* Start working on subscription processor
* Work on new scheduler
* Test fixes
* Scheduler refactoring
* Fix test failure
* One more test fix
* Updates to scheduler
* More scheduler work
* Tests now all passing
* Ongoing work on export
* Ongoing scheduler work
* Ongoing testing
* Work on export task
* Sync master
* Ongoing work
* Bump xml patch version
* Work on provider
* Work on bulk
* Work on export scheduler
* More test fies
* More test fixes
* Compile fix
* Reduce logging
* Improve logging
* Reuse bulk export jobs
* Export provider
* Improve logging in bulk export
* Work on bulk export service
* One more bugfix
* Ongoing work on Bulk Data
* Add changelog
* Ongoing work on core structures suport
* Ongoing work on this merge
* Ongoing work to clean up structures
* Credit for #1179
* Address compile issues
* Latest changes
* Work on parallel execution
* Work on getting tests passing
* Add H2 to migrator
* Test fix
* Start working on cascading deleets
* WOrk on cascading deletes
* Work on cascading deletes
* Enable cascading deletes
* Cascade deletes almost working
* FIx test
* Test fix
* Make IBundle list interface more flexible
* Make matchResourceLinks visible
* Make matchResourceLinks visible
* Pull in-memory matching functionality out of subscription module into dedicated classes in searchparam module
* Added more accessors to version independent resource classes
* improved forcedid matching (needs to be verified by James)
filter out nulls from HashMapProvider.searchAll() (hopefully no tests depend on this)
* add AnyComposition
* cleanup of Any classes
* create searchparam config so people can use that module without using subscriptions
* create searchparam config so people can use that module without using subscriptions
* fix broken test
* added derivedFrom
* renamed daoRegistry.getResourceDaoIfExists to getResourceDaoOrNull
also added method to add a new dao to the registry and retrieve registered types
* TokenParam.equals
* add error check to searchable hashmap resource provider
* added support for R4
* add successor to measure
* delete conflict hook
start with failing test
* test passes.
hook is called.
* added delete conflict hook
all delete conflict tests pass
* added javadoc
* Performance test
* test bad interceptor case
* typo
* rename criteria matcher
* added support for multiple topics
* fix method name
* add effective period
* Temporarily excluding StressTest from build to try to get Travis under time limit
* change log
* review feedback
* review feedback
* review feedback
* move expunge out
* move to prototype
* turns parameters into fields
* cleanup control flow
* moved transactions outside of loops
* add expungeeverything
* moved daos out
* add expungeThreadCount
* add expungeBatchSize
* added partition runner to run in separate threads
* all done. just need to consolidate test code.
* Moar tests
* consolidated pointcutlatch into hapi-fhir-jpaserver-model
* final cleanup
* update javadoc
* change log
* failing test
* added delete
also @Transactional
* remove unused parameter
* fix compile
* race condition
(cherry picked from commit e1940d2fb20838859a205f2b96c833b0ce9f05eb)
# Conflicts:
# hapi-fhir-jpaserver-base/src/test/java/ca/uhn/fhir/jpa/dao/expunge/PartitionRunnerTest.java
* ja feedback
* ja feedback
* fix bug in websocket subscription (It wasn't destroying the channel when there are no subscribers)
* add support for removing channel. Also synchronize removal (there was a race condition between sync and queue)
* keep deprecated method for backwards compatibility
* make websocket endpoint configurable
* make websocket context path configurable
* make websocket context path configurable
* trying mvn clean test instead of mvn clean install to see if the build goes faster
* that didn't work at all. reverting.
* change log
* subscription loader now only pulls active subscriptions
only allow websocket connections to subscriptions of type websocket
* Added a "flag for deletion" to ActiveSubscription in the SubscriptionRegistry to handle the race condition of a scheduled sync overlapping with a subscription creation. We could have used a package-scoped semaphore or a pre-remove FHIR read, but this seemed like the safest, simplest and most performant way to handle it.
* ActiveSubscriptionCacheTest
* WebsocketConnectionValidatorTest
* fix compile error in jpa example
* might keep this
* might keep this
* undoing start in interface
* fixed bug that only supported resources with ResourceType/id prefixes
* added standalone subscription support for Dstu2. untested.
* failing test
* fixed cannonicalsubscription.equals for email
* final cleanup
* fix bug in ResourceIndexSearchParamQuantity
Move transaction boundary around SubscriptionMatcherInterceptor so it can be overridden in subclass
Lots of CRLF changes from a mvn -P DIST
* FIXME
* Avoid paging links when offset is artificially set very high
* Ongoing fixes
* Adding logging and clean up transaction processing
* Add changelog
* More work on queries
* Keep working on getting tests passing
* More test fixes
* More work on tests
* More test works
* One more test attempt
* Ongoing fixes
* One more change
* FIxes to search algorithm
* Add changelog
* Clean up for review
* Fixed NPE in token matcher when system is not null and value is null
* test opposite (null system not null value)
* added null protection to all matchers
* Address review comments
* Address broken test
* Trying to resolve DB ordering issues
* Starting work on this
* Working on narrative templates
* Work on new narrative system
* Account for some failing tests due to refactoring of narrative mode
* Another test fix
* More test fixes
* One more test fix
* Work on searches
* Address review comments
Squashed commit adding subscription debug logger
* Work on subscription debug logger
* Add subscription logging interceptor
* Remove unneeded variables
* A bit of cleanup
* Tweak the output logs a bit
* Address review comments
* Fixed a bug in standalone subscription subscriber: It was adding REQUESTED subscriptions to the active subscription registry. (Only ACTIVE subscriptions should be added.)
* Start work on this
* Work on interceptors
* Attempt fix
* Avoid environment dependency
* Test fixes
* One more test fix
* One more build tweak
* Lots of cleanup
* A bit more cleanup
* Still more cleanup
* Some test fixes
* Add legacy methods temporarily
* Don't auto-scan interceptor beans
* One more test fix
* rsolve merge conflicts
* Address review comments
SearchParamRegistry and SubscriptionRegistry now both poll retrying connection to the server. This will help in particular in the scenario where the subscriptions are managed in a standalone server that depends on a FHIR Server being available to pull search parameters and subscriptions from.
* Reorganizing packages and dependencies to support standalone subscription running within a CDR container where all hapi modules are on the classpath.
* EXPERIMENTAL: Moved Subscription registry out of interceptor and introduced ISubscriptionLoader that will be either a Database or FhirClient loader.
5 tests fail. Looks like we're getting too many matches--likely because there is now just one list of subscriptions instead of one list per interceptor.
* Created ActiveSubscription and moved cache bits into it
* Compiles. Next step is get app context to load.
* Application context loads. Now fix NullPointer.
* All subscription tests pass
* FIXME cleanup
* jpa-subscription tests
* fixed config so other module tests work
* MAJOR MILESTONE: All hapi-fhir tests pass.
updated READMEs in example projects
* Moved ExecutorQueue stuff out into its own class
* Organize Imports
* FIXME cleanup
* Null check -> Optional
* Add test and supporting code to validate SubscriptionConstants.EXT_SUBSCRIPTION_RESTHOOK_DELIVER_LATEST_VERSION behaviour
* Added SubscriptionCheckingSubscriber test that works without a database
* Moved a few beans to @ComponentScan
* Replaced use of beanFactory with concrete factory classes
* Switched test to use subscribablechannel
* Added SubscriptionLoaderFhirClientTest
* Undid changes that caused SearchParamProviderFhirClientTest to revert to the Database version. It's now calling the FhirClient version again. (oops)
* Confirm that our SubscriptionProviderFhirClient works with a live fhir client
* Organize imports
* Organize imports
* Register interceptors with DaoConfig instead of RestServer.
Also, Rename @VisibleForTesting methods with ForUnitTest
* Ready to go
* organize imports
* add processing queue
* Fixed interface implementation names
* Fix triggering service so it uses new subscriptionmatcherinterceptor
* fixed example
* Renamed "Database" classes to "Dao"
* Tightened up StoppableSubscriber API
* final code review
* processing -> matching naming change
* fix required by CDR
* oops
* Updated changes.xml
* Renamed subscriptioncheckingsubscriber to subscriptionmatchingsubscriber
* Renamed subscriptioncheckingsubscriber to subscriptionmatchingsubscriber
* CDR integration now works
* Changing subscribable channel abstraction layer in preparation for supporting more types of subscribable channels
* Add interface over both types of messages put on queues
* cosmetic change
* added subscription support for r4
* fixed again
* oops
* self code-review
* Reorganizing packages and dependencies to support standalone subscription running within a CDR container where all hapi modules are on the classpath.
Moved Subscription registry out of interceptor and introduced SubscriptionLoader
* Created ActiveSubscription and moved cache bits into it
* Moved ExecutorQueue stuff out into its own class
* Add test and supporting code to validate SubscriptionConstants.EXT_SUBSCRIPTION_RESTHOOK_DELIVER_LATEST_VERSION behaviour
* Added SubscriptionCheckingSubscriber
* Moved a few beans to @ComponentScan
* Replaced use of beanFactory with concrete factory classes
* Switched test to use subscribablechannel
* Added SubscriptionLoaderFhirClientTest
* Confirm that our SubscriptionProviderFhirClient works with a live fhir client
* Register interceptors with DaoConfig instead of RestServer.
Also, Rename @VisibleForTesting methods with ForUnitTest
* Fix triggering service so it uses new subscriptionmatcherinterceptor
* Renamed "Database" classes to "Dao"
* processing -> matching naming change
Ken Stevens
jamesagnew
Sean McIlvenna
Stig Rohde Døssing
Ken Stevens
Ken Stevens
Volker Schmidt