[
https://issues.apache.org/jira/browse/PHOENIX-1409?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14270613#comment-14270613
]
Samarth Jain commented on PHOENIX-1409:
---------------------------------------
Does this test need to be run in it's own cluster, as it's not overriding any
properties? If not, let's keep it as-is.
This test suite is kinda unique in the sense that we are verifying hbase level
properties and hence we can't have overlapping table names for some tests.
Optionally I could set the DROP_METADATA_ATTRIB to true but then that makes the
test suite execution much slower since each test then disables and drops the
underlying hbase table. Just having non overlapping table names and extending
BaseHBaseManagedTimedTest doesn't help the test execution time either because
we end up disabling and dropping all the tables at the end of test suite. Both
the approaches turn out to be slower by over 4 minutes than having the test run
in its own cluster which doesn't have to pay the penalty of disabling and
dropping tables.
Shouldn't the following old test still pass? If so, let's leave as-is and
create a new one for other stuff we want to test. In general, a change of a
unit test is a potential red flag for a change in behavior IMO.
The older test case is now covered under
testSetPropertyAndAddColumnForNewColumnFamily and
testSetPropertyAndAddColumnForDefaultColumnFamily.
The behavior at CREATE time for this is that it'll end up as an
HTableDescriptor property. Can we add a test for this and make sure that ALTER
TABLE matches this behavior?
The same behavior happens for ALTER TABLE too. Will add a test to verify the
same.
> Allow ALTER TABLE <table> SET command to update HTableDescriptor and
> HColumnDescriptor properties
> -------------------------------------------------------------------------------------------------
>
> Key: PHOENIX-1409
> URL: https://issues.apache.org/jira/browse/PHOENIX-1409
> Project: Phoenix
> Issue Type: Improvement
> Affects Versions: 4.2
> Reporter: James Taylor
> Assignee: Alicia Ying Shu
> Attachments: PHOENIX-1409-v3.patch, PHOENIX-1409-v6.patch,
> Phoenix-1409-v1.patch, Phoenix-1409-v4-2.patch, Phoenix-1409-v4.patch,
> Phoenix-1409-v5.patch, Phoenix-1409.patch, WIP.patch, phoenix-1409-v2.patch
>
>
> Once PHOENIX-1408 is fixed, we should allow HTableDescriptor and
> HColumnDescriptor properties through the ALTER TABLE <table> SET command.
> It'd just be a matter of passing these properties through the existing
> methods, as we support this for CREATE TABLE.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)