Fix link hashes

Author: dstufft

src/pip/_internal/index/collector.py

@@ -34,7 +34,7 @@
 from pip._vendor.requests.exceptions import RetryError, SSLError
 
 from pip._internal.exceptions import NetworkConnectionError
-from pip._internal.models.link import SUPPORTED_HASHES, Link
+from pip._internal.models.link import Link
 from pip._internal.models.search_scope import SearchScope
 from pip._internal.network.session import PipSession
 from pip._internal.network.utils import raise_for_status
@@ -274,11 +274,6 @@ def _clean_link(url: str) -> str:
     return urllib.parse.urlunparse(result._replace(path=path))
 
 
-_HASH_RE = re.compile(
-    r"({choices})=([a-f0-9]+)".format(choices="|".join(SUPPORTED_HASHES))
-)
-
-
 def _create_link_from_element(
     element_attribs: Dict[str, Optional[str]],
     page_url: str,
@@ -295,17 +290,11 @@ def _create_link_from_element(
     pyrequire = element_attribs.get("data-requires-python")
     yanked_reason = element_attribs.get("data-yanked")
 
-    hashes = {}
-    hm = _HASH_RE.search(url)
-    if hm is not None:
-        hashes[hm.group(1).lower()] = hm.group(2)
-
     link = Link(
         url,
         comes_from=page_url,
         requires_python=pyrequire,
         yanked_reason=yanked_reason,
-        hashes=hashes,
     )
 
     return link

src/pip/_internal/models/link.py

@@ -33,7 +33,7 @@
 
 # Order matters, earlier hashes have a precedence over later hashes for what
 # we will pick to use.
-SUPPORTED_HASHES = ("sha512", "sha384", "sha256", "sha224", "sha1", "md5")
+_SUPPORTED_HASHES = ("sha512", "sha384", "sha256", "sha224", "sha1", "md5")
 
 
 class Link(KeyBasedCompareMixin):
@@ -179,18 +179,32 @@ def subdirectory_fragment(self) -> Optional[str]:
             return None
         return match.group(1)
 
+    _hash_re = re.compile(
+        r"({choices})=([a-f0-9]+)".format(choices="|".join(_SUPPORTED_HASHES))
+    )
+
     @property
     def hash(self) -> Optional[str]:
-        for hashname in SUPPORTED_HASHES:
+        for hashname in _SUPPORTED_HASHES:
             if hashname in self._hashes:
                 return self._hashes[hashname]
+
+        match = self._hash_re.search(self._url)
+        if match:
+            return match.group(2)
+
         return None
 
     @property
     def hash_name(self) -> Optional[str]:
-        for hashname in SUPPORTED_HASHES:
+        for hashname in _SUPPORTED_HASHES:
             if hashname in self._hashes:
                 return hashname
+
+        match = self._hash_re.search(self._url)
+        if match:
+            return match.group(1)
+
         return None
 
     @property
@@ -284,7 +298,7 @@ def _clean_link(link: Link) -> _CleanResult:
         subdirectory = ""
     # If there are multiple hash values under the same algorithm, use the
     # first one. This matches the behavior of Link.hash_value.
-    hashes = {k: fragment[k][0] for k in SUPPORTED_HASHES if k in fragment}
+    hashes = {k: fragment[k][0] for k in _SUPPORTED_HASHES if k in fragment}
     return _CleanResult(
         parsed=parsed._replace(netloc=netloc, query="", fragment=""),
         query=urllib.parse.parse_qs(parsed.query),

tests/unit/test_collector.py

@@ -201,7 +201,7 @@ def test_get_simple_response_no_head(
                 "Cache-Control": "max-age=0",
             },
         ),
-        mock.call().headers.get("Content-Type", ""),
+        mock.call().headers.get("Content-Type", "Unknown"),
     ]
     mock_raise_for_status.assert_called_once_with(resp)
 
@@ -667,7 +667,7 @@ def test_get_index_content_invalid_content_type(
     assert (
         "pip._internal.index.collector",
         logging.WARNING,
-        "Skipping page {} because the GET request got Content-Type: {}."
+        "Skipping page {} because the GET request got Content-Type: {}. "
         "The only supported Content-Types are application/vnd.pypi.simple.v1+json, "
         "application/vnd.pypi.simple.v1+html, and text/html'".format(url, content_type),
     ) in caplog.record_tuples