rules and fuzzy match fix:

- rules: fix rule from regex '~' switch, add test
- fuzzymatch filters: use set instead of list to avoid dupes

pywb/rules.yaml

@@ -303,7 +303,7 @@ rules:
                   #- mime
 
           filter:
-              - '~urlkey:{0}'
+              - 'urlkey:{0}'
               - '!mimetype:text/plain'
 
           type: 'domain'

pywb/warcserver/index/fuzzymatcher.py

@@ -64,7 +64,7 @@ def parse_fuzzy_rule(self, rule):
         return FuzzyRule(url_prefix, regex, replace_after, filter_str, match_type)
 
     def get_fuzzy_match(self, urlkey, params):
-        filters = []
+        filters = set()
         matched_rule = None
 
         for rule in self.rules:
@@ -78,7 +78,7 @@ def get_fuzzy_match(self, urlkey, params):
             matched_rule = rule
             for g in m.groups():
                 for f in matched_rule.filter_str:
-                    filters.append(f.format(g))
+                    filters.add(f.format(g))
 
             break
 

pywb/warcserver/index/test/test_fuzzymatcher.py

@@ -36,7 +36,7 @@ def get_params(self, url, actual_url, mime='text/html'):
         return params
 
     def get_expected(self, url, mime='text/html', filters=None):
-        filters = filters or ['urlkey:']
+        filters = filters or {'urlkey:'}
         exp = [{'filter': filters,
                'is_fuzzy': True,
                'urlkey': canonicalize(url),
@@ -102,12 +102,23 @@ def test_fuzzy_jquery_2(self):
         cdx_iter, errs = self.fuzzy(self.source, params)
         assert list(cdx_iter) == self.get_expected(actual_url)
 
-    def test_fuzzy_custom_rule(self):
+    def test_fuzzy_custom_rule_yt(self):
         url = 'http://youtube.com/get_video_info?a=b&html5=true&___abc=123&video_id=ABCD&id=1234'
         actual_url = 'http://youtube.com/get_video_info?a=d&html5=true&___abc=125&video_id=ABCD&id=1234'
         params = self.get_params(url, actual_url)
         cdx_iter, errs = self.fuzzy(self.source, params)
-        filters = ['urlkey:html5=true', 'urlkey:video_id=abcd']
+        filters = {'urlkey:html5=true', 'urlkey:video_id=abcd'}
+        assert list(cdx_iter) == self.get_expected(url=actual_url, filters=filters)
+
+    def test_fuzzy_custom_rule_yt_2(self):
+        url = 'https://r1---sn-xyz.googlevideo.com/videoplayback?id=ABCDEFG&itag=22&food=abc'
+        actual_url = 'https://r1---sn-abcdefg.googlevideo.com/videoplayback?id=ABCDEFG&itag=22&foo=abc&_1=2'
+        params = self.get_params(url, actual_url)
+        cdx_iter, errs = self.fuzzy(self.source, params)
+        filters = {'urlkey:id=abcdefg',
+                   'urlkey:itag=22',
+                   '!mimetype:text/plain'}
+
         assert list(cdx_iter) == self.get_expected(url=actual_url, filters=filters)
 
     def test_no_fuzzy_custom_rule_video_id_diff(self):
@@ -159,3 +170,4 @@ def test_no_fuzzy_jquery_other_arg_mismatch(self):
         cdx_iter, errs = self.fuzzy(self.source, params)
         assert list(cdx_iter) == []
 
+