From 449b7d60803da1e9dc05bd92f20cb2bb7ba71c68 Mon Sep 17 00:00:00 2001
From: Dan Scales <dan.scales@gmail.com>
Date: Thu, 26 Dec 2024 06:13:29 -0800
Subject: [PATCH] GTC-3077 Give NoIntersectionError in AFi if geometry too
 small

In the AFi analysis, we currently don't return any row at all for a
location whose geometry does not intersect the centroid of any pixel
(mostly because the geometry is small compared to the pixel size). Add
similar code as in FCD to detect this case and return
NoIntersectionError.

Same change for GFWProDashboard as well, though need to check if that is
really what we want.
---
 .../summarystats/afi/AFiAnalysis.scala             | 12 +++++++++++-
 .../gfwpro_dashboard/GfwProDashboardAnalysis.scala | 14 +++++++++++++-
 2 files changed, 24 insertions(+), 2 deletions(-)

diff --git a/src/main/scala/org/globalforestwatch/summarystats/afi/AFiAnalysis.scala b/src/main/scala/org/globalforestwatch/summarystats/afi/AFiAnalysis.scala
index 5b7a0c6e..041de6ab 100644
--- a/src/main/scala/org/globalforestwatch/summarystats/afi/AFiAnalysis.scala
+++ b/src/main/scala/org/globalforestwatch/summarystats/afi/AFiAnalysis.scala
@@ -11,6 +11,7 @@ import org.apache.spark.storage.StorageLevel
 import org.apache.spark.sql.functions._
 import scala.collection.immutable.SortedMap
 import io.circe.syntax._
+import cats.data.Validated.{Invalid, Valid}
 
 object AFiAnalysis extends SummaryAnalysis {
 
@@ -42,13 +43,22 @@ object AFiAnalysis extends SummaryAnalysis {
 
     import spark.implicits._
 
+    // If a location has no AFiDataGroup entries, then the geometry must not have
+    // intersected the centroid of any pixels, so report the location as
+    // NoIntersectionError.
+     val summary1RDD = summaryRDD.map {
+       case Valid(Location(fid, data)) if data.isEmpty =>
+         Invalid(Location(fid, NoIntersectionError))
+       case data => data
+     }
+
     // Null out gadm_id for all non-dissolved rows and then aggregate all results for
     // each unique (list_id, location_id, gadm_id, loss_year). Need to combine first
     // with key including loss_year, so we don't have duplicate loss year entries
     // when we create the map of loss years.
     val summary1DF = AFiAnalysis.aggregateByLossYear(
         AFiDF
-        .getFeatureDataFrame(summaryRDD, spark)
+        .getFeatureDataFrame(summary1RDD, spark)
         .withColumn(
           "gadm_id", when(col("location_id") =!= -1, lit("") ).otherwise(col("gadm_id"))
         )
diff --git a/src/main/scala/org/globalforestwatch/summarystats/gfwpro_dashboard/GfwProDashboardAnalysis.scala b/src/main/scala/org/globalforestwatch/summarystats/gfwpro_dashboard/GfwProDashboardAnalysis.scala
index 54e08847..7f2b2aa9 100644
--- a/src/main/scala/org/globalforestwatch/summarystats/gfwpro_dashboard/GfwProDashboardAnalysis.scala
+++ b/src/main/scala/org/globalforestwatch/summarystats/gfwpro_dashboard/GfwProDashboardAnalysis.scala
@@ -18,6 +18,7 @@ import org.globalforestwatch.util.GeotrellisGeometryValidator.makeValidGeom
 import scala.collection.JavaConverters._
 import java.time.LocalDate
 import org.globalforestwatch.util.IntersectGeometry
+import cats.data.Validated.{Invalid, Valid}
 
 object GfwProDashboardAnalysis extends SummaryAnalysis {
 
@@ -109,7 +110,18 @@ object GfwProDashboardAnalysis extends SummaryAnalysis {
           val validatedSummaryStatsRdd = GfwProDashboardRDD(tmp,
                GfwProDashboardGrid.blockTileGrid,
                kwargs + ("getRasterGadm" -> !doGadmIntersect))
-          ValidatedWorkflow(validatedSummaryStatsRdd).mapValid { summaryStatsRDD =>
+          validatedSummaryStatsRdd.collect().foreach(println)
+
+          // If a location has no GfwProDashboardRawDataGroup entries, then the
+          // geometry must not have intersected the centroid of any pixels, so report
+          // the location as NoIntersectionError.
+          val validatedSummaryStatsRdd1 = validatedSummaryStatsRdd.map {
+            case Valid(Location(fid, data)) if data.isEmpty =>
+              Invalid(Location(fid, NoIntersectionError))
+            case data => data
+          }
+
+          ValidatedWorkflow(validatedSummaryStatsRdd1).mapValid { summaryStatsRDD =>
             summaryStatsRDD
               .flatMap { case (CombinedFeatureId(fid@GfwProFeatureId(listId, locationId), gadmId), summary) =>
                 // For non-dissolved locations or vector gadm intersection, merge all