FHIR Terminology Ecosystem IG
1.11.1-snapshot - Release International flag

FHIR Terminology Ecosystem IG, published by HL7 FHIR Product Director. This guide is not an authorized publication; it is the continuous build for version 1.11.1-snapshot built by the FHIR (HL7® FHIR® Standard) CI Build. This version is based on the current content of https://github.com/HL7/fhir-tx-ecosystem-ig/ and changes regularly. See the Directory of published versions

Change Log

Changelog

All notable changes to this IG will be documented in this file. The format is based on Keep a Changelog.

Up to 1.10.0, this history covered the terminology tests, and was maintained on the Test Case History page in the IG; before 1.7.5, no reliable version history was maintained.

[Unreleased]

Ecosystem and IG
  • Resolution: when a code system resolution finds no server, the registry may say that it knows of no code system at all on the host of the URL (unknown-host), so that clients don't have to ask about every other code system from that host (see Unknown Hosts). formatVersion is unchanged: older clients ignore the new property
  • Registrations: a registry (in the master registration file) or a server (in a server registry) can be marked "disabled" : true, which takes it out of the ecosystem without deleting its registration (see Disabling)
New Tests
  • A supplement with the url of the code system it supplements, and a version of its own (new supplement-url suite): this is not valid - a supplement needs a canonical url of its own - but it is found in published packages, so servers answer as usual, and warn about the supplement (CODESYSTEM_CS_SUPP_SAME_URL) whenever they look up the code system it has the url of. The supplement's version is not a version of the code system: a value set that asks for it can't be expanded (supplement-url-expand-v2) or validated against (supplement-url-validate-v2), and the valid versions named are only the code system's; a value set that asks for the code system's own version still expands (supplement-url-expand-v1), and the supplement is still used when asked for (supplement-url-expand-v1-use-supplement). New setup resources supplement-url/codesystem-supplement-url-base.json, codesystem-supplement-url-supplement.json, valueset-supplement-url-v1.json and valueset-supplement-url-v2.json
  • A versioned coding against a value set whose include names an unknown code system without a version (errors suite, unknown-system-version): the code system is unknown, and nothing else - there is no version in the include for the coding's version to differ from, so no version mismatch is reported
  • An in filter with an invalid code in its list (snomed suite): concept in 50325005,999999006 (999999006 is not in the edition, though its check digit is right) is treated as is-a on an invalid code is (snomed-expand-isa-invalid): the server may refuse to expand the value set (4xx), or expand it with none of the filter's codes - not even the valid one - and an expansion-issue naming the filter (snomed-expand-in-invalid); and $validate-code of an unknown code against it is not valid, with an error saying the value set is not valid, and why (snomed-validate-in-invalid). New setup resource sct/valueset-filter-in-invalid.json
  • Add test to ensure that if CodeSytem is versionNeeded = true, a warning is returned from $validate-code if there is no explicit version
  • Text filters and supplements (search suite): a supplement in use whose text is what a text filter matched is a used-supplement, whether or not the displays shown are its own. With displayLanguage=es, an unreferenced global language pack's Spanish displays match (search-lang-supplement-display); a code that matches only on the pack's Spanish designation is in the expansion with the code system's own Spanish designation as its display, and the pack is still reported as used (search-lang-supplement-only); when the code system's own designation is what matched, reporting the pack is optional (search-lang-own). The filters are whole words, as given, so the results are the same however a server matches text. New setup resources search/codesystem-search-lang.json, search/codesystem-search-lang-es.json and search/valueset-search-lang-all.json
  • LOINC property in filters (loinc suite): core's [property] in [string,string...] is a property value equal to one of the strings, so CLASSTYPE in 3,4 (the numbers, as for =), CLASSTYPE in Claims attachment,Surveys and ORDER_OBS in Order,Both are the union of = each value, and ORDER_OBS in order,BOTH the same (LOINC property values are not case sensitive). $validate-code of a code with each listed value and one with another (loinc-validate-filter-classtype-in-*, loinc-validate-filter-order-obs-in-*), and the sizes of the expansions (loinc-expand-filter-classtype-in-total, -in-names-total, loinc-expand-prop-order-obs-in-total, -in-case-total). New setup resources loinc/valueset-filter-classtype-in.json, -classtype-in-names.json, -prop-order-obs-in.json and -prop-order-obs-in-case.json

  • ConceptMaps that name value sets in place of codes (translate suite, R5 and later; see Value sets in place of codes): a target that names a value set maps to every code in it (translate-vs-target); a group's unmapped (fixed) that names a value set maps every unmapped code to every code in it (translate-vs-unmapped); and in reverse, an element that names a value set maps from every code in it (translate-vs-reverse-element). Matching the code being translated against an element's value set (translate-vs-element), or in reverse a target's (translate-vs-reverse-target), is optional, and those two tests are tx.fhir.org mode. New setup resources translate/codesystem-vs-source.json, codesystem-vs-target.json, valueset-vs-t12.json, valueset-vs-s34.json, valueset-vs-t46.json and ConceptMap-valuesets.json
  • LOINC answer codes and SNOMED CT (loinc suite): $translate with LOINC's implicit map http://loinc.org/cm/sct, from an answer code (LA32-8 to 373067005, loinc-translate-answer-sct) and to one (260413007 to LA137-2, loinc-translate-sct-answer)
  • A new translate-check suite (tx.fhir.org mode): not a requirement - tx.fhir.org using the ConceptMaps it holds to comment, in $validate-code, on the codings of a CodeableConcept: a hint where a map relates a code only to codes other than those it is with, and a warning where a map says it has no mapping (noMap) or is not related to one of them. Ten tests use the translate fixtures, including maps that name value sets (R5 and later); two use LOINC's mappings to SNOMED CT (and need loinc mode as well). New setup resource translate-check/valueset-translate-all.json
Corrections and New Tests
  • Invalid codes in an enumerated include: a value set that lists a code its code system doesn't have can't be expanded, or can be expanded without the code, saying so. These tests let the server choose ("http-code" : "200|4xx", which needs runner 7.0.6, so the simple-cases, loinc and snomed suites now give "runner" : "7.0.6"): either a 4xx with an OperationOutcome naming the code, or a 200 whose expansion leaves the code out and has an expansion-issue extension referring to a contained OperationOutcome with a warning naming it. New tests simple-expand-enum-invalid (xxxx), snomed-expand-enum-invalid (999999006, not in the edition but with a correct check digit) and snomed-expand-enum-invalid-check-digit (999999007), loinc-expand-enum-invalid (99999-5, not in LOINC but with a correct check digit) and loinc-expand-enum-invalid-check-digit (99999-6), each with a valid code beside it; the two kinds of invalid code are separate tests, since a server may treat them differently. New setup resources simple/valueset-enumerated-invalid.json, sct/valueset-enumerated-invalid.json, sct/valueset-enumerated-invalid-check-digit.json, loinc/valueset-enumerated-invalid.json and loinc/valueset-enumerated-invalid-check-digit.json. simple-expand-enum-bad and loinc-expand-enum-bad expected the invalid codes to be left out silently; they now follow the same pattern
  • Invalid codes in an include filter: the same choice for a filter that selects by concept (concept is-a <code>) on a code the code system doesn't have - a 4xx, or a 200 whose expansion has none of the filter's codes and an expansion-issue naming the code. New tests simple-expand-isa-invalid, snomed-expand-isa-invalid, snomed-expand-isa-invalid-check-digit, loinc-expand-isa-invalid and loinc-expand-isa-invalid-check-digit (the same codes as the enumerated tests, each beside a valid enumerated code), with new setup resources simple/valueset-filter-isa-invalid.json, sct/valueset-filter-isa-invalid.json, sct/valueset-filter-isa-invalid-check-digit.json, loinc/valueset-filter-isa-invalid.json and loinc/valueset-filter-isa-invalid-check-digit.json
  • Existing tests brought into line (so the overload and icd-11 suites now give "runner" : "7.0.6" too): expand-enum-bad (overload: code3 is not in version 2.0.0), expand-adhoc-enum-uri and expand-adhoc-bad-code (icd-11) expected the invalid codes to be left out silently, and expand-adhoc-isa-uri and expand-pcs-uri-form (icd-11, a filter on entity URIs, which are not MMS codes) required a 4xx; all five now accept either. The message id in the 4xx responses of the last two is now VALUESET_EXPANSION_FILTER_UNKNOWN_CODE_ERROR
  • $validate-code against those invalid value sets: an unknown code (correct check digit - yyyy, 999998003, 99973-0) validated against each of the ten new value sets is not valid, and the OperationOutcome has, as well as the usual two errors (not in the value set, unknown code), a third: the value set is not valid (vs-invalid), saying why. The setting a server may have for expansions (refuse, or warn) doesn't change this. New tests simple-validate-enum-invalid, simple-validate-isa-invalid, snomed-validate-enum-invalid, snomed-validate-enum-invalid-check-digit, snomed-validate-isa-invalid, snomed-validate-isa-invalid-check-digit, and the same four for loinc
  • LOINC answer list filters, following the agreement with LOINC (UP-645): answers-for and answer-list are the defined filters. The LIST filter tests (loinc-expand-filter-list-request-parameters, loinc-validate-filter-list-type-good, loinc-validate-filter-list-bad) now run only in tx.fhir.org mode: tx.fhir.org only supports LIST for backwards compatibility. New answers-for equivalents of the LIST validation tests (loinc-validate-filter-answers-for3-good, loinc-validate-filter-answers-for3-bad, new setup resource loinc/valueset-filter-answers-for3.json). loinc-expand-filter-answer-list never tested answer-list - its value set was a copy of the LIST one; it now expands answer-list = LA9040-2 (the answer lists LL451-6 and LL6136-7), with new loinc-validate-filter-answer-list-good and -bad. The requirements say what servers must support for answer lists
  • translate-scope-1 to translate-scope-5 (translate suite, R5 and later): $translate with targetScope and no targetSystem. The maps into the code systems the target value set draws on apply even when they don't name it as their targetScope; only target codes in the value set are returned (a map's unmapped included); and when no map applies, the message names the source's code system and version - the version given, or the one the server took the source to be. New setup resource translate/valueset-target-some.json
  • snomed-translate-edition-*: the target of an edition to edition translation may be given as the release of the International Edition the national edition was built on (.../version/20250801), or - by a server that doesn't hold that release on its own - as the edition (http://snomed.info/sct/900000000000207008), so that the code it gives out is one it can validate

[1.11.0] - 2026-11-06

Testing Infrastructure
  • Runner version: a suite can now give runner, the oldest version of the test runner (the validator) that can run its tests (see Test Cases). It is 7.0.4 for this release: an older runner doesn't know some of the tests, and reports cs-batch-validate, for example, as "Unknown Operation". The runner doesn't check it (yet)
  • Messages: more expected responses give message text as $external (PR #91, and the SNOMED CT MRCM messages in the validate-code-pc-mrcm-* tests), and the externals file for tx.health-samurai.io has been extended. The operationoutcome-message-id and other issue extensions are now $optional$ in another 35 expected responses (PR #90)
  • expansion.property: an expansion that returns no codes - count=0, asking only for the total - needs no property declarations, but may have them (a server may declare the properties of the whole value set, e.g. status). The expected responses for totals accept either, and the requirements say so
  • translate-reverse-r5+ and translate-reverse-r5+-a are renamed translate-reverse-r5plus and translate-reverse-r5plus-a
  • exclude-gender2 no longer depends on the order the server sorts the codes in, with an exact variant for tx.fhir.org (PR #94); fixes to the notSelectable unprop tests (PR #88); byte order marks removed from 53 test files
Ecosystem and IG
  • The IG now separates the ecosystem framework from HL7's own instance of it: other authorities and networks can run their own ecosystems with the same framework. The registration process for HL7's ecosystem is documented in the ig-registry, and the coordination server has its own page
  • A new page documents the coordination server's public API: discovery (listing known servers), resolution (recommending a server for a code system), and the optional server choice (choosing one server for a request that involves several code systems)
  • The home page says which test cases are jointly maintained with the terminology authorities: the SNOMED CT tests with SNOMED International, and the ICD-11 tests with WHO
  • Approved servers: the HELEX Terminology Server and the HELEX Terminology Aggregator have been added, and FHIRsmith now passes 1.11.0
  • The change history of the test cases is now kept in this changelog, which replaces the Test Case History page
Corrections and New Tests
  • Global language packs: the global-lang-pack code introduced in 1.10.0 is withdrawn. A global language pack is a lang-pack supplement (codesystem-supplement-type) that is also marked with the new extension codesystem-supplement-global = true - a lang-pack supplement without it is not global - and servers have to support global language packs. A new global-lang-pack suite (mode general) checks that a global language pack passed in tx-resource is used for the code system it supplements without being referenced - in $validate-code, $lookup and $expand, where it is reported as a used-supplement - and only for that code system; and that any other supplement passed in tx-resource - including a language pack that is not marked global - is only used once it is referenced, in $validate-code, $lookup and $expand (glp-*-nl-unreferenced). Until now this was only tested in the tx.fhir.org mode
  • Global language packs and the code system's own designations: where the code system has its own designation in the language, it is the display for that code, and the pack provides the displays for the other codes; the pack's display stays a valid display, its designation is still listed, and the pack is still a used-supplement (glp-own-expand-de, glp-own-lookup-de, glp-own-validate-pack-display-de)
  • Supplement versions: a supplement only applies to the version of the code system it names - a global language pack for version 0.9.0 is not used on 1.0.0 (glp-expand-pinned-other-version), and requiring it with useSupplement fails (glp-expand-pinned-other-version-required). When two versions of the same supplement are in scope, only one applies: the version asked for by version (useSupplement url|version), and used-supplement reports that version (expand-supplement-version)
  • Filters on supplement properties: a value set filter on a property that a supplement in use defines, not the code system, selects the codes the supplement gives that value, alone or together with filters on the code system's own properties (expand-supplement-prop, expand-supplement-prop-base, validate-supplement-prop-base-code4, validate-supplement-prop-base-code5)
  • Batch validation: new tests of CodeSystem/$batch-validate-code (cs-batch-validate, cs-batch-validate-bad, cs-batch-validate-codings), with a new test operation cs-batch-validate. A top-level url names the code system for every entry, and an entry with no url is validated against the code system its coding names. Batches with no value set belong here, not on ValueSet/$batch-validate-code: the IG Publisher used to send them to ValueSet/$batch-validate-code with no url, which servers rightly rejected, and now sends them to CodeSystem/$batch-validate-code
  • Extension-codes supplements: new extension-codes suite (mode general, 20 tests) - a code system with two reserved code spaces (one with a default-display) (reserved-code-space), and a global and a non-global extension-codes supplement that define codes in it - testing $lookup, $validate-code (displays, wrong displays, a code nothing defines, a code outside the code space) and $expand (listed codes, and the whole code system). The same for ISO 3166's user-assigned codes in the extension-codes-3166 suite (mode tx.fhir.org, 6 tests), for ICF's qualified (post-coordinated) codes in the icf-ec suite (mode icf, 13 tests), and for SNOMED CT post-coordinated expressions in the sct-ec suite (mode snomed, 13 tests) - suites of their own, since their global supplements would change the icf and snomed suites' other tests. The requirements now describe reserved code spaces and extension-codes supplements
  • Supplements are used once: a global supplement that is also referenced (useSupplement) gives each display and designation once, and is one used-supplement (ec-validate-global-referenced-wrong-display, ec-lookup-global-referenced in extension-codes); with two supplements referenced, the child and grandchild concepts of a code system with a hierarchy get each supplement's designation once (lookup-supplements-hierarchy-child, lookup-supplements-hierarchy-grandchild, validate-supplements-grandchild-wrong-display in extensions)
  • Reserved code spaces: an expansion of all of a code system that has reserved code spaces (codesystem-reservedCodeSpace) is marked unclosed, since $validate-code accepts codes in those spaces that the expansion can't list (ec-expand-all). In a value set, a reserved code is a member of an include of all of the code system (ec-vs-validate-all-undefined) and of one that lists it (ec-vs-validate-listed-undefined), but a filter never selects one, even a concept in filter that lists it: reserved codes have no hierarchy or properties (ec-vs-expand-* and ec-vs-validate-*-undefined for concept in, concept not-in, code not-in, is-not-a and code regex); 13 new tests in extension-codes
  • Supplements with no language: a supplement that has no CodeSystem.language gives displays (its concepts' display, and designations with no language) that are correct in any language, while its designations with a language are only displays in that language (validate-supplement-nolang-display, validate-supplement-nolang-designation, validate-supplement-nolang-other-lang in extensions). Which display is returned for the requested language - the supplement's or the code system's own - is not fixed by these tests
  • not-in filters: concept not-in and code not-in exclude only the codes they list - a child of a listed code is still in the value set (simple-expand-concept-not-in, simple-expand-code-not-in, validation-simple-concept-not-in-listed, validation-simple-concept-not-in-child-of-listed, validation-simple-code-not-in-listed, validation-simple-code-not-in-other)
  • regex filters match the whole value - the code, the property value or the display - for every code system: the pattern is ^(?:pattern)$, so code2|code3 selects code2 and code3, and not the codes that start with code2 (simple-expand-regex-alt)
  • The designation parameter is a token: a bare code (no system) matches a designation whose language is that code, or whose use has that code in any system, and designation=es is the same as urn:ietf:bcp:47|es. A language matches exactly: de-CH does not select de designations (language-echo-en-designation-bare)
  • Text filters ($expand filter) search one language: the first language the request asks for that the code system has text in, else the code system's own language. All the displays and designations in that language are searched, including those from supplements in use; designations in other languages are not (search-all-de-yes, search-all-de-no-language, search-enum-de-yes, search-filter-de-yes, search-all-supplement-fr-yes, search-all-supplement-fr-no-language)
  • Properties in $expand: properties asked for with the property parameter, or by ValueSet.compose.property, are returned on each code in the expansion and declared in expansion.property - the same values $lookup returns - for SNOMED CT, LOINC, OMOP and UCUM (snomed-expand-property, loinc-expand-property, loinc-expand-compose-property, omop-expand-property, expand-ucum-property)
  • Stored expansions: a value set imported by another that carries an expansion made from a different version of the code system than the one the request uses is expanded from its compose, not from the stale expansion (vs-expand-stale-import, and vs-expand-stale-import-force with force-system-version)
  • Modifier extensions: a value set filter with a modifier extension the server doesn't know can't be used, and that is an error in the request (4xx), not the server (filter-modifier-expand)
  • inferSystem with a value set that includes all of a code system the server does not know - directly or through an imported value set - is an error naming the unknown system, since the code might be in that code system (infer-unknown-system, infer-unknown-system-import)
  • activeOnly: a code is validated against the value set as expanded with activeOnly, so an inactive code is not in the value set whether the value set lists it or selects it with a filter (inactive-enum-activeonly-validate, inactive-filter-activeonly-validate); without activeOnly, an inactive code a value set lists is in it, with an inactive warning (inactive-enum-validate)
Terminology Specific Testing
SNOMED CT (snomed and sct-ecl suites)
  • The test edition's extension module (TEST1000003) has new content: Dutch descriptions and a Dutch language reference set, a simple map and an extended map reference set, decimal, string and boolean concrete values, and inactive relationship rows. Servers need to reload the test edition, and some counts change (e.g. snomed-expand-ecl-concept-filter-inactive)
  • Inactive relationships are ignored, in property filters, ECL refinements, is-a, $lookup and $subsumes (snomed-expand-property-inactive-rel, snomed-lookup-inactive-rel, snomed-lookup-inactive-isa-child, snomed-expand-isa-inactive-isa, subsumes-inactive-isa, snomed-expand-ecl-refinement-inactive-rel, snomed-expand-ecl-descendants-inactive-isa)
  • $lookup returns the properties the SNOMED CT on FHIR page defines. property=* returns all of them - parent, child, the defining relationships, effectiveTime, inactive, moduleId, sufficientlyDefined and semanticTag - except the normal forms, which are only returned when asked for by name (snomed-lookup-all-properties, snomed-lookup-normal-form). Concrete values are properties named by the attribute's concept id, typed by the attribute's range in the MRCM: valueDecimal for dec(), valueInteger for int(), and valueString and valueBoolean (snomed-lookup-concrete)
  • Implicit value sets: the ECL in ?fhir_vs=ecl/ is URL-encoded (snomed-expand-implicit-ecl); a URL that names an edition uses the latest version of that edition, and one that names a version uses that version (snomed-expand-implicit-edition, snomed-expand-implicit-version); the generic form uses the edition and version the request asks for with system-version and the like (snomed-expand-implicit-system-version); and ?fhir_vs=refset/ can name a map reference set (snomed-validate-implied-refset-map)
  • Filters: concept in takes a list of concepts (snomed-expand-concept-in, snomed-validate-concept-in), and code is the same filter property as concept, since value sets in production use both (snomed-expand-code-in, snomed-validate-code-in); descendent-leaf (snomed-expand-descendent-leaf) and is-not-a (snomed-expand-is-not-a)
  • The display filter with regex, which published value sets (VSAC among them) use: a concept matches when one of its active descriptions in one language matches the whole term, case-sensitively. The language is the first one the request asks for that the edition has, else the edition's default language (snomed-expand-display-regex, snomed-validate-display-regex, snomed-expand-display-regex-nl)
  • Expanding a compose with is-a includes and is-a excludes, including an exclude that removes a whole include (snomed-expand-include-exclude, snomed-expand-include-exclude-multi). No SNOMED test had an exclude, so a server that combines the parts into one ECL expression was never checked for bracketing them
  • ECL: 96 new tests, most of them based on Snow Owl's ECL test suite - hierarchy operators and member-of applied to *, nested memberOf, and/or chains and ambiguous precedence, attribute and group cardinality, reverse and dotted attributes, !=, concept and description filters (including the optional D, wild terms, and filters on each side of and, or and minus), history supplements (+ HISTORY and its profiles), and alternate identifiers (LOINC#code). Also concrete values (integer, decimal, string and boolean, grouped, and type mismatches; snomed-expand-ecl-concrete-*, snomed-validate-ecl-concrete-*), map reference sets as memberOf and in member filters (snomed-expand-ecl-memberOf-*-map, snomed-expand-ecl-member-filter-map-target), and description filters in Dutch (snomed-expand-ecl-description-filter-wild-*)
LOINC (loinc suite)
  • The LOINC tests have moved out of the tx.fhir.org suite into their own loinc suite and mode, and their files from tests/tx.fhir.org to tests/loinc
  • Property values and filters (see Requirements): the six axes take the Part code, and CLASS the class name. $lookup now returns CLASS as a string (CHEM), optionally with the Part code as partCode. New tests filter SCALE_TYP and METHOD_TYP by Part code (loinc-expand-scale-type-part, loinc-validate-filter-scale-type-part-*, loinc-expand-prop-method-part), and CLASS = PULM with SCALE_TYP = LP7753-9 (loinc-expand-class-scale, loinc-validate-filter-class-scale-*, from the MII lung function value set). The legacy forms - an axis filtered by name (DOC, Glucose), CLASS filtered by Part code (loinc-expand-class-part-scale, loinc-validate-filter-class-part-scale-good) - stay, and all servers are required to support them. loinc-validate-filter-scale-type-bad used the wrong value set, and loinc-expand-prop-method is now documented as deliberately filtering on the wrong axis
  • Every request now names LOINC version 2.83: version in $lookup, $subsumes and $validate-code on the code system, systemVersion in $validate-code on a value set, system-version in $expand (echoed in expansion.parameter), and the version of the source Coding in $translate (echoed in sourceConcept)
  • Large expansions: which codes a page of an expansion holds is up to the server, so the tests that check a page are now only for tx.fhir.org, and every one has a -total variant (count=0) that checks the size of the expansion
  • Hierarchy and implicit value sets: servers have to support the ancestor filter on the multi-axial hierarchy, and LOINC's implicit value sets http://loinc.org/vs/[partcode] (loinc-expand-implicit-part); a part's parents are each listed once, even when it is under the same parent on two paths (loinc-lookup-part-parent)
  • Languages: the LinguisticVariant translations of a term's axes are designations of the Parts (loinc-lookup-part-languages, loinc-lookup-code-languages), and text filters search the names in the language asked for (loinc-expand-search-de, loinc-expand-search-no-language, loinc-expand-search-filter-de)
  • CLASSTYPE 3 (Claims attachment) is reported like the other class types (loinc-lookup-classtype)

[1.10.0] - 2026-10-25

This is a big release that covers a lot of decision making and development associated with the cluster of terminology meetings in Sydney in October 2026.

Testing Infrastructure
  • Features: a suite or test can now depend on what the server declares in its CapabilityStatement, with feature - the operations it implements (expand, validate, lookup, subsumes, translate, closure, related), and tx-resource for the CodeSystemAsParameter feature. See Features. This needs a test runner that knows about features
  • Messages: about 440 expected responses now give message text as $external, so that a server can provide its own wording in an externals file instead of matching tx.fhir.org's (see Test Cases). An externals file for tx.health-samurai.io has been added
New Functionality
  • $closure: a new closure suite (11 tests) builds a closure table one concept at a time - initialising it, adding codes, resynchronising from a version, and disposing of it (closure-dispose). The tests depend on each other, so the suite runs in order, on one thread, after the other suites. It needs the closure feature: a server runs it when it declares $closure, whatever modes the tests are run with. R4 servers can pass it, with the R4 equivalence codes
Corrections
  • $lookup: name may be either the CodeSystem name or its title. When a code is not selectable, $lookup always returns both abstract (true) and the notSelectable property (true), whether or not they were asked for; when it is selectable, it returns neither unless asked. A displayLanguage that the code system has no content in (zxx) gets the default display and a warning (simple-lookup-zxx, lookup-mms-zxx)
  • Contained value sets: a #id reference is resolved against the contained resources of the containing resource, so a contained value set can import another value set contained in the same resource (simple-expand-contained-nested, validation-contained-nested); contained value sets that import each other are a circular reference (simple-expand-contained-circular). The requirements now say explicitly that value sets contained in value sets are the only use of contained resources servers are required to support
  • Excluding a whole value set - an exclude with only a valueSet - had no test of its own (exclude-combo excludes a value set together with a code system and concepts); exclude-valueset excludes one by url, and exclude-contained a contained one by #id. An excluded value set is reported as a used-valueset, as an imported one is
  • $translate: the server checks each target concept it returns against its code system. A target that is not valid is still returned as the map states it, with an error parameter saying that the target code is not valid (translate-2b, translate-5, translate-5b, translate-ciel4); a target from a code system the server does not know carries a note. A target that is not in its code system's normal form also gets a normalized-concept, and a source code is matched both as given and in its normal form; the map's own codes are not normalised
  • Global language packs: the codesystem-supplement-type code lang-pack is renamed global-lang-pack, and servers have to support global language packs. A new global-lang-pack suite (mode general) checks that a global language pack passed in tx-resource is used for the code system it supplements without being referenced - in $validate-code, $lookup and $expand, where it is reported as a used-supplement - and only for that code system; and that any other supplement passed in tx-resource is only used once it is referenced. Until now this was only tested in the tx.fhir.org mode
  • Unknown properties: a property the server does not know, asked for with the property parameter (x-unknown-property), is ignored - not an error or a warning - by $lookup (simple-lookup-unknown-property, simple-lookup-only-unknown-property) and $expand (simple-expand-unknown-property, where it is not declared in expansion.property). The first test also checks that a property asked for by its own code is returned
  • Inactive codes: in a CodeableConcept, the inactive warning names the inactive coding, not the first one, and a later active coding does not suppress it (inactive-4-validate, inactive-5-validate)
  • Requirements: $validate-code supports valueset-membership-only (this replaces the mode/valueSetMode requirement); a value set that imports itself, directly or through other value sets, is an error, but one that is imported more than once by different routes is not
  • Smaller changes: a flat-mode expected response for vs-expand-versionless, a fix to the R4 expected response for omop-translate-loinc-implicit, and corrected messages in validation-simple-code-implied-bad-code. In five of the 1.9.5 missing-code tests (validation-missing-*), $optional$ was the string "true", which is read as a mode name and so made the content required for every server: the operationoutcome-message-id extension is now optional for every server except tx.fhir.org, as elsewhere, and the echoed code in validation-missing-cs-coding-no-system is optional
Terminology Specific Testing
ICD-11 (icd-11 suite)

82 new tests, and the code form has changed: in the MMS and ICF the code is the bare entity id, and the full entity URI is no longer a code (lookup-mms-uri and cs-validate-uri now expect it not to be found; lookup-mms-id and cs-validate-id use the id). New tests cover:

  • the classificationCode extension on a Coding, in $validate-code against the code system and value sets (cs-validate-coding-cc-*, vs-validate-coding-cc-*)
  • the normal form of a postcoordinated expression (cs-validate-nf-*): qualifiers in the order of the root's axes, several values on one axis sorted by entity id, cluster members ordered by axis, duplicates removed, and normalized-code in the id form
  • $lookup of postcoordinated expressions (clusters, injuries, retinopathy, pneumonia), with and without metadata, and of chapters, windows, residual categories, scales, sources and languages
  • $expand totals for MMS, ICF, the Foundation, the Extensions and the Medicaments, ad hoc value sets in the id form, and $subsumes
  • $translate with ConceptMaps to and from CIEL that use codes and classificationCode expressions (translate-ciel*)
  • abstract is required in $lookup of a concept that is not selectable, and the expansion timestamp is optional
SNOMED CT (snomed suite)
  • the test edition now includes a fragment of the LOINC Extension (module LO1010000), and new tests cover it: $lookup returns the LOINC code as a property, by default too, and a concept with no LOINC code has none (snomed-lookup-loinc*); $expand of LOINC Extension concepts, with and without a LOINC code (snomed-expand-loinc*); and $translate between SNOMED CT and LOINC in both directions (snomed-translate-loinc*)
  • translating from a national edition to the International Edition, with no ConceptMap (snomed-translate-edition-*), using extension concepts added to the test edition (module TEST1000003): an extension concept equivalent to an International concept, a defined concept that becomes a post-coordinated expression (grouped and ungrouped), a primitive concept that translates to its nearest International ancestor, an inactive concept with and without a replacement, and concepts with no useful translation (directly under a top-level hierarchy, or with no International ancestor at all). Where a post-coordinated expression is produced, it uses only International Edition concepts
  • the concepts added to the test edition change the counts in snomed-expand-count-all and the ECL wildcard expansions, so servers need to reload the test edition
  • ECL (sct-ecl suite): 34 new tests - member filters (including dates, and a misplaced filter), top and bottom of set, operators applied to sets and to memberOf, or and and in refinements and inside groups, attribute names given as expressions, sets and wildcards, dotted attributes, comments, concept filters (definitionStatus, effectiveTime, active), description filters (term, term sets and wildcards, typeId, language, active and its default, and two filter blocks), and memberOf with descriptions and field selection
  • an attribute the MRCM marks as grouped, but written outside a relationship group (the Close To User Form of the SNOMED postcoordination guide), is no longer an error. The server applies the "Adding a Self-grouped Attribute" transformation, so 40468003:363698007=10200004 is equivalent to 40468003:{363698007=10200004} for validation, rendering and subsumption (subsumes-expr-grouped-ungrouped and subsumes-expr-ungrouped-morphology now expect a result, not an error). The requirements page also records the scope of post-coordination support agreed by the SNOMED Computable Languages Group; the parts that are not yet tested are marked
ICF (Pre ICD-11) (mode icf)

A new suite and mode icf that tests the ICF qualifier grammar - . then one qualifier digit per position, where the positions depend on the component: extent for body functions (b); extent, nature and localization for body structures (s); performance, capacity, and optionally capacity with assistance and performance without assistance, for activities and participation (d); and one for environmental factors (e), a barrier after . or a facilitator after + in place of the . (no +9). 8 (not specified) fills a position and is left out of the generated display.

  • $validate-code of the
  • The tests cover bad qualifiers, too many qualifiers, headings and the generated displays
  • $expand of an enumerated value set of qualified codes
  • and value sets that control qualifiers with filters: each qualifier exists true/false (a qualifier of 8 does not count as present) and qualified = false (no qualified codes at all)
LOINC (tx.fhir.org suite)
  • Updated to 2.83, and the expected responses updated to match
  • is-a includes the concept itself, as FHIR defines it, so loinc-expand-concept-is-a now includes LP14082-9 (431 codes, not 430). The LOINC hierarchy is read from ComponentHierarchyBySystem path by path: a code is a descendant of the parts on its own PATH_TO_ROOT, and the children a part has under one parent are not inherited by its other parents
  • a part's mappings to other code systems (PartRelatedCodeMapping) are Coding properties named for the map - sct, ncbi, chebi - with the equivalence as a subproperty, and are returned by default (loinc-lookup-part-mappings); an answer's external code from the CDC answer list is reported in the CDC race and ethnicity system as the property race (loinc-lookup-answer-race)
  • a filter for terms with no method - METHOD_TYP exists false - in $expand and $validate-code (loinc-expand-filter-no-method, loinc-validate-filter-no-method-*)
  • $translate between LOINC and other code systems using the maps that LOINC publishes: parts to and from SNOMED CT (loinc-translate-part-sct, -sct-url, loinc-translate-sct-part), and codes to ISO/IEEE 11073 MDC (loinc-translate-code-mdc, where the MDC codes are narrower than the LOINC code)

[1.9.6] - 2026-09-29

This release mostly corrects expected responses that were wrong, but were not being checked properly.

  • Test runner: when two strings did not match, the runner fell back to comparing them as base64, and the base64 decoder skips any character outside the base64 alphabet and stops at the first =. So any string containing an = - most SNOMED CT expression messages, and anything naming an implicit value set (?fhir_vs=) - was only compared up to its first =. The runner now only compares strings as base64 when they really are base64 (this is fixed in the test runner, not the test cases; runners up to 6.10.5 have the bug). Correcting it exposed the wrong expected responses listed below
  • SNOMED CT: an unknown code message names the edition, and the test edition is the Test Edition, not the International Edition (validate-code-procedure-pc-bad1 and -bad2, validate-code-pc-unbracketed, and the validate-code-pc-mrcm-* tests)
  • SNOMED CT: validate-code-implied-2 named the wrong value set in its message - it is the implicit value set that was asked for (http://snomed.info/sct?fhir_vs=refset/734139008), not sct-pc-none; and the display of 80891009 in bugs/validate-sct-parse-pc is Heart structure
  • Smaller message corrections that had been hidden the same way: the display-comment messages in the inactive display tests (validate-code-inactive-display-*, both SNOMED CT and extensions), the list of matching systems in errors-combination-bad, the quoted code system URL in the unknown code system messages (errors-unknown-system2, validation-simple-coding-bad-system), and a stray double space in the related simple-all message
  • Other expected content corrected: the ECL expansion responses had a stray } at the end of the value set URL; the overload tests expect the display Display #2; the OMOP used-conceptmap has no trailing |; and the LOINC supplement in the tx.fhir.org $lookup tests is version 2.82
  • ValueSet $validate-code: the code system version is systemVersion, not version (which is the CodeSystem/$validate-code parameter, and is not defined for ValueSet/$validate-code). Seven SNOMED CT and ICD-11 requests used version, which servers were right to ignore; they now use systemVersion, and validate-code-pc-none now states it too
  • validate-code-pc-none: when the code is not in the value set, the version reported is the one the value set selects (compose.include.version, here the test edition), not the server's default edition
  • $translate: result can only be true if at least one match is an actual translation. A match that says the concept is not related (not-related-to, or disjoint / unmatched in R4), or that there is no mapping (noMap), is still returned, but result is false (translate-2b, translate-4)
  • Messages that legitimately vary between servers are now checked for their content rather than their wording: the expression error in validate-code-procedure-pc-bad1 and validate-code-pc-unbracketed must mention expression (and the code at fault), and the not-in-value-set message in validate-code-pc-none must name the value set and the expression
  • displayLanguage is written without spaces (en,*;q=0, not en, *; q=0), in the requests and in the expansion parameter a server echoes back (language-xform-*-hard, validation-simple-coding-bad-language). The Accept-Language header in validation-simple-coding-bad-language-header keeps its spaces, which are legal in a header

[1.9.5] - 2026-09-26

  • $validate-code with something missing: eighteen new tests in the validation suite cover a request with no code at all (a code with no system and no inferSystem, a system with no code, and nothing at all — all request errors), and a Coding or CodeableConcept that is missing its code or its system, against both a value set and a code system. A Coding or CodeableConcept that was supplied but has bad content is a validation outcome (result = false), not a request error, and a missing code is reported against the coding itself, as a warning with issue type invalid and a tx-issue-type of invalid-data (the data supplied was invalid; there is no code for invalid-code to be about), with no companion "not in the value set" issue for a code that was never provided
  • A Coding with neither a system nor a code - a display on its own, say - is one problem, not two. The server reports a single warning (Coding_has_no_system_or_code__cannot_validate) saying that neither was provided, rather than one message about the missing system and another about the missing code: there is nothing there to validate, and saying so twice does not make it clearer
  • A missing Coding.system or Coding.code is a warning, not an error: a code with no system has no defined meaning and cannot be validated, but only the caller knows whether that is acceptable where the coding came from — a validator has no way to tell, and has to report it as a warning either way, so the server reports it as one. The code still does not validate, so result is false. Two consequences: a result = false response does not always carry an error-level issue, and a CodeableConcept with a systemless coding alongside a coding that is in the value set still validates (validation-missing-vs-cc-mixed). inferSystem is defined only for the code parameter - it does not apply to a Coding, nor to the codings of a CodeableConcept - and two tests pin that: with it set, a systemless coding is reported exactly as it is without it
  • The uuid parameter has been removed from the profiles the tests send (parameters-default.json, the two version profiles, and the CDA profile). Nothing ever read it: it was sent on every request, but no server or client uses it, and it is not part of any operation's definition
  • $subsumes with no code system (simple-subsumes-no-system, simple-subsumes-no-system-coding): codeA and codeB with no system parameter, or a codingA and codingB with no system, is a problem with the request, not a code system that cannot be found - 400, with issue type invalid, a tx-issue-type of invalid-data, and the message id SUBSUMPTION_NO_SYSTEM
  • SNOMED CT: an attribute value that is itself refined must be bracketed. The compositional grammar allows only a bare concept reference as an attribute value; anything more is a bracketed sub-expression, so the correct form is 367430006:{405813007=(85562004:272741003=24028007)}. Without the brackets the expression is ambiguous as well as invalid - in A:{B=C:D=E,F=G} nothing says whether F=G refines C or A. lookup-pc, validate-code-pc-nested-good, validate-code-implied-2 and validate-code-implied-2b used the unbracketed form, and have been corrected (including the expected displays); two new tests check that the unbracketed form fails (lookup-pc-unbracketed, validate-code-pc-unbracketed)
  • SNOMED CT: the comma before an attribute group is optional in the compositional grammar, so an attribute set followed directly by a group has to be accepted, and normalises to the form with the comma (validate-code-pc-scg-no-comma, validate-code-pc-scg-comma)
  • SNOMED CT: an attribute with more than one domain in the concept model is valid in any of them, and not outside all of them (validate-code-pc-mrcm-domain-finding, -event and -multi)
  • Mime types: subsumption where a parameter has a default value. Bare text/plain means charset=us-ascii (RFC 6657) and format=fixed (RFC 3676), so writing the default out gives the same media type, and a different value contradicts the bare form rather than narrowing it - this is where the rule that a parameter narrows a media type does not hold. For a text/* type whose registration does not say how its charset is determined, a server cannot tell whether a missing charset means any charset or US-ASCII, so the relationship cannot be determined; and an unknown parameter carried identically by both codes does not stop the server deciding
  • $translate: ConceptMap.sourceScope limits the scope of a map, so a code outside it has no applicable map, and the map's unmapped does not apply either (translate-6a, which has been corrected); naming the map explicitly does not put the code into its scope (translate-6b)
  • Draft content: a reference to a draft code system from an active value set is now reported as MSG_DRAFT_SRC_STATUS, and the message names the value set as well as the code system
  • Smaller changes to expected responses: the SNOMED CT $lookup name is the edition name (SNOMED CT Test Edition), not system|version; validate-sct-display-2 expects the US English display now that refset language handling is fixed; the ICD-11 cs-validate-uri issue is CODE_NOT_IN_NORMAL_FORM rather than CODE_CASE_DIFFERENCE; the capability statement must declare $subsumes; and bugs/validate-no-system and validation-simple-coding-no-system follow the missing-system rules above

[1.9.4]

Not published as a separate IG release; these changes were first published in 1.9.5.

This release is mostly about $subsumes, which previously had almost no tests, and adds suites for mime types, ICD-11 and ValueSet.compose.property.

  • $subsumes against a CodeSystem resource: the simple-subsumes-* tests cover parent, child, equivalent, transitive, siblings, unrelated roots and an unknown code, each with both codeA/codeB and codingA/codingB. The tho suite does the same for act-class, where the hierarchy is stated in subsumedBy properties rather than by nesting, including a code with two parents - a relationship nesting cannot express at all - and two parents of the same code, which are not thereby related to each other
  • Filters on a hierarchy stated by properties rather than nesting: is-a, descendent-of, is-not-a and child-of against act-class, and child-of in the simple code system (validation-simple-child-of-child, -grandchild). descendent-of excludes the code named in the filter, which is what distinguishes it from is-a; child-of excludes it too, and stops at the direct children, which is what distinguishes it from descendent-of
  • SNOMED CT $subsumes: simple codes (including an unknown code, a malformed code, an unknown edition and an unknown code system), and 17 tests for post-coordinated expressions - refined expressions against ancestors and their own focus concept, attribute values, extra and different attributes, a primitive focus concept, a concept against its own normal form, grouping as the MRCM requires it, terms, conjunctions, and expressions against precoordinated concepts
  • $subsumes can fail to decide. It has no outcome for "unknown", so where a server cannot determine the relationship it returns an error with the tx-issue-type cannot-determine rather than claiming not-subsumed. This applies to SNOMED CT expressions where the structure finds no relationship (which is not a proof that none exists - a server that can classify may answer not-subsumed instead), and to mime types with a parameter the server does not know
  • $subsumes for LOINC (in the tx.fhir.org suite) and UCUM. UCUM has no hierarchy: the same unit written two ways is equivalent, but a scaled unit, or a unit with the same canonical unit and a different magnitude, is not subsumed
  • Language codes (BCP 47): validation against the subtag registry (private-use region ranges, grandfathered tags, and the Prefix rules for extlang and variant subtags); case - a tag in the wrong case is valid, since BCP 47 is case-insensitive, but the server returns normalized-code; subsumption, following RFC 4647 extended filtering; language, region and script filters, where an absent component does not match a fixed one; and expansion of those filters, which is paged for a language, too costly for a region, and not possible for a script alone
  • Mime types: a new mimetypes suite (mode mimetypes). Type and subtype have no hierarchy, but parameters narrow a media type, and a structured suffix is tested too. The base filter selects by type or type/subtype and the registered filter by presence in the IANA registry; only registered = true can be expanded, and that expansion is marked unclosed
  • ICD-11: the icd-11 suite (mode icd-11) now has tests for $lookup (MMS short codes and entity URIs, groupers, residual categories, the Foundation, ICF, languages and postcoordination), $validate-code against the code system and against value sets, and $expand of the WHO postcoordination scale value sets and of client-supplied value sets. Several are expected to fail against the current WHO ICD-API; tests/icd-11/doco.txt says what each one asserts and why
  • ValueSet.compose.property: a new properties suite that expands value sets asking for all properties, with a wildcard and by enumerating them
  • validation-simple-codeableconcept-unknown-system: a CodeableConcept with a coding from an unknown code system alongside a coding that is in the value set. The unknown system cannot be checked, but the other coding satisfies membership, so the result is true, with a warning
  • Inactive codes are always reported as inactive, in the LOINC and SNOMED CT $validate-code tests in the tx.fhir.org suite
  • OperationOutcome: diagnostics is no longer asserted in any expected response - it is for server-specific detail - and every issue now carries a tx-issue-type coding (too-costly and not-supported were the ones missing)
  • ConceptMap-novs has been given its own id: it had the same id as ConceptMap-full

[1.9.3] - 2026-09-04

This release is dominated by a rework of the $translate tests, and by settling how servers report inactive concepts.

  • $translate: the source concept may now be named with sourceCode + sourceSystem, with sourceCoding, or with sourceCodeableConcept, and R4 servers are tested with the R4 spellings of the same thing (code + system, coding, codeableConcept, and targetsystem)
  • $translate: tests for the full range of relationship types, including not-related-to, which is returned as a match like any other; and for ConceptMap.group.element.comment and .target.comment, returned as sourceComment and targetComment (element.comment is preadopted from R6 using a cross-version extension)
  • $translate: tests for noMap — stated as element.noMap in R5, and as a target with no code in R4, but reported the same way either way — and for ConceptMap.group.unmapped in all three modes (use-source-code, fixed and other-map), including chains of otherMap maps and the detection of circular references
  • $translate: the response reports originMap, the concept map the chain of maps started from, and a used-conceptmap for every other map that contributed to it (originMap was introduced in 1.9.2; it is now settled as the start of the chain of maps, not the end)
  • $translate: the R4 reverse parameter is tested — an R4 server reverses the source and target parameters and translates forwards, an R5 or later server returns an error — along with the R5+ way of asking the same question, which is to name the target concept as the source. The translate tests are split into two suites so that the unmapped tests see only the concept maps they are about
  • Display validation: validate-code-inactive-display and validation-simple-code-bad-display are each split into a lenient and a not-lenient variant, driven by a new lenient-display property on the test case
  • Test cases can be restricted to particular FHIR versions with a version property (4.0, or !4.0 for "any version but R4"), and $optional$ accepts a version filter (version:4) so that one expected response can cover both R4 and R5 servers where they legitimately differ. Expected content can also be marked $only$ — required in the nominated versions or modes, and prohibited in all the others; this is the counterpart of $optional$, which only ever relaxes a requirement
  • Inactive concepts: a server that knows a concept is inactive must say why, so the status property is no longer optional in an expansion, and the expected responses have been updated to require it. Also, a status of deprecated no longer makes a concept inactive — inactive and retired do (see the requirements page)
  • A filter that matches no codes: an include whose only filter selects nothing expands to nothing, and must not be treated as an include with no filter, which would select every code in the code system (simple-expand-regex-none, validation-simple-code-regex-none)
  • Repeating concept properties: CodeSystem.concept.property is 0.., so a filter selects a concept when *any of its values match — not just the first — and expansion.contains.property reports all of them (simple-expand-repeating-prop, simple-expand-repeating-prop-values)
  • SNOMED CT: the test subontology has been regenerated from the International August 2025 release with effective time 20250909, and now includes the MRCM attribute domain and attribute range reference sets, the lateralizable body structure reference set, and a simple reference set. The relationship module in the subset has been corrected, and the loading and regeneration instructions in tx-source/readme.md have been rewritten
  • SNOMED CT: new postcoordination tests for concept model (MRCM) validation — attribute domain, attribute range, cardinality, grouping and laterality — and for concrete values, both valid and invalid (out of range, a decimal where an integer is required, a concept where a concrete value is required and the reverse); plus an expression whose attribute value is itself refined
  • SNOMED CT: a constraint = * filter returns every concept, including inactive concepts and module concepts, so the ECL wildcard expansion and snomed-expand-count-all now agree (2259 codes)
  • Fix the SNOMED CT test set URL: correct the edition/version identifier used across the SNOMED test cases to the terminology-ecosystem test edition (http://snomed.info/xsct/31000003106/version/20250909), updating the affected requests and expected responses (display names and version URIs) to match, and correct the type of the system-version parameter from string to uri
  • OperationOutcome.issue.location is optional in the expected responses, and the operationoutcome-message-id extension is optional for every server except tx.fhir.org; the optional version OUT parameter is allowed in the six codeableconcept-*-vs1wb version tests
  • Duplicate canonical URLs resolved: simple-all and simple-enumerated were each defined twice with different content. The permutations suite now uses the shared simple/valueset-all.json, and its own enumerated value set has been renamed to valueset-simple-enumerated-codes.json
  • Housekeeping: remove test files that were no longer referenced by any test case (the openEHR set, the explicit-version OMOP translate tests, the value set import tests and others), de-duplicate code-vnn-vsmix-2, remove a stray uuid parameter and stale $optional$ markers, fix JSON typing errors (a boolean written as a string), add the total OUT parameter to the CPT expansion test now that CPT iteration is fixed, and add the missing flat-format expected response for search-expand-all-yes

[1.9.2] - 2026-07-28

This is preparatory for settling the SNOMED CT tests, and then it will be labelled as 2.0.0

  • Extensive SNOMED CT ECL testing: reorganised and greatly expanded the ECL tests — operator grouping and precedence (grouped-or / grouped-and, ambiguous precedence), cardinality and role-group occurrence counting, complex refinements and exclusions, and wildcard expansions; require valueset-unclosed on filter-based ECL expansions; and tests for enumerating grammar-based code systems (including count=0 counts)
  • Add test cases for the new $compare operation (previously named $related)
  • Translate improvements: reverse translation, and originMap (replacing sourceMap)
  • More code system version-control tests: value sets including/excluding different versions of the same code system, version pinning across all SNOMED requests, and version-dependent translate tests
  • Tests for complex exclusions, contained value sets, child-of, and regex filters — including a regex denial-of-service case and improved regex error messages
  • Search $expand: flat result format (response:flat) and support for the total OUT parameter
  • Warnings for status / retired / inactive codes, with fixes to the reported location
  • Display-name consistency, additional language and secondary-display tests, and allow designation.use.display in responses
  • Validate a CodeableConcept using the first matching code rather than the last
  • Content refresh: LOINC 2.82, updated SNOMED test load-set and version URIs, and updated VSAC content
  • Numerous renames, parameter-name and typo fixes for consistency, unique value set ids, and removal of stale parameters (non-standard limit replaced by count, bad cache-id)

[1.9.0] - 2026-02-16

  • Add a set of tests for controlling the version of CodeSystems (ValueSet.compose.include.version, with wildcards, and the expansion parameters system-version etc.)
  • Revisit the tests to ensure that they are correct
  • Remove the deprecated version parameter that is no longer used or supported (after 1 year of grace)
  • Various improvements to error messages and test descriptions

[1.7.7-SNAPSHOT]

Not published as an IG release.

  • Define hierarchyMeaning for the simple CodeSystem

[1.7.6] - 2025-02-26

  • Allow expansion.id & expansion.offset in many tests
  • Rename valueset-version to default-valueset-version
  • tx.fhir.org only: Add supplement test cases and LOINC tests for CLASSTYPE, answers-for, and answer-list

[1.7.5] - 2025-02-07

  • Add test for validation of displayLanguage

[1.7.4]

  • Add message id

[1.7.3]

  • Use systemVersion instead of version

[1.7.2]

  • Add message-id extension to tests

[1.7.1]

  • Remove language weights except where it matters
  • Add x-caused-by-unknown-system when a supplement is not found
  • Remove R4 variants

[1.7.0]

  • Move master tests to tx-ecosystem-ig instead of general test cases

[1.6.6]

  • Fix references to wrong value set in supplement tests

[1.6.2]

  • Add specific tests for the interaction between ValueSet.compose.inactive and the activeOnly parameter
  • Add tests for correctly populated CapabilityStatement and TerminologyCapabilities resources

[1.6.0]

There is no specific history for the terminology test cases prior to version 1.6.0. The only history notes available are mixed in with all the other kinds of tests in the GitHub release notes.