aboutsummaryrefslogtreecommitdiffstats
path: root/scalding/src/test/scala
diff options
context:
space:
mode:
authorEllen Spertus <ellen.spertus@gmail.com>2018-07-30 11:55:19 -0700
committerEllen Spertus <ellen.spertus@gmail.com>2018-07-30 11:55:19 -0700
commit81dbd0e05653682dccb8bc74b39067b4ee7ac1f2 (patch)
tree657118763cc81f1f1ae5538bed1b18c8d82f8f6f /scalding/src/test/scala
parentdd0df0fe3574352011d6a0fe3c12e59b0a4b8259 (diff)
downloadsandcrawler-81dbd0e05653682dccb8bc74b39067b4ee7ac1f2.tar.gz
sandcrawler-81dbd0e05653682dccb8bc74b39067b4ee7ac1f2.zip
Changed scoring, including adding code to compute string differences. Turned off line length checking.
New scores: ['(583,sha1:K2DKSSVTXWPRMFDTWSTCQW3RVWRIOV3Q,DOI-0,'title 1','title 1: tng')'] ['(500,sha1:K2DKSSVTXWPRMFDTWSTCQW3RVWRIOV3Q,DOI-0.5,'title 1','title 1: tng 2')'] ['(500,sha1:K2DKSSVTXWPRMFDTWSTCQW3RVWRIOV3Q,DOI-0.75,'title 1','title 1: tng 3')'] ['(588,sha1:C3YNNEGH5WAG5ZAAXWAEBNXJWT6CZ3WU,DOI-1,'title 2: tng','title 2: rebooted')']
Diffstat (limited to 'scalding/src/test/scala')
-rw-r--r--scalding/src/test/scala/sandcrawler/HBaseCrossrefScoreTest.scala49
1 files changed, 48 insertions, 1 deletions
diff --git a/scalding/src/test/scala/sandcrawler/HBaseCrossrefScoreTest.scala b/scalding/src/test/scala/sandcrawler/HBaseCrossrefScoreTest.scala
index 655dda1..e6ff4a8 100644
--- a/scalding/src/test/scala/sandcrawler/HBaseCrossrefScoreTest.scala
+++ b/scalding/src/test/scala/sandcrawler/HBaseCrossrefScoreTest.scala
@@ -188,6 +188,53 @@ class HBaseCrossrefScoreTest extends FlatSpec with Matchers {
HBaseCrossrefScore.removeAccents("SØREN") shouldBe "SOREN"
}
+ // Tests adapted from https://oldfashionedsoftware.com/2009/11/19/string-distance-and-refactoring-in-scala/
+ "stringDistance" should "work on empty strings" in {
+ HBaseCrossrefScore.stringDistance("", "") shouldBe 0
+ HBaseCrossrefScore.stringDistance("a", "") shouldBe 1
+ HBaseCrossrefScore.stringDistance("", "a") shouldBe 1
+ HBaseCrossrefScore.stringDistance("abc", "") shouldBe 3
+ HBaseCrossrefScore.stringDistance("", "abc") shouldBe 3
+ }
+
+ it should "work on equal strings" in {
+ HBaseCrossrefScore.stringDistance("", "") shouldBe 0
+ HBaseCrossrefScore.stringDistance("a", "a") shouldBe 0
+ HBaseCrossrefScore.stringDistance("abc", "abc") shouldBe 0
+ }
+
+ it should "work where only inserts are needed" in {
+ HBaseCrossrefScore.stringDistance("", "a") shouldBe 1
+ HBaseCrossrefScore.stringDistance("a", "ab") shouldBe 1
+ HBaseCrossrefScore.stringDistance("b", "ab") shouldBe 1
+ HBaseCrossrefScore.stringDistance("ac", "abc") shouldBe 1
+ HBaseCrossrefScore.stringDistance("abcdefg", "xabxcdxxefxgx") shouldBe 6
+ }
+
+ it should "work where only deletes are needed" in {
+ HBaseCrossrefScore.stringDistance( "a", "") shouldBe 1
+ HBaseCrossrefScore.stringDistance( "ab", "a") shouldBe 1
+ HBaseCrossrefScore.stringDistance( "ab", "b") shouldBe 1
+ HBaseCrossrefScore.stringDistance("abc", "ac") shouldBe 1
+ HBaseCrossrefScore.stringDistance("xabxcdxxefxgx", "abcdefg") shouldBe 6
+ }
+
+ it should "work where only substitutions are needed" in {
+ HBaseCrossrefScore.stringDistance( "a", "b") shouldBe 1
+ HBaseCrossrefScore.stringDistance( "ab", "ac") shouldBe 1
+ HBaseCrossrefScore.stringDistance( "ac", "bc") shouldBe 1
+ HBaseCrossrefScore.stringDistance("abc", "axc") shouldBe 1
+ HBaseCrossrefScore.stringDistance("xabxcdxxefxgx", "1ab2cd34ef5g6") shouldBe 6
+ }
+
+ it should "work where many operations are needed" in {
+ HBaseCrossrefScore.stringDistance("example", "samples") shouldBe 3
+ HBaseCrossrefScore.stringDistance("sturgeon", "urgently") shouldBe 6
+ HBaseCrossrefScore.stringDistance("levenshtein", "frankenstein") shouldBe 6
+ HBaseCrossrefScore.stringDistance("distance", "difference") shouldBe 5
+ HBaseCrossrefScore.stringDistance("java was neat", "scala is great") shouldBe 7
+ }
+
// Pipeline tests
val output = "/tmp/testOutput"
@@ -227,7 +274,7 @@ class HBaseCrossrefScoreTest extends FlatSpec with Matchers {
// "Title 1: TNG", "Title 1: TNG 2", "Title 1: TNG 3", "Title 2 Rebooted"
// Join should have 3 "Title 1" slugs and 1 "Title 2" slug
outputBuffer =>
- it should "return a 4-element list" in {
+ "The pipeline" should "return a 4-element list" in {
outputBuffer should have length 4
}