AssessmentWorkflow.staff_score_exists() always returns False (looks up get_latest_assessment; the staff API defines get_latest_staff_assessment), so the ORA Summary Report's is_staff_grade_received is always 0

[shaurya416]

AssessmentWorkflow.staff_score_exists() always returns False (looks up get_latest_assessment; the staff API defines get_latest_staff_assessment), so the ORA Summary Report's is_staff_grade_received is always 0

openassessment/workflow/models.py:558-571 (read at eee6b7b, current master)

def staff_score_exists(self):
    """
    Check if a staff score exists for this submission.
    """
    steps = self._get_steps()
    step_for_name = {step.name: step for step in steps}
    staff_step = step_for_name.get(self.STAFF_STEP_NAME)
    if staff_step is not None:
        get_latest_func = getattr(staff_step.api(), 'get_latest_assessment', None)   # :566
        if get_latest_func is not None:
            staff_assessment = get_latest_func(self.submission_uuid)
            if staff_assessment is not None:
                return True
    return False

For the staff step, api() resolves to openassessment.assessment.api.staff (models.py:934-936). _get_steps() always creates that step (models.py:472-505), so this path runs for every AssessmentWorkflow. The staff module defines get_latest_staff_assessment (staff.py:163) and no get_latest_assessment, and it has no alias or module-level __getattr__. teams.py is the same (teams.py:157). getattr(..., None) therefore returns None on every call, and the function falls through to return False.

Repo-wide, get_latest_assessment appears only at models.py:566 and as test method names in assessment/test/test_team.py:152,160, which call get_latest_staff_assessment. The lookup has been mismatched since the method was introduced in #758, the same PR that added get_latest_staff_assessment to staff.py.

Two things read this function:

• OraAggregateData.collect_ora2_summary() (data.py:816): is_staff_grade_received = 1 if aw.staff_score_exists() else 0. This backs the instructor-dashboard ORA Summary Report (upload_ora2_summary in openedx-platform lms/djangoapps/instructor_task/tasks_helper/misc.py:282-291). Per the code path, the is_staff_grade_received column is 0 for every row, including learners who have a staff assessment.
• AssessmentWorkflow.set_score() (models.py:551): if not self.staff_score_exists(): sub_api.set_score(...). The guard never blocks, but its practical reach is narrow. update_from_assessments (models.py:453-456) already skips set_score when the score carries a staff_id, and get_score (models.py:254-293) returns the staff score first whenever one exists. The only call where the guard would change behaviour is cancel() (models.py:605-607).

Measured

This was a modeled reproduction, not an observation on a running Open edX instance. No repository code was imported, installed or run. The function text of staff_score_exists, set_score and cancel was copied verbatim from models.py (ast segments) and executed in a plain Python process with no network. Substitutions used:

• a SimpleNamespace in place of the workflow instance
• a Step class whose api() returns a stand-in module
• a stand-in module exposing exactly the top-level def names that ast finds in api/staff.py (only get_latest_staff_assessment has behaviour)
• a recorder in place of sub_api.set_score
• a logger stub

The data.py:816 line was executed as read from the file.

Case	staff_score_exists()	is_staff_grade_received	sub_api.set_score writes
A. staff assessment exists, real staff.py names	False	0	1 (guard did not block)
B. no staff assessment, real staff.py names	False	0	1
C. staff assessment exists, module exposes get_latest_assessment	True	1	0
D. no staff assessment, module exposes get_latest_assessment	False	0	1

A and B are indistinguishable. C and D are the controls: they show the function body behaves correctly once the name resolves.

cancel() (models.py:605-607) with a 7/10 staff score present, using the copied cancel text:

Surface	points written to the submissions API	status
real staff.py names (as shipped)	[0]	cancelled
module exposing get_latest_assessment	[]	cancelled

Consequence

• The ORA Summary Report's is_staff_grade_received column cannot report a staff grade. It is 0 for staff-graded learners as well as ungraded ones.
• The existing test_collect_ora2_summary (openassessment/tests/test_data.py:914) expects is_staff_grade_received=0 on both rows. Its fixture has only peer assessments, so 0 is correct there; the test does not cover a staff-graded submission and cannot catch this.
• The set_score guard has never been live. cancel() currently zeroes a staff-scored submission only because the guard is dead; its docstring says "Set the points earned to 0 and workflow status to cancelled." Correcting the name alone would make set_score refuse inside cancel(), so a cancelled submission would keep its staff points in the score table.

Suggested fix

Look up get_latest_staff_assessment, the name already used by api/staff.py:60, api/teams.py:55 and xblock/apis/assessments/staff_assessment_api.py:51. Because the guard has never run, the rename should not land alone. Land it together with:

1. cancel() writing the zero directly through sub_api.set_score (or a force argument on set_score), so that cancellation still zeroes a staff score.
2. Tests:
   • staff_score_exists: True with a staff assessment, False without.
   • cancel() on a staff-scored workflow: points end at 0.
   • A staff-graded row in test_collect_ora2_summary asserting is_staff_grade_received=1.