Use 1/2 of rule len as LicenseMatch max merge dist

 * 1/5th was too small
 * do not break out of merge loop if two matches are not seq matches

Signed-off-by: Philippe Ombredanne <[email protected]>

Author: pombredanne

src/licensedcode/match.py

@@ -437,7 +437,7 @@ def merge_matches(matches, max_dist=MAX_DIST):
     for rid, rule_matches in matches_by_rule:
         if TRACE_MERGE: logger_debug('merge_matches: processing rule:', rid)
         rlen = rule_matches[0].rule.length
-        max_rlen_dist = min((rlen // 5) or 1, MAX_DIST)
+        max_rlen_dist = min((rlen // 2) or 1, MAX_DIST)
 
         # compare two matches in the sorted sequence: current and next
         i = 0
@@ -450,13 +450,14 @@ def merge_matches(matches, max_dist=MAX_DIST):
                 if TRACE_MERGE: logger_debug('---> merge_matches: next:   ', next_match)
 
                 # two exact matches can never be merged as they will not be overlapping
-                if current_match.matcher != MATCH_SEQ and next_match.matcher != MATCH_SEQ:
-                    if TRACE_MERGE: logger_debug('    ---> ###merge_matches: both matches are EXACT_MATCHES, skipping')
-                    break
+                # only sequence matches for the same rule can be merged
+                #if current_match.matcher != MATCH_SEQ and next_match.matcher != MATCH_SEQ:
+                #    if TRACE_MERGE: logger_debug('    ---> ###merge_matches: both matches are EXACT_MATCHES, skipping')
+                #    break
 
                 # FIXME: also considers the match length!
                 # stop if we exceed max dist
-                # or distance over 1/5 of rule length
+                # or distance over 1/2 of rule length
                 if (current_match.qdistance_to(next_match) > max_rlen_dist
                 or current_match.idistance_to(next_match) > max_rlen_dist):
                     if TRACE_MERGE: logger_debug('    ---> ###merge_matches: MAX_DIST reached, breaking')

tests/licensedcode/test_match.py

@@ -280,11 +280,24 @@ def test_merge_does_merge_overlapping_matches_of_same_rules_if_in_sequence(self)
 
     def test_merge_does_not_merge_overlapping_matches_of_same_rules_if_in_sequence_with_gaps(self):
         r1 = Rule(text_file='r1', licenses=['apache-2.0', 'gpl'])
+        r1.length = 50
 
         m1 = LicenseMatch(rule=r1, qspan=Span(1, 3), ispan=Span(1, 3))
         m2 = LicenseMatch(rule=r1, qspan=Span(14, 20), ispan=Span(4, 10))
 
-        assert [LicenseMatch(rule=r1, qspan=Span(1, 3) | Span(14, 20), ispan=Span(1, 10))] == merge_matches([m1, m2])
+        expected = [LicenseMatch(rule=r1, qspan=Span(1, 3) | Span(14, 20), ispan=Span(1, 10))]
+        results = merge_matches([m1, m2])
+        assert expected == results
+
+    def test_merge_does_not_merge_overlapping_matches_of_same_rules_if_in_sequence_with_gaps_for_long_match(self):
+        r1 = Rule(text_file='r1', licenses=['apache-2.0', 'gpl'])
+        r1.length = 20
+        m1 = LicenseMatch(rule=r1, qspan=Span(1, 10), ispan=Span(1, 10))
+        m2 = LicenseMatch(rule=r1, qspan=Span(14, 20), ispan=Span(14, 20))
+
+        expected = [LicenseMatch(rule=r1, qspan=Span(1, 10) | Span(14, 20), ispan=Span(1, 10) | Span(14, 20))]
+        results = merge_matches([m1, m2])
+        assert expected == results
 
     def test_merge_does_not_merge_overlapping_matches_of_same_rules_if_in_not_sequence(self):
         r1 = Rule(text_file='r1', licenses=['apache-2.0', 'gpl'])