Moves dump_to_neo4j cache backend to neo4j instead of memcached (SUST-76)

Also
* Instead of caching when a course is last published, we get this information
  from the CourseStructure table
* This commit introduces a mock py2neo Graph to be used for testing

openedx/core/djangoapps/coursegraph/apps.py

@@ -12,9 +12,3 @@ class CoursegraphConfig(AppConfig):
     AppConfig for courseware app
     """
     name = 'openedx.core.djangoapps.coursegraph'
-
-    def ready(self):
-        """
-        Import signals on startup
-        """
-        from openedx.core.djangoapps.coursegraph import signals  # pylint: disable=unused-variable

openedx/core/djangoapps/coursegraph/management/commands/dump_to_neo4j.py

@@ -7,18 +7,14 @@ from __future__ import unicode_literals, print_function
 import logging
 
 from django.core.management.base import BaseCommand
-from django.utils import six
+from django.utils import six, timezone
 from opaque_keys.edx.keys import CourseKey
-from py2neo import Graph, Node, Relationship, authenticate
+from py2neo import Graph, Node, Relationship, authenticate, NodeSelector
 from py2neo.compat import integer, string, unicode as neo4j_unicode
 from request_cache.middleware import RequestCache
 from xmodule.modulestore.django import modulestore
 
-from openedx.core.djangoapps.coursegraph.utils import (
-    CommandLastRunCache,
-    CourseLastPublishedCache,
-)
-
+from openedx.core.djangoapps.content.course_structures.models import CourseStructure
 
 log = logging.getLogger(__name__)
 
@@ -29,9 +25,6 @@ bolt_log.setLevel(logging.ERROR)
 
 PRIMITIVE_NEO4J_TYPES = (integer, string, neo4j_unicode, float, bool)
 
-COMMAND_LAST_RUN_CACHE = CommandLastRunCache()
-COURSE_LAST_PUBLISHED_CACHE = CourseLastPublishedCache()
-
 
 class ModuleStoreSerializer(object):
     """
@@ -45,8 +38,10 @@ class ModuleStoreSerializer(object):
         If that parameter isn't furnished, loads all course_keys from the
         modulestore.
         Filters out course_keys in the `skip` parameter, if provided.
-        :param courses: string serialization of course keys
-        :param skip: string serialization of course keys
+        Args:
+            courses: A list of string serializations of course keys.
+                For example, ["course-v1:org+course+run"].
+            skip: Also a list of string serializations of course keys.
         """
         if courses:
             course_keys = [CourseKey.from_string(course.strip()) for course in courses]
@@ -67,7 +62,7 @@ class ModuleStoreSerializer(object):
 
         Returns:
             fields: a dictionary of an XBlock's field names and values
-            label: the name of the XBlock's type (i.e. 'course'
+            block_type: the name of the XBlock's type (i.e. 'course'
             or 'problem')
         """
         # convert all fields to a dict and filter out parent and children field
@@ -88,25 +83,27 @@ class ModuleStoreSerializer(object):
         fields['course_key'] = six.text_type(course_key)
         fields['location'] = six.text_type(item.location)
 
-        label = item.scope_ids.block_type
+        block_type = item.scope_ids.block_type
 
-        # prune some fields
-        if label == 'course':
+        if block_type == 'course':
+            # prune the checklists field
             if 'checklists' in fields:
                 del fields['checklists']
 
-        return fields, label
+            # record the time this command was run
+            fields['time_last_dumped_to_neo4j'] = six.text_type(timezone.now())
+
+        return fields, block_type
 
     def serialize_course(self, course_id):
         """
+        Serializes a course into py2neo Nodes and Relationships
         Args:
             course_id: CourseKey of the course we want to serialize
 
         Returns:
             nodes: a list of py2neo Node objects
             relationships: a list of py2neo Relationships objects
-
-        Serializes a course into Nodes and Relationships
         """
         # create a location to node mapping we'll need later for
         # writing relationships
@@ -116,12 +113,12 @@ class ModuleStoreSerializer(object):
         # create nodes
         nodes = []
         for item in items:
-            fields, label = self.serialize_item(item)
+            fields, block_type = self.serialize_item(item)
 
             for field_name, value in six.iteritems(fields):
                 fields[field_name] = self.coerce_types(value)
 
-            node = Node(label, 'item', **fields)
+            node = Node(block_type, 'item', **fields)
             nodes.append(node)
             location_to_node[item.location] = node
 
@@ -168,44 +165,92 @@ class ModuleStoreSerializer(object):
             transaction.create(entity)
 
     @staticmethod
-    def should_dump_course(course_key):
+    def get_command_last_run(course_key, graph):
+        """
+        This information is stored on the course node of a course in neo4j
+        Args:
+            course_key: a CourseKey
+            graph: a py2neo Graph
+
+        Returns: The datetime that the command was last run, converted into
+            text, or None, if there's no record of this command last being run.
+
+        """
+        selector = NodeSelector(graph)
+        course_node = selector.select(
+            "course",
+            course_key=six.text_type(course_key)
+        ).first()
+
+        last_this_command_was_run = None
+        if course_node:
+            last_this_command_was_run = course_node['time_last_dumped_to_neo4j']
+
+        return last_this_command_was_run
+
+    @staticmethod
+    def get_course_last_published(course_key):
+        """
+        We use the CourseStructure table to get when this course was last
+        published.
+        Args:
+            course_key: a CourseKey
+
+        Returns: The datetime the course was last published at, converted into
+            text, or None, if there's no record of the last time this course
+            was published.
+        """
+        try:
+            structure = CourseStructure.objects.get(course_id=course_key)
+            course_last_published_date = six.text_type(structure.modified)
+        except CourseStructure.DoesNotExist:
+            course_last_published_date = None
+
+        return course_last_published_date
+
+    def should_dump_course(self, course_key, graph):
         """
         Only dump the course if it's been changed since the last time it's been
         dumped.
-        :param course_key: a CourseKey object.
-        :return: bool. Whether or not this course should be dumped to neo4j.
+        Args:
+            course_key: a CourseKey object.
+            graph: a py2neo Graph object.
+
+        Returns: bool of whether this course should be dumped to neo4j.
         """
 
-        last_this_command_was_run = COMMAND_LAST_RUN_CACHE.get(course_key)
-        last_course_had_published_event = COURSE_LAST_PUBLISHED_CACHE.get(
-            course_key
-        )
+        last_this_command_was_run = self.get_command_last_run(course_key, graph)
 
-        # if we have no record of this course being serialized, serialize it
+        course_last_published_date = self.get_course_last_published(course_key)
+
+        # if we don't have a record of the last time this command was run,
+        # we should serialize the course and dump it
         if last_this_command_was_run is None:
             return True
 
         # if we've serialized the course recently and we have no published
-        # events, we can skip re-serializing it
-        if last_this_command_was_run and last_course_had_published_event is None:
+        # events, we will not dump it, and so we can skip serializing it
+        # again here
+        if last_this_command_was_run and course_last_published_date is None:
             return False
 
-        # otherwise, serialize if the command was run before the course's last
-        # published event
-        return last_this_command_was_run < last_course_had_published_event
+        # otherwise, serialize and dump the course if the command was run
+        # before the course's last published event
+        return last_this_command_was_run < course_last_published_date
 
     def dump_courses_to_neo4j(self, graph, override_cache=False):
         """
-        Parameters
-        ----------
+        Method that iterates through a list of courses in a modulestore,
+        serializes them, then writes them to neo4j
+        Args:
             graph: py2neo graph object
             override_cache: serialize the courses even if they'be been recently
                 serialized
 
-        Returns two lists: one of the courses that were successfully written
-          to neo4j, and one of courses that were not.
-        -------
+        Returns: two lists--one of the courses that were successfully written
+            to neo4j and one of courses that were not.
         """
+
         total_number_of_courses = len(self.course_keys)
 
         successful_courses = []
@@ -222,7 +267,7 @@ class ModuleStoreSerializer(object):
                 total_number_of_courses,
             )
 
-            if not (override_cache or self.should_dump_course(course_key)):
+            if not (override_cache or self.should_dump_course(course_key, graph)):
                 log.info("skipping dumping %s, since it hasn't changed", course_key)
                 continue
 
@@ -258,7 +303,6 @@ class ModuleStoreSerializer(object):
                 unsuccessful_courses.append(course_string)
 
             else:
-                COMMAND_LAST_RUN_CACHE.set(course_key)
                 successful_courses.append(course_string)
 
         return successful_courses, unsuccessful_courses

openedx/core/djangoapps/coursegraph/management/commands/tests/utils.py

+"""
+Utilities for testing the dump_to_neo4j management command
+"""
+from __future__ import unicode_literals
+
+from py2neo import Node
+
+
+class MockGraph(object):
+    """
+    A stubbed out version of py2neo's Graph object, used for testing.
+    Args:
+        transaction_errors: a bool for whether transactions should throw
+        an error.
+    """
+    def __init__(self, transaction_errors=False, **kwargs):  # pylint: disable=unused-argument
+        self.nodes = set()
+        self.number_commits = 0
+        self.number_rollbacks = 0
+        self.transaction_errors = transaction_errors
+
+    def begin(self):
+        """
+        A stub of the method that generates transactions
+        Returns: a MockTransaction object (instead of a py2neo Transaction)
+        """
+        return MockTransaction(self)
+
+
+class MockTransaction(object):
+    """
+    A stubbed out version of py2neo's Transaction object, used for testing.
+    """
+    def __init__(self, graph):
+        self.temp = set()
+        self.graph = graph
+
+    def run(self, query):
+        """
+        Deletes all nodes associated with a course. Normally `run` executes
+        an arbitrary query, but in our code, we only use it to delete nodes
+        associated with a course.
+        Args:
+            query: query string to be executed (in this case, to delete all
+            nodes associated with a course)
+        """
+        start_string = "WHERE n.course_key='"
+        start = query.index(start_string) + len(start_string)
+        query = query[start:]
+        end = query.find("'")
+        course_key = query[:end]
+
+        self.graph.nodes = set([
+            node for node in self.graph.nodes if node['course_key'] != course_key
+        ])
+
+    def create(self, element):
+        """
+        Adds elements to the transaction's temporary backend storage
+        Args:
+            element: a py2neo Node object
+        """
+        if isinstance(element, Node):
+            self.temp.add(element)
+
+    def commit(self):
+        """
+        Takes elements in the transaction's temporary storage and adds them
+        to the mock graph's storage. Throws an error if the graph's
+        transaction_errors param is set to True.
+        """
+        if self.graph.transaction_errors:
+            raise Exception("fake exception while trying to commit")
+        for element in self.temp:
+            self.graph.nodes.add(element)
+        self.temp.clear()
+        self.graph.number_commits += 1
+
+    def rollback(self):
+        """
+        Clears the transactions temporary storage
+        """
+        self.temp.clear()
+        self.graph.number_rollbacks += 1
+
+
+class MockNodeSelector(object):
+    """
+    Mocks out py2neo's NodeSelector class. Used to select a node from a graph.
+    py2neo's NodeSelector expects a real graph object to run queries against,
+    so, rather than have to mock out MockGraph to accommodate those queries,
+    it seemed simpler to mock out NodeSelector as well.
+    """
+    def __init__(self, graph):
+        self.graph = graph
+
+    def select(self, label, course_key):
+        """
+        Selects nodes that match a label and course_key
+        Args:
+            label: the string of the label we're selecting nodes by
+            course_key: the string of the course key we're selecting node by
+
+        Returns: a MockResult of matching nodes
+        """
+        nodes = []
+        for node in self.graph.nodes:
+            if node.has_label(label) and node["course_key"] == course_key:
+                nodes.append(node)
+        return MockNodeSelection(nodes)
+
+
+class MockNodeSelection(list):
+    """
+    Mocks out py2neo's NodeSelection class: this is the type of what
+    MockNodeSelector's `select` method returns.
+    """
+    def first(self):
+        """
+        Returns: the first element of a list if the list has elements.
+            Otherwise, None.
+        """
+        return self[0] if self else None

openedx/core/djangoapps/coursegraph/signals.py

-"""
-Signal handlers for the CourseGraph application
-"""
-from django.dispatch.dispatcher import receiver
-from xmodule.modulestore.django import SignalHandler
-
-from openedx.core.djangoapps.coursegraph.utils import CourseLastPublishedCache
-
-
-@receiver(SignalHandler.course_published)
-def _listen_for_course_publish(sender, course_key, **kwargs):  # pylint: disable=unused-argument
-    """
-    Register when the course was published on a course publish event
-    """
-    CourseLastPublishedCache().set(course_key)

openedx/core/djangoapps/coursegraph/tests/test_signals.py

-"""
-Tests for coursegraph's signal handler on course publish
-"""
-from __future__ import unicode_literals
-
-from opaque_keys.edx.keys import CourseKey
-
-from openedx.core.djangoapps.coursegraph.signals import _listen_for_course_publish
-from openedx.core.djangoapps.coursegraph.utils import CourseLastPublishedCache
-from openedx.core.djangolib.testing.utils import CacheIsolationTestCase
-
-
-class TestCourseGraphSignalHandler(CacheIsolationTestCase):
-    """
-    Tests for the course publish course handler
-    """
-    ENABLED_CACHES = ['default']
-
-    def test_cache_set_on_course_publish(self):
-        """
-        Tests that the last published cache is set on course publish
-        """
-        course_key = CourseKey.from_string('course-v1:org+course+run')
-        last_published_cache = CourseLastPublishedCache()
-        self.assertIsNone(last_published_cache.get(course_key))
-        _listen_for_course_publish(None, course_key)
-        self.assertIsNotNone(last_published_cache.get(course_key))

openedx/core/djangoapps/coursegraph/utils.py

-"""
-Helpers for the CourseGraph app
-"""
-from django.core.cache import cache
-from django.utils import timezone
-
-
-class TimeRecordingCacheBase(object):
-    """
-    A base class for caching the current time for some key.
-    """
-    # cache_prefix should be defined in children classes
-    cache_prefix = None
-    _cache = cache
-
-    def _key(self, course_key):
-        """
-        Make a cache key from the prefix and a course_key
-        :param course_key: CourseKey object
-        :return: a cache key
-        """
-        return self.cache_prefix + unicode(course_key)
-
-    def get(self, course_key):
-        """
-        Gets the time value associated with the CourseKey.
-        :param course_key: a CourseKey object.
-        :return: the time the key was last set.
-        """
-        return self._cache.get(self._key(course_key))
-
-    def set(self, course_key):
-        """
-        Sets the current time for a CourseKey key.
-        :param course_key: a CourseKey object.
-        """
-        return self._cache.set(self._key(course_key), timezone.now())
-
-
-class CourseLastPublishedCache(TimeRecordingCacheBase):
-    """
-    Used to record the last time that a course had a publish event run on it.
-    """
-    cache_prefix = u'course_last_published'
-
-
-class CommandLastRunCache(TimeRecordingCacheBase):
-    """
-    Used to record the last time that the dump_to_neo4j command was run on a
-    course.
-    """
-    cache_prefix = u'dump_to_neo4j_command_last_run'