Taught export to insure de-duplicated Artifact.json.

Along the way taught export to operate on a QuerySet of Artifacts
instead of (prematurely) hydrating all affected Artifacts into
a list.

fixes #4159.

(cherry picked from commit 6178887)

CHANGES/4159.bugfix

@@ -0,0 +1,4 @@
+Taught the Artifact.json of an export to hold minimum-unique-set of Artifact entries.
+
+In highly-duplicated-content export scenarios, this can mean a significant decrease
+in export-size, and significant improvement in import-performance.

pulpcore/app/importexport.py

@@ -94,14 +94,14 @@ def export_artifacts(export, artifacts):
 
     Args:
         export (django.db.models.PulpExport): export instance that's doing the export
-        artifacts (django.db.models.Artifacts): list of artifacts in all repos being exported
+        artifacts (django.db.models.Artifacts): QuerySet of artifacts in all repos being exported
 
     Raises:
         ValidationError: When path is not in the ALLOWED_EXPORT_PATHS setting
     """
     data = dict(message="Exporting Artifacts", code="export.artifacts", total=len(artifacts))
     with ProgressReport(**data) as pb:
-        for artifact in pb.iter(artifacts):
+        for artifact in artifacts.iterator():  # chunk_size= defaults to 2000 at a fetch
             dest = artifact.file.name
             if settings.DEFAULT_FILE_STORAGE != "pulpcore.app.models.storage.FileSystem":
                 with tempfile.TemporaryDirectory(dir=".") as temp_dir:
@@ -112,6 +112,7 @@ def export_artifacts(export, artifacts):
                         export.tarfile.add(temp_file.name, dest)
             else:
                 export.tarfile.add(artifact.file.path, dest)
+            pb.increment()
 
     resource = ArtifactResource()
     resource.queryset = artifacts

pulpcore/app/tasks/export.py

@@ -492,7 +492,7 @@ def _do_export(pulp_exporter, tar, the_export):
     starting_versions = _get_starting_versions(do_incremental, pulp_exporter, the_export)
     vers_match = _version_match(ending_versions, starting_versions)
     # Gather up versions and artifacts
-    artifacts = []
+    artifacts = None  # Will be a QuerySet selecting the Artifacts that need to be exported
     for version in ending_versions:
         # Check version-content to make sure we're not being asked to export
         # an on_demand repo
@@ -501,10 +501,15 @@ def _do_export(pulp_exporter, tar, the_export):
             raise RuntimeError(_("Remote artifacts cannot be exported."))
 
         if do_incremental:
-            vers_artifacts = version.artifacts.difference(vers_match[version].artifacts).all()
+            vers_artifacts = version.artifacts.difference(vers_match[version].artifacts)
         else:
-            vers_artifacts = version.artifacts.all()
-        artifacts.extend(vers_artifacts)
+            vers_artifacts = version.artifacts
+
+        if artifacts:
+            artifacts.union(vers_artifacts)
+        else:
+            artifacts = vers_artifacts
+
     # export plugin-version-info
     export_versions(the_export, plugin_version_info)
     # Export the top-level entities (artifacts and repositories)