music_kraken/utils/string_processing.py

@@ -129,15 +129,27 @@ UNIFY_TO = " "
 ALLOWED_LENGTH_DISTANCE = 20
 
 
-def unify_punctuation(to_unify: str) -> str:
+def unify_punctuation(to_unify: str, unify_to: str = UNIFY_TO) -> str:
     for char in string.punctuation:
-        to_unify = to_unify.replace(char, UNIFY_TO)
+        to_unify = to_unify.replace(char, unify_to)
     return to_unify
 
 def hash_url(url: Union[str, ParseResult]) -> str:
     if isinstance(url, str): 
         url = urlparse(url)
 
+    unify_to = "-"
+
+    def unify_part(part: str) -> str:
+        nonlocal unify_to
+        return unify_punctuation(part.lower(), unify_to=unify_to).strip(unify_to)
+
+    # netloc
+    netloc = unify_part(url.netloc)
+    if netloc.startswith("www" + unify_to):
+        netloc = netloc[3 + len(unify_to):]
+
+    # query
     query = url.query
     query_dict: Optional[dict] = None
     try:
@@ -150,9 +162,9 @@ def hash_url(url: Union[str, ParseResult]) -> str:
         # sort keys alphabetically
         query = ""
         for key, value in sorted(query_dict.items(), key=lambda i: i[0]):
-            query += f"_{key.strip()}_{''.join(i.strip() for i in value)}"
+            query += f"{key.strip()}-{''.join(i.strip() for i in value)}"
 
-    r = f"{url.netloc}_{url.path.replace('/', '_')}{query}"
+    r = f"{netloc}_{unify_part(url.path)}_{unify_part(query)}"
     r = r.lower().strip()
     return r
 

tests/test_hash_url.py

@@ -0,0 +1,35 @@
+import unittest
+
+from music_kraken.utils.string_processing import hash_url
+
+
+class TestCollection(unittest.TestCase):
+    def test_remove_schema(self):
+        self.assertFalse(hash_url("https://www.youtube.com/watch?v=3jZ_D3ELwOQ").startswith("https"))
+        self.assertFalse(hash_url("ftp://www.youtube.com/watch?v=3jZ_D3ELwOQ").startswith("https"))
+        self.assertFalse(hash_url("sftp://www.youtube.com/watch?v=3jZ_D3ELwOQ").startswith("https"))
+        self.assertFalse(hash_url("http://www.youtube.com/watch?v=3jZ_D3ELwOQ").startswith("https"))
+
+    def test_no_punctuation(self):
+        self.assertNotIn(hash_url("https://www.you_tube.com/watch?v=3jZ_D3ELwOQ"), "you_tube")
+        self.assertNotIn(hash_url("https://docs.gitea.com/next/install.ation/comparison"), ".")
+
+    def test_three_parts(self):
+        """
+        The url is parsed into three parts [netloc; path; query]
+        Which are then appended to each other with an underscore between.
+        """
+
+        self.assertTrue(hash_url("https://duckduckgo.com/?t=h_&q=dfasf&ia=web").count("_") == 2)
+
+    def test_sort_query(self):
+        """
+        The query is sorted alphabetically
+        """
+        hashed = hash_url("https://duckduckgo.com/?t=h_&q=dfasf&ia=web")
+        sorted_keys = ["ia-", "q-", "t-"]
+
+        self.assertTrue(hashed.index(sorted_keys[0]) < hashed.index(sorted_keys[1]) < hashed.index(sorted_keys[2]))
+
+if __name__ == "__main__":
+    unittest.main()