ConformanceReport

The result of a CDISC conformance validation run.

Usage

ConformanceReport(
    validations=dict(),
    package=None,
    agency=None,
    core=None,
    native_result=None
)

A ConformanceReport is returned by validate_sdtmig() and SubmissionPackage.validate_conformance(). It exists in one of two forms depending on the engine used:

Built-in rules engine (is_rules is True) — produced by Pointblank’s SDTMIG rule catalog. Each rule is evaluated against the supplied datasets and receives one of five statuses: "pass", "fail", "error", "not_applicable", or "not_supported". Row-level findings (the individual failing records) are collected for RECORD_CHECK rules and accessible via findings_df() and get_findings_table().
CDISC CORE (is_core is True) — produced by the external CDISC CORE command-line engine. Rule-keyed findings and run provenance are exposed via findings() and rules().

In a Jupyter or Quarto notebook the report renders automatically as a color-coded rule summary table (calling _repr_html_() is equivalent to get_tabular_report()._repr_html_()).

Parameters

validations: dict[str, Validate] = dict(): Reserved for legacy use; not populated by the built-in engine.
package: SubmissionPackage | None = None: The SubmissionPackage the report was produced from, if any.
agency: str | None = None: The agency rule-set selector used for the run (None for CDISC base rules).
core: ParsedCoreReport | None = None: The parsed CDISC CORE report (CORE form only). None for built-in engine reports.
native_result: NativeConformanceResult | None = None: The NativeConformanceResult produced by the rules engine (native form only).

Attributes

Name	Description
is_core	Whether this report wraps CDISC CORE engine results (vs. built-in engine results).
is_rules	Whether this report was produced by Pointblank’s built-in rule-based conformance engine.
n_datasets	Number of datasets validated.

is_core

Whether this report wraps CDISC CORE engine results (vs. built-in engine results).

is_core: bool

is_rules

Whether this report was produced by Pointblank’s built-in rule-based conformance engine.

is_rules: bool

n_datasets

Number of datasets validated.

n_datasets: int

Methods

Name	Description
all_passed()	Whether the run reported no conformance failures.
findings()	Return the row-level findings.
findings_df()	Return all row-level findings as a Polars DataFrame.
from_core_report()	Build a CORE-backed ConformanceReport from a CDISC CORE JSON report.
get_findings_table()	Build a record-level findings table as a styled Great Tables object.
get_tabular_report()	Build a rule-level conformance summary table as a styled Great Tables object.
get_validation()	Get the Validate object for a single dataset (case-insensitive).
issues()	Return the conformance issues found.
rules()	Return the per-rule run results.
summary()	Return a summary of the conformance run.
to_excel()	Save the conformance report as an Excel workbook.
to_json()	Save the conformance report as a JSON file.

all_passed()

Whether the run reported no conformance failures.

Usage

Source

all_passed()

For built-in engine reports, this is True when every check in every dataset passed with no failing test units. For CORE reports, this is True when no rule reported an issue or execution error.

findings()

Return the row-level findings.

Usage

Source

findings()

For CORE reports, returns CoreFinding objects from CORE’s Issue_Details. For built-in engine reports, returns NativeRowFinding objects. For Validate-based reports, returns an empty list.

findings_df()

Return all row-level findings as a Polars DataFrame.

Usage

Source

findings_df()

Each row represents one failing record captured during the conformance run. Use this method for programmatic analysis (filtering by rule, grouping by subject, exporting to CSV, or joining back to the source datasets to investigate root causes).

Only RECORD_CHECK and DATASET_CONTENTS_CHECK rules produce row-level findings; rules that check metadata or domain presence (e.g., VARIABLE_METADATA_CHECK, DOMAIN_PRESENCE_CHECK) report a finding count in get_tabular_report() but do not appear here. To see the visual findings table call get_findings_table() instead.

Findings are capped at 100 rows per rule to bound memory use on large datasets. The n_issues value shown in get_tabular_report() always reflects the true total count for a rule, even when more than 100 records failed.

Returns

polars.DataFrame

One row per captured finding with the following columns:

rule_id: CDISC CORE rule identifier (e.g., "SDTM-007").
dataset: The SDTM domain the failing record belongs to (e.g., "AE").
row_index: 0-based row position of the failing record in the source dataset.
usubjid: Unique Subject Identifier from the "USUBJID" column, if present.
checked_column: The specific variable that violated the rule (e.g., "SEX").
checked_value: The actual value of checked_column in that row.
description: Human-readable rule description. Derived first from the rule’s operations; falls back to the conditions tree for rules with no explicit operations (e.g., range checks like AGE < 0).
checked_value: The actual value of checked_column in that row.
description: Human-readable rule description.

Returns an empty DataFrame (with the same schema) when all rules pass.

Raises

TypeError: If called on a CDISC CORE-backed report. Use findings() instead, which returns a list of CoreFinding objects.

from_core_report()

Build a CORE-backed ConformanceReport from a CDISC CORE JSON report.

Usage

Source

from_core_report(report, package=None, agency=None)

Parameters

report: dict | ParsedCoreReport: Either a raw CORE JSON report (dict, as produced by core validate -of JSON) or an already-parsed ParsedCoreReport.
package: SubmissionPackage | None = None: The SubmissionPackage the run was produced from, if any.
agency: str | None = None: The agency rule-set selector used for the run.

Returns

ConformanceReport: A report in CORE form (is_core is True).

get_findings_table()

Build a record-level findings table as a styled Great Tables object.

Usage

Source

get_findings_table()

Returns one row per failing record captured by Pointblank’s built-in rules engine. This is the drill-down companion to get_tabular_report(): where the tabular report shows one row per rule with an aggregate issue count, the findings table shows the individual offending records so reviewers can trace violations back to specific subjects and variables.

Table Layout

The table has two column spanners:

Rule: Domain and Description identify which rule fired and in which domain.
Finding: USUBJID, Column, Row, and Value identify the specific record.
- USUBJID: the unique subject identifier (e.g., "CDISCPILOT01-01-001").
- Column: the variable that violated the rule (e.g., "SEX").
- Row: 1-based row number of the failing record in the source domain dataset.
- Value: the actual value found in Column for that row.

The header shows the standard and version (e.g., SDTMIG 3-4) alongside a breakdown of how many rules passed, failed, and were not applicable across the full run.

A narrow red bar on the left edge of each row marks it as a failure, consistent with the color coding in get_tabular_report().

Findings Cap

At most 100 findings per rule are shown. When a rule has more than 100 failing records the table shows the first 100; the true total is always visible in get_tabular_report().

Returns

GT: A styled great_tables.GT object. Renders automatically in Jupyter and Quarto notebooks.

Raises

TypeError: If called on a CDISC CORE-backed report. The findings table is only available for built-in engine results.
ValueError: If there are no row-level findings to display (i.e., all applicable rules passed).

get_tabular_report()

Build a rule-level conformance summary table as a styled Great Tables object.

Usage

Source

get_tabular_report()

Returns one row per rule in the catalog, summarizing whether each rule passed, failed, was not applicable, or could not be evaluated. This is the high-level overview; use get_findings_table() or findings_df() to drill into the individual failing records.

Table Layout

Each row contains:

A colored status bar on the left edge: green for pass, red for fail, amber for error, and grey for not-applicable or not-supported.
Rule — CDISC CORE rule identifier (e.g., "SDTM-007").
Domain — The SDTM domain(s) the rule targets. Rules that apply to every domain show a comma-separated list; rules targeting all SUPP– datasets show "SUPP--".
Type — The rule category: Record, Variable, Metadata, Domain, Dataset, Define, or Codelist.
Issues — Count of failing records or dataset-level violations. Shown in bold red when non-zero. This count always reflects the true total, even when the findings table caps display at 100 rows per rule.
Description — Human-readable explanation of what the rule checks.

Rows are sorted by severity: failing rules appear first, followed by errors, passing rules, not-applicable rules, and unsupported rule types.

The table header shows "CDISC Conformance" with a PASS or FAIL badge, and a subtitle line with the standard, version, and a count breakdown (e.g., SDTMIG 3-4 · 410 passed · 4 failed · 12 n/a).

Returns

GT: A styled great_tables.GT object set in IBM Plex Sans / IBM Plex Mono. Renders automatically in Jupyter and Quarto notebooks; call ._repr_html_() to get the HTML string directly. This is the same object produced by _repr_html_().

Raises

TypeError: If called on a CDISC CORE-backed report. The tabular report is only available for built-in engine results.

get_validation()

Get the Validate object for a single dataset (case-insensitive).

Usage

Source

get_validation(name)

issues()

Return the conformance issues found.

Usage

Source

issues(severity=None, status=None)

Parameters

severity: str | None = None: (Built-in engine reports only.) Optional severity filter: "warning", "error", or "critical". Requires thresholds to have been set on the run. If None, all steps with failing test units are returned.
status: str | None = None: (CORE reports only.) Optional rule-status filter, e.g. "ISSUE REPORTED" or "EXECUTION ERROR". If None, all reported issues are returned.

Returns

list[dict]: For a built-in engine report, one dict per failing step, with keys dataset, step, assertion, column, n_failed, n, and severity.
For a CORE report, one dict per (dataset, rule) with reported issues, with keys dataset, rule_id, message, issues (count), and status.

rules()

Return the per-rule run results.

Usage

Source

rules(status=None)

For CORE reports, returns CoreRuleResult objects. For built-in engine reports, returns NativeRuleResult objects. For Validate-based reports, returns an empty list.

Parameters

status: str | None = None: Optional status filter. For CORE: e.g. "SUCCESS", "SKIPPED". For built-in engine reports: "pass", "fail", "error", "not_applicable", "not_supported".

summary()

Return a summary of the conformance run.

Usage

Source

summary()

Returns

dict: For a built-in engine report, a mapping of dataset name to a dict with keys n_steps, n_steps_failed, n_failed (failing test units), and all_passed.
For a CORE report, a single dict with keys standard, version, engine_version, n_rules, status_counts (rule counts by run status), n_issues (total reported issues), n_datasets, and all_passed.

to_excel()

Save the conformance report as an Excel workbook.

Usage

Source

to_excel(path)

For CORE reports the workbook contains sheets Issue_Summary, Issue_Details, Rules_Report, and Conformance_Details. For built-in engine reports the workbook contains Issues and Summary.

Requires the openpyxl package (pip install openpyxl or pip install 'pointblank[excel]').

Parameters

path: str | Path: Destination path (including filename). Parent directories are created if needed.

Returns

Path: The path written.

Raises

ImportError: If openpyxl or pandas are not installed.

to_json()

Save the conformance report as a JSON file.

Usage

Source

to_json(path)

For CORE reports the output mirrors the original CORE JSON structure (Conformance_Details, Dataset_Details, Issue_Summary, Issue_Details, Rules_Report), making the file readable by anything that parses a standard CORE report. For built-in engine reports the file contains summary and issues keys.

Parameters

path: str | Path: Destination path (including filename). Parent directories are created if needed.

Returns

Path: The path written.