As per documentation, conflict is handling by updating current winning revision document and deleting conflicted revision
https://developer.couchbase.com/documentation/mobile/current/guides/sync-gateway/resolving-conflicts/index.html
will conflict resolve with deleting conflict branch only instead of updating current winning revision?
Hi - Can you elaborate your question with an example .
Suppose documents has two conflict branches below
"103-8e795f19581d47e4da620ec7988e8e52" and “60-1e795f19581d47e4de620ec7988e8e52”,
Winning revision is: "103-8e795f19581d47e4da620ec7988e8e52"
I have resolved conflict by making _deleted=true to conflicted branch(“60-1e795f19581d47e4de620ec7988e8e52”) only(Not updating current winning revision i.e “103-8e795f19581d47e4da620ec7988e8e52”)
hence we are getting below _sync metadata
{
"_sync": {
“rev”: “103-8e795f19581d47e4da620ec7988e8e52”,
“new_rev”: “61-5384584a19a24f5ea6a65e1c5c7f5053”,
“flags”: 20,
“sequence”: 133961,
“recent_sequences”: [
130029,
130133,
130222,
130515,
Is this valid way of conflict resolution?
Yes, the idea is that you want to tell others about the _deleted:true on that branch. Just make sure all the people who read and write documents in your system follow the some conflict resolution policy deterministically.
PRO TIP - Try to stay away from conflict resolution policy/rules based on timestamps inside documents. In the real world not every body is connected to the same atomic NTP server.
Thanks @househippo, we are identifying conflicts and deleting conflicts using documentation only.
will resolving conflict by only deleting conflicted branch(not updating current winning revision) be replicated by sg-replicator?
Yes, make sure that the rev_limit is not to short. Example the default is 1,000 that handles nearly most people history of changes in a document including _delete:true of a branch.