Thanks for your reply, friends!
Base on the discussion, we'll choose option c. I'll continue to implement
Hive's get_json_object function in IMPALA-376. For implementing the ANSI
JSON functions, I created a new JIRA (IMPALA-7609) to track it.
If anyone has more comments, feel free to reply in
+1 for option c
On Mon, Sep 10, 2018 at 10:34 AM, Tim Armstrong
wrote:
> I think I agree with Lars. Having the (mostly) Hive-compatible version is
> useful if there are shared views between Hive/Impala and for people
> migrating queries from Hive or some of the Impala JSON UDFs that I've seen
>
I think I agree with Lars. Having the (mostly) Hive-compatible version is
useful if there are shared views between Hive/Impala and for people
migrating queries from Hive or some of the Impala JSON UDFs that I've seen
floating around.
On Mon, Sep 10, 2018 at 8:20 AM, Lars Volker wrote:
> Thanks
Thanks for the comprehensive summary, Quanlong.
I'm in favor of (c) but I don't feel strongly about the order in which we
should prioritize them. Being compatible to Hive and following the SQL
standard will help users to switch more easily between systems. I'm not
worried about the confusing