From the HL7 Connectathon research (HL7 Connectathon/RESEARCH_FINDINGS_2026-07-15.md §2.2):
cqframework/cql-tests = ~1,731 language-level conformance tests (16 XML files, no retrieves/data model/terminology — our translator's QICore limitation is irrelevant here). Our exact runtime (cql-execution 3.3.x) has public posted results: 1,533 pass / 81 fail / 113 skip / 4 error — but that run used the Java translator. A WorkWell run measures the unpublished delta: what @cqframework/cql 4.0.0-beta.1 (our JS translator) breaks that the Java translator doesn't.
Phase 1 (~1–2 days): in-repo dev harness — parse the test XMLs, wrap each expression as a define, translate via our existing cql-translator.ts, execute via cql-execution, compare (evaluate the expected literal through the engine and use Equivalent to avoid writing a CQL literal parser; honor invalid="true|semantic" as expect-translation-error). Emit results in the cql-tests-runner JSON schema. Build a documented SkipList (known upstream clusters: Long type, LowBoundary/HighBoundary, decimal precision, Quantity mod/div, interval Expand — none used by our 14 runnable measures).
Phase 2 (optional, ~3–4 days): dev-only $cql FHIR operation so cql-tests-runner runs stock → eligibility to post vendor results to cqframework/cql-tests-results (credibility artifact for MIE).
Refs: https://github.com/cqframework/cql-tests · /cql-tests-runner · /cql-tests-results · https://hl7.org/fhir/uv/cql/OperationDefinition-cql-cql.html
From the HL7 Connectathon research (
HL7 Connectathon/RESEARCH_FINDINGS_2026-07-15.md§2.2):cqframework/cql-tests= ~1,731 language-level conformance tests (16 XML files, no retrieves/data model/terminology — our translator's QICore limitation is irrelevant here). Our exact runtime (cql-execution 3.3.x) has public posted results: 1,533 pass / 81 fail / 113 skip / 4 error — but that run used the Java translator. A WorkWell run measures the unpublished delta: what@cqframework/cql4.0.0-beta.1 (our JS translator) breaks that the Java translator doesn't.Phase 1 (~1–2 days): in-repo dev harness — parse the test XMLs, wrap each expression as a define, translate via our existing
cql-translator.ts, execute via cql-execution, compare (evaluate the expected literal through the engine and useEquivalentto avoid writing a CQL literal parser; honorinvalid="true|semantic"as expect-translation-error). Emit results in the cql-tests-runner JSON schema. Build a documented SkipList (known upstream clusters: Long type, LowBoundary/HighBoundary, decimal precision, Quantity mod/div, interval Expand — none used by our 14 runnable measures).Phase 2 (optional, ~3–4 days): dev-only
$cqlFHIR operation socql-tests-runnerruns stock → eligibility to post vendor results tocqframework/cql-tests-results(credibility artifact for MIE).Refs: https://github.com/cqframework/cql-tests · /cql-tests-runner · /cql-tests-results · https://hl7.org/fhir/uv/cql/OperationDefinition-cql-cql.html