drill-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Sean Hsuan-Yi Chu <hsua...@usc.edu>
Subject Re: Populating Operator Table
Date Tue, 05 Apr 2016 07:23:31 GMT
Sudheesh, I think you are right. Whenever a query is started,
DrillOperatorTable
is created again.

This setting has been there for a long time. One "possible" explanation I
can think of is for query option. For example, if Decimal type is turned
on, the behavior of some functions/ operators could be different.

I agree this seems adding some overhead, but maybe there are some other
reasons?






On Mon, Apr 4, 2016 at 10:23 PM, Sudheesh Katkam <skatkam@maprtech.com>
wrote:

> Hey y’all,
>
> It seems an instance of DrillOperatorTable is created per query <
> https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_apache_drill_blob_master_exec_java-2Dexec_src_main_java_org_apache_drill_exec_ops_QueryContext.java-23L89&d=CwIFaQ&c=clK7kQUTWtAVEOVIgvi0NU5BOUHhpN0H8p7CSfnc_gI&r=jUnl_PVGM7il9n5EkUTnEA&m=iZWEIm96V0s4T6zVJvwv3Z9sUhwYe7rAlifI8Th88Yw&s=5eJEEd_QfCObl-IeZJrTcKpYQJPxT8LLvCpVMkw_7Xo&e=
> >. With type inference changes, this adds to the query startup time (not
> part of query time, rather “run query” request handling time). This object
> could be initialized once in DrillbitContext, and used across queries. So
> is there anything query specific that requires this object to initialized
> within QueryContext?
>
> Thank you,
> Sudheesh

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message