sc updated the revision "
HBASE-5515 [jira] Add a processRow API that supports atomic multiple reads and writes on a row".
Reviewers: tedyu, dhruba, JIRA
Addressed Lars and Ted's comments
@tedyu: I changed reference count to 31 and long count to 6.
This is because we have added the following two things
@lhofhansl: I have made most of the changes you suggested. Thanks!
I tried to put RowProcessor into ProcessRowEndpoint.
With that approach, we will need multiple CoprocessorProtocol. One for each implementation of RowProcessor.
Because the implementation of the RowProcessor is now passed by the protocol class name instead of the Writable.
For example, in our unit test, we have to do
FriendsOfFriendsProtocol p =
And the unit test needs to provide both FriendsOfFriendsProtocol and FriendsOfFriendsEndpoint.
I feel this is not very convenient. I still prefer to pass the RowProcessor information via a Writable.
The other thought is that the user of RowProcessor is actually HRegion.
But ProcessRowEndpoint is used by the client via coprocessor.
I feel that it would be nice to have them separated instead of listing their methods together.
What do you think?