fix(ocl): auto-register factories for post-startup sources + expose concept extras in $lookup by italomacedo · Pull Request #266 · HealthIntersections/FHIRsmith

italomacedo · 2026-07-10T14:44:40Z

Two OCL fixes, verified against the current main base (9 OCL test suites / 80 tests passing). Scoped entirely to tx/ocl/ plus tests.

1. Auto-register factories for OCL sources discovered post-startup

New OCL sources created after the server starts were silently falling back to FhirCodeSystemProvider, which reads the raw content field and returns contentMode() = 'not-present', causing ValueSet expansions to fail with "The code system definition has no content".

Root cause: the refresh cycle detected new sources via getCodeSystemChanges() and called addCodeSystem(), but never created an OCLSourceCodeSystemFactory — so getCodeSystemProvider() never found a factory and fell through to the raw CodeSystem.

Fix (entirely within tx/ocl/):

OCLSourceCodeSystemFactory stores the first i18n instance in a static field so discovery-time factory creation can reuse it.
New static createForDiscoveredSource(httpClient, meta) creates and auto-registers a factory for a newly detected source.
#scheduleRefresh now creates factories for entries in changes.added that don't yet have one.
New patchProviderForOCLFactorySync (in shared/patches.cjs) syncs any factory not yet in provider.codeSystemFactories, following the same pattern already used for the SearchWorker and ValueSetExpander patches.

2. Expose concept extras as properties in `$lookup`

Carry OCL concept extras through toConceptContext (non-empty plain objects only) and implement extendLookup on OCLSourceCodeSystemProvider to emit each entry as a FHIR property parameter with typed value[x] (boolean/integer/decimal/string; arrays and objects JSON-stringified). Honors the $lookup property filter per key.

🤖 Generated with Claude Code

…artup New OCL sources created after the server starts were silently falling back to FhirCodeSystemProvider, which reads the raw 'content' field and returns contentMode() = 'not-present', causing ValueSet expansions to fail with "The code system definition has no content". Root cause: the 60-second refresh cycle detected new sources via getCodeSystemChanges() and called addCodeSystem(), but never created an OCLSourceCodeSystemFactory — so getCodeSystemProvider() never found a factory and fell through to the raw CodeSystem. Fix (entirely within tx/ocl/): - OCLSourceCodeSystemFactory stores the first i18n instance in a static field (#sharedI18n) so discovery-time factory creation can reuse it without requiring library.js to pass i18n again - Added static createForDiscoveredSource(httpClient, meta) that creates and auto-registers a factory for a newly detected source; factories self-register in the static factoriesByKey map on construction - #scheduleRefresh now calls createForDiscoveredSource for each entry in changes.added that does not yet have a factory - patchProviderForOCLFactorySync (new patch in shared/patches.cjs) monkey-patches Provider.prototype.updateCodeSystemList to sync any factory in factoriesByKey that is not yet in provider.codeSystemFactories, following the same idiomatic pattern used for SearchWorker and ValueSetExpander patches No files outside tx/ocl/ were modified. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

Carry OCL concept extras through toConceptContext (non-empty plain objects only) and implement extendLookup on OCLSourceCodeSystemProvider to emit each entry as a FHIR property parameter with typed value[x] (boolean/integer/decimal/string; arrays and objects JSON-stringified). Honors the $lookup property filter per key. Provider now extends BaseCSServices for the _hasProp helper and is exported for tests. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

grahamegrieve · 2026-08-26T06:36:27Z

@italomacedo one general issue with this PR: console.log/console.error - please use the module logger instead

italomacedo · 2026-08-26T13:23:58Z

@italomacedo one general issue with this PR: console.log/console.error - please use the module logger instead

Working on it

Replace all console.log/console.error calls across the OCL module with a child logger from library/logger (Logger.getInstance().child({ module })), per review feedback on PR #266. - cs-ocl.cjs, background-queue.cjs, cache-utils.cjs -> module 'ocl' - vs-ocl.cjs -> module 'ocl-vs' (preserves the former [OCL-ValueSet] split) - drop the redundant [OCL]/[OCL-ValueSet] message prefixes (the child logger now tags each line with {module}) - fetchAllPages in #fetchAllPages now receives the child logger with an empty loggerPrefix instead of console - getColdCacheAgeMs: drop the now-unused logPrefix parameter No behavior change beyond log routing; all 80 OCL tests pass. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

italomacedo · 2026-08-26T13:38:03Z

Good catch — done in cd8bb7b. I converted the whole OCL module to the module logger (Logger.getInstance().child({ module })), not just the new line: all 32 console.log/console.error calls across cs-ocl, vs-ocl, background-queue and cache-utils now go through it. Dropped the redundant [OCL]/[OCL-ValueSet] message prefixes since the child logger tags each line with the module. All 80 OCL tests still pass.

grahamegrieve

ok

italomacedo · 2026-07-10T14:44:40Z

Two OCL fixes, verified against the current main base (9 OCL test suites / 80 tests passing). Scoped entirely to tx/ocl/ plus tests.

1. Auto-register factories for OCL sources discovered post-startup

New OCL sources created after the server starts were silently falling back to FhirCodeSystemProvider, which reads the raw content field and returns contentMode() = 'not-present', causing ValueSet expansions to fail with "The code system definition has no content".

Root cause: the refresh cycle detected new sources via getCodeSystemChanges() and called addCodeSystem(), but never created an OCLSourceCodeSystemFactory — so getCodeSystemProvider() never found a factory and fell through to the raw CodeSystem.

Fix (entirely within tx/ocl/):

OCLSourceCodeSystemFactory stores the first i18n instance in a static field so discovery-time factory creation can reuse it.
New static createForDiscoveredSource(httpClient, meta) creates and auto-registers a factory for a newly detected source.
#scheduleRefresh now creates factories for entries in changes.added that don't yet have one.
New patchProviderForOCLFactorySync (in shared/patches.cjs) syncs any factory not yet in provider.codeSystemFactories, following the same pattern already used for the SearchWorker and ValueSetExpander patches.

2. Expose concept extras as properties in `$lookup`

Carry OCL concept extras through toConceptContext (non-empty plain objects only) and implement extendLookup on OCLSourceCodeSystemProvider to emit each entry as a FHIR property parameter with typed value[x] (boolean/integer/decimal/string; arrays and objects JSON-stringified). Honors the $lookup property filter per key.

🤖 Generated with Claude Code

…artup New OCL sources created after the server starts were silently falling back to FhirCodeSystemProvider, which reads the raw 'content' field and returns contentMode() = 'not-present', causing ValueSet expansions to fail with "The code system definition has no content". Root cause: the 60-second refresh cycle detected new sources via getCodeSystemChanges() and called addCodeSystem(), but never created an OCLSourceCodeSystemFactory — so getCodeSystemProvider() never found a factory and fell through to the raw CodeSystem. Fix (entirely within tx/ocl/): - OCLSourceCodeSystemFactory stores the first i18n instance in a static field (#sharedI18n) so discovery-time factory creation can reuse it without requiring library.js to pass i18n again - Added static createForDiscoveredSource(httpClient, meta) that creates and auto-registers a factory for a newly detected source; factories self-register in the static factoriesByKey map on construction - #scheduleRefresh now calls createForDiscoveredSource for each entry in changes.added that does not yet have a factory - patchProviderForOCLFactorySync (new patch in shared/patches.cjs) monkey-patches Provider.prototype.updateCodeSystemList to sync any factory in factoriesByKey that is not yet in provider.codeSystemFactories, following the same idiomatic pattern used for SearchWorker and ValueSetExpander patches No files outside tx/ocl/ were modified. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

Carry OCL concept extras through toConceptContext (non-empty plain objects only) and implement extendLookup on OCLSourceCodeSystemProvider to emit each entry as a FHIR property parameter with typed value[x] (boolean/integer/decimal/string; arrays and objects JSON-stringified). Honors the $lookup property filter per key. Provider now extends BaseCSServices for the _hasProp helper and is exported for tests. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

grahamegrieve · 2026-08-26T06:36:27Z

@italomacedo one general issue with this PR: console.log/console.error - please use the module logger instead

italomacedo · 2026-08-26T13:23:58Z

@italomacedo one general issue with this PR: console.log/console.error - please use the module logger instead

Working on it

Replace all console.log/console.error calls across the OCL module with a child logger from library/logger (Logger.getInstance().child({ module })), per review feedback on PR #266. - cs-ocl.cjs, background-queue.cjs, cache-utils.cjs -> module 'ocl' - vs-ocl.cjs -> module 'ocl-vs' (preserves the former [OCL-ValueSet] split) - drop the redundant [OCL]/[OCL-ValueSet] message prefixes (the child logger now tags each line with {module}) - fetchAllPages in #fetchAllPages now receives the child logger with an empty loggerPrefix instead of console - getColdCacheAgeMs: drop the now-unused logPrefix parameter No behavior change beyond log routing; all 80 OCL tests pass. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

italomacedo · 2026-08-26T13:38:03Z

Good catch — done in cd8bb7b. I converted the whole OCL module to the module logger (Logger.getInstance().child({ module })), not just the new line: all 32 console.log/console.error calls across cs-ocl, vs-ocl, background-queue and cache-utils now go through it. Dropped the redundant [OCL]/[OCL-ValueSet] message prefixes since the child logger tags each line with the module. All 80 OCL tests still pass.

grahamegrieve

ok

italomacedo · 2026-07-10T14:44:40Z

Two OCL fixes, verified against the current main base (9 OCL test suites / 80 tests passing). Scoped entirely to tx/ocl/ plus tests.

1. Auto-register factories for OCL sources discovered post-startup

New OCL sources created after the server starts were silently falling back to FhirCodeSystemProvider, which reads the raw content field and returns contentMode() = 'not-present', causing ValueSet expansions to fail with "The code system definition has no content".

Root cause: the refresh cycle detected new sources via getCodeSystemChanges() and called addCodeSystem(), but never created an OCLSourceCodeSystemFactory — so getCodeSystemProvider() never found a factory and fell through to the raw CodeSystem.

Fix (entirely within tx/ocl/):

OCLSourceCodeSystemFactory stores the first i18n instance in a static field so discovery-time factory creation can reuse it.
New static createForDiscoveredSource(httpClient, meta) creates and auto-registers a factory for a newly detected source.
#scheduleRefresh now creates factories for entries in changes.added that don't yet have one.
New patchProviderForOCLFactorySync (in shared/patches.cjs) syncs any factory not yet in provider.codeSystemFactories, following the same pattern already used for the SearchWorker and ValueSetExpander patches.

2. Expose concept extras as properties in `$lookup`

Carry OCL concept extras through toConceptContext (non-empty plain objects only) and implement extendLookup on OCLSourceCodeSystemProvider to emit each entry as a FHIR property parameter with typed value[x] (boolean/integer/decimal/string; arrays and objects JSON-stringified). Honors the $lookup property filter per key.

🤖 Generated with Claude Code

…artup New OCL sources created after the server starts were silently falling back to FhirCodeSystemProvider, which reads the raw 'content' field and returns contentMode() = 'not-present', causing ValueSet expansions to fail with "The code system definition has no content". Root cause: the 60-second refresh cycle detected new sources via getCodeSystemChanges() and called addCodeSystem(), but never created an OCLSourceCodeSystemFactory — so getCodeSystemProvider() never found a factory and fell through to the raw CodeSystem. Fix (entirely within tx/ocl/): - OCLSourceCodeSystemFactory stores the first i18n instance in a static field (#sharedI18n) so discovery-time factory creation can reuse it without requiring library.js to pass i18n again - Added static createForDiscoveredSource(httpClient, meta) that creates and auto-registers a factory for a newly detected source; factories self-register in the static factoriesByKey map on construction - #scheduleRefresh now calls createForDiscoveredSource for each entry in changes.added that does not yet have a factory - patchProviderForOCLFactorySync (new patch in shared/patches.cjs) monkey-patches Provider.prototype.updateCodeSystemList to sync any factory in factoriesByKey that is not yet in provider.codeSystemFactories, following the same idiomatic pattern used for SearchWorker and ValueSetExpander patches No files outside tx/ocl/ were modified. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

Carry OCL concept extras through toConceptContext (non-empty plain objects only) and implement extendLookup on OCLSourceCodeSystemProvider to emit each entry as a FHIR property parameter with typed value[x] (boolean/integer/decimal/string; arrays and objects JSON-stringified). Honors the $lookup property filter per key. Provider now extends BaseCSServices for the _hasProp helper and is exported for tests. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

grahamegrieve · 2026-08-26T06:36:27Z

@italomacedo one general issue with this PR: console.log/console.error - please use the module logger instead

italomacedo · 2026-08-26T13:23:58Z

@italomacedo one general issue with this PR: console.log/console.error - please use the module logger instead

Working on it

Replace all console.log/console.error calls across the OCL module with a child logger from library/logger (Logger.getInstance().child({ module })), per review feedback on PR #266. - cs-ocl.cjs, background-queue.cjs, cache-utils.cjs -> module 'ocl' - vs-ocl.cjs -> module 'ocl-vs' (preserves the former [OCL-ValueSet] split) - drop the redundant [OCL]/[OCL-ValueSet] message prefixes (the child logger now tags each line with {module}) - fetchAllPages in #fetchAllPages now receives the child logger with an empty loggerPrefix instead of console - getColdCacheAgeMs: drop the now-unused logPrefix parameter No behavior change beyond log routing; all 80 OCL tests pass. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

italomacedo · 2026-08-26T13:38:03Z

Good catch — done in cd8bb7b. I converted the whole OCL module to the module logger (Logger.getInstance().child({ module })), not just the new line: all 32 console.log/console.error calls across cs-ocl, vs-ocl, background-queue and cache-utils now go through it. Dropped the redundant [OCL]/[OCL-ValueSet] message prefixes since the child logger tags each line with the module. All 80 OCL tests still pass.

grahamegrieve

ok

italomacedo · 2026-07-10T14:44:40Z

Two OCL fixes, verified against the current main base (9 OCL test suites / 80 tests passing). Scoped entirely to tx/ocl/ plus tests.

1. Auto-register factories for OCL sources discovered post-startup

New OCL sources created after the server starts were silently falling back to FhirCodeSystemProvider, which reads the raw content field and returns contentMode() = 'not-present', causing ValueSet expansions to fail with "The code system definition has no content".

Root cause: the refresh cycle detected new sources via getCodeSystemChanges() and called addCodeSystem(), but never created an OCLSourceCodeSystemFactory — so getCodeSystemProvider() never found a factory and fell through to the raw CodeSystem.

Fix (entirely within tx/ocl/):

OCLSourceCodeSystemFactory stores the first i18n instance in a static field so discovery-time factory creation can reuse it.
New static createForDiscoveredSource(httpClient, meta) creates and auto-registers a factory for a newly detected source.
#scheduleRefresh now creates factories for entries in changes.added that don't yet have one.
New patchProviderForOCLFactorySync (in shared/patches.cjs) syncs any factory not yet in provider.codeSystemFactories, following the same pattern already used for the SearchWorker and ValueSetExpander patches.

2. Expose concept extras as properties in `$lookup`

Carry OCL concept extras through toConceptContext (non-empty plain objects only) and implement extendLookup on OCLSourceCodeSystemProvider to emit each entry as a FHIR property parameter with typed value[x] (boolean/integer/decimal/string; arrays and objects JSON-stringified). Honors the $lookup property filter per key.

🤖 Generated with Claude Code

…artup New OCL sources created after the server starts were silently falling back to FhirCodeSystemProvider, which reads the raw 'content' field and returns contentMode() = 'not-present', causing ValueSet expansions to fail with "The code system definition has no content". Root cause: the 60-second refresh cycle detected new sources via getCodeSystemChanges() and called addCodeSystem(), but never created an OCLSourceCodeSystemFactory — so getCodeSystemProvider() never found a factory and fell through to the raw CodeSystem. Fix (entirely within tx/ocl/): - OCLSourceCodeSystemFactory stores the first i18n instance in a static field (#sharedI18n) so discovery-time factory creation can reuse it without requiring library.js to pass i18n again - Added static createForDiscoveredSource(httpClient, meta) that creates and auto-registers a factory for a newly detected source; factories self-register in the static factoriesByKey map on construction - #scheduleRefresh now calls createForDiscoveredSource for each entry in changes.added that does not yet have a factory - patchProviderForOCLFactorySync (new patch in shared/patches.cjs) monkey-patches Provider.prototype.updateCodeSystemList to sync any factory in factoriesByKey that is not yet in provider.codeSystemFactories, following the same idiomatic pattern used for SearchWorker and ValueSetExpander patches No files outside tx/ocl/ were modified. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

Carry OCL concept extras through toConceptContext (non-empty plain objects only) and implement extendLookup on OCLSourceCodeSystemProvider to emit each entry as a FHIR property parameter with typed value[x] (boolean/integer/decimal/string; arrays and objects JSON-stringified). Honors the $lookup property filter per key. Provider now extends BaseCSServices for the _hasProp helper and is exported for tests. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

grahamegrieve · 2026-08-26T06:36:27Z

@italomacedo one general issue with this PR: console.log/console.error - please use the module logger instead

italomacedo · 2026-08-26T13:23:58Z

@italomacedo one general issue with this PR: console.log/console.error - please use the module logger instead

Working on it

Replace all console.log/console.error calls across the OCL module with a child logger from library/logger (Logger.getInstance().child({ module })), per review feedback on PR #266. - cs-ocl.cjs, background-queue.cjs, cache-utils.cjs -> module 'ocl' - vs-ocl.cjs -> module 'ocl-vs' (preserves the former [OCL-ValueSet] split) - drop the redundant [OCL]/[OCL-ValueSet] message prefixes (the child logger now tags each line with {module}) - fetchAllPages in #fetchAllPages now receives the child logger with an empty loggerPrefix instead of console - getColdCacheAgeMs: drop the now-unused logPrefix parameter No behavior change beyond log routing; all 80 OCL tests pass. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

italomacedo · 2026-08-26T13:38:03Z

Good catch — done in cd8bb7b. I converted the whole OCL module to the module logger (Logger.getInstance().child({ module })), not just the new line: all 32 console.log/console.error calls across cs-ocl, vs-ocl, background-queue and cache-utils now go through it. Dropped the redundant [OCL]/[OCL-ValueSet] message prefixes since the child logger tags each line with the module. All 80 OCL tests still pass.

grahamegrieve

ok

italomacedo · 2026-07-10T14:44:40Z

Two OCL fixes, verified against the current main base (9 OCL test suites / 80 tests passing). Scoped entirely to tx/ocl/ plus tests.

1. Auto-register factories for OCL sources discovered post-startup

New OCL sources created after the server starts were silently falling back to FhirCodeSystemProvider, which reads the raw content field and returns contentMode() = 'not-present', causing ValueSet expansions to fail with "The code system definition has no content".

Root cause: the refresh cycle detected new sources via getCodeSystemChanges() and called addCodeSystem(), but never created an OCLSourceCodeSystemFactory — so getCodeSystemProvider() never found a factory and fell through to the raw CodeSystem.

Fix (entirely within tx/ocl/):

OCLSourceCodeSystemFactory stores the first i18n instance in a static field so discovery-time factory creation can reuse it.
New static createForDiscoveredSource(httpClient, meta) creates and auto-registers a factory for a newly detected source.
#scheduleRefresh now creates factories for entries in changes.added that don't yet have one.
New patchProviderForOCLFactorySync (in shared/patches.cjs) syncs any factory not yet in provider.codeSystemFactories, following the same pattern already used for the SearchWorker and ValueSetExpander patches.

2. Expose concept extras as properties in `$lookup`

Carry OCL concept extras through toConceptContext (non-empty plain objects only) and implement extendLookup on OCLSourceCodeSystemProvider to emit each entry as a FHIR property parameter with typed value[x] (boolean/integer/decimal/string; arrays and objects JSON-stringified). Honors the $lookup property filter per key.

🤖 Generated with Claude Code

…artup New OCL sources created after the server starts were silently falling back to FhirCodeSystemProvider, which reads the raw 'content' field and returns contentMode() = 'not-present', causing ValueSet expansions to fail with "The code system definition has no content". Root cause: the 60-second refresh cycle detected new sources via getCodeSystemChanges() and called addCodeSystem(), but never created an OCLSourceCodeSystemFactory — so getCodeSystemProvider() never found a factory and fell through to the raw CodeSystem. Fix (entirely within tx/ocl/): - OCLSourceCodeSystemFactory stores the first i18n instance in a static field (#sharedI18n) so discovery-time factory creation can reuse it without requiring library.js to pass i18n again - Added static createForDiscoveredSource(httpClient, meta) that creates and auto-registers a factory for a newly detected source; factories self-register in the static factoriesByKey map on construction - #scheduleRefresh now calls createForDiscoveredSource for each entry in changes.added that does not yet have a factory - patchProviderForOCLFactorySync (new patch in shared/patches.cjs) monkey-patches Provider.prototype.updateCodeSystemList to sync any factory in factoriesByKey that is not yet in provider.codeSystemFactories, following the same idiomatic pattern used for SearchWorker and ValueSetExpander patches No files outside tx/ocl/ were modified. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

Carry OCL concept extras through toConceptContext (non-empty plain objects only) and implement extendLookup on OCLSourceCodeSystemProvider to emit each entry as a FHIR property parameter with typed value[x] (boolean/integer/decimal/string; arrays and objects JSON-stringified). Honors the $lookup property filter per key. Provider now extends BaseCSServices for the _hasProp helper and is exported for tests. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

grahamegrieve · 2026-08-26T06:36:27Z

@italomacedo one general issue with this PR: console.log/console.error - please use the module logger instead

italomacedo · 2026-08-26T13:23:58Z

@italomacedo one general issue with this PR: console.log/console.error - please use the module logger instead

Working on it

Replace all console.log/console.error calls across the OCL module with a child logger from library/logger (Logger.getInstance().child({ module })), per review feedback on PR #266. - cs-ocl.cjs, background-queue.cjs, cache-utils.cjs -> module 'ocl' - vs-ocl.cjs -> module 'ocl-vs' (preserves the former [OCL-ValueSet] split) - drop the redundant [OCL]/[OCL-ValueSet] message prefixes (the child logger now tags each line with {module}) - fetchAllPages in #fetchAllPages now receives the child logger with an empty loggerPrefix instead of console - getColdCacheAgeMs: drop the now-unused logPrefix parameter No behavior change beyond log routing; all 80 OCL tests pass. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

italomacedo · 2026-08-26T13:38:03Z

Good catch — done in cd8bb7b. I converted the whole OCL module to the module logger (Logger.getInstance().child({ module })), not just the new line: all 32 console.log/console.error calls across cs-ocl, vs-ocl, background-queue and cache-utils now go through it. Dropped the redundant [OCL]/[OCL-ValueSet] message prefixes since the child logger tags each line with the module. All 80 OCL tests still pass.

grahamegrieve

ok

italomacedo · 2026-07-10T14:44:40Z

Two OCL fixes, verified against the current main base (9 OCL test suites / 80 tests passing). Scoped entirely to tx/ocl/ plus tests.

1. Auto-register factories for OCL sources discovered post-startup

New OCL sources created after the server starts were silently falling back to FhirCodeSystemProvider, which reads the raw content field and returns contentMode() = 'not-present', causing ValueSet expansions to fail with "The code system definition has no content".

Root cause: the refresh cycle detected new sources via getCodeSystemChanges() and called addCodeSystem(), but never created an OCLSourceCodeSystemFactory — so getCodeSystemProvider() never found a factory and fell through to the raw CodeSystem.

Fix (entirely within tx/ocl/):

OCLSourceCodeSystemFactory stores the first i18n instance in a static field so discovery-time factory creation can reuse it.
New static createForDiscoveredSource(httpClient, meta) creates and auto-registers a factory for a newly detected source.
#scheduleRefresh now creates factories for entries in changes.added that don't yet have one.
New patchProviderForOCLFactorySync (in shared/patches.cjs) syncs any factory not yet in provider.codeSystemFactories, following the same pattern already used for the SearchWorker and ValueSetExpander patches.

2. Expose concept extras as properties in `$lookup`

Carry OCL concept extras through toConceptContext (non-empty plain objects only) and implement extendLookup on OCLSourceCodeSystemProvider to emit each entry as a FHIR property parameter with typed value[x] (boolean/integer/decimal/string; arrays and objects JSON-stringified). Honors the $lookup property filter per key.

🤖 Generated with Claude Code

…artup New OCL sources created after the server starts were silently falling back to FhirCodeSystemProvider, which reads the raw 'content' field and returns contentMode() = 'not-present', causing ValueSet expansions to fail with "The code system definition has no content". Root cause: the 60-second refresh cycle detected new sources via getCodeSystemChanges() and called addCodeSystem(), but never created an OCLSourceCodeSystemFactory — so getCodeSystemProvider() never found a factory and fell through to the raw CodeSystem. Fix (entirely within tx/ocl/): - OCLSourceCodeSystemFactory stores the first i18n instance in a static field (#sharedI18n) so discovery-time factory creation can reuse it without requiring library.js to pass i18n again - Added static createForDiscoveredSource(httpClient, meta) that creates and auto-registers a factory for a newly detected source; factories self-register in the static factoriesByKey map on construction - #scheduleRefresh now calls createForDiscoveredSource for each entry in changes.added that does not yet have a factory - patchProviderForOCLFactorySync (new patch in shared/patches.cjs) monkey-patches Provider.prototype.updateCodeSystemList to sync any factory in factoriesByKey that is not yet in provider.codeSystemFactories, following the same idiomatic pattern used for SearchWorker and ValueSetExpander patches No files outside tx/ocl/ were modified. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

Carry OCL concept extras through toConceptContext (non-empty plain objects only) and implement extendLookup on OCLSourceCodeSystemProvider to emit each entry as a FHIR property parameter with typed value[x] (boolean/integer/decimal/string; arrays and objects JSON-stringified). Honors the $lookup property filter per key. Provider now extends BaseCSServices for the _hasProp helper and is exported for tests. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

grahamegrieve · 2026-08-26T06:36:27Z

@italomacedo one general issue with this PR: console.log/console.error - please use the module logger instead

italomacedo · 2026-08-26T13:23:58Z

@italomacedo one general issue with this PR: console.log/console.error - please use the module logger instead

Working on it

Replace all console.log/console.error calls across the OCL module with a child logger from library/logger (Logger.getInstance().child({ module })), per review feedback on PR #266. - cs-ocl.cjs, background-queue.cjs, cache-utils.cjs -> module 'ocl' - vs-ocl.cjs -> module 'ocl-vs' (preserves the former [OCL-ValueSet] split) - drop the redundant [OCL]/[OCL-ValueSet] message prefixes (the child logger now tags each line with {module}) - fetchAllPages in #fetchAllPages now receives the child logger with an empty loggerPrefix instead of console - getColdCacheAgeMs: drop the now-unused logPrefix parameter No behavior change beyond log routing; all 80 OCL tests pass. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

italomacedo · 2026-08-26T13:38:03Z

Good catch — done in cd8bb7b. I converted the whole OCL module to the module logger (Logger.getInstance().child({ module })), not just the new line: all 32 console.log/console.error calls across cs-ocl, vs-ocl, background-queue and cache-utils now go through it. Dropped the redundant [OCL]/[OCL-ValueSet] message prefixes since the child logger tags each line with the module. All 80 OCL tests still pass.

grahamegrieve

ok

italomacedo · 2026-07-10T14:44:40Z

Two OCL fixes, verified against the current main base (9 OCL test suites / 80 tests passing). Scoped entirely to tx/ocl/ plus tests.

1. Auto-register factories for OCL sources discovered post-startup

New OCL sources created after the server starts were silently falling back to FhirCodeSystemProvider, which reads the raw content field and returns contentMode() = 'not-present', causing ValueSet expansions to fail with "The code system definition has no content".

Root cause: the refresh cycle detected new sources via getCodeSystemChanges() and called addCodeSystem(), but never created an OCLSourceCodeSystemFactory — so getCodeSystemProvider() never found a factory and fell through to the raw CodeSystem.

Fix (entirely within tx/ocl/):

OCLSourceCodeSystemFactory stores the first i18n instance in a static field so discovery-time factory creation can reuse it.
New static createForDiscoveredSource(httpClient, meta) creates and auto-registers a factory for a newly detected source.
#scheduleRefresh now creates factories for entries in changes.added that don't yet have one.
New patchProviderForOCLFactorySync (in shared/patches.cjs) syncs any factory not yet in provider.codeSystemFactories, following the same pattern already used for the SearchWorker and ValueSetExpander patches.

2. Expose concept extras as properties in `$lookup`

Carry OCL concept extras through toConceptContext (non-empty plain objects only) and implement extendLookup on OCLSourceCodeSystemProvider to emit each entry as a FHIR property parameter with typed value[x] (boolean/integer/decimal/string; arrays and objects JSON-stringified). Honors the $lookup property filter per key.

🤖 Generated with Claude Code

…artup New OCL sources created after the server starts were silently falling back to FhirCodeSystemProvider, which reads the raw 'content' field and returns contentMode() = 'not-present', causing ValueSet expansions to fail with "The code system definition has no content". Root cause: the 60-second refresh cycle detected new sources via getCodeSystemChanges() and called addCodeSystem(), but never created an OCLSourceCodeSystemFactory — so getCodeSystemProvider() never found a factory and fell through to the raw CodeSystem. Fix (entirely within tx/ocl/): - OCLSourceCodeSystemFactory stores the first i18n instance in a static field (#sharedI18n) so discovery-time factory creation can reuse it without requiring library.js to pass i18n again - Added static createForDiscoveredSource(httpClient, meta) that creates and auto-registers a factory for a newly detected source; factories self-register in the static factoriesByKey map on construction - #scheduleRefresh now calls createForDiscoveredSource for each entry in changes.added that does not yet have a factory - patchProviderForOCLFactorySync (new patch in shared/patches.cjs) monkey-patches Provider.prototype.updateCodeSystemList to sync any factory in factoriesByKey that is not yet in provider.codeSystemFactories, following the same idiomatic pattern used for SearchWorker and ValueSetExpander patches No files outside tx/ocl/ were modified. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

Carry OCL concept extras through toConceptContext (non-empty plain objects only) and implement extendLookup on OCLSourceCodeSystemProvider to emit each entry as a FHIR property parameter with typed value[x] (boolean/integer/decimal/string; arrays and objects JSON-stringified). Honors the $lookup property filter per key. Provider now extends BaseCSServices for the _hasProp helper and is exported for tests. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

grahamegrieve · 2026-08-26T06:36:27Z

@italomacedo one general issue with this PR: console.log/console.error - please use the module logger instead

italomacedo · 2026-08-26T13:23:58Z

@italomacedo one general issue with this PR: console.log/console.error - please use the module logger instead

Working on it

Replace all console.log/console.error calls across the OCL module with a child logger from library/logger (Logger.getInstance().child({ module })), per review feedback on PR #266. - cs-ocl.cjs, background-queue.cjs, cache-utils.cjs -> module 'ocl' - vs-ocl.cjs -> module 'ocl-vs' (preserves the former [OCL-ValueSet] split) - drop the redundant [OCL]/[OCL-ValueSet] message prefixes (the child logger now tags each line with {module}) - fetchAllPages in #fetchAllPages now receives the child logger with an empty loggerPrefix instead of console - getColdCacheAgeMs: drop the now-unused logPrefix parameter No behavior change beyond log routing; all 80 OCL tests pass. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

italomacedo · 2026-08-26T13:38:03Z

Good catch — done in cd8bb7b. I converted the whole OCL module to the module logger (Logger.getInstance().child({ module })), not just the new line: all 32 console.log/console.error calls across cs-ocl, vs-ocl, background-queue and cache-utils now go through it. Dropped the redundant [OCL]/[OCL-ValueSet] message prefixes since the child logger tags each line with the module. All 80 OCL tests still pass.

grahamegrieve

ok

italomacedo · 2026-07-10T14:44:40Z

Two OCL fixes, verified against the current main base (9 OCL test suites / 80 tests passing). Scoped entirely to tx/ocl/ plus tests.

1. Auto-register factories for OCL sources discovered post-startup

New OCL sources created after the server starts were silently falling back to FhirCodeSystemProvider, which reads the raw content field and returns contentMode() = 'not-present', causing ValueSet expansions to fail with "The code system definition has no content".

Root cause: the refresh cycle detected new sources via getCodeSystemChanges() and called addCodeSystem(), but never created an OCLSourceCodeSystemFactory — so getCodeSystemProvider() never found a factory and fell through to the raw CodeSystem.

Fix (entirely within tx/ocl/):

OCLSourceCodeSystemFactory stores the first i18n instance in a static field so discovery-time factory creation can reuse it.
New static createForDiscoveredSource(httpClient, meta) creates and auto-registers a factory for a newly detected source.
#scheduleRefresh now creates factories for entries in changes.added that don't yet have one.
New patchProviderForOCLFactorySync (in shared/patches.cjs) syncs any factory not yet in provider.codeSystemFactories, following the same pattern already used for the SearchWorker and ValueSetExpander patches.

2. Expose concept extras as properties in `$lookup`

Carry OCL concept extras through toConceptContext (non-empty plain objects only) and implement extendLookup on OCLSourceCodeSystemProvider to emit each entry as a FHIR property parameter with typed value[x] (boolean/integer/decimal/string; arrays and objects JSON-stringified). Honors the $lookup property filter per key.

🤖 Generated with Claude Code

…artup New OCL sources created after the server starts were silently falling back to FhirCodeSystemProvider, which reads the raw 'content' field and returns contentMode() = 'not-present', causing ValueSet expansions to fail with "The code system definition has no content". Root cause: the 60-second refresh cycle detected new sources via getCodeSystemChanges() and called addCodeSystem(), but never created an OCLSourceCodeSystemFactory — so getCodeSystemProvider() never found a factory and fell through to the raw CodeSystem. Fix (entirely within tx/ocl/): - OCLSourceCodeSystemFactory stores the first i18n instance in a static field (#sharedI18n) so discovery-time factory creation can reuse it without requiring library.js to pass i18n again - Added static createForDiscoveredSource(httpClient, meta) that creates and auto-registers a factory for a newly detected source; factories self-register in the static factoriesByKey map on construction - #scheduleRefresh now calls createForDiscoveredSource for each entry in changes.added that does not yet have a factory - patchProviderForOCLFactorySync (new patch in shared/patches.cjs) monkey-patches Provider.prototype.updateCodeSystemList to sync any factory in factoriesByKey that is not yet in provider.codeSystemFactories, following the same idiomatic pattern used for SearchWorker and ValueSetExpander patches No files outside tx/ocl/ were modified. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

Carry OCL concept extras through toConceptContext (non-empty plain objects only) and implement extendLookup on OCLSourceCodeSystemProvider to emit each entry as a FHIR property parameter with typed value[x] (boolean/integer/decimal/string; arrays and objects JSON-stringified). Honors the $lookup property filter per key. Provider now extends BaseCSServices for the _hasProp helper and is exported for tests. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

grahamegrieve · 2026-08-26T06:36:27Z

@italomacedo one general issue with this PR: console.log/console.error - please use the module logger instead

italomacedo · 2026-08-26T13:23:58Z

@italomacedo one general issue with this PR: console.log/console.error - please use the module logger instead

Working on it

Replace all console.log/console.error calls across the OCL module with a child logger from library/logger (Logger.getInstance().child({ module })), per review feedback on PR #266. - cs-ocl.cjs, background-queue.cjs, cache-utils.cjs -> module 'ocl' - vs-ocl.cjs -> module 'ocl-vs' (preserves the former [OCL-ValueSet] split) - drop the redundant [OCL]/[OCL-ValueSet] message prefixes (the child logger now tags each line with {module}) - fetchAllPages in #fetchAllPages now receives the child logger with an empty loggerPrefix instead of console - getColdCacheAgeMs: drop the now-unused logPrefix parameter No behavior change beyond log routing; all 80 OCL tests pass. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

italomacedo · 2026-08-26T13:38:03Z

Good catch — done in cd8bb7b. I converted the whole OCL module to the module logger (Logger.getInstance().child({ module })), not just the new line: all 32 console.log/console.error calls across cs-ocl, vs-ocl, background-queue and cache-utils now go through it. Dropped the redundant [OCL]/[OCL-ValueSet] message prefixes since the child logger tags each line with the module. All 80 OCL tests still pass.

grahamegrieve

ok

italomacedoand others added 2 commits July 10, 2026 11:33

italomacedo requested a review from grahamegrieve July 23, 2026 14:35