Create and expose metadata file

fixes #1047

Author: jobselko

CHANGES/1047.feature

@@ -0,0 +1 @@
+Added exposure of metadata file to Simple API (PEP 658)

pulp_python/app/management/commands/repair-python-metadata.py

@@ -24,8 +24,14 @@ def repair_metadata(content):
     set_of_update_fields = set()
     total_repaired = 0
     for package in immediate_content.prefetch_related("_artifacts").iterator(chunk_size=1000):
+        # Get the main artifact
+        main_artifact = (
+            package.contentartifact_set.exclude(relative_path__endswith=".metadata")
+            .first()
+            .artifact
+        )
         new_data = artifact_to_python_content_data(
-            package.filename, package._artifacts.get(), package.pulp_domain
+            package.filename, main_artifact, package.pulp_domain
         )
         changed = False
         for field, value in new_data.items():

pulp_python/app/serializers.py

@@ -1,5 +1,6 @@
 import logging
 import os
+import tempfile
 from gettext import gettext as _
 from django.conf import settings
 from django.db.utils import IntegrityError
@@ -22,6 +23,7 @@
 )
 from pulp_python.app.utils import (
     DIST_EXTENSIONS,
+    artifact_to_metadata_artifact,
     artifact_to_python_content_data,
     get_project_metadata_from_file,
     parse_project_metadata,
@@ -93,11 +95,31 @@ class Meta:
         model = python_models.PythonDistribution
 
 
+class PythonSingleContentArtifactField(core_serializers.SingleContentArtifactField):
+    """
+    Custom field with overridden get_attribute method. Meant to be used only in
+    PythonPackageContentSerializer to handle possible existence of metadata artifact.
+    """
+
+    def get_attribute(self, instance):
+        # When content has multiple artifacts (wheel + metadata), return the main one
+        if instance._artifacts.count() > 1:
+            for ca in instance.contentartifact_set.all():
+                if not ca.relative_path.endswith(".metadata"):
+                    return ca.artifact
+
+        return super().get_attribute(instance)
+
+
 class PythonPackageContentSerializer(core_serializers.SingleArtifactContentUploadSerializer):
     """
     A Serializer for PythonPackageContent.
     """
 
+    artifact = PythonSingleContentArtifactField(
+        help_text=_("Artifact file representing the physical content"),
+    )
+
     # Core metadata
     # Version 1.0
     author = serializers.CharField(
@@ -386,8 +408,21 @@ def deferred_validate(self, data):
         if attestations := data.pop("attestations", None):
             data["provenance"] = self.handle_attestations(filename, data["sha256"], attestations)
 
+        # Create metadata artifact for wheel files
+        if filename.endswith(".whl"):
+            if metadata_artifact := artifact_to_metadata_artifact(filename, artifact):
+                data["metadata_artifact"] = metadata_artifact
+                data["metadata_sha256"] = metadata_artifact.sha256
+
         return data
 
+    def get_artifacts(self, validated_data):
+        artifacts = super().get_artifacts(validated_data)
+        if metadata_artifact := validated_data.pop("metadata_artifact", None):
+            relative_path = f"{validated_data['filename']}.metadata"
+            artifacts[relative_path] = metadata_artifact
+        return artifacts
+
     def retrieve(self, validated_data):
         content = python_models.PythonPackageContent.objects.filter(
             sha256=validated_data["sha256"], _pulp_domain=get_domain()
@@ -419,6 +454,7 @@ def create(self, validated_data):
 
     class Meta:
         fields = core_serializers.SingleArtifactContentUploadSerializer.Meta.fields + (
+            "artifact",
             "author",
             "author_email",
             "description",
@@ -514,6 +550,15 @@ def validate(self, data):
             data["provenance"] = self.handle_attestations(
                 filename, data["sha256"], attestations, offline=True
             )
+        # Create metadata artifact for wheel files
+        if filename.endswith(".whl"):
+            with tempfile.TemporaryDirectory(dir=settings.WORKING_DIRECTORY) as temp_dir:
+                if metadata_artifact := artifact_to_metadata_artifact(
+                    filename, artifact, tmp_dir=temp_dir
+                ):
+                    data["metadata_artifact"] = metadata_artifact
+                    data["metadata_sha256"] = metadata_artifact.sha256
+
         return data
 
     class Meta(PythonPackageContentSerializer.Meta):

pulp_python/app/tasks/repair.py

@@ -95,9 +95,13 @@ def repair_metadata(content: QuerySet[PythonPackageContent]) -> tuple[int, set[s
     progress_report.save()
     with progress_report:
         for package in progress_report.iter(immediate_content.iterator(chunk_size=BULK_SIZE)):
-            new_data = artifact_to_python_content_data(
-                package.filename, package._artifacts.get(), domain
+            # Get the main artifact
+            main_artifact = (
+                package.contentartifact_set.exclude(relative_path__endswith=".metadata")
+                .first()
+                .artifact
             )
+            new_data = artifact_to_python_content_data(package.filename, main_artifact, domain)
             total_repaired += update_package_if_needed(
                 package, new_data, batch, set_of_update_fields
             )
@@ -113,7 +117,11 @@ def repair_metadata(content: QuerySet[PythonPackageContent]) -> tuple[int, set[s
             grouped_by_url = defaultdict(list)
 
             for package in group_set:
-                for ra in package.contentartifact_set.get().remoteartifact_set.all():
+                for ra in (
+                    package.contentartifact_set.exclude(relative_path__endswith=".metadata")
+                    .first()
+                    .remoteartifact_set.all()
+                ):
                     grouped_by_url[ra.remote.url].append((package, ra))
 
             # Prioritize the URL that can serve the most packages

pulp_python/app/tasks/sync.py

@@ -229,11 +229,15 @@ async def create_content(self, pkg):
         create a Content Unit to put into the pipeline
         """
         declared_contents = {}
+        page = await aget_remote_simple_page(pkg.name, self.remote)
+        upstream_pkgs = {pkg.filename: pkg for pkg in page.packages}
+
         for version, dists in pkg.releases.items():
             for package in dists:
                 entry = parse_metadata(pkg.info, version, package)
                 url = entry.pop("url")
                 size = package["size"] or None
+                d_artifacts = []
 
                 artifact = Artifact(sha256=entry["sha256"], size=size)
                 package = PythonPackageContent(**entry)
@@ -245,11 +249,29 @@ async def create_content(self, pkg):
                     remote=self.remote,
                     deferred_download=self.deferred_download,
                 )
-                dc = DeclarativeContent(content=package, d_artifacts=[da])
+                d_artifacts.append(da)
+
+                if upstream_pkg := upstream_pkgs.get(entry["filename"]):
+                    if upstream_pkg.has_metadata:
+                        url = upstream_pkg.metadata_url
+                        md_sha256 = upstream_pkg.metadata_digests.get("sha256")
+                        package.metadata_sha256 = md_sha256
+                        artifact = Artifact(sha256=md_sha256)
+
+                        metadata_artifact = DeclarativeArtifact(
+                            artifact=artifact,
+                            url=url,
+                            relative_path=f"{entry['filename']}.metadata",
+                            remote=self.remote,
+                            deferred_download=self.deferred_download,
+                        )
+                        d_artifacts.append(metadata_artifact)
+
+                dc = DeclarativeContent(content=package, d_artifacts=d_artifacts)
                 declared_contents[entry["filename"]] = dc
                 await self.python_stage.put(dc)
 
-        if pkg.releases and (page := await aget_remote_simple_page(pkg.name, self.remote)):
+        if pkg.releases and page:
             if self.remote.provenance:
                 await self.sync_provenance(page, declared_contents)
 

pulp_python/app/tasks/upload.py

@@ -15,7 +15,7 @@
     Provenance,
     verify_provenance,
 )
-from pulp_python.app.utils import artifact_to_python_content_data
+from pulp_python.app.utils import artifact_to_metadata_artifact, artifact_to_python_content_data
 
 
 def upload(artifact_sha256, filename, attestations=None, repository_pk=None):
@@ -97,6 +97,11 @@ def create_content(artifact_sha256, filename, domain):
     def create():
         content = PythonPackageContent.objects.create(**data)
         ContentArtifact.objects.create(artifact=artifact, content=content, relative_path=filename)
+
+        if metadata_artifact := artifact_to_metadata_artifact(filename, artifact):
+            ContentArtifact.objects.create(
+                artifact=metadata_artifact, content=content, relative_path=f"{filename}.metadata"
+            )
         return content
 
     new_content = create()

pulp_python/app/utils.py

@@ -1,4 +1,5 @@
 import hashlib
+import logging
 import pkginfo
 import re
 import shutil
@@ -8,14 +9,19 @@
 from aiohttp.client_exceptions import ClientError
 from collections import defaultdict
 from django.conf import settings
+from django.db.utils import IntegrityError
 from django.utils import timezone
 from jinja2 import Template
 from packaging.utils import canonicalize_name
 from packaging.requirements import Requirement
 from packaging.version import parse, InvalidVersion
 from pypi_simple import ACCEPT_JSON_PREFERRED, ProjectPage
-from pulpcore.plugin.models import Remote
+from pulpcore.plugin.models import Artifact, Remote
 from pulpcore.plugin.exceptions import TimeoutException
+from pulpcore.plugin.util import get_domain
+
+
+log = logging.getLogger(__name__)
 
 
 PYPI_LAST_SERIAL = "X-PYPI-LAST-SERIAL"
@@ -206,25 +212,34 @@ def get_project_metadata_from_file(filename):
     return metadata
 
 
-def compute_metadata_sha256(filename: str) -> str | None:
+def extract_wheel_metadata(filename: str) -> bytes | None:
     """
-    Compute SHA256 hash of the metadata file from a Python package.
+    Extract the metadata file content from a wheel file.
 
-    Returns SHA256 hash or None if metadata cannot be extracted.
+    Returns the raw metadata content as bytes or None if metadata cannot be extracted.
     """
     if not filename.endswith(".whl"):
         return None
     try:
         with zipfile.ZipFile(filename, "r") as f:
             for file_path in f.namelist():
                 if file_path.endswith(".dist-info/METADATA"):
-                    metadata_content = f.read(file_path)
-                    return hashlib.sha256(metadata_content).hexdigest()
-    except (zipfile.BadZipFile, KeyError, OSError):
-        pass
+                    return f.read(file_path)
+    except (zipfile.BadZipFile, KeyError, OSError) as e:
+        log.warning(f"Failed to extract metadata file from {filename}: {e}")
     return None
 
 
+def compute_metadata_sha256(filename: str) -> str | None:
+    """
+    Compute SHA256 hash of the metadata file from a Python package.
+
+    Returns SHA256 hash or None if metadata cannot be extracted.
+    """
+    metadata_content = extract_wheel_metadata(filename)
+    return hashlib.sha256(metadata_content).hexdigest() if metadata_content else None
+
+
 def artifact_to_python_content_data(filename, artifact, domain=None):
     """
     Takes the artifact/filename and returns the metadata needed to create a PythonPackageContent.
@@ -233,6 +248,7 @@ def artifact_to_python_content_data(filename, artifact, domain=None):
     # because pkginfo validates that the filename has a valid extension before
     # reading it
     with tempfile.NamedTemporaryFile("wb", dir=".", suffix=filename) as temp_file:
+        artifact.file.seek(0)
         shutil.copyfileobj(artifact.file, temp_file)
         temp_file.flush()
         metadata = get_project_metadata_from_file(temp_file.name)
@@ -245,6 +261,42 @@ def artifact_to_python_content_data(filename, artifact, domain=None):
     return data
 
 
+def artifact_to_metadata_artifact(
+    filename: str, artifact: Artifact, tmp_dir: str = "."
+) -> Artifact | None:
+    """
+    Creates artifact for metadata from the provided wheel artifact.
+    """
+    if not filename.endswith(".whl"):
+        return None
+
+    with tempfile.NamedTemporaryFile("wb", dir=tmp_dir, suffix=filename, delete=False) as temp_file:
+        temp_wheel_path = temp_file.name
+        artifact.file.seek(0)
+        shutil.copyfileobj(artifact.file, temp_file)
+        temp_file.flush()
+
+    metadata_content = extract_wheel_metadata(temp_wheel_path)
+    if not metadata_content:
+        return None
+
+    with tempfile.NamedTemporaryFile(
+        "wb", dir=tmp_dir, suffix=".metadata", delete=False
+    ) as temp_md:
+        temp_metadata_path = temp_md.name
+        temp_md.write(metadata_content)
+        temp_md.flush()
+
+    metadata_artifact = Artifact.init_and_validate(temp_metadata_path)
+    try:
+        metadata_artifact.save()
+    except IntegrityError:
+        metadata_artifact = Artifact.objects.get(
+            sha256=metadata_artifact.sha256, pulp_domain=get_domain()
+        )
+    return metadata_artifact
+
+
 def fetch_json_release_metadata(name: str, version: str, remotes: set[Remote]) -> dict:
     """
     Fetches metadata for a specific release from PyPI's JSON API. A release can contain
@@ -408,7 +460,9 @@ def find_artifact():
             _art = models.RemoteArtifact.objects.filter(content_artifact=content_artifact).first()
         return _art
 
-    content_artifact = content.contentartifact_set.first()
+    content_artifact = content.contentartifact_set.exclude(
+        relative_path__endswith=".metadata"
+    ).first()
     artifact = find_artifact()
     origin = settings.CONTENT_ORIGIN or settings.PYPI_API_HOSTNAME or ""
     origin = origin.strip("/")

pulp_python/tests/functional/api/test_crud_content_unit.py

@@ -10,7 +10,10 @@
     PYTHON_EGG_FILENAME,
     PYTHON_EGG_URL,
     PYTHON_SM_FIXTURE_CHECKSUMS,
+    PYTHON_WHEEL_FILENAME,
+    PYTHON_WHEEL_URL,
 )
+from pulp_python.tests.functional.utils import ensure_metadata
 
 
 def test_content_crud(
@@ -179,3 +182,22 @@ def test_upload_metadata_24_spec(python_content_factory):
                 assert content.license_expression == "MIT"
                 assert content.license_file == '["LICENSE"]'
                 break
+
+
+@pytest.mark.parallel
+def test_package_creation_with_metadata(
+    pulp_content_url,
+    python_content_factory,
+    python_distribution_factory,
+    python_repo,
+):
+    """
+    Test that the creation of a Python wheel package creates a metadata artifact.
+    """
+    python_content_factory(
+        repository=python_repo, relative_path=PYTHON_WHEEL_FILENAME, url=PYTHON_WHEEL_URL
+    )
+    distro = python_distribution_factory(repository=python_repo)
+
+    # Test that metadata is accessible
+    ensure_metadata(pulp_content_url, distro.base_path, PYTHON_WHEEL_FILENAME)