hbase-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Chia-Ping Tsai (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HBASE-19437) Batch operation can't handle the null result for Append/Increment
Date Wed, 06 Dec 2017 13:06:00 GMT

    [ https://issues.apache.org/jira/browse/HBASE-19437?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16280151#comment-16280151
] 

Chia-Ping Tsai commented on HBASE-19437:
----------------------------------------

bq. In this case, what we were returning? Null or empty? (In 1.x)..
The cases impacted by {{isReturnResults}} are shown below.
1) all done and {{isReturnResults}} is false
||branch-1||master||
|return null|return empty|
The change is made by HBASE-16283. It tried to fix the bug about that the batch operation
can' handle the null. (the case differs from this issue)
2) preAppendAfterRowLock return stuff and {{isReturnResults}} is false
||branch-1||master||
|doesn't work|return null|
The change is made by HBASE-15285

bq.  If we change, that will be a BC break?
User won't receive the null value from append/increment after this fix. I don't think it is
a BC break as the empty is what we should return to user when there is nothing to return.

> Batch operation can't handle the null result for Append/Increment
> -----------------------------------------------------------------
>
>                 Key: HBASE-19437
>                 URL: https://issues.apache.org/jira/browse/HBASE-19437
>             Project: HBase
>          Issue Type: Bug
>            Reporter: Chia-Ping Tsai
>            Assignee: Chia-Ping Tsai
>             Fix For: 2.0.0
>
>         Attachments: HBASE-19437.v0.patch
>
>
> But the Table#append and #increment can handle the null result...that is an inconsistent
behavior for user.
> I have noticed two scenarios that server will return null result to user.
> # postAppend/postIncrement return null
> # mutation.isReturnResults() is false and preIncrementAfterRowLock/preAppendAfterRowLock
doesn't return null
> We should wrap the null to empty result on server side. CP user should throw Exception
rather than return null if they intend to say something is broken.



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)

Mime
View raw message