[
https://issues.apache.org/jira/browse/CALCITE-492?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14233549#comment-14233549
]
Julian Hyde commented on CALCITE-492:
-------------------------------------
I was initially thinking we would have create a Phoenix-specific extension to
the parser. But since UPSERT is identical to INSERT except for the first word,
that would seem to be overkill. I think it would work to add a new reserved
word, parse UPSERT as INSERT, and add a boolean flag to SqlInsert AST node.
That flag would be carried into the TableModify relational expression, perhaps
an extra value in its enum Operation { INSERT, UPDATE, DELETE, MERGE }. It's
TBD whether UPSERT behaves more like INSERT or MERGE during optimization. It's
clear that it's more like INSERT at parse time.
> Support UPSERT statement in parser
> ----------------------------------
>
> Key: CALCITE-492
> URL: https://issues.apache.org/jira/browse/CALCITE-492
> Project: Calcite
> Issue Type: Improvement
> Reporter: James Taylor
> Assignee: Julian Hyde
> Labels: phoenix
>
> Support an UPSERT VALUES and UPSERT SELECT statement in Calcite. These have
> the same syntax as INSERT VALUES and INSERT SELECT, but they insert the row
> if it doesn't already exist and update it otherwise. This is a pretty common
> pattern for big data applications that want to have an option to write
> without reading first. See
> http://phoenix.apache.org/language/index.html#upsert_values and
> http://phoenix.apache.org/language/index.html#upsert_select for syntax.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
