It's August, so it's time for our next quarterly relese: HAPI FHIR 5.2.0 (Codename: Numbat).
Security Notice:
Major New Features:
The JPA SearchBuilder (which turns FHIR searches into SQL statements to be executed by the database) has been completely rewritten to not use Hibernate. This allows for much more efficient SQL to be generated in some cases. For some specific queries on a very large test repository running on Postgresql this new search builder performed 10x faster. Note that this new module is enabled by default in HAPI FHIR 5.2.0 but can be disabled via a DaoConfig setting. It is disabled by default in Smile CDR 2020.11.R01 but will be enabled by default in the next major release.
Support for RDF Turtle encoding has been added, finally bringing native support for the 3rd official FHIR encoding to HAPI FHIR. This support was contributed by Josh Collins and Eric Prud'hommeaux of the company Janeiro Digital. We greatly appreciate the contribution! To see an example of the RDF encoding: hapi.fhir.org/baseR4/Patient?_format=rdf
Terminology Enhancements:
Integration with remote terminology services has been improved so that required bindings to closed valuesets are no longer delegated to the remote terminology server. This improves performance since there is no need for remote services in this case.
The CodeSystem/$validate-code
operation has been implemented for R4+ JPA servers.
The JPA Terminology Server is now version aware, meaning that multiple versions of a single CodeSystem can now be stored in a single FHIR terminology server repository. ValueSet expansion, CodeSystem lookup, and ConceptMap translation are all now fully version aware. Note that implementing support for fully versioned terminology is mostly complete, but some validation operations may still not work. This should be completed by our next major release.
ValueSet expansion with filtering (e.g. using the filter
parameter on the $expand
operation) has now been implemented in such a way that it fully supports filtering on pre-expanded ValueSets, including using offsets and counts. This is a major improvement for people building picker UIs leveraging the $expand operation.
EMPI Improvements:
Identifier matchers have been added, providing native FHIR support for matching on resource identifiers
The $empi-clear
operation performance has been greatly improved
Other Notable Improvements:
A new combined "delete+expunge" mode has been added to the DELETE operation in the JPA server. This mode deletes resources and expunges (physically deletes) them in a single fast operation. Note that with this mode must be enabled, and completely bypasses interceptor hooks notifying registered listeners that data is being deletes and expunged. It is several orders of magnitude faster when deleting large sets of data, and is generally intended for test scenarios.
The Package Server module now supports installing non-conformance resources from packages.
The _typeFilter
parameter has been implemented for the $bulk-export module.
As always, see the changelog for a full list of changes.
Thanks to everyone who contributed to this release!
It's August, so it's time for our next quarterly relese: HAPI FHIR 5.2.0 (Codename: Manticore).
Notable changes in this release include:
An XSS vulnerability has been fixed in the testpage overlay project. This issue affects only the testpage overlay module, but users of this module should upgrade immediately. A CVE number for this issue has been requested and will be updated here when it is assigned.
Support for the new FHIR NPM Package spec has been added. Currently this support is limited to JPA servers, and support should be added to plain servers in the next release. Packages can be imported on startup, either by supplying NPM files locally or by downloading them automatically from an NPM server such as packages.fhir.org. Package contents (the StructureDefinition, CodeSystem, ValueSet, etc. resources in the package) can be installed into the repository, or can be stored in a dedicated set of tables and made available to the validator without actually being installed in the repository.
Support for the Observation/$lastn
operation has been implemented thanks to a partnership with LHNCBC/NIH. This operation uses ElasticSearch to support querying for recent Observations over a set of test codes for one or more patients in a very efficient way.
The FHIR PATCH operation now supports FHIRPatch in addition to the already supported XML and JSON Patch specs. FHIRPatch is a very expressive mechanism for creating patches and can be used to supply very precise patches.
A new operatiion called $diff
has been added. Diff can be used too generate a FHIRPatch diff between two resrouces, or between two versions of the same resource. For example: http://hapi.fhir.org/baseR4/Patient/example/$diff
Several performance problems and occasional failures in the resource expunge operation have been corrected
The memory use for Subscription delivery queues has been reduced
Snapshot generaton now uses a single snapshot generator codebase for generating snapshots across all versions of FHIR. This makes ongoing maintenance much easier and fixes a number of version specific bugs.
The maximum cascade depth for cascading deletes is now configurable.
AuthorizationInterceptor can now fully authorize GraphQL calls, including allowing/blocking individual resources returned by the graph.
GraphQL now supports POST form (thanks to Kholilul Islam!)
The LOINC uploader now supports LOINC 2.68
A new batch job framework has been introduced, leveraging the Spring Batch library. Initial jobs to use this new framework are the Bulk Export and EMPI modules, but eventually all long processes will be adapted to use this new framework.
TThe HAPI FHIR built-in Terminology Server now includes support for validating UCUM (units of measure), BCP-13 (mimetypes), ISO 4217 (currencies), ISO 3166 (countries), and USPS State Codes.
It is now possible to disable referential integrity for delete operations for speciific reference paths.
A regression has been fixed that significantly degraded validation performance in the JPA server for validation of large numbers of resources.
Unit tests have been migrated to JUnit 5. This change has no user visible impacts, but will help us continue to improve ongoing maintenance of our test suites.
As always, see the changelog for a full list of changes.
Thanks to everyone who contributed to this release!
A second point release of the HAPI FHIR 5.0 (Labrador) release cycle has been pushed to Maven Central.
This release corrects only two issues:
A snapshot dependency was accidentally left into the POM for HAPI FHIR 5.0.1. This has been corrected.
The default setting for the new partitioning feature "Add Partition to Search Indexes" was incorrectly set to enabled (true). It has been set to false, which was the intended default for this setting.