Hi Alexander,

Not sure if it's the problem you have, but it might be worth checking if
your ArchiveEnumerableSort contains "collation" only, i.e. offset==null &&
fetch==null. The SortRemoveRule deals with sort-only Sort rels, if either
offset or fetch is not null this rule won't apply. Let me know if it helps.


Thanks,
Maryann

On Tue, Nov 3, 2015 at 11:09 AM, Alexander Reshetov <
[email protected]> wrote:

> Thanks Maryann.
>
> Trying to debug what's going on I created ArchiveConverterRule (and
> registered this rule).
> Rule implements ConverterRule and add convert(), which returns
> ArchiveEnumerableSort.
>
> So basically ArchiveEnumerableSort should be my replacement of
> EnumerableSort.
> But when I call explain plan for select I get EnumerableSort anyway,
> while I expect to see
> implemented ArchiveEnumerableSort.
>
> Should it work as I expect, or there is something other that I need to do?
>
> On Sat, Oct 31, 2015 at 3:26 AM, Maryann Xue <[email protected]>
> wrote:
> > Oh, it's not on master branch yet. It's the "calcite" branch, and for
> > PhoenixTable specifically, you can go:
> >
> https://git1-us-west.apache.org/repos/asf?p=phoenix.git;a=blob;f=phoenix-core/src/main/java/org/apache/phoenix/calcite/PhoenixTable.java;h=4be745036335443bc1bcaae07acf1f32fda6c59a;hb=refs/heads/calcite
> > .
> >
> > On Fri, Oct 30, 2015 at 4:36 PM, Alexander Reshetov <
> > [email protected]> wrote:
> >
> >> Could you please point me to PhoenixTable in repo? Unfortunately I
> >> could not find it.
> >>
> >> On Fri, Oct 30, 2015 at 10:33 PM, Maryann Xue <[email protected]>
> >> wrote:
> >> > SortRemoveRule.INSTANCE is already in the default rule set. I think
> what
> >> > you should do is just make your table return the right collation
> trait.
> >> For
> >> > example in Phoenix we have a Table implementation called PhoenixTable
> >> which
> >> > overrides getStatistic() and implements getCollations() of Statistic.
> >> >
> >> > On Fri, Oct 30, 2015 at 1:40 PM, Alexander Reshetov <
> >> > [email protected]> wrote:
> >> >
> >> >> Is it correct that I need to override
> >> >> org.apache.calcite.rel.AbstractRelNode.register and add rule
> >> >> SortRemoveRule.INSTANCE ?
> >> >>
> >> >> But while it checks RelCollation, then I should also somehow set it.
> >> >> But I could not find how to do it.
> >> >>
> >> >> On Fri, Oct 30, 2015 at 4:23 PM, Alexander Reshetov
> >> >> <[email protected]> wrote:
> >> >> > Here some clarification of my intends.
> >> >> >
> >> >> > Calcite choosing next plan:
> >> >> >
> >> >> > +------+
> >> >> > | PLAN |
> >> >> > +------+
> >> >> > | EnumerableSort(sort0=[$0], dir0=[DESC])
> >> >> >   ArchiveTableScan(table=[[MAIN, SUB]], fields=[[0, 1, 2, 3, 4, 5,
> 6,
> >> >> > 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21]])
> >> >> >  |
> >> >> > +------+
> >> >> >
> >> >> >
> >> >> > The problem is invocation of EnumerableSort. Because data already
> >> >> > sorted by one column. So I need to force exclusion of
> EnumerableSort
> >> >> > from planner.
> >> >> >
> >> >> > Do I need implement some interface or override some method to
> achieve
> >> >> this?
> >> >> >
> >> >> > On Fri, Oct 30, 2015 at 9:51 AM, Julian Hyde <[email protected]>
> >> wrote:
> >> >> >> Yes, this is possible. In fact Phoenix does exactly this.
> >> >> >>
> >> >> >> Take a look at https://issues.apache.org/jira/browse/CALCITE-818
> and
> >> >> https://issues.apache.org/jira/browse/CALCITE-71, which were fixed
> >> along
> >> >> the way.
> >> >> >>
> >> >> >>> On Oct 29, 2015, at 1:16 AM, Alexander Reshetov <
> >> >> [email protected]> wrote:
> >> >> >>>
> >> >> >>> Hello,
> >> >> >>>
> >> >> >>> I have kinda theoretical question. But it should help to
> understand
> >> >> >>> overall Calcite's possibility.
> >> >> >>>
> >> >> >>> Suppose there is some abstract format/table, which supports only
> >> >> >>> sorting by one column.
> >> >> >>>
> >> >> >>> Let's say there is next query passed to calcite:
> >> >> >>>
> >> >> >>>    select * from table order by id
> >> >> >>>
> >> >> >>> Is it possible to rewrite query tree so it will change this
> request
> >> >> into this?
> >> >> >>>
> >> >> >>>    select * from table
> >> >> >>>
> >> >> >>> In this way after logical query rewrite Calcite should not be
> >> >> >>> concerned about sorting at all.
> >> >> >>
> >> >>
> >>
>

Reply via email to