[GitHub] [hive] kgyrtkirk commented on a change in pull request #1031: HIVE-23462 rewrite cume_dist

[GitBox <gi...@apache.org>]

kgyrtkirk commented on a change in pull request #1031:
URL: https://github.com/apache/hive/pull/1031#discussion_r432310866



##########
File path: ql/src/java/org/apache/hadoop/hive/ql/optimizer/calcite/rules/HiveRewriteToDataSketchesRules.java
##########
@@ -368,4 +389,216 @@ void rewrite(AggregateCall aggCall) {
       }
     }
   }
+
+  /**
+   * Generic support for rewriting Windowing expression into a different form usually using joins.
+   */
+  private static abstract class WindowingToProjectAggregateJoinProject extends RelOptRule {
+
+    protected final String sketchType;
+
+    public WindowingToProjectAggregateJoinProject(String sketchType) {
+      super(operand(HiveProject.class, any()));
+      this.sketchType = sketchType;
+    }
+
+    @Override
+    public void onMatch(RelOptRuleCall call) {
+
+      final Project project = call.rel(0);
+
+      VbuilderPAP vb = buildProcessor(call);
+      RelNode newProject = vb.processProject(project);
+
+      if (newProject instanceof Project && ((Project) newProject).getChildExps().equals(project.getChildExps())) {
+        return;
+      } else {
+        call.transformTo(newProject);
+      }
+    }
+
+    protected abstract VbuilderPAP buildProcessor(RelOptRuleCall call);
+
+
+    protected static abstract class VbuilderPAP {
+      private final String sketchClass;
+      protected final RelBuilder relBuilder;
+      protected final RexBuilder rexBuilder;
+
+      protected VbuilderPAP(String sketchClass, RelBuilder relBuilder) {
+        this.sketchClass = sketchClass;
+        this.relBuilder = relBuilder;
+        rexBuilder = relBuilder.getRexBuilder();
+      }
+
+      final class ProcessShuttle extends RexShuttle {
+        public RexNode visitOver(RexOver over) {
+          return processCall(over);
+        }
+      };
+
+      protected RelNode processProject(Project project) {
+        relBuilder.push(project.getInput());
+        RexShuttle shuttle = new ProcessShuttle();
+        List<RexNode> newProjects = new ArrayList<RexNode>();
+        for (RexNode expr : project.getChildExps()) {
+                newProjects.add(expr.accept(shuttle));
+        }
+        relBuilder.project(newProjects);
+        return relBuilder.build();
+      }
+
+      private final RexNode processCall(RexNode expr) {
+        if (expr instanceof RexOver) {
+          RexOver over = (RexOver) expr;
+          if (isApplicable(over)) {
+            return rewrite(over);
+          }
+        }
+        return expr;
+      }
+
+      protected final SqlOperator getSqlOperator(String fnName) {
+        UDFDescriptor fn = DataSketchesFunctions.INSTANCE.getSketchFunction(sketchClass, fnName);
+        if (!fn.getCalciteFunction().isPresent()) {
+          throw new RuntimeException(fn.toString() + " doesn't have a Calcite function associated with it");
+        }
+        return fn.getCalciteFunction().get();
+      }
+
+      abstract RexNode rewrite(RexOver expr);
+
+      abstract boolean isApplicable(RexOver expr);
+
+    }
+
+  }
+
+  public static class CumeDistRewrite extends WindowingToProjectAggregateJoinProject {
+
+    public CumeDistRewrite(String sketchType) {
+      super(sketchType);
+    }
+
+    @Override
+    protected VbuilderPAP buildProcessor(RelOptRuleCall call) {
+      return new VB(sketchType, call.builder());
+    }
+
+    private static class VB extends VbuilderPAP {
+
+      protected VB(String sketchClass, RelBuilder relBuilder) {
+        super(sketchClass, relBuilder);
+      }
+
+      @Override
+      boolean isApplicable(RexOver over) {
+        SqlAggFunction aggOp = over.getAggOperator();
+        RexWindow window = over.getWindow();
+        if (aggOp.getName().equalsIgnoreCase("cume_dist") && window.orderKeys.size() == 1
+            && window.getLowerBound().isUnbounded() && window.getUpperBound().isUnbounded()) {
+          return true;
+        }
+        return false;
+      }
+
+      @Override
+      RexNode rewrite(RexOver over) {
+
+        over.getOperands();
+        RexWindow w = over.getWindow();
+
+        RexFieldCollation orderKey = w.orderKeys.get(0);
+        // we don't really support nulls in aggregate/etc...they are actually ignored
+        // so some hack will be needed for NULLs anyway..
+        ImmutableList<RexNode> partitionKeys = w.partitionKeys;
+
+        relBuilder.push(relBuilder.peek());

Review comment:
       yes; this is the point where the other side of the join is getting started to being built
   added some explanation to the #rewrite method about this




----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

For queries about this service, please contact Infrastructure at:
users@infra.apache.org



---------------------------------------------------------------------
To unsubscribe, e-mail: gitbox-unsubscribe@hive.apache.org
For additional commands, e-mail: gitbox-help@hive.apache.org