A Persistence Framework for Scala and NoSQL
TestDataGenerator and
CustomGeneratorPool by wrapping them in longevity
classes. Slightly simplified the API for adding a custom test data
generators.LongevityContext constructors and apply methods
that take either a LongevityConfig or a Typesafe config.@subdomain in package
longevity.subdomain.annotations.Subdomain constructor and apply method that
collect all the PTypes and CTypes by package scanning.@component,
@derivedComponent, @derivedPersistent, @keyVal, @mprops,
@persistent, @polyComponent, and @polyPersistent in package
longevity.subdomain.annotations.PType inner objects keys and
indexes to look for keys and indexes. Users must now define
PType.keySet, and override PType.indexSet, to declare keys and
indexes. We made this change since the object scanning was
superfluous, and complicates the API. (In contrast, scanning for
properties is useful, as users need to be able to call properties by
name. However, users never really need to call keys and indexes by
name, so there is no advantage to naming them in an inner object.)PType.emptyKeySet.PType.partitionKey methods return Key[P]
instead of PartitionKey[P]. this is for convenience of Scala 2.11
users, so they dont have to declare the full type of their keySet.longevity.subdomain.Key and longevity.subdomain.PartitionKey.
Remove types longevity.subdomain.AnyKey and
longevity.subdomain.AnyPartitionKey, as they no longer serve any
purpose.longevity.subdomain.KeyVal.longevity.subdomain.EType (for "embeddable
type") to longevity.subdomain.CType (for "component type").longevity.subdomain.Embeddable and
longevity.subdomain.Persistent. Users no longer need to extend
their subdomain classes with these empty marker traits.RepoCrudSpec from a FeatureSpec to a
FlatSpec. Users using LongevityContext.repoCrudSpec will notice
significantly less verbose test output.DateTime to UTC time zone. This
seems like the best approach right now, as both Cassandra and
MongoDB back ends support timestamps without time zone
information. It also resolves issues with sorting dates.KeyVal.key and change KeyVal from an
abstract class into a trait. To migrate existing code, you will need
to remove the Key argument supplied to each of your KeyVal types.longevity.subdomain.query. This
should not affect you if you are only using the query DSL.Please refer to the user manual for the latest on the query API.
In brief, we have collapsed packages longevity.subdomain.embeddable and longevity.subdomain.persistent into longevity.subdomain. We have moved a number of classes that were nothing more than DDD synonyms for existing longevity terms into a new side-project called longevity-ddd.
In effect we are cleaning out the namespace and removing some of the clutter of our DDD roots in our continued efforts to develop a general-purpose persistence framework.
The bloody details:
DerivedType to DerivedEType. Rename
PolyType to PolyEType.CoreDomain, GenericSubdomain, and
SupportingSubdomain from package longevity.subdomain to
longevity.ddd.subdomain.Entity, EntityType, ValueObject, and
ValueType from package longevity.subdomain.embeddable to
longevity.ddd.subdomain.Event, Root, and ViewItem from package
longevity.subdomain.persistent to package
longevity.ddd.subdomain.EventType, RootType, and View from package
longevity.subdomain.ptype to package longevity.ddd.subdomain.DerivedEType, EType, ETypePool,
Embeddable, and PolyEType from package
longevity.subdomain.embeddable to longevity.subdomain.Persistent from package
longevity.subdomain.persistent to longevity.subdomain.DerivedPType, PType, PTypePool, and
PolyPType from package longevity.subdomain.ptype to
longevity.subdomain.longevity.ddd.subdomain into a separate
project called longevity-ddd. If you want to continue using the
wrapper classes found there, please add the following dependency to
your project: libraryDependencies += "org.longevityframework" %% "longevity-ddd" % "x.y.z".There was a bug in our use of Scala reflection. In brief, we were using the class loader (i.e., scala reflection Mirror) that was used to load the longevity library. This is bogus, as we are reflecting against user classes! We changed things to reflect on the mirror of the TypeTags (they get wrapped in TypeKeys) that the user library provides to use.
End result is that projects that do funky things with class loaders will not get reflection exceptions when using longevity.
Some odds and ends that have been accumulating in the backlog.
PersistenceStrategy to BackEnd. Move
BackEnd from being a separate argument to LongevityContext
creator methods, to being part of the config, under config property
longevity.backEnd.OPState to go along with FPState and FOPState.LongevityContext.jsonMarshaller and
LongevityContext.jsonUnmarshaller.Some odds and ends that have been accumulating in the backlog. The next release, Jetsam, will be more of the same.
PState.dirty. We are taking this out because
there we may decide to stop keeping track of the original version
of the persistent object, in order to reduce memory usage.RepoPool.createSchema() and configuration flag
autocreateSchema.Repo methods and database calls.RepoPool.closeSession(). This was
added because leaving the Cassandra session open can cause user
programs to fail to terminate under certain circumstances, If your
main program is hanging when using Cassandra, please call this
method at the end of your program.the previous fix in 0.11.3 was bogus. one of those days!
this bug is a NullPointerE in Cassandra create or update methods. It occurred when optimistic locking was turned off. i need to expand my test suite to cover different configs...