Merge pull request #702 from smblott-github/relevancy-fix

Improve consistency and fix bug in  relevancy calculation.

2 files changed, 28 insertions, 7 deletions

diff --git a/background_scripts/completion.coffee b/background_scripts/completion.coffee
index 79f31073..cf946b06 100644
--- a/background_scripts/completion.coffee
+++ b/background_scripts/completion.coffee
@@ -249,12 +249,16 @@ class MultiCompleter
 
 # Utilities which help us compute a relevancy score for a given item.
 RankingUtils =
-  # Whether the given URL or title match any one of the query terms. This is used to prune out irrelevant
-  # suggestions before we try to rank them.
-  matches: (queryTerms, url, title) ->
+  # Whether the given things (usually URLs or titles) match any one of the query terms.
+  # This is used to prune out irrelevant suggestions before we try to rank them, and for calculating word relevancy.
+  # Every term must match at least one thing.
+  matches: (queryTerms, things...) ->
     for term in queryTerms
       regexp = RegexpCache.get(term)
-      return false unless title.match(regexp) || url.match(regexp)
+      matchedTerm = false
+      for thing in things
+        matchedTerm = matchedTerm || (thing?.match && thing.match regexp)
+      return false unless matchedTerm
     true
 
   # Returns a number between [0, 1] indicating how often the query terms appear in the url and title.
@@ -264,8 +268,8 @@ RankingUtils =
     titleScore = 0.0
     for term in queryTerms
       queryLength += term.length
-      urlScore += 1 if url.indexOf(term) >= 0
-      titleScore += 1 if title && title.indexOf(term) >= 0
+      urlScore += 1 if RankingUtils.matches [term], url
+      titleScore += 1 if RankingUtils.matches [term], title
     urlScore = urlScore / queryTerms.length
     urlScore = urlScore * RankingUtils.normalizeDifference(queryLength, url.length)
     if title
diff --git a/tests/unit_tests/completion_test.coffee b/tests/unit_tests/completion_test.coffee
index f3ed4e61..32b62a28 100644
--- a/tests/unit_tests/completion_test.coffee
+++ b/tests/unit_tests/completion_test.coffee
@@ -154,8 +154,25 @@ context "suggestions",
 
 context "RankingUtils",
   should "do a case insensitive match", ->
-    assert.isTrue RankingUtils.matches(["maRiO"], "MARIO", "MARIo")
+    assert.isTrue RankingUtils.matches(["aRi"], "MARIO", "MARio")
 
+  should "do a case insensitive match on full term", ->
+    assert.isTrue RankingUtils.matches(["MaRiO"], "MARIO", "MARio")
+
+  should "do a case insensitive match on more than just two terms", ->
+    assert.isTrue RankingUtils.matches(["aRi"], "DOES_NOT_MATCH", "DOES_NOT_MATCH_EITHER", "MARio")
+
+  should "do case insensitive word relevancy (matching)", ->
+    assert.isTrue RankingUtils.wordRelevancy(["aRi"], "MARIO", "MARio") > 0.0
+
+  should "do case insensitive word relevancy (not matching)", ->
+    assert.isTrue RankingUtils.wordRelevancy(["DOES_NOT_MATCH"], "MARIO", "MARio") == 0.0
+
+  should "every term must match at least one thing (matching)", ->
+    assert.isTrue RankingUtils.matches(["cat", "dog"], "catapult", "hound dog")
+
+  should "every term must match at least one thing (not matching)", ->
+    assert.isTrue not RankingUtils.matches(["cat", "dog", "wolf"], "catapult", "hound dog")
 
 # A convenience wrapper around completer.filter() so it can be called synchronously in tests.
 filterCompleter = (completer, queryTerms) ->