...
How will it be implemented?
| Code Block |
|---|
def merge(conflictResolver: ConflictResolver[C])(parentHash: Blake2b256Hash, hashA: Blake2b256Hash, hashB: Blake2b256Hash): Option[Blake2b256Hash] |
Conflict Resolution
The current implementation does a mapping:
| Code Block |
|---|
private def buildBlockAncestorChannels[F[_]: Monad: BlockStore]( blockAncestorsMeta: List[BlockMetadata] ): F[Set[ByteString]] = for { maybeAncestors <- blockAncestorsMeta.traverse( blockAncestorMeta => BlockStore[F].get(blockAncestorMeta.blockHash) ) ancestors = maybeAncestors.flatten ancestorEvents = ancestors.flatMap(_.getBody.deploys.flatMap(_.deployLog)) ancestorChannels = ancestorEvents.flatMap { case Event(Produce(produce: ProduceEvent)) => Set(produce.channelsHash) case Event(Consume(consume: ConsumeEvent)) => consume.channelsHashes.toSet case Event(Comm(CommEvent(Some(consume: ConsumeEvent), produces))) => consume.channelsHashes.toSet ++ produces.map(_.channelsHash).toSet case _ => throw new RuntimeException("incorrect ancestor events") }.toSet } yield ancestorChannels |
This throws away all of the additional information carried in Event.
A plan to enable a denser perceived DAG is to introduce more introspection and a broader set of conflict resolution rules.
Two rules that will be implemented soon are:
- peeking values
- invoking persistent continuations
These do not require additional context information and can be resolved at channel level.
IOW the planned changes to conflict resolution should not block the future work on trie merging.
It would make sense to move the mechanism itself to a dedicated interface which will be used in merge.
Open topics
- is it possible to use this mechanism to merge multiple tries at once