PERF: 5% faster pip install. algorithmic bug at the end of pip instal…

…l in the code to display installed packages. O(n^2) to enumerate installed packages due to accidental loop in loop.

get_distribution(package_name) does a loop over all installed packages. that is quite surprising and unexpected. do not use inside a loop.

on a pip install run that takes 11 seconds:
the loop takes 0.735960 seconds on main branch.
the loop takes 0.064672 seconds with this fix.

news/bugfix.12791.rst

@@ -0,0 +1,2 @@
+Improve performance of pip install. Fix code to display installed packages at the end of pip install,
+was O(n^2) to enumerate packages due to accidental loop in loop.

src/pip/_internal/commands/install.py

@@ -7,6 +7,7 @@
 from optparse import SUPPRESS_HELP, Values
 from typing import List, Optional
 
+from pip._vendor.packaging.utils import canonicalize_name
 from pip._vendor.rich import print_json
 
 from pip._internal.cache import WheelCache
@@ -472,17 +473,18 @@ def run(self, options: Values, args: List[str]) -> int:
             )
             env = get_environment(lib_locations)
 
+            # Display a summary of installed packages, with extra care to
+            # display a package name as it was requested by the user.
             installed.sort(key=operator.attrgetter("name"))
             items = []
-            for result in installed:
-                item = result.name
-                try:
-                    installed_dist = env.get_distribution(item)
-                    if installed_dist is not None:
-                        item = f"{item}-{installed_dist.version}"
-                except Exception:
-                    pass
-                items.append(item)
+            expected_items = {
+                canonicalize_name(item.name): item.name for item in installed
+            }
+            for distribution in env.iter_all_distributions():
+                display_name = expected_items.get(distribution.canonical_name, None)
+                if display_name is not None:
+                    item = f"{display_name}-{distribution.version}"
+                    items.append(item)
 
             if conflicts is not None:
                 self._warn_about_conflicts(

src/pip/_internal/metadata/importlib/_envs.py

@@ -181,9 +181,10 @@ def _iter_distributions(self) -> Iterator[BaseDistribution]:
             yield from finder.find_linked(location)
 
     def get_distribution(self, name: str) -> Optional[BaseDistribution]:
+        canonical_name = canonicalize_name(name)
         matches = (
             distribution
             for distribution in self.iter_all_distributions()
-            if distribution.canonical_name == canonicalize_name(name)
+            if distribution.canonical_name == canonical_name
         )
         return next(matches, None)