Reject a Key That Depends on a Later Column

Review ask

As a clinical trial statistician or programmer, challenge anything in this benchmark: describe a problem in Comments, or propose a fix with the Edit button on any source file.

Task: Is this a specification error a study team could realistically make, and should the build stop on it rather than continue?
Inputs: Do they trigger the failure for the reason the summary gives?
Expected failure: Is this the right error, raised at the right point, and would it tell the programmer what to fix?

The other tabs beside Comments are background for assessment, not under review.

Summary

EditUnder reviewWhat this benchmark means

README.md

Goal: carry the analysis date (ADT), study day (ADY), and measured value (AVAL) through from a pre-derived ADVS slice (ADVS_RAW), placing each record in the analysis window its study day falls in.

Input: pre-derived ADVS slice carrying analysis date (ADT), study day (ADY), and measured value (AVAL).

Variables:

AVISIT would be the analysis window the study day falls in: SCREENING before day 0, BASELINE on days 0 and 1, WEEK 2 on days 2 to 21, WEEK 4 on days 22 to 42, and POST-TREATMENT from day 43; blank when the study day is missing, so the record belongs to no window.

Note: a record is identified here by study, subject, parameter, and analysis window. With nothing else choosing the rows, those identities are the rows, so they are settled before any other value is computed. The window is placed from the study day, which is one of those other values, so the run is rejected before any data is read and no artifact is accepted. Even if the window read the slice's study day directly, a record with no study day would have no window and so no identity. The expected file shows the intended rows, not an accepted result.

How to fix

Decide which keys the dataset is on. If it is one record per analysis visit, every record must fall in a window. Recover the analysis date in the governed source where it is available. Where it is not, keep the record out of the dataset rather than give it a place it does not have. A row template selects the records that become rows, and each row then comes from one kept record, so its window can be placed from its study day. Move one column's derivation into it and leave the rest where they are:

columns:
  - name: AVAL
    type: float

rows:
  - id: windowed
    filter: "ADVS_RAW.ADY IS NOT NULL"
    derivations:
      AVAL:
        source: ADVS_RAW.AVAL

If it is one record per collected record, identify rows by the collected record instead, and a record with no window keeps an empty one:

keys: [STUDYID, USUBJID, PARAMCD, VSSEQ]

Do not label the record into a placeholder window such as NOT ASSIGNED. That keeps the row by asserting an analysis visit the data does not support, and a second such record collides with the first.

Expected failure

Editthe run is rejected; no artifact is accepted

expected/error.yaml

phase: validation
condition: key_dependency
requirement: REQ-0074

expected/error.yaml

phase: validation
condition: key_dependency
spec_paths:
  - columns.AVISIT.derivation
requirement: REQ-0074
context:
  column: AVISIT
  dependency: ADY

YAML specification

1 spec file

spec.yaml

Edit61 lines

1schema_version: "1.0"2domain: ADVS3keys: [STUDYID, USUBJID, PARAMCD, AVISIT]4input:5  ADVS_RAW: {path: input/advs_raw.csv, types: {ADT: date}}67output:8  path: advs.csv9  columns: [STUDYID, USUBJID, PARAMCD, VSSEQ, ADT, ADY, AVAL, AVISIT]1011columns:12  - name: STUDYID13    type: str14    label: Study Identifier15    derivation: ADVS_RAW.STUDYID1617  - name: USUBJID18    type: str19    label: Unique Subject Identifier20    derivation: ADVS_RAW.USUBJID2122  - name: PARAMCD23    type: str24    label: Parameter Code25    derivation: ADVS_RAW.PARAMCD2627  - name: VSSEQ28    type: int29    label: Sequence Number30    derivation: ADVS_RAW.VSSEQ3132  - name: ADT33    type: date34    label: Analysis Date35    derivation: ADVS_RAW.ADT3637  - name: ADY38    type: int39    label: Analysis Relative Day40    derivation: ADVS_RAW.ADY4142  - name: AVAL43    type: float44    label: Analysis Value45    derivation: ADVS_RAW.AVAL4647  # AVISIT is a key, and with no row templates the rows are the distinct48  # key values, settled before any other column. Reading the ADY column49  # here is therefore rejected before any data is read. Windows: SCREENING50  # before day 0, BASELINE days 0-1, WEEK 2 days 2-21, WEEK 4 days 22-42,51  # POST-TREATMENT from day 43.52  - name: AVISIT53    type: str54    label: Analysis Visit55    derivation:56      cut:57        source: ADY58        breaks: [0, 2, 22, 43]59        labels: [SCREENING, BASELINE, WEEK 2, WEEK 4, POST-TREATMENT]60        right: false61        missing: null

Inputs

Under review1 source file

input/advs_raw.csv

Edit9 rows

input/advs_raw.csv
STUDYID	PARAMCD	VSSEQ	ADT	ADY	AVAL
YAMAA-01	SYSBP	1	2025-01-08	1	132
YAMAA-01	SYSBP	2	2025-01-22	15	138
YAMAA-01	SYSBP	3	2025-02-05	29	130
YAMAA-01	SYSBP	4	2025-02-19	43	128
YAMAA-01	SYSBP	1	2025-02-08	1	118
YAMAA-01	SYSBP	2	2025-02-21	14	122
YAMAA-01	SYSBP	3	—	—	120
YAMAA-01	SYSBP	1	2025-01-02	-5	140
YAMAA-01	SYSBP	2	2025-01-07	0	136

Unexpected Output

Under review9 rows presented to the failing check - not an accepted artifact

expected/advs.csv

Edit9 rows

expected/advs.csv
STUDYID	PARAMCD	VSSEQ	ADT	ADY	AVAL	AVISIT
YAMAA-01	SYSBP	1	2025-01-08	1	132	BASELINE
YAMAA-01	SYSBP	2	2025-01-22	15	138	WEEK 2
YAMAA-01	SYSBP	3	2025-02-05	29	130	WEEK 4
YAMAA-01	SYSBP	4	2025-02-19	43	128	POST-TREATMENT
YAMAA-01	SYSBP	1	2025-02-08	1	118	BASELINE
YAMAA-01	SYSBP	2	2025-02-21	14	122	WEEK 2
YAMAA-01	SYSBP	3	—	—	120	—
YAMAA-01	SYSBP	1	2025-01-02	-5	140	SCREENING
YAMAA-01	SYSBP	2	2025-01-07	0	136	BASELINE

Shaded columns are derived— = missing value

Comments and assessment materials

Dataset Name	Variable Name	Variable Label	Type	Codelist/Controlled Terms	Computational Method	Origin
ADVS	STUDYID	Study Identifier	character		Copy value from ADVS_RAW.STUDYID.	Predecessor
ADVS	USUBJID	Unique Subject Identifier	character		Copy value from ADVS_RAW.USUBJID.	Predecessor
ADVS	PARAMCD	Parameter Code	character		Copy value from ADVS_RAW.PARAMCD.	Predecessor
ADVS	VSSEQ	Sequence Number	numeric		Copy value from ADVS_RAW.VSSEQ.	Predecessor
ADVS	ADT	Analysis Date	numeric	ISO 8601	Copy value from ADVS_RAW.ADT.	Predecessor
ADVS	ADY	Analysis Relative Day	numeric		Copy value from ADVS_RAW.ADY.	Predecessor
ADVS	AVAL	Analysis Value	numeric		Copy value from ADVS_RAW.AVAL.	Predecessor
ADVS	AVISIT	Analysis Visit	character		Band ADY at 0, 2, 22, 43 into "SCREENING", "BASELINE", "WEEK 2", "WEEK 4" and "POST-TREATMENT", intervals closed on the left; a missing source → blank.	Derived

Version	Date	Author	Description
1.3	2026-09-29	yamaa docs build	Show study codelist names, external dictionaries, and ISO 8601 formats in variable rows; list the document's codelist values and metadata
1.2	2026-09-29	yamaa docs build	Classify ADaM copies from a declared input as Predecessor and recodes as Derived, and SDTM recodes of locally derived columns as Derived
1.1	2026-09-29	yamaa docs build	Resolve lookups into the donor record they select, and give every derivation shape a plain-language rule instead of its raw YAML
1.0		yamaa docs build	Initial generation from spec.yaml

How to fix

input/advs_raw.csv

expected/advs.csv

Comments

Mapping spec