Break out Assessment into parts, Rubrics into their various pieces, with serializers.

apps/openassessment/peer/admin.py

 from django.contrib import admin
 
-from openassessment.peer.models import PeerEvaluation
+from openassessment.peer.models import Assessment
 
-admin.site.register(PeerEvaluation)
+admin.site.register(Assessment)

apps/openassessment/peer/api.py

@@ -6,15 +6,12 @@ the workflow for a given submission.
 """
 import copy
 import logging
+import math
 
 from django.db import DatabaseError
-import math
-from openassessment.peer.models import PeerEvaluation
 
-from openassessment.peer.serializers import PeerAssessmentSerializer
-from openassessment.peer.serializers import (
-    PeerEvaluationSerializer, content_hash_for_rubric_dict
-)
+from openassessment.peer.models import Assessment
+from openassessment.peer.serializers import AssessmentSerializer
 from submissions import api as submission_api
 from submissions.models import Submission, StudentItem, Score
 from submissions.serializers import SubmissionSerializer, StudentItemSerializer
@@ -73,7 +70,7 @@ def create_assessment(
         required_assessments_for_student,
         required_assessments_for_submission,
         assessment_dict,
-        rubric_dict,
+ #       rubric_dict,
         scored_at=None):
     """Creates an assessment on the given submission.
 
@@ -137,7 +134,8 @@ def create_assessment(
         if scored_at:
             peer_assessment["scored_at"] = scored_at
 
-        peer_serializer = PeerAssessmentSerializer(data=peer_assessment)
+        peer_serializer = AssessmentSerializer(data=peer_evaluation)
+
         if not peer_serializer.is_valid():
             raise PeerAssessmentRequestError(peer_serializer.errors)
         peer_serializer.save()
@@ -199,7 +197,7 @@ def _check_if_finished_and_create_score(student_item,
         student_item.student_id,
         required_assessments_for_student
     )
-    assessments = PeerEvaluation.objects.filter(submission=submission)
+    assessments = Assessment.objects.filter(submission=submission)
     submission_finished = assessments.count() >= required_assessments_for_submission
     scores = []
     for assessment in assessments:
@@ -264,7 +262,7 @@ def has_finished_required_evaluating(student_id, required_assessments):
     if required_assessments < 0:
         raise PeerAssessmentRequestError(
             "Required Assessment count must be a positive integer.")
-    return PeerEvaluation.objects.filter(
+    return Assessment.objects.filter(
         scorer_id=student_id
     ).count() >= required_assessments
 
@@ -312,8 +310,8 @@ def get_assessments(submission_id):
     """
     try:
         submission = Submission.objects.get(uuid=submission_id)
-        assessments = PeerEvaluation.objects.filter(submission=submission)
-        serializer = PeerAssessmentSerializer(assessments, many=True)
+        assessments = Assessment.objects.filter(submission=submission)
+        serializer = AssessmentSerializer(assessments, many=True)
         return serializer.data
     except DatabaseError:
         error_message = (
@@ -395,10 +393,10 @@ def _get_first_submission_not_evaluated(student_items, student_id, required_num_
         "-attempt_number"
     )
     for submission in submissions:
-        assessments = PeerEvaluation.objects.filter(submission=submission)
+        assessments = Assessment.objects.filter(submission=submission)
         if assessments.count() < required_num_assessments:
             already_evaluated = False
             for assessment in assessments:
                 already_evaluated = already_evaluated or assessment.scorer_id == student_id
             if not already_evaluated:
-                return submission
+                return submission
\ No newline at end of file

apps/openassessment/peer/models.py

+# coding=utf-8
 """
-These Models have to capture not only the state of evaluations made for certain
-submissions, but also the state of the specific rubrics at the time those
-evaluations were made. This means we have a number of little models, and that
-much of this data is immutable once created, so that we don't lose historical
-information. This also means that if you change the Rubric in a problem and
-this system is seeing that new Rubric for the first time, we're going to be
-writing a whole little tree of objects into the database. Fortunately, we only
-need to write this when we see a changed problem (rare). Performance concerns
-when reading this data is mitigated by the fact that it's easy to cache the
-entire tree of objects (since everything is immutable).
+These models have to capture not only the state of assessments made for certain
+submissions, but also the state of the rubrics at the time those assessments
+were made.
 """
+from copy import deepcopy
 from hashlib import sha1
 import json
 
@@ -19,43 +14,49 @@ from django.utils.timezone import now
 from submissions.models import Submission
 
 
-class PeerEvaluation(models.Model):
-    submission = models.ForeignKey(Submission)
-    points_earned = models.PositiveIntegerField(default=0)
-    points_possible = models.PositiveIntegerField(default=0)
-    scored_at = models.DateTimeField(default=now, db_index=True)
-    scorer_id = models.CharField(max_length=255, db_index=True)
-    score_type = models.CharField(max_length=2)
-    feedback = models.TextField(max_length=10000, default="")
-
-    def __repr__(self):
-        return repr(dict(
-            submission=self.submission,
-            points_earned=self.points_earned,
-            points_possible=self.points_possible,
-            scored_at=self.scored_at,
-            scorer_id=self.scorer_id,
-            score_type=self.score_type,
-            feedback=self.feedback,
-        ))
-
-    class Meta:
-        ordering = ["-scored_at"]
-
-
 class Rubric(models.Model):
-    """
-    A Rubric
+    """A Rubric contains the guidelines on how to assess a submission.
+
+    Rubrics are composed of :class:`Criterion` objects which are in turn
+    composed of :class:`CriterionOption` objects.
+
+    This model is a bit unusual in that it is the representation of the rubric
+    that an assessment was made with *at the time of assessment*. The source
+    rubric data lives in the problem definition, which is in the
+    :class:`OpenAssessmentBlock`. When an assessment is made, the XBlock passes
+    that rubric information along as well. When this Django app records the
+    :class:`Assessment`, we do a lookup to see if the Rubric model object
+    already exists (using hashing). If the Rubric is not found, we create a new
+    one with the information OpenAssessmentBlock has passed in.
+
+    .. warning::
+       Never change Rubric model data after it's written!
+
+    The little tree of objects that compose a Rubric is meant to be immutable —
+    once created, they're never updated. When the problem changes, we end up
+    creating a new Rubric instead. This makes it easy to cache and do hash-based
+    lookups.
     """
     # SHA1 hash
     content_hash = models.CharField(max_length=40)
 
-    # This is actually the prompt for the whole question, which may be a
-    # complex, nested XML structure.
-    prompt = models.TextField(max_length=10000)
-
+    @property
     def points_possible(self):
-        return sum(crit.points_possible() for crit in self.criteria.all())
+        return sum(crit.points_possible for crit in self.criteria.all())
+
+    @staticmethod
+    def content_hash_for_rubric_dict(rubric_dict):
+        """
+
+        """
+        rubric_dict = deepcopy(rubric_dict)
+        # Neither "id" nor "content_hash" would count towards calculating the
+        # content_hash.
+        rubric_dict.pop("id", None)
+        rubric_dict.pop("content_hash", None)
+
+        canonical_form = json.dumps(rubric_dict, sort_keys=True)
+        return sha1(canonical_form).hexdigest()
 
 
 class Criterion(models.Model):
@@ -72,6 +73,7 @@ class Criterion(models.Model):
         ordering = ["rubric", "order_num"]
 
 
+    @property
     def points_possible(self):
         return max(option.points for option in self.options.all())
 
@@ -109,3 +111,38 @@ class CriterionOption(models.Model):
     def __unicode__(self):
         return repr(self)
 
+
+class Assessment(models.Model):
+    submission = models.ForeignKey(Submission)
+    rubric = models.ForeignKey(Rubric)
+
+    scored_at = models.DateTimeField(default=now, db_index=True)
+    scorer_id = models.CharField(max_length=40, db_index=True)
+    score_type = models.CharField(max_length=2)
+
+    # TODO: move this to its own model
+    feedback = models.TextField(max_length=10000, default="")
+
+    class Meta:
+        ordering = ["-scored_at"]
+
+    @property
+    def points_earned(self):
+        return sum(part.points_earned for part in self.parts.all())
+
+    @property
+    def points_possible(self):
+        return self.rubric.points_possible
+
+
+class AssessmentPart(models.Model):
+    assessment = models.ForeignKey(Assessment, related_name='parts')
+    option = models.ForeignKey(CriterionOption) # TODO: no reverse
+
+    @property
+    def points_earned(self):
+        return self.option.points
+
+    @property
+    def points_possible(self):
+        return self.option.criterion.points_possible

apps/openassessment/peer/serializers.py

@@ -8,61 +8,118 @@ import json
 
 from rest_framework import serializers
 from openassessment.peer.models import (
-    Criterion, CriterionOption, PeerEvaluation, Rubric
+    Assessment, AssessmentPart, Criterion, CriterionOption, Rubric
 )
 
+class InvalidRubric(Exception):
+    def __init__(self, errors):
+        Exception.__init__(self, repr(errors))
+        self.errors = deepcopy(errors)
 
-class PeerAssessmentSerializer(serializers.ModelSerializer):
-    class Meta:
-        model = PeerEvaluation
-        fields = (
-            'submission',
-            'points_earned',
-            'points_possible',
-            'scored_at',
-            'scorer_id',
-            'score_type',
-            'feedback',
-        )
-
-
-class CriterionOptionSerializer(serializers.ModelSerializer):
+
+class NestedModelSerializer(serializers.ModelSerializer):
+    """Model Serializer that supports arbitrary nesting.
+
+    The Django REST Framework does not currently support deserialization more
+    than one level deep (so a parent and children). We want to be able to
+    create a Rubric -> Criterion -> CriterionOption hierarchy.
+
+    Much of the base logic already "just works" and serialization of arbritrary
+    depth is supported. So we just override the save_object method to
+    recursively link foreign key relations instead of doing it one level deep.
+
+    We don't touch many-to-many relationships because we don't need to for our
+    purposes.
+    """
+
+    def recursively_link_related(self, obj, **kwargs):
+        if getattr(obj, '_related_data', None):
+            for accessor_name, related in obj._related_data.items():
+                setattr(obj, accessor_name, related)
+                for related_obj in related:
+                    self.recursively_link_related(related_obj, **kwargs)
+            del(obj._related_data)
+
+
+    def save_object(self, obj, **kwargs):
+        obj.save(**kwargs)
+
+        if getattr(obj, '_m2m_data', None):
+            for accessor_name, object_list in obj._m2m_data.items():
+                setattr(obj, accessor_name, object_list)
+            del(obj._m2m_data)
+
+        self.recursively_link_related(obj, **kwargs)
+
+
+class CriterionOptionSerializer(NestedModelSerializer):
     class Meta:
         model = CriterionOption
         fields = ('order_num', 'points', 'name', 'explanation')
 
 
-class CriterionSerializer(serializers.ModelSerializer):
-    options = CriterionOptionSerializer(many=True)
+class CriterionSerializer(NestedModelSerializer):
+    options = CriterionOptionSerializer(required=True, many=True)
 
     class Meta:
         model = Criterion
         fields = ('order_num', 'prompt', 'options')
 
 
-class RubricSerializer(serializers.ModelSerializer):
-    criteria = CriterionSerializer(many=True)
+    def validate_options(self, attrs, source):
+        options = attrs[source]
+        if not options:
+            raise serializers.ValidationError(
+                "Criterion must have at least one option."
+            )
+        return attrs
+
+
+class RubricSerializer(NestedModelSerializer):
+    criteria = CriterionSerializer(required=True, many=True)
 
     class Meta:
         model = Rubric
-        fields = ('id', 'content_hash', 'prompt', 'criteria')
+        fields = ('id', 'content_hash', 'criteria')
 
 
-def content_hash_for_rubric_dict(rubric_dict):
-    """
-    It's passing in the results from a RubricSerializer, so we just have to get
-    rid of the content_hash.
-    """
-    rubric_dict = deepcopy(rubric_dict)
-    # Neither "id" nor "content_hash" would count towards calculating the
-    # content_hash.
-    rubric_dict.pop("id", None)
-    rubric_dict.pop("content_hash", None)
+    def validate_criteria(self, attrs, source):
+        criteria = attrs[source]
+        if not criteria:
+            raise serializers.ValidationError("Must have at least one criterion")
+        return attrs
+
+    #def validate(self, attrs):
+        #total_possible = sum(
+        #    max(option.get("points", 0) for option in criterion["options"])
+        #    for criterion in attrs["criteria"]
+        #)
+    #    total_possible = sum(crit.points_possible() for crit in attrs['criteria'])
 
-    canonical_form = json.dumps(rubric_dict, sort_keys=True)
-    return sha1(canonical_form).hexdigest()
+    #    if total_possible <= 0:
+    #        raise serializers.ValidationError(
+    #            "Rubric must have > 0 possible points."
+    #        )
+
+
+class AssessmentPartSerializer(serializers.ModelSerializer):
+    option = CriterionOptionSerializer()
+
+    class Meta:
+        model = AssessmentPart
+        fields = ('option',)
 
-def rubric_id_for(rubric_dict):
+
+class AssessmentSerializer(serializers.ModelSerializer):
+    parts = AssessmentPartSerializer(required=True, many=True)
+
+    class Meta:
+        model = Assessment
+        fields = ('submission', 'rubric', 'scored_at', 'scorer_id', 'score_type')
+
+
+
+def rubric_from_dict(rubric_dict):
     """Given a rubric_dict, return the rubric ID we're going to submit against.
 
     This will create the Rubric and its children if it does not exist already.
@@ -70,7 +127,7 @@ def rubric_id_for(rubric_dict):
     rubric_dict = deepcopy(rubric_dict)
 
     # Calculate the hash based on the rubric content...
-    content_hash = content_hash_for_rubric_dict(rubric_dict)
+    content_hash = Rubric.content_hash_for_rubric_dict(rubric_dict)
 
     try:
         rubric = Rubric.objects.get(content_hash=content_hash)
@@ -78,8 +135,7 @@ def rubric_id_for(rubric_dict):
         rubric_dict["content_hash"] = content_hash
         rubric_serializer = RubricSerializer(data=rubric_dict)
         if not rubric_serializer.is_valid():
-            raise ValueError("Some better Exception here")
-
+            raise InvalidRubric(rubric_serializer.errors)
         rubric = rubric_serializer.save()
 
-    return rubric.id
+    return rubric

apps/openassessment/peer/test/rubric_data/empty_criteria.json

+{
+  "prompt": "Create a plan to deliver edx-tim!",
+  "criteria": [
+  ]
+}

apps/openassessment/peer/test/rubric_data/empty_options.json

+{
+  "prompt": "Create a plan to deliver edx-tim!",
+  "criteria": [
+    {
+      "order_num": 0,
+      "prompt": "Is the deadline realistic?",
+      "options": [
+        {
+          "order_num": 0,
+          "points": 0,
+          "name": "No",
+          "explanation": ""
+        },
+        {
+          "order_num": 1,
+          "points": 2,
+          "name": "Maybe",
+          "explanation": ""
+        },
+        {
+          "order_num": 2,
+          "points": 4,
+          "name": "Yes",
+          "explanation": ""
+        }
+      ]
+    },
+    {
+      "order_num": 1,
+      "prompt": "Describe the architecture.",
+      "options": [
+      ]
+    }
+  ]
+}

apps/openassessment/peer/test/rubric_data/missing_criteria.json

+{
+  "prompt": "Create a plan to deliver edx-tim!"
+}

apps/openassessment/peer/test/rubric_data/missing_options.json

+{
+  "prompt": "Create a plan to deliver edx-tim!",
+  "criteria": [
+    {
+      "order_num": 0,
+      "prompt": "Is the deadline realistic?"
+    },
+    {
+      "order_num": 1,
+      "prompt": "Describe the architecture.",
+      "options": [
+        {
+          "order_num": 0,
+          "points": 0,
+          "name": "Crazy",
+          "explanation": ""
+        },
+        {
+          "order_num": 1,
+          "points": 1,
+          "name": "Plausible",
+          "explanation": ""
+        },
+        {
+          "order_num": 2,
+          "points": 2,
+          "name": "Solid",
+          "explanation": ""
+        }
+      ]
+    }
+  ]
+}

apps/openassessment/peer/test/rubric_data/no_points.json

+{
+  "prompt": "Create a plan to deliver edx-tim!",
+  "criteria": [
+    {
+      "order_num": 0,
+      "prompt": "Is the deadline realistic?",
+      "options": [
+        {
+          "order_num": 0,
+          "points": 0,
+          "name": "No",
+          "explanation": ""
+        },
+        {
+          "order_num": 1,
+          "points": 0,
+          "name": "Maybe",
+          "explanation": ""
+        },
+        {
+          "order_num": 2,
+          "points": 0,
+          "name": "Yes",
+          "explanation": ""
+        }
+      ]
+    },
+    {
+      "order_num": 1,
+      "prompt": "Describe the architecture.",
+      "options": [
+        {
+          "order_num": 0,
+          "points": 0,
+          "name": "Crazy",
+          "explanation": ""
+        },
+        {
+          "order_num": 1,
+          "points": 0,
+          "name": "Plausible",
+          "explanation": ""
+        },
+        {
+          "order_num": 2,
+          "points": 0,
+          "name": "Solid",
+          "explanation": ""
+        }
+      ]
+    }
+  ]
+}

apps/openassessment/peer/test/test_api.py

@@ -8,7 +8,7 @@ from nose.tools import raises
 from mock import patch
 
 from openassessment.peer import api
-from openassessment.peer.models import PeerEvaluation
+from openassessment.peer.models import Assessment
 from submissions import api as sub_api
 from submissions.models import Submission
 from submissions.tests.test_api import STUDENT_ITEM, ANSWER_ONE
@@ -168,7 +168,7 @@ class TestApi(TestCase):
             MONDAY
         )
 
-    @patch.object(PeerEvaluation.objects, 'filter')
+    @patch.object(Assessment.objects, 'filter')
     @raises(sub_api.SubmissionInternalError)
     def test_error_on_get_evaluation(self, mock_filter):
         submission = sub_api.create_submission(STUDENT_ITEM, ANSWER_ONE)
@@ -205,4 +205,4 @@ class TestApi(TestCase):
         self.assertIsNotNone(evaluation)
         self.assertEqual(evaluation["points_earned"], sum(points_earned))
         self.assertEqual(evaluation["points_possible"], points_possible)
-        self.assertEqual(evaluation["feedback"], feedback)
+        self.assertEqual(evaluation["feedback"], feedback)
\ No newline at end of file

apps/openassessment/peer/test/test_serializers.py

@@ -5,32 +5,85 @@ from ddt import ddt, file_data
 from django.test import TestCase
 
 from openassessment.peer.models import Criterion, CriterionOption, Rubric
-from openassessment.peer.serializers import rubric_id_for
+from openassessment.peer.serializers import (
+    InvalidRubric, RubricSerializer, rubric_from_dict
+)
 
 def json_data(filename):
     curr_dir = os.path.dirname(__file__)
     with open(os.path.join(curr_dir, filename), "rb") as json_file:
         return json.load(json_file)
 
-class TestPeerSerializers(TestCase):
 
-    def test_repeat_data(self):
+class TestRubricDeserialization(TestCase):
+
+    def test_rubric_only_created_once(self):
+        # Make sure sending the same Rubric data twice only creates one Rubric,
+        # and returns a reference to it the next time.
         rubric_data = json_data('rubric_data/project_plan_rubric.json')
 
-        rubric_id1 = rubric_id_for(rubric_data)
-        rubric_id2 = rubric_id_for(rubric_data)
+        r1 = rubric_from_dict(rubric_data)
 
-        self.assertEqual(rubric_id1, rubric_id2)
+        with self.assertNumQueries(1):
+            # Just the select -- shouldn't need the create queries
+            r2 = rubric_from_dict(rubric_data)
 
-        Rubric.objects.get(id=rubric_id1).delete()
+        self.assertEqual(r1.id, r2.id)
+        r1.delete()
 
-    def test_db_access(self):
-        rubric_data = json_data('rubric_data/project_plan_rubric.json')
+    def test_rubric_requires_positive_score(self):
+        with self.assertRaises(InvalidRubric):
+            rubric_from_dict(json_data('rubric_data/no_points.json'))
 
-        with self.assertNumQueries(4):
-            rubric_id1 = rubric_id_for(rubric_data)
+    def test_dont_accept_content_hash(self):
+        # we should always calculate this ourselves and not trust the user.
+        pass
 
-        with self.assertNumQueries(1):
-            rubric_id2 = rubric_id_for(rubric_data)
 
-        Rubric.objects.get(id=rubric_id1).delete()
+
\ No newline at end of file
+class TestCriterionDeserialization(TestCase):
+
+    def test_empty_criteria(self):
+        with self.assertRaises(InvalidRubric) as cm:
+            rubric_from_dict(json_data('rubric_data/empty_criteria.json'))
+        self.assertEqual(
+            cm.exception.errors,
+            {'criteria': [u'Must have at least one criterion']}
+        )
+
+    def test_missing_criteria(self):
+        with self.assertRaises(InvalidRubric) as cm:
+            rubric_from_dict(json_data('rubric_data/missing_criteria.json'))
+        self.assertEqual(
+            cm.exception.errors,
+            {'criteria': [u'This field is required.']}
+        )
+
+class TestCriterionOptionDeserialization(TestCase):
+
+    def test_empty_options(self):
+        with self.assertRaises(InvalidRubric) as cm:
+            rubric_from_dict(json_data('rubric_data/empty_options.json'))
+        self.assertEqual(
+            cm.exception.errors,
+            {
+                'criteria': [
+                    {},  # There are no errors in the first criterion
+                    {'options': [u'Criterion must have at least one option.']}
+                ]
+            }
+        )
+
+    def test_missing_options(self):
+        with self.assertRaises(InvalidRubric) as cm:
+            rubric_from_dict(json_data('rubric_data/missing_options.json'))
+        self.assertEqual(
+            cm.exception.errors,
+            {
+                'criteria': [
+                    {'options': [u'This field is required.']},
+                    {}  # No errors in second criterion
+                ]
+            }
+        )
+

doc/source/api.rst

@@ -48,3 +48,6 @@ Models
 ++++++
 .. automodule:: submissions.models
    :members:
+
+.. automodule:: openassessment.peer.models
+   :members: