Merge branch 'refactoring' into 'master'

Refactoring to add, use, and test class HBaseBuilder to eliminate duplicated code and facilitate HBaseSource creation

See merge request webgroup/sandcrawler!1

2 files changed, 46 insertions, 11 deletions

diff --git a/scalding/src/test/scala/sandcrawler/HBaseBuilderTest.scala b/scalding/src/test/scala/sandcrawler/HBaseBuilderTest.scala
new file mode 100644
index 0000000..b45751d
--- /dev/null
+++ b/scalding/src/test/scala/sandcrawler/HBaseBuilderTest.scala
@@ -0,0 +1,43 @@
+package example
+
+import cascading.tuple.Fields
+import org.scalatest._
+import sandcrawler.HBaseBuilder
+
+class HBaseBuilderTest extends FlatSpec with Matchers {
+  "parseColSpec()" should "work on legal nontrivial input" in {
+    val (fams, fields) = HBaseBuilder.parseColSpec(List("file:size", "file:cdx", "match0:status"))
+    fams should have length 2
+    fields should have length 2
+    val fileIndex = fams.indexOf("file")
+    fileIndex should not be -1
+    fields(fileIndex) should be (new Fields("size", "cdx"))
+    val match0Index = fams.indexOf("match0")
+    match0Index should not be -1
+    fields(match0Index) should be (new Fields("status"))
+  }
+
+  it should "work on empty input" in {
+    val (fams, fields) = HBaseBuilder.parseColSpec(List())
+    fams should have length 0
+    fields should have length 0
+  }
+
+  it should "throw IllegalArgumentException on malformed input" in {
+    a [IllegalArgumentException] should be thrownBy {
+      HBaseBuilder.parseColSpec(List("file_size"))
+    }
+  }
+
+  it should "throw IllegalArgumentException on nonexistent family" in {
+    a [IllegalArgumentException] should be thrownBy {
+      HBaseBuilder.parseColSpec(List("foo:bar"))
+    }
+  }
+
+  it should "throw IllegalArgumentException on nonexistent column" in {
+    a [IllegalArgumentException] should be thrownBy {
+      HBaseBuilder.parseColSpec(List("file:bar"))
+    }
+  }
+}
diff --git a/scalding/src/test/scala/sandcrawler/HBaseRowCountTest.scala b/scalding/src/test/scala/sandcrawler/HBaseRowCountTest.scala
index 94b3740..abb017c 100644
--- a/scalding/src/test/scala/sandcrawler/HBaseRowCountTest.scala
+++ b/scalding/src/test/scala/sandcrawler/HBaseRowCountTest.scala
@@ -12,6 +12,7 @@ import scala._
 import com.twitter.scalding.Tsv
 import parallelai.spyglass.hbase.HBaseSource
 import parallelai.spyglass.hbase.HBaseConstants.SourceMode
+import sandcrawler.HBaseRowCountJob
 
 /**
  * Example of how to define tests for HBaseSource
@@ -39,18 +40,9 @@ class HBaseRowCountTest extends FunSpec with TupleConversions {
     .arg("app.conf.path", "app.conf")
     .arg("output", output)
     .arg("debug", "true")
-    .source[Tuple](
-    new HBaseSource(
-      //"table_name",
-      //"quorum_name:2181",
-      "wbgrp-journal-extract-0-qa",
-      "mtrcs-zk1.us.archive.org:2181",
-      new Fields("key"),
-      List("file"),
-      List(new Fields("size", "mimetype")),
-      sourceMode = SourceMode.SCAN_ALL),
+    .source[Tuple](HBaseRowCountJob.getHBaseSource,
       sampleData.map(l => new Tuple(l.map(s => {new ImmutableBytesWritable(Bytes.toBytes(s))}):_*)))
-    .sink[Tuple](Tsv(output)) {
+      .sink[Tuple](Tsv(output)) {
       outputBuffer =>
 
         it("should return the test data provided.") {