Replace self._val with separate properties

CHANGES/1397.misc.rst

@@ -0,0 +1 @@
+1396.misc.rst

tests/test_pickle.py

@@ -6,7 +6,7 @@
 
 
 def test_pickle():
-    u1 = URL("test")
+    u1 = URL("picklepickle")
     hash(u1)
     v = pickle.dumps(u1)
     u2 = pickle.loads(v)
@@ -16,9 +16,45 @@ def test_pickle():
 
 
 def test_default_style_state():
-    u = URL("test")
-    hash(u)
-    val = ("test", "test", "test", "test", "test")
+    u = object.__new__(URL)
+    val = ("set_state", "set_state", "set_state", "set_state", "set_state")
     u.__setstate__((None, {"_val": val, "_strict": False, "_cache": {"hash": 1}}))
-    assert not u._cache
     assert u._val == val
+    assert hash(u) != 1
+
+
+def test_empty_url_is_not_cached():
+    u = URL.__new__(URL)
+    val = ("set_state", "set_state", "set_state", "set_state", "set_state")
+    u.__setstate__((None, {"_val": val, "_strict": False, "_cache": {"hash": 1}}))
+    assert u._val == val
+    assert hash(u) != 1
+
+
+def test_pickle_does_not_pollute_cache():
+    """Verify the unpickling does not pollute the cache.
+
+    Since unpickle will call URL.__new__ with default
+    args, we need to make sure that default args never
+    end up in the pre_encoded_url or encode_url cache.
+    """
+    u1 = URL.__new__(URL)
+    u1._scheme = "this"
+    u1._netloc = "never.appears.any.where.else.in.tests"
+    u1._path = ""
+    u1._query = ""
+    u1._fragment = ""
+    hash(u1)
+    v = pickle.dumps(u1)
+    u2: URL = pickle.loads(v)
+    assert u1._cache
+    assert hash(u1) == hash(u2)
+    assert u2._scheme == "this"
+    assert u2._netloc == "never.appears.any.where.else.in.tests"
+    assert u2._path == ""
+    assert u2._query == ""
+    assert u2._fragment == ""
+    # Verify unpickling did not the cache wrong scheme
+    # for empty args.
+    assert URL().scheme == ""
+    assert URL("").scheme == ""

tests/test_url_benchmarks.py

@@ -188,7 +188,7 @@ def test_raw_host_empty_cache(benchmark: BenchmarkFixture) -> None:
     @benchmark
     def _run() -> None:
         for _ in range(100):
-            url._cache = {}
+            url._cache.pop("raw_host", None)
             url.raw_host
 
 

yarl/_parse.py

@@ -3,7 +3,7 @@
 import re
 import unicodedata
 from functools import lru_cache
-from typing import Final, Union
+from typing import Union
 from urllib.parse import scheme_chars, uses_netloc
 
 from ._quoters import QUOTER
@@ -20,11 +20,6 @@
 USES_AUTHORITY = frozenset(uses_netloc)
 
 SplitURL = tuple[str, str, str, str, str]
-SCHEME: Final[int] = 0
-NETLOC: Final[int] = 1
-PATH: Final[int] = 2
-QUERY: Final[int] = 3
-FRAGMENT: Final[int] = 4
 
 
 def split_url(url: str) -> SplitURL: