[
https://issues.apache.org/jira/browse/PHOENIX-2056?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
James Taylor updated PHOENIX-2056:
----------------------------------
Description:
Given that we allow a nullable PK column to be added to a base table and push
it to all views (PHOENIX-1504), we need to ensure that the PK column is also
added to any indexes on the view. Without this, indexes on views would no
longer maintain their 1:1 correspondence with data rows.
Since the pushing of the new column is done under lock in MetaDataEndPointImpl
and we have a PTable instantiated for each view, we can just iterate through
its indexes (view.getIndexes()) and generate a couple of additional Puts: one
for the PK column row in the index and one for the header row of the index
table to increment the column count. The index table header row needs to be
added to the list of rows to be invalidated as well.
was:
Given that we allow a nullable PK column to be added to a base table and push
it to all views (PHOENIX-1504), we need to ensure that the PK column is also
added to any indexes on the view. Without this, indexes on views would no
longer maintain their 1:1 correspondence with data rows.
> Ensure PK column from base table is added to any indexes on views
> ------------------------------------------------------------------
>
> Key: PHOENIX-2056
> URL: https://issues.apache.org/jira/browse/PHOENIX-2056
> Project: Phoenix
> Issue Type: Bug
> Reporter: James Taylor
> Assignee: Samarth Jain
>
> Given that we allow a nullable PK column to be added to a base table and push
> it to all views (PHOENIX-1504), we need to ensure that the PK column is also
> added to any indexes on the view. Without this, indexes on views would no
> longer maintain their 1:1 correspondence with data rows.
> Since the pushing of the new column is done under lock in
> MetaDataEndPointImpl and we have a PTable instantiated for each view, we can
> just iterate through its indexes (view.getIndexes()) and generate a couple of
> additional Puts: one for the PK column row in the index and one for the
> header row of the index table to increment the column count. The index table
> header row needs to be added to the list of rows to be invalidated as well.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
