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. >> >> >> >> >> >>
