Thanks Julian! I managed to inject my code by setting a ThreadLocal attribute in RelMetadataQuery, as follows,
RelMetadataQuery.THREAD_PROVIDERS.set( JaninoRelMetadataProvider.of(new MyRelMetadataProvider())); Now it works. However, does that mean RelOptCluster#setMetadataProvider is deprecated? On Tue, Mar 29, 2016 at 12:26 PM Julian Hyde <jh...@apache.org> wrote: > Did you take a look at https://issues.apache.org/jira/browse/CALCITE-1147 > <https://issues.apache.org/jira/browse/CALCITE-1147>? It helps make > RelMetadataQuery pluggable. > > > On Mar 28, 2016, at 9:21 PM, Ted Xu <frank...@gmail.com> wrote: > > > > Hi all, > > > > After [CALCITE-604 <https://issues.apache.org/jira/browse/CALCITE-604>], > I > > found the methods of RelMetadataQuery is not pluggable. By 'pluggable', I > > mean RelMetadataQuery's behaviour can be re-defined by registering a > > RelMetadataProvider in RelOptCluster. > > > > For example, if I defined a RelMetadataProvider as follows and register > it > > using RelOptCluster#setMetadataProvider, > > > > public class MyRelMetadataProvider extends ChainedRelMetadataProvider { > > public MyRelMetadataProvider() { > > super( > > ImmutableList.of( > > MyRelMdRowCount.SOURCE)); > > } > > } > > > > I expect RelMetadataQuery#getRowCount will invoke MyRelMdRowCount.SOURCE. > > However I noticed that after CALCITE-604 it always use > > DefaultRelMetadataProvider. > > > > I do noticed an example of overriding RelMetadataQuery in > RelMetadataTest, > > but I guess by now I can only add metadata support rather than override > > existing metadata? > > > > I'm not sure if I'm using it in the right way, please correct me if I'm > > wrong. Thanks! > > > > Regards, > > Ted Xu. > >
