Normalize wheel distribution filenames parsed from wheels

Allowing us to unpack wheels with non-normalized filenames.

Part of pypa#97, but doesn't retrieve the names from distribution metadata,
yet.

Fixes: pypa#134

src/installer/utils.py

@@ -106,7 +106,9 @@ def parse_wheel_filename(filename: str) -> WheelFilename:
     wheel_info = _WHEEL_FILENAME_REGEX.match(filename)
     if not wheel_info:
         raise ValueError(f"Not a valid wheel filename: {filename}")
-    return WheelFilename(*wheel_info.groups())
+    parsed = wheel_info.groups()
+    normalized_name = normalize_distribution_name(parsed[0])
+    return WheelFilename(normalized_name, *parsed[1:])
 
 
 def copyfileobj_with_hashing(

tests/test_utils.py

@@ -93,6 +93,11 @@ class TestParseWheelFilename:
                 "tensorflow-2.3.0-cp38-cp38-win_amd64.whl",
                 WheelFilename("tensorflow", "2.3.0", None, "cp38-cp38-win_amd64"),
             ),
+            # Non-canonicalized real wheel names
+            (
+                "Quart-0.18.0-py3-none-any.whl",
+                WheelFilename("quart", "0.18.0", None, "py3-none-any"),
+            ),
         ],
     )
     def test_valid_cases(self, string, expected):