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] [Resolved] (PHOENIX-809) Metadata is not getting updated for dynamically created columns
Date Fri, 02 May 2014 04:42:20 GMT

     [ https://issues.apache.org/jira/browse/PHOENIX-809?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]

James Taylor resolved PHOENIX-809.
----------------------------------

    Resolution: Invalid

WAD

> Metadata is not getting updated for dynamically created columns
> ---------------------------------------------------------------
>
>                 Key: PHOENIX-809
>                 URL: https://issues.apache.org/jira/browse/PHOENIX-809
>             Project: Phoenix
>          Issue Type: Task
>            Reporter: alex kamil
>
> After creating columns dynamically with upsert, users need to keep track of  the newly
created metadata (column names/types) themselves, as the metadata for dynamically added columns
is not retrievable via standard JDBC DatabaseMetaData calls.
> for example:
> create table with one column:
> CREATE TABLE IF NOT EXISTS my_test_table (rowkey VARCHAR NOT NULL PRIMARY KEY)
> add two more columns dynamically:
> UPSERT INTO my_test_table (rowkey, col1 VARCHAR, col2 VARCHAR)  VALUES ('rk', 'value1',
'value2');
> Now the only way to see the new column values is by defining column names/types every
time at runtime:
> select * from my_test_table (rowkey VARCHAR, col1 VARCHAR, col2 VARCHAR);
> i.e. simple "select * from my_test_table;" would only show the values of the columns
defined using stanard ddl (such as "rowkey" column defined in "create table" statment above,
or "alter table"). The col1,col2 metadata is not accessible via DatabaseMetaData calls (such
as getColumns).
> While dynamically defined columns seem to go beyond standard SQL functionality , this
is one of the key features of hbase/phoenix that makes it useful for us.
> current workarounds are ugly - e.g either spawn a separate thread updating metadata via
"ALTER TABLE" asynchronously (decoupled from the thread doing upserts) in order not to slow
down the upserts; or use a local metadata cache which introduces consistency issues when syncing
multiple phoenix clients.



--
This message was sent by Atlassian JIRA
(v6.2#6252)

Mime
View raw message