#1274: Add inline compared syntax for N-way set-based output variable comparison

[RakeshBobba03]

#1274: Implements explicit inline comparison syntax for output variables in rules via a new compared: dict block within Output Variables, enabling N-way set-based comparisons (len >= 2) where the first variable serves as baseline and subsequent variables are compared against it. The implementation flattens all variables (siblings + compared children) for UI display while isolating comparison logic to only variables within compared blocks, always uses set-based (order-independent) comparison. Reporting now shows formatted comparison summaries (missing/extra items) followed by raw variable lists, with Excel multi-line rendering support.

Attached are the Rule and dataset used for testing:

CORE-000334.yaml
unit-test-coreid-CG0016-negative.xlsx

[RamilCDISC]

I ran a validation using dev editor and updated the rule to have USUBJID also in output variables like following

Outcome:
  Message: At least one expected variable is missing from dataset
  Output Variables:
    - USUBJID
    - compared:
        - $dataset_variables
        - $expected_variables
    

The AE dataset in attached excel file has USUBJID column but engine returns column not found.

[
  {
    "executionStatus": "success",
    "dataset": "ae.xpt",
    "domain": "AE",
    "variables": [
      "$dataset_variables",
      "$expected_variables",
      "USUBJID"
    ],
    "message": "At least one expected variable is missing from dataset",
    "errors": [
      {
        "value": {
          "$dataset_variables": [
            "STUDYID",
            "DOMAIN",
            "USUBJID",
            "AESEQ",
            "AELNKID",
            "AETERM",
            "AELLT",
            "AELLTCD",
            "AEDECOD",
            "AEPTCD",
            "AEHLT",
            "AEHLTCD",
            "AEHLGT",
            "AEHLGTCD",
            "AEBDSYCD",
            "AESOC",
            "AESOCCD",
            "AESEV",
            "AEACN",
            "AEREL",
            "AEOUT",
            "AESCAN",
            "AESCONG",
            "AESDISAB",
            "AESDTH",
            "AESHOSP",
            "AESLIFE",
            "AESOD",
            "EPOCH",
            "AESTDTC",
            "AEENDTC",
            "AESTDY",
            "AEENDY",
            "AEENRTPT",
            "AEENTPT"
          ],
          "$expected_variables": [
            "AELLT",
            "AELLTCD",
            "AEPTCD",
            "AEHLT",
            "AEHLTCD",
            "AEHLGT",
            "AEHLGTCD",
            "AEBODSYS",
            "AEBDSYCD",
            "AESOC",
            "AESOCCD",
            "AESER",
            "AEACN",
            "AEREL",
            "AESTDTC",
            "AEENDTC"
          ],
          "USUBJID": "Not in dataset"
        },
        "dataset": "ae.xpt"
      }
    ],
    "compare_groups": [
      [
        "$dataset_variables",
        "$expected_variables"
      ]
    ]
  }
]

[RakeshBobba03]

I ran a validation using dev editor and updated the rule to have USUBJID also in output variables

Thanks for bringing this to my attention @RamilCDISC

USUBJID is record-level data, not metadata. When it's included in Output Variables for a Variable Metadata Check rule, the rule operates on a metadata dataset, not the original data records. Since USUBJID is a data column and not part of the metadata structure, it doesn't exist in the metadata dataset. That's why it was showing "Not in dataset".

Sam and I agreed that the message "Not in dataset" is misleading in this context, since USUBJID actually exists in the original dataset, it's just not available in the metadata context for Variable Metadata Check rules. We decided to change the error message to "not available in metadata context" for Variable Metadata Check rules and Dataset Metadata Check rules (and their variants) to better reflect what's happening.

I updated the error message logic in actions.py to return "not available in metadata context" for metadata check rule types instead of "Not in dataset".

and

I updated the documentation in Rule_Type.md to clarify that output variables must match their respective rule types, and that for Variable Metadata Check and Dataset Metadata Check rules, variables not available in the metadata context will display "not available in metadata context" instead of the variable value.

[RamilCDISC]

If the rule has only one variable in comparison block then the report like

"Output_Variables": [
            "variable_name",
            {
                "compared": [
                    "$dataset_variables"
                ]
            }
        ]

the first row in the issue details sheet of report correctly outputs both dataset_variable and variable_name. But for all remaining rows only variable_name is put in the report. Please see attached report.

CORE-Report-2026-01-23T16-55-16.xlsx

[gerrycampion]

Sorry to throw this in here, but after seeing this PR and giving this some thought, I don't think this should be a change in the reporting mechanism. I think we should just have a new operation like minus. Then we could have a rule like:

Operations:
  - id: $expected_variables
    operator: expected_variables
  - id: $dataset_variables
    operator: get_column_order_from_dataset
  - id: $expected_minus_dataset
    name: $expected_variables
    operator: minus
    value: $dataset_variables
Check:
  all:
    - name: $expected_minus_dataset
      operator: non_empty
Outcome:
  Message: At least one expected variable is missing from dataset
  Output Variables:
    - $expected_variables
    - $dataset_variables
    - $expected_minus_dataset

Would this work?

[gerrycampion]

Sorry to throw this in here, but after seeing this PR and giving this some thought, I don't think this should be a change in the reporting mechanism. I think we should just have a new operation like minus. Then we could have a rule like:

Would this work?

btw if we go this route, it's probably easiest to just create a new branch