phoenix-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "James Taylor (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (PHOENIX-2975) CSVBulkLoad for local index only fails.
Date Mon, 06 Mar 2017 22:50:33 GMT

    [ https://issues.apache.org/jira/browse/PHOENIX-2975?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15898310#comment-15898310
] 

James Taylor commented on PHOENIX-2975:
---------------------------------------

That's a fair point, [~sergey.soldatov]. Want to file a separate JIRA and we can discuss if
we should just not support index building directly on CSV?

> CSVBulkLoad for local index only fails.
> ---------------------------------------
>
>                 Key: PHOENIX-2975
>                 URL: https://issues.apache.org/jira/browse/PHOENIX-2975
>             Project: Phoenix
>          Issue Type: Bug
>    Affects Versions: 4.8.0
>            Reporter: Sergey Soldatov
>            Assignee: Sergey Soldatov
>             Fix For: 4.8.0
>
>         Attachments: PHOENIX-2975.patch
>
>
> When we use bulk load for local index only we have a problem: in mapper we generate values
using upsert statement, so uncommitted changes contains cells for both table and ALL local
indexes. There is no easy way to separate them from each others. I.e. to check that we got
KV for local index we may check that CF is "L#0" (can it be something else?), but if we have
two local indexes, to find whether cell belongs to particular one  we need to look into rowkey
values. 
> And so, here is the question - is there any real use case to load local index separately
from the parent table using csv bulk load? I can't find any. Even if user forgot to create
index before using csv bulk load, there is  indextool to load indexes which is doing it in
the right way. 
> [~jamestaylor], [~rajeshbabu] any thoughts?
>  



--
This message was sent by Atlassian JIRA
(v6.3.15#6346)

Mime
View raw message