Merge branch 'bnewbold-scoring-patches' into 'master'

Patches on top of scoring-refactor branch (Crossref/GROBID matching work)

See merge request webgroup/sandcrawler!15

1 files changed, 57 insertions, 0 deletions

diff --git a/scalding/src/test/scala/sandcrawler/ScorableFeaturesTest.scala b/scalding/src/test/scala/sandcrawler/ScorableFeaturesTest.scala
new file mode 100644
index 0000000..80d92aa
--- /dev/null
+++ b/scalding/src/test/scala/sandcrawler/ScorableFeaturesTest.scala
@@ -0,0 +1,57 @@
+package sandcrawler
+
+import org.scalatest._
+
+// scalastyle:off null
+class ScorableFeaturesTest extends FlatSpec with Matchers {
+  private def titleToSlug(s : String) : String = {
+    new ScorableFeatures(title = s).toSlug
+  }
+
+  "toMapFeatures()" should "work with gnarly inputs" in {
+    new ScorableFeatures(title = null).toMapFeatures
+    new ScorableFeatures(title = "something", doi = null, sha1 = null, year = 123).toMapFeatures
+  }
+
+  "mapToSlug()" should "extract the parts of titles before a colon" in {
+    titleToSlug("HELLO:there") shouldBe "hellothere"
+  }
+
+  it should "extract an entire colon-less string" in {
+    titleToSlug("hello THERE") shouldBe "hellothere"
+  }
+
+  it should "return Scorable.NoSlug if given empty string" in {
+    titleToSlug("") shouldBe Scorable.NoSlug
+  }
+
+  it should "return Scorable.NoSlug if given null" in {
+    titleToSlug(null) shouldBe Scorable.NoSlug
+  }
+
+  it should "strip punctuation" in {
+    titleToSlug("HELLO!:the:re") shouldBe "hellothere"
+    titleToSlug("a:b:c") shouldBe "abc"
+    titleToSlug(
+      "If you're happy and you know it, clap your hands!") shouldBe "ifyourehappyandyouknowitclapyourhands"
+    titleToSlug(":;\"\'") shouldBe Scorable.NoSlug
+  }
+
+  it should "filter stub titles" in {
+    titleToSlug("abstract") shouldBe Scorable.NoSlug
+    titleToSlug("title!") shouldBe Scorable.NoSlug
+    titleToSlug("a real title which is not on blacklist") shouldBe "arealtitlewhichisnotonblacklist"
+  }
+
+  it should "strip special characters" in {
+    titleToSlug(":;!',|\"\'`.#?!-@*/\\=+~%$^{}()[]<>-_") shouldBe Scorable.NoSlug
+    // TODO: titleToSlug("©™₨№…") shouldBe Scorable.NoSlug
+    // TODO: titleToSlug("πµΣσ") shouldBe Scorable.NoSlug
+  }
+
+  it should "remove whitespace" in {
+    titleToSlug("foo bar : baz ::") shouldBe "foobarbaz"
+    titleToSlug("\na\t:b:c") shouldBe "abc"
+    titleToSlug("\n \t \r  ") shouldBe Scorable.NoSlug
+  }
+}