graphframes · SemyonSinchenko · Mar 31, 2026 · Mar 30, 2026 · Mar 30, 2026 · Mar 30, 2026
diff --git a/.github/workflows/scala-ci.yml b/.github/workflows/scala-ci.yml
@@ -6,6 +6,8 @@ jobs:
       fail-fast: false
       matrix:
         include:
+          - spark-version: 3.5.7
+            java-version: 8
           - spark-version: 3.5.7
             java-version: 11
           - spark-version: 3.5.7

diff --git a/benchmarks/src/main/scala/org/graphframes/benchmarks/LDBCBenchmarkBase.scala b/benchmarks/src/main/scala/org/graphframes/benchmarks/LDBCBenchmarkBase.scala
@@ -8,6 +8,7 @@ import org.openjdk.jmh.annotations.*
 
 import java.io.File
 import java.nio.file.Path
+import java.nio.file.Paths
 
 trait LDBCBenchmarkBase {
   @Param(Array("wiki-Talk"))
@@ -19,7 +20,7 @@ trait LDBCBenchmarkBase {
   var spark: SparkSession = _
   var graph: GraphFrame = _
 
-  protected def cacheDir: Path = Path.of(new File("target").toURI).resolve("ldbc-cache")
+  protected def cacheDir: Path = Paths.get(new File("target").toURI).resolve("ldbc-cache")
 
   @Setup(Level.Trial)
   def setup(): Unit = {

diff --git a/benchmarks/src/main/scala/org/graphframes/benchmarks/LDBCBenchmarkSuite.scala b/benchmarks/src/main/scala/org/graphframes/benchmarks/LDBCBenchmarkSuite.scala
@@ -14,6 +14,7 @@ import org.openjdk.jmh.infra.Blackhole
 import java.io.File
 import java.nio.file.Files
 import java.nio.file.Path
+import java.nio.file.Paths
 import java.util.Properties
 import java.util.concurrent.TimeUnit
 
@@ -85,7 +86,7 @@ class LDBCBenchmarkSuite {
 
   private def caseRoot: Path = resourcesPath.resolve(benchmarkGraphName)
 
-  private def resourcesPath = Path.of(new File("target").toURI)
+  private def resourcesPath = Paths.get(new File("target").toURI)
 
   @Benchmark
   def benchmarkSP(blackhole: Blackhole): Unit = {

diff --git a/benchmarks/src/main/scala/org/graphframes/benchmarks/ParquetDataLoader.scala b/benchmarks/src/main/scala/org/graphframes/benchmarks/ParquetDataLoader.scala
@@ -1,12 +1,11 @@
 package org.graphframes.benchmarks
 
-import java.net.URL
 import java.nio.file.Files
 import java.nio.file.Path
+import scala.sys.process.*
 
 class ParquetDataLoader(cacheDir: Path) {
   private val LDBC_PARQUET_URL_PREFIX = "https://datasets.ldbcouncil.org/graphalytics-parquet/"
-  private val bufferSize = 8192
 
   def downloadParquetIfNotExists(graphName: String): Unit = {
     val graphDir = cacheDir.resolve(graphName)
@@ -33,21 +32,31 @@ class ParquetDataLoader(cacheDir: Path) {
     }
   }
 
+  // Use curl instead of Java's URLConnection because the LDBC CDN (Cloudflare)
+  // rejects Java 8's TLS fingerprint with HTTP 403.
+  // TODO: restore URLConnection after Spark 3.5.x EOL (~April 2026) when JDK 8 can be dropped:
+  //   private def downloadFile(url: String, dest: Path): Unit = {
+  //     val connection = new java.net.URL(url).openConnection()
+  //     connection.setConnectTimeout(30000)
+  //     connection.setReadTimeout(30000)
+  //     val inputStream = connection.getInputStream
+  //     val outputStream = Files.newOutputStream(dest)
+  //     val buffer = new Array[Byte](8192)
+  //     var bytesRead = 0
+  //     try {
+  //       while ({ bytesRead = inputStream.read(buffer); bytesRead } != -1) {
+  //         outputStream.write(buffer, 0, bytesRead)
+  //       }
+  //     } finally {
+  //       inputStream.close()
+  //       outputStream.close()
+  //     }
+  //     println(s"Downloaded $url to $dest")
+  //   }
   private def downloadFile(url: String, dest: Path): Unit = {
-    val connection = new URL(url).openConnection()
-    connection.setConnectTimeout(30000)
-    connection.setReadTimeout(30000)
-    val inputStream = connection.getInputStream
-    val outputStream = Files.newOutputStream(dest)
-    val buffer = new Array[Byte](bufferSize)
-    var bytesRead = 0
-    try {
-      while ({ bytesRead = inputStream.read(buffer); bytesRead } != -1) {
-        outputStream.write(buffer, 0, bytesRead)
-      }
-    } finally {
-      inputStream.close()
-      outputStream.close()
+    val curlExit = s"curl -fSL -o ${dest.toString} $url".!
+    if (curlExit != 0) {
+      throw new RuntimeException(s"Failed to download $url (curl exit code: $curlExit)")
     }
     println(s"Downloaded $url to $dest")
   }

diff --git a/core/src/main/scala/org/apache/spark/sql/graphframes/expressions/KMinSampling.scala b/core/src/main/scala/org/apache/spark/sql/graphframes/expressions/KMinSampling.scala
@@ -146,7 +146,7 @@ object KMinSampling extends Serializable {
     // That is very stupid way actually. But it is the only way with public API
     spark
       .createDataFrame(
-        java.util.List.of[Row](),
+        java.util.Collections.emptyList[Row](),
         StructType(
           StructField(colNames(0), dataType) :: StructField(colNames(1), LongType) :: Nil))
       .encoder

diff --git a/core/src/main/scala/org/graphframes/examples/LDBCUtils.scala b/core/src/main/scala/org/graphframes/examples/LDBCUtils.scala
@@ -1,12 +1,10 @@
 package org.graphframes.examples
 
-import java.net.URL
 import java.nio.file.*
 import scala.sys.process.*
 
 object LDBCUtils {
   private val LDBC_URL_PREFIX = "https://datasets.ldbcouncil.org/graphalytics/"
-  private val bufferSize = 8192 // 8Kb
 
   val TEST_BFS_DIRECTED = "test-bfs-directed"
   val TEST_BFS_UNDIRECTED = "test-bfs-undirected"
@@ -51,7 +49,7 @@ object LDBCUtils {
     CIT_PATENTS,
     WIKI_TALKS)
 
-  private def ldbcURL(caseName: String): URL = new URL(s"${LDBC_URL_PREFIX}${caseName}.tar.zst")
+  private def ldbcURL(caseName: String): String = s"${LDBC_URL_PREFIX}${caseName}.tar.zst"
 
   private def checkZSTD(): Unit = {
     try {
@@ -81,16 +79,24 @@ object LDBCUtils {
         Files.createDirectories(dir)
       }
       val archivePath = path.resolve(s"${name}.tar.zst")
-      val connection = ldbcURL(name).openConnection()
-      val inputStream = connection.getInputStream
-      val outputStream = Files.newOutputStream(archivePath)
-      val buffer = new Array[Byte](bufferSize)
-      var bytesRead = 0
-      while ({ bytesRead = inputStream.read(buffer); bytesRead } != -1) {
-        outputStream.write(buffer, 0, bytesRead)
+      // Use curl instead of Java's URLConnection because the LDBC CDN (Cloudflare)
+      // rejects Java 8's TLS fingerprint with HTTP 403.
+      // TODO: restore URLConnection after Spark 3.5.x EOL (~April 2026) when JDK 8 can be dropped:
+      //   val connection = new java.net.URL(ldbcURL(name)).openConnection()
+      //   val inputStream = connection.getInputStream
+      //   val outputStream = Files.newOutputStream(archivePath)
+      //   val buffer = new Array[Byte](8192)
+      //   var bytesRead = 0
+      //   while ({ bytesRead = inputStream.read(buffer); bytesRead } != -1) {
+      //     outputStream.write(buffer, 0, bytesRead)
+      //   }
+      //   inputStream.close()
+      //   outputStream.close()
+      val curlExit = s"curl -fSL -o ${archivePath.toString} ${ldbcURL(name)}".!
+      if (curlExit != 0) {
+        throw new RuntimeException(
+          s"Failed to download ${ldbcURL(name)} (curl exit code: $curlExit)")
       }
-      inputStream.close()
-      outputStream.close()
       println(s"Uncompressing ${archivePath.toString} to ${dir.toString}...")
       s"zstd -d ${archivePath.toString} -o ${archivePath.toString.replace(".zst", "")}".!
       s"tar -xf ${archivePath.toString.replace(".zst", "")} -C ${dir.toString}".!