You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@calcite.apache.org by Florent Martineau <ft...@gmail.com> on 2021/11/26 14:50:37 UTC
Error: Need to implement org.apache.calcite.adapter.jdbc.JdbcToEnumerableConverter
Hi everyone,
I'm having troubles transforming a RelNode to SQL. When running this code:
RelToSqlConverter converter = new
RelToSqlConverter(SqlDialect.DatabaseProduct.CALCITE.getDialect());
SqlImplementor.Result result = converter.visitRoot(relNode);
I'm getting the following error on visitRoot:
Need to implement org.apache.calcite.adapter.jdbc.JdbcToEnumerableConverter
As the RelToSqlConverter only takes a dialect as an argument, and visitRoot
a RelNode, I suspect the error comes from the planning.
To build my RelNode, I'm running the optimizer. It works without error and
seems to optimize the statement, but this RelNode seems not suitable for
the RelToSqlConverter later on.
Here is how I run the optimizer:
RelOptCluster cluster = rel.getCluster();
RelOptPlanner planner = cluster.getPlanner();
// JdbcConvention convention = new JdbcConvention(SqlDialect
.DatabaseProduct.CALCITE.getDialect(), new
DefaultExpression(RelNode.class), "my_schema");
RelTraitSet desiredTraits =
cluster.traitSet().plus(EnumerableConvention.INSTANCE); //.plus(convention);
RelNode newRoot = planner.changeTraits(rel, desiredTraits);
planner.setRoot(newRoot);
RelNode optimizedNode = planner.findBestExp();
As you can see commented out, I've tried adding the JdbcConvention but it
doesn't work, and I don't get how to include the JdbcToEnumerableConverter.
If you have any hints as to where the error could come from, this would be
truly awesome!!
Have a great day,
Florent
Re: Error: Need to implement org.apache.calcite.adapter.jdbc.JdbcToEnumerableConverter
Posted by Florent Martineau <ft...@gmail.com>.
Thank you very much Stamatis !
On Mon, Nov 29, 2021, 9:37 AM Stamatis Zampetakis <za...@gmail.com> wrote:
> Hi Florent,
>
> First when you post an error/exception it is helpful to have the full stack
> trace to understand better the general context.
>
> Second when you suspect the problem is related to the plan it is also
> a good idea to share the plan in question;
> use
> to
> org.apache.calcite.plan.RelOptUtil#toString(org.apache.calcite.rel.RelNode)
> or another similar API to print the plan.
>
> As you said the reason that you are getting the error is most likely the
> input plan; the relNode that you are passing in the converter.
> I guess the plan contains a RelNode which cannot be handled by the
> `RelToSqlConverter`.
> Usually when this class is used by Calcite all the operators are of the
> form Jdbc* (e.g., JdbcJoin, JdbcFilter, JdbcProject, etc.).
>
> To understand how Calcite transforms operators in SQL, run some test cases
> from JdbcAdapterTest in debug mode and put breakpoints in
> JdbcToEnumerableConverter [1].
>
> Based on this you may need to replace converter.visitRoot with
> converter.visitInput and possibly modify the traitset to have only the
> JdbcConvention.
>
> RelTraitSet desiredTraits = cluster.traitSet().replace(convention);
>
> You can probably use the `RelToSqlConverter` with just Logical or
> Enumerable operators but you may find out that some operators are not
> supported either because it doesn't make sense or because nobody
> implemented support for those till now.
>
> Best,
> Stamatis
>
> [1]
>
> https://github.com/apache/calcite/blob/7c423ef23878271b1c50c03629ebfff674985681/core/src/main/java/org/apache/calcite/adapter/jdbc/JdbcToEnumerableConverter.java#L346
>
>
> On Fri, Nov 26, 2021 at 3:51 PM Florent Martineau <ftecmartineau@gmail.com
> >
> wrote:
>
> > Hi everyone,
> >
> > I'm having troubles transforming a RelNode to SQL. When running this
> code:
> >
> > RelToSqlConverter converter = new
> > RelToSqlConverter(SqlDialect.DatabaseProduct.CALCITE.getDialect());
> > SqlImplementor.Result result = converter.visitRoot(relNode);
> >
> > I'm getting the following error on visitRoot:
> > Need to implement
> org.apache.calcite.adapter.jdbc.JdbcToEnumerableConverter
> >
> > As the RelToSqlConverter only takes a dialect as an argument, and
> visitRoot
> > a RelNode, I suspect the error comes from the planning.
> >
> > To build my RelNode, I'm running the optimizer. It works without error
> and
> > seems to optimize the statement, but this RelNode seems not suitable for
> > the RelToSqlConverter later on.
> >
> > Here is how I run the optimizer:
> >
> > RelOptCluster cluster = rel.getCluster();
> > RelOptPlanner planner = cluster.getPlanner();
> > // JdbcConvention convention = new JdbcConvention(SqlDialect
> > .DatabaseProduct.CALCITE.getDialect(), new
> > DefaultExpression(RelNode.class), "my_schema");
> > RelTraitSet desiredTraits =
> > cluster.traitSet().plus(EnumerableConvention.INSTANCE);
> > //.plus(convention);
> > RelNode newRoot = planner.changeTraits(rel, desiredTraits);
> > planner.setRoot(newRoot);
> > RelNode optimizedNode = planner.findBestExp();
> >
> > As you can see commented out, I've tried adding the JdbcConvention but it
> > doesn't work, and I don't get how to include the
> JdbcToEnumerableConverter.
> >
> > If you have any hints as to where the error could come from, this would
> be
> > truly awesome!!
> >
> > Have a great day,
> >
> > Florent
> >
>
Re: Error: Need to implement org.apache.calcite.adapter.jdbc.JdbcToEnumerableConverter
Posted by Stamatis Zampetakis <za...@gmail.com>.
Hi Florent,
First when you post an error/exception it is helpful to have the full stack
trace to understand better the general context.
Second when you suspect the problem is related to the plan it is also
a good idea to share the plan in question;
use
to org.apache.calcite.plan.RelOptUtil#toString(org.apache.calcite.rel.RelNode)
or another similar API to print the plan.
As you said the reason that you are getting the error is most likely the
input plan; the relNode that you are passing in the converter.
I guess the plan contains a RelNode which cannot be handled by the
`RelToSqlConverter`.
Usually when this class is used by Calcite all the operators are of the
form Jdbc* (e.g., JdbcJoin, JdbcFilter, JdbcProject, etc.).
To understand how Calcite transforms operators in SQL, run some test cases
from JdbcAdapterTest in debug mode and put breakpoints in
JdbcToEnumerableConverter [1].
Based on this you may need to replace converter.visitRoot with
converter.visitInput and possibly modify the traitset to have only the
JdbcConvention.
RelTraitSet desiredTraits = cluster.traitSet().replace(convention);
You can probably use the `RelToSqlConverter` with just Logical or
Enumerable operators but you may find out that some operators are not
supported either because it doesn't make sense or because nobody
implemented support for those till now.
Best,
Stamatis
[1]
https://github.com/apache/calcite/blob/7c423ef23878271b1c50c03629ebfff674985681/core/src/main/java/org/apache/calcite/adapter/jdbc/JdbcToEnumerableConverter.java#L346
On Fri, Nov 26, 2021 at 3:51 PM Florent Martineau <ft...@gmail.com>
wrote:
> Hi everyone,
>
> I'm having troubles transforming a RelNode to SQL. When running this code:
>
> RelToSqlConverter converter = new
> RelToSqlConverter(SqlDialect.DatabaseProduct.CALCITE.getDialect());
> SqlImplementor.Result result = converter.visitRoot(relNode);
>
> I'm getting the following error on visitRoot:
> Need to implement org.apache.calcite.adapter.jdbc.JdbcToEnumerableConverter
>
> As the RelToSqlConverter only takes a dialect as an argument, and visitRoot
> a RelNode, I suspect the error comes from the planning.
>
> To build my RelNode, I'm running the optimizer. It works without error and
> seems to optimize the statement, but this RelNode seems not suitable for
> the RelToSqlConverter later on.
>
> Here is how I run the optimizer:
>
> RelOptCluster cluster = rel.getCluster();
> RelOptPlanner planner = cluster.getPlanner();
> // JdbcConvention convention = new JdbcConvention(SqlDialect
> .DatabaseProduct.CALCITE.getDialect(), new
> DefaultExpression(RelNode.class), "my_schema");
> RelTraitSet desiredTraits =
> cluster.traitSet().plus(EnumerableConvention.INSTANCE);
> //.plus(convention);
> RelNode newRoot = planner.changeTraits(rel, desiredTraits);
> planner.setRoot(newRoot);
> RelNode optimizedNode = planner.findBestExp();
>
> As you can see commented out, I've tried adding the JdbcConvention but it
> doesn't work, and I don't get how to include the JdbcToEnumerableConverter.
>
> If you have any hints as to where the error could come from, this would be
> truly awesome!!
>
> Have a great day,
>
> Florent
>