aboutsummaryrefslogtreecommitdiffstats
path: root/scalding/src/main/scala
diff options
context:
space:
mode:
Diffstat (limited to 'scalding/src/main/scala')
-rw-r--r--scalding/src/main/scala/sandcrawler/HBaseBuilder.scala8
-rw-r--r--scalding/src/main/scala/sandcrawler/HBaseRowCountJob.scala2
2 files changed, 7 insertions, 3 deletions
diff --git a/scalding/src/main/scala/sandcrawler/HBaseBuilder.scala b/scalding/src/main/scala/sandcrawler/HBaseBuilder.scala
index 35aeb46..eb108a2 100644
--- a/scalding/src/main/scala/sandcrawler/HBaseBuilder.scala
+++ b/scalding/src/main/scala/sandcrawler/HBaseBuilder.scala
@@ -15,8 +15,12 @@ object HBaseBuilder {
// Code from https://stackoverflow.com/a/50595189/6310511
val inverseSchema = for ((k, vs) <- schema; v <- vs) yield (k + ":" + v, k)
+ // The argument should be a string with a comma-separated list of family:column,
+ // such as "f:c, file:size". Spaces after the comma are optional.
@throws(classOf[IllegalArgumentException])
- def parseColSpec(colSpecs: List[String]) : (List[String], List[Fields]) = {
+ def parseColSpec(colSpec: String) : (List[String], List[Fields]) = {
+ val colSpecs = (if (colSpec.trim.length == 0) List() else colSpec.split(", *").toList)
+
// Verify that all column specifiers are legal.
for (colSpec <- colSpecs) {
if (!(inverseSchema contains colSpec)) {
@@ -40,7 +44,7 @@ object HBaseBuilder {
(families, groupedColNames.map({fields => new Fields(fields : _*)}))
}
- def build(table: String, server: String, colSpec: List[String], sourceMode: SourceMode, keyList: List[String] = List("key")) = {
+ def build(table: String, server: String, colSpec: String, sourceMode: SourceMode, keyList: List[String] = List("key")) = {
val (families, fields) = parseColSpec(colSpec)
new HBaseSource(table, server, new Fields("key"), families, fields, sourceMode = sourceMode, keyList = keyList)
}
diff --git a/scalding/src/main/scala/sandcrawler/HBaseRowCountJob.scala b/scalding/src/main/scala/sandcrawler/HBaseRowCountJob.scala
index 9d074ab..041df4d 100644
--- a/scalding/src/main/scala/sandcrawler/HBaseRowCountJob.scala
+++ b/scalding/src/main/scala/sandcrawler/HBaseRowCountJob.scala
@@ -27,6 +27,6 @@ object HBaseRowCountJob {
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:size", "file:mime"),
+ "file:size, file:mime",
SourceMode.SCAN_ALL)
}