make col counter generic

1 files changed, 37 insertions, 0 deletions

diff --git a/scalding/src/main/scala/sandcrawler/HBaseColCountJob.scala b/scalding/src/main/scala/sandcrawler/HBaseColCountJob.scala
new file mode 100644
index 0000000..a007339
--- /dev/null
+++ b/scalding/src/main/scala/sandcrawler/HBaseColCountJob.scala
@@ -0,0 +1,37 @@
+package sandcrawler
+
+import java.util.Properties
+
+import cascading.property.AppProps
+import cascading.tuple.Fields
+import com.twitter.scalding._
+import parallelai.spyglass.base.JobBase
+import parallelai.spyglass.hbase.HBaseConstants.SourceMode
+import parallelai.spyglass.hbase.HBasePipeConversions
+import parallelai.spyglass.hbase.HBaseSource
+
+class HBaseColCountJob(args: Args) extends JobBase(args) with HBasePipeConversions {
+
+  val output = args("output")
+
+  HBaseColCountJob.getHBaseSource(
+    args("hbase-table"),
+    args("zookeeper-hosts"),
+    args("column")
+    .read
+    .debug
+    .groupAll { _.size('count) }
+    .write(Tsv(output))
+}
+
+object HBaseColCountJob {
+
+  // eg, "wbgrp-journal-extract-0-qa",7 "mtrcs-zk1.us.archive.org:2181"
+  def getHBaseSource(hbaseTable: String, zookeeperHosts: String, col: String) : HBaseSource = {
+    HBaseBuilder.build(
+      hbaseTable,
+      zookeeperHosts,
+      List(col),
+      SourceMode.SCAN_ALL)
+  }
+}