Copilot commented on code in PR #2953:
URL: https://github.com/apache/sedona/pull/2953#discussion_r3249892784
##########
spark/common/src/main/scala/org/apache/spark/sql/sedona_sql/strategy/join/TraitJoinQueryBase.scala:
##########
@@ -49,8 +51,7 @@ trait TraitJoinQueryBase {
spatialRdd.setRawSpatialRDD(
rdd
.map { x =>
- val shape =
-
GeometrySerializer.deserialize(shapeExpression.eval(x).asInstanceOf[Array[Byte]])
+ val shape = TraitJoinQueryBase.shapeToGeometry(shapeExpression, x)
shape.setUserData(x.copy)
shape
Review Comment:
`shapeToGeometry` can return null (e.g., when the shape column is NULL), but
`toSpatialRDD` immediately calls `shape.setUserData(...)`, which will throw
NPE. Previously this path used `GeometrySerializer.deserialize(...)`, which
converts null bytes to an empty geometry collection, avoiding a crash. Please
either filter out null shapes here (flatMap) or map null to a non-null empty
geometry consistently.
##########
spark/common/src/main/scala/org/apache/spark/sql/sedona_sql/strategy/join/TraitJoinQueryBase.scala:
##########
@@ -178,3 +178,36 @@ trait TraitJoinQueryBase {
}
}
}
+
+object TraitJoinQueryBase {
+
+ /**
+ * Materialise a shape column value as a JTS [[Geometry]]. Box2D-typed
columns are turned into
+ * the closed rectangular polygon implied by their `(xmin, ymin, xmax,
ymax)` bounds; all other
+ * shape columns are deserialised from the Sedona geometry binary form.
+ *
+ * Producing a JTS rectangle here lets the rest of the join machinery —
partitioner, R-tree
+ * `IndexBuilder`, refine evaluator — stay shape-agnostic. JTS already
short-circuits
+ * rectangle-rectangle predicates (`Polygon.isRectangle` triggers
`RectangleIntersects` /
+ * `RectangleContains`), so a `ST_BoxIntersects` join naturally pays only
the four-double
+ * envelope comparison at refine time.
+ */
+ def shapeToGeometry(shapeExpression: Expression, row: InternalRow): Geometry
= {
+ val evaluated = shapeExpression.eval(row)
+ if (evaluated == null) {
+ null
+ } else
+ shapeExpression.dataType match {
+ case _: Box2DUDT =>
+ val box = evaluated.asInstanceOf[InternalRow]
+ Constructors
+ .polygonFromEnvelope(
+ box.getDouble(0),
+ box.getDouble(1),
+ box.getDouble(2),
+ box.getDouble(3))
Review Comment:
In the Box2D branch, this builds a polygon directly from the four doubles
without enforcing the ordered-bound contract (`xmin <= xmax` and `ymin <=
ymax`). The scalar `ST_BoxIntersects/ST_BoxContains` path explicitly throws on
inverted bounds (see
`org.apache.sedona.common.Predicates.boxIntersects/boxContains`), so join
planning should reject the same inputs to avoid silently returning misleading
results. Also consider using `Constructors.geomFromBox2D` (point/linestring for
degenerate boxes) to avoid creating invalid zero-area polygons.
##########
spark/common/src/test/scala/org/apache/sedona/sql/Box2DJoinSuite.scala:
##########
@@ -0,0 +1,147 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements. See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership. The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License. You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing,
+ * software distributed under the License is distributed on an
+ * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+ * KIND, either express or implied. See the License for the
+ * specific language governing permissions and limitations
+ * under the License.
+ */
+package org.apache.sedona.sql
+
+import org.apache.spark.sql.DataFrame
+import org.apache.spark.sql.functions.{broadcast, expr}
+import org.apache.spark.sql.sedona_sql.strategy.join.{BroadcastIndexJoinExec,
RangeJoinExec}
+
+class Box2DJoinSuite extends TestBaseScala {
+
+ import Box2DJoinSuite.TestBox
+
+ /**
+ * Three left-side boxes and three right-side boxes wired so we can predict
exact result sizes:
+ *
+ * - L1=(0,0,10,10) R1=(5,5,15,15) — overlapping
+ * - L1=(0,0,10,10) R2=(2,2,8,8) — R2 fully inside L1
+ * - L2=(0,0,10,10) R1=(5,5,15,15) — overlapping
+ * - L2=(0,0,10,10) R2=(2,2,8,8) — R2 fully inside L2
+ * - L3 and R3 are disjoint from everything else; (L3,R3) is itself
disjoint.
+ *
+ * Intersection-pair count: 4. Containment-pair count: 2 (L1⊇R2, L2⊇R2).
+ */
+ private def leftBoxes: DataFrame = {
+ import sparkSession.implicits._
+ Seq(TestBox(1, 0, 0, 10, 10), TestBox(2, 0, 0, 10, 10), TestBox(3, 20, 20,
30, 30))
+ .toDF("id", "xmin", "ymin", "xmax", "ymax")
+ .selectExpr("id", "ST_MakeBox2D(ST_Point(xmin, ymin), ST_Point(xmax,
ymax)) AS box")
+ }
+
+ private def rightBoxes: DataFrame = {
+ import sparkSession.implicits._
+ Seq(TestBox(11, 5, 5, 15, 15), TestBox(12, 2, 2, 8, 8), TestBox(13, 40,
40, 50, 50))
+ .toDF("id", "xmin", "ymin", "xmax", "ymax")
+ .selectExpr("id", "ST_MakeBox2D(ST_Point(xmin, ymin), ST_Point(xmax,
ymax)) AS box")
+ }
+
+ describe("Box2D spatial join") {
+
+ it("ST_BoxIntersects: broadcast index join produces correct pairs") {
+ val df = leftBoxes
+ .alias("L")
+ .join(broadcast(rightBoxes.alias("R")), expr("ST_BoxIntersects(L.box,
R.box)"))
+ val plan = df.queryExecution.sparkPlan
Review Comment:
This suite covers broadcast/range planning and basic result counts, but it
doesn’t exercise Box2D error/NULL edge cases that are part of the Box2D
predicate contract (e.g., inverted bounds should throw; NULL boxes should not
crash and should produce no matches). Adding targeted tests for those cases
would help lock in the intended join semantics, especially given the new Box2D
materialization path in the join executors.
--
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.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
