[
https://issues.apache.org/jira/browse/PHOENIX-1705?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14352422#comment-14352422
]
Dumindu Buddhika commented on PHOENIX-1705:
-------------------------------------------
[~jamestaylor], [~sergey.b] Thank you very much for the feedback.
Hi, I think ARRAY_APPEND should not handle the case where the appending element
itself is an array. If you look at PostgreSQL array functions They have a
separate function for concatenating two arrays called ARRAY_CAT,
http://www.postgresql.org/docs/9.2/static/functions-array.html
If we concatenate two arrays in the ARRAY_APPEND function it carries a wrong
meaning,
For example if someone use the function in the following way,
{code:sql}
ARRAY_APPEND(array[1,2,3], array [4,5])
{code}
He is expecting a result like this,
{code:sql}
ARRAY[1,2,3,ARRAY[4,5]]
{code}
But since we do not support multi-dimensional arrays(to my knowledge) we
shouldn't handle this case here.
I think a separate ARRAY_CAT built in function should be implemented for
concatenation of arrays.
What do you think about that?
[~jamestaylor] I will remove the copying of bytes to a new array and improve
the function to support variable length arrays. Thanks for the info on
coerceBytes :).
I am not very clear about the third point. So should we check types in the
constructor of ArrayAppendFunction class?
at here,
{code:java}
public ArrayAppendFunction(List<Expression> children) throws SQLException {
super(children);
}
{code}
Thanks.
> implement ARRAY_APPEND built in function
> ----------------------------------------
>
> Key: PHOENIX-1705
> URL: https://issues.apache.org/jira/browse/PHOENIX-1705
> Project: Phoenix
> Issue Type: Sub-task
> Reporter: Dumindu Buddhika
> Assignee: Dumindu Buddhika
> Attachments:
> PHOENIX-1705_implement_ARRAY_APPEND_built_in_function.patch
>
>
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
