, 0 deletions
diff --git a/scalding/src/main/scala/sandcrawler/HBaseMimeCountJob.scala b/scalding/src/main/scala/sandcrawler/HBaseMimeCountJob.scala
new file mode 100644
index 0000000..819a652
--- /dev/null
+++ b/scalding/src/main/scala/sandcrawler/HBaseMimeCountJob.scala
@@ -0,0 +1,28 @@
+package sandcrawler
+
+import cascading.property.AppProps
+import cascading.tuple.Fields
+import com.twitter.scalding._
+import java.util.Properties
+import parallelai.spyglass.base.JobBase
+import parallelai.spyglass.hbase.{HBaseSource, HBasePipeConversions}
+import parallelai.spyglass.hbase.HBaseConstants.SourceMode
+
+class HBaseMimeCountJob(args: Args) extends JobBase(args) with HBasePipeConversions {
+ val output = args("output")
+
+ HBaseMimeCountJob.getHBaseSource
+ .read
+ .fromBytesWritable(List('mime))
+ .debug
+ .groupBy('mime){group => group.size('count)}
+ .write(Tsv(output))
+}
+
+object HBaseMimeCountJob {
+ def getHBaseSource = HBaseBuilder.build(
+ "wbgrp-journal-extract-0-qa", // HBase Table Name
+ "mtrcs-zk1.us.archive.org:2181", // HBase Zookeeper server (to get runtime config info; can be array?)
+ List("file:mime"),
+ SourceMode.SCAN_ALL)
+}