DDF00046: grouped distinct crashes on empty ScheduledDecisionInstance reference subset

[Panikos]

Standard

CDISC USDM

Reference Rule ID(s)

DDF00046

Conformance Rule ID(s) (if published in CORE)

CORE-000804

CLI Command Used

core validate -s usdm -v 4-0 -dp /path/to/core-input.json -of JSON -o /path/to/core-report.json -p disabled -l error

CORE engine version: 0.16.0.

Rule YAML / relevant executable contract

The cached CORE contract for DDF00046 includes ScheduledActivityInstance and
ScheduledDecisionInstance, filters relation rows to Timing references, and then
runs grouped distinct(parent_id):

Core:
  Id: CORE-000804
  Version: '1'
Description: A timing must only be specified as being relative to/from a scheduled activity/decision instance that is defined within the same timeline as the timing.
Entities:
  Include:
    - ScheduledActivityInstance
    - ScheduledDecisionInstance
Operations:
  - id: $parent_of_scheduled_instance
    name: parent_id
    operator: distinct
    group:
      - id
      - rel_type
    group_aliases:
      - id
      - rel_type.Timing
Check:
  all:
    - get_dataset.instanceType is ScheduledActivityInstance or ScheduledDecisionInstance
    - get_dataset.rel_type equals reference
    - get_dataset.parent_rel is relativeToScheduledInstanceId or relativeFromScheduledInstanceId
    - get_dataset.rel_type.Timing equals definition
    - parent_id.Timing does not contain $parent_of_scheduled_instance

Minimal test data / precondition

Use a valid USDM schedule timeline containing:

• one ScheduledDecisionInstance definition in ScheduleTimeline.instances;
• no Timing.relativeFromScheduledInstanceId or
  Timing.relativeToScheduledInstanceId reference to that decision;
• optionally, valid Timing references to a ScheduledActivityInstance in the
  same timeline.

After DDF00046 conditions are applied for the ScheduledDecisionInstance
domain, the evaluation dataset is empty. This is a valid non-applicable subset:
USDM does not require every scheduled decision instance to be referenced by a
Timing.

A direct unit-level reproducer is to extend the existing grouped Distinct
test with a filter that matches zero rows while retaining the expected columns:

data = PandasDataset.from_dict(
    {
        "parent_id": ["timeline_1"],
        "id": ["decision_1"],
        "rel_type": ["definition"],
        "parent_rel": ["instances"],
    }
)
# Apply a condition equivalent to rel_type == "reference", yielding zero rows,
# then execute grouped Distinct over parent_id.

Actual output

The rule is reported as EXECUTION ERROR:

Failed to execute rule operation. Operation: distinct, Target: parent_id,
Domain: ScheduledDecisionInstance, Error: single positional indexer is out-of-bounds

The failure occurs in
BaseOperation._expand_operation_results_in_grouping when it evaluates
operation_col.iloc[0] on the empty post-condition series.

Expected output

DDF00046 should be non-applicable for the empty ScheduledDecisionInstance
reference subset and complete without an execution error or issue. Other
non-empty entity subsets must continue to be evaluated normally.

Suggested fix

Guard the empty series before reading its first value:

def _expand_operation_results_in_grouping(self, grouping_list):
    expanded = []
    for item in grouping_list:
        if item in self.evaluation_dataset.columns:
            operation_col = self.evaluation_dataset[item]
            if operation_col.empty:
                expanded.append(item)
                continue
            first_val = operation_col.iloc[0]
            # existing list/tuple expansion follows

The grouped operation should then return an empty evaluation result.

Regression coverage requested

• grouped distinct after filtering to zero rows;
• DDF00046 with activity-only Timing references and an unreferenced decision;
• decision-only references;
• mixed activity and decision references;
• valid same-timeline and invalid cross-timeline references.

This report is based on sanitized structural evidence only. No sponsor or
protocol content is included.