Commits on Dec 12, 2013

1. Merge CRDTs inside riak_object merge …

   Why? Well it means no CRDT siblings on disk. Previously CRDTs
   were merged only when a CRDT Operation was performed from the CRDT
   API endpoints. This means read repair, AAE, and repl, could all cause
   siblings on disk.
   
   With the addition of DVV style causality, we need to make a choice
   about CRDTs. We don't have DVVSets and the anonymous list. So CRDTs
   either have to merge Dots into a mini-version-vector, or not have dots.
   
   If CRDTs don't have dots, so their values always survive a merge,
   then they are subject to repl induced sibling-explosion (possibly)
   
   Merging in riak_object:merge solves that issue.
   As an added benefit it also means that Yokozuna indexing of CRDTs
   will just work.
   
   I'd really appreciate a performance comparision between this and
   the plain DVV branch, and the pre-DVV stuff for 2.0.

   

   russelldb committed Dec 12, 2013
   Configuration menu

   • View commit details
   • Copy full SHA for 18fae05
   • Browse repository at this point

   Copy the full SHA

   18fae05 View commit details

   Browse the repository at this point in the history