Associate PublishableEntities with Collections [FC-0062]

openedx_learning/__init__.py

@@ -1,4 +1,4 @@
 """
 Open edX Learning ("Learning Core").
 """
-__version__ = "0.11.2"
+__version__ = "0.11.3"

openedx_learning/apps/authoring/collections/api.py

@@ -4,19 +4,25 @@
 from __future__ import annotations
 
 from django.db.models import QuerySet
+from django.db.transaction import atomic
+from django.utils import timezone
 
-from .models import Collection
+from ..publishing.models import PublishableEntity
+from .models import Collection, CollectionObject
 
 # The public API that will be re-exported by openedx_learning.apps.authoring.api
 # is listed in the __all__ entries below. Internal helper functions that are
 # private to this module should start with an underscore. If a function does not
 # start with an underscore AND it is not in __all__, that function is considered
 # to be callable only by other apps in the authoring package.
 __all__ = [
+    "add_to_collections",
     "create_collection",
     "get_collection",
     "get_collections",
     "get_learning_package_collections",
+    "get_object_collections",
+    "remove_from_collections",
     "update_collection",
 ]
 
@@ -26,16 +32,27 @@ def create_collection(
     title: str,
     created_by: int | None,
     description: str = "",
+    contents_qset: QuerySet[PublishableEntity] = PublishableEntity.objects.none(),  # default to empty set,
 ) -> Collection:
     """
     Create a new Collection
     """
-    collection = Collection.objects.create(
-        learning_package_id=learning_package_id,
-        title=title,
-        created_by_id=created_by,
-        description=description,
-    )
+
+    with atomic():
+        collection = Collection.objects.create(
+            learning_package_id=learning_package_id,
+            title=title,
+            created_by_id=created_by,
+            description=description,
+        )
+
+        added = add_to_collections(
+            Collection.objects.filter(id=collection.id),
+            contents_qset,
+        )
+        if added:
+            collection.refresh_from_db()  # fetch updated modified date
+
     return collection
 
 
@@ -70,6 +87,90 @@ def update_collection(
     return collection
 
 
+def add_to_collections(
+    collections_qset: QuerySet[Collection],
+    contents_qset: QuerySet[PublishableEntity],
+) -> int:
+    """
+    Adds a QuerySet of PublishableEntities to a QuerySet of Collections.
+
+    Records are created in bulk, and so integrity errors are deliberately ignored: they indicate that the content(s)
+    have already been added to the collection(s).
+
+    Returns the number of entities added (including any that already exist).
+    """
+    collection_objects = []
+    object_ids = contents_qset.values_list("pk", flat=True)
+    collection_ids = collections_qset.values_list("pk", flat=True)
+
+    for collection_id in collection_ids:
+        for object_id in object_ids:
+            collection_objects.append(
+                CollectionObject(
+                    collection_id=collection_id,
+                    object_id=object_id,
+                )
+            )
+
+    created = []
+    if collection_objects:
+        created = CollectionObject.objects.bulk_create(
+            collection_objects,
+            ignore_conflicts=True,
+        )
+
+        # Update the modified date for all the provided Collections.
+        # (Ignoring conflicts means we don't know which ones were modified.)
+        collections_qset.update(modified=timezone.now())
+
+    return len(created)
+
+
+def remove_from_collections(
+    collections_qset: QuerySet,
+    contents_qset: QuerySet,
+) -> int:
+    """
+    Removes a QuerySet of PublishableEntities from a QuerySet of Collections.
+
+    PublishableEntities are deleted from each Collection, in bulk.
+
+    Collections which had objects removed are marked with modified=now().
+
+    Returns the total number of entities deleted.
+    """
+    total_deleted = 0
+    object_ids = contents_qset.values_list("pk", flat=True)
+    collection_ids = collections_qset.values_list("pk", flat=True)
+    modified_collection_ids = []
+
+    for collection_id in collection_ids:
+        num_deleted, _ = CollectionObject.objects.filter(
+            collection_id=collection_id,
+            object_id__in=object_ids,
+        ).delete()
+
+        if num_deleted:
+            modified_collection_ids.append(collection_id)
+
+        total_deleted += num_deleted
+
+    # Update the modified date for the affected Collections
+    Collection.objects.filter(id__in=modified_collection_ids).update(modified=timezone.now())
+
+    return total_deleted
+
+
+def get_object_collections(object_key: str) -> QuerySet[Collection]:
+    """
+    Get all collections associated with a given PublishableEntity.key.
+
+    Only enabled collections are returned.
+    """
+    entity = PublishableEntity.objects.get(key=object_key)
+    return entity.collections.filter(enabled=True).order_by("pk")
+
+
 def get_learning_package_collections(learning_package_id: int) -> QuerySet[Collection]:
     """
     Get all collections for a given learning package

openedx_learning/apps/authoring/collections/migrations/0003_collection_contents.py

@@ -0,0 +1,32 @@
+# Generated by Django 4.2.14 on 2024-08-21 07:15
+
+import django.db.models.deletion
+from django.db import migrations, models
+
+
+class Migration(migrations.Migration):
+
+    dependencies = [
+        ('oel_publishing', '0002_alter_learningpackage_key_and_more'),
+        ('oel_collections', '0002_remove_collection_name_collection_created_by_and_more'),
+    ]
+
+    operations = [
+        migrations.CreateModel(
+            name='CollectionObject',
+            fields=[
+                ('id', models.BigAutoField(auto_created=True, primary_key=True, serialize=False, verbose_name='ID')),
+                ('collection', models.ForeignKey(on_delete=django.db.models.deletion.CASCADE, to='oel_collections.collection')),
+                ('object', models.ForeignKey(on_delete=django.db.models.deletion.CASCADE, to='oel_publishing.publishableentity')),
+            ],
+        ),
+        migrations.AddField(
+            model_name='collection',
+            name='contents',
+            field=models.ManyToManyField(related_name='collections', through='oel_collections.CollectionObject', to='oel_publishing.publishableentity'),
+        ),
+        migrations.AddConstraint(
+            model_name='collectionobject',
+            constraint=models.UniqueConstraint(fields=('collection', 'object'), name='oel_collections_cpe_uniq_col_obj'),
+        ),
+    ]

openedx_learning/apps/authoring/collections/models.py

@@ -1,5 +1,68 @@
 """
 Core models for Collections
+
+TLDR Guidelines:
+1. DO NOT modify these models to store full version snapshots.
+2. DO NOT use these models to try to reconstruct historical versions of
+   Collections for fast querying.
+
+If you're trying to do either of these things, you probably want a new model or
+app. For more details, read on.
+
+The goal of these models is to provide a lightweight method of organizing
+PublishableEntities. The first use case for this is modeling the structure of a
+v1 Content Library within a LearningPackage. This is what we'll use the
+Collection model for.
+
+An important thing to note here is that Collections are *NOT* publishable
+entities themselves. They have no "Draft" or "Published" versions. Collections
+are never "published", though the things inside of them are.
+
+When a LibraryContentBlock makes use of a Content Library, it copies all of
+the items it will use into the Course itself. It will also store a version
+on the LibraryContentBlock -- this is a MongoDB ObjectID in v1 and an integer in
+v2 Libraries. Later on, the LibraryContentBlock will want to check back to see
+if any updates have been made, using its version as a key. If a new version
+exists, the course team has the option of re-copying data from the Library.
+
+ModuleStore based v1 Libraries and Blockstore-based v2 libraries both version
+the entire library in a series of snapshots. This makes it difficult to have
+very large libraries, which is an explicit goal for Modular Learning. In
+Learning Core, we've moved to tracking the versions of individual Components to
+address this issue. But that means we no longer have a single version indicator
+for "has anything here changed"?
+
+We *could* have put that version in the ``publishing`` app's PublishLog, but
+that would make it too broad. We want the ability to eventually collapse many v1
+Libraries into a single Learning Core backed v2 Library. If we tracked the
+versioning in only a central location, then we'd have many false positives where
+the version was bumped because something else in the Learning Package changed.
+So instead, we're creating a new Collection model inside the LearningPackage to
+track that concept.
+
+A critical takeaway is that we don't have to store snapshots of every version of
+a Collection, because that data has been copied over by the LibraryContentBlock.
+We only need to store the current state of the Collection, and increment the
+version numbers when changes happen. This will allow the LibraryContentBlock to
+check in and re-copy over the latest version if the course team desires.
+
+That's why these models only store the current state of a Collection. Unlike the
+``components`` app,  ``collections`` does not store fully materialized snapshots
+of past versions. This is done intentionally in order to save space and reduce
+the cost of writes. Collections may grow to be very large, and we don't want to
+be writing N rows with every version, where N is the number of
+PublishableEntities in a Collection.
+
+MVP of these models does not store changesets, but we can add this when there's a
+use case for it. The number of rows in these changesets would grow in proportion
+to the number of things that are actually changing (instead of copying over
+everything on every version). This is could be used to make it easier to figure out
+what changed between two given versions of a Collection. A LibraryContentBlock
+in a course would have stored the version number of the last time it copied data
+from the Collection, and we can eventually surface this data to the user. Note that
+while it may be possible to reconstruct past versions of Collections based off of
+this changeset data, it's going to be a very slow process to do so, and it is
+strongly discouraged.
 """
 from __future__ import annotations
 
@@ -9,10 +72,11 @@
 
 from ....lib.fields import MultiCollationTextField, case_insensitive_char_field
 from ....lib.validators import validate_utc_datetime
-from ..publishing.models import LearningPackage
+from ..publishing.models import LearningPackage, PublishableEntity
 
 __all__ = [
     "Collection",
+    "CollectionObject",
 ]
 
 
@@ -79,6 +143,12 @@ class Collection(models.Model):
         ],
     )
 
+    contents: models.ManyToManyField[PublishableEntity, "CollectionObject"] = models.ManyToManyField(
+        PublishableEntity,
+        through="CollectionObject",
+        related_name="collections",
+    )
+
     class Meta:
         verbose_name_plural = "Collections"
         indexes = [
@@ -96,3 +166,30 @@ def __str__(self) -> str:
         User-facing string representation of a Collection.
         """
         return f"<{self.__class__.__name__}> ({self.id}:{self.title})"
+
+
+class CollectionObject(models.Model):
+    """
+    Collection -> PublishableEntity association.
+    """
+    collection = models.ForeignKey(
+        Collection,
+        on_delete=models.CASCADE,
+    )
+    object = models.ForeignKey(
+        PublishableEntity,
+        on_delete=models.CASCADE,
+    )
+
+    class Meta:
+        constraints = [
+            # Prevent race conditions from making multiple rows associating the
+            # same Collection to the same Entity.
+            models.UniqueConstraint(
+                fields=[
+                    "collection",
+                    "object",
+                ],
+                name="oel_collections_cpe_uniq_col_obj",
+            )
+        ]

openedx_learning/apps/authoring/collections/readme.rst

@@ -12,3 +12,12 @@ had to create many small libraries.
 
 For the Libraries Relaunch ("v2"), we want to encourage people to create large libraries with lots of content,
 and organize that content using tags and collections.
+
+Architecture Guidelines
+-----------------------
+
+Things to remember:
+
+* Collections may grow very large.
+* Collections are not publishable in and of themselves.
+* Collections link to PublisableEntity records, **not** PublishableEntityVersion records.