Support StringDecode for GBK encoding

docs/additional-functionality/advanced_configs.md

@@ -171,6 +171,7 @@ Name | Description | Default Value | Applicable at
 <a name="sql.window.range.int.enabled"></a>spark.rapids.sql.window.range.int.enabled|When the order-by column of a range based window is int type and the range boundary calculated for a value has overflow, CPU and GPU will get the different results. When set to false disables the range window acceleration for the int type order-by column|true|Runtime
 <a name="sql.window.range.long.enabled"></a>spark.rapids.sql.window.range.long.enabled|When the order-by column of a range based window is long type and the range boundary calculated for a value has overflow, CPU and GPU will get the different results. When set to false disables the range window acceleration for the long type order-by column|true|Runtime
 <a name="sql.window.range.short.enabled"></a>spark.rapids.sql.window.range.short.enabled|When the order-by column of a range based window is short type and the range boundary calculated for a value has overflow, CPU and GPU will get the different results. When set to false disables the range window acceleration for the short type order-by column|false|Runtime
+
 ## Supported GPU Operators and Fine Tuning
 _The RAPIDS Accelerator for Apache Spark_ can be configured to enable or disable specific
 GPU accelerated expressions.  Enabled expressions are candidates for GPU execution. If the
@@ -385,6 +386,7 @@ Name | SQL Function(s) | Description | Default Value | Notes
 <a name="sql.expression.Sqrt"></a>spark.rapids.sql.expression.Sqrt|`sqrt`|Square root|true|None|
 <a name="sql.expression.Stack"></a>spark.rapids.sql.expression.Stack|`stack`|Separates expr1, ..., exprk into n rows.|true|None|
 <a name="sql.expression.StartsWith"></a>spark.rapids.sql.expression.StartsWith| |Starts with|true|None|
+<a name="sql.expression.StringDecode"></a>spark.rapids.sql.expression.StringDecode| |Decodes binary data from a charset to a UTF-8 string|true|None|
 <a name="sql.expression.StringInstr"></a>spark.rapids.sql.expression.StringInstr|`instr`|Instr string operator|true|None|
 <a name="sql.expression.StringLPad"></a>spark.rapids.sql.expression.StringLPad| |Pad a string on the left|true|None|
 <a name="sql.expression.StringLocate"></a>spark.rapids.sql.expression.StringLocate|`locate`, `position`|Substring search operator|true|None|

docs/supported_ops.md

@@ -8267,7 +8267,7 @@ are limited.
 <td> </td>
 <td> </td>
 <td> </td>
-<td><em>PS<br/>UTC is only supported TZ for child TIMESTAMP;<br/>unsupported child types BINARY, CALENDAR, UDT, DAYTIME, YEARMONTH</em></td>
+<td><em>PS<br/>UTC is only supported TZ for child TIMESTAMP;<br/>unsupported child types CALENDAR, UDT, DAYTIME, YEARMONTH</em></td>
 <td> </td>
 <td> </td>
 <td> </td>
@@ -8290,7 +8290,7 @@ are limited.
 <td> </td>
 <td> </td>
 <td> </td>
-<td><em>PS<br/>UTC is only supported TZ for child TIMESTAMP;<br/>unsupported child types BINARY, CALENDAR, UDT, DAYTIME, YEARMONTH</em></td>
+<td><em>PS<br/>UTC is only supported TZ for child TIMESTAMP;<br/>unsupported child types CALENDAR, UDT, DAYTIME, YEARMONTH</em></td>
 <td> </td>
 <td> </td>
 <td> </td>
@@ -16773,6 +16773,80 @@ are limited.
 <th>YEARMONTH</th>
 </tr>
 <tr>
+<td rowSpan="3">StringDecode</td>
+<td rowSpan="3"> </td>
+<td rowSpan="3">Decodes binary data from a charset to a UTF-8 string</td>
+<td rowSpan="3">None</td>
+<td rowSpan="3">project</td>
+<td>bin</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>S</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+</tr>
+<tr>
+<td>charset</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td><em>PS<br/>Literal value only</em></td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+</tr>
+<tr>
+<td>result</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>S</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+</tr>
+<tr>
 <td rowSpan="3">StringInstr</td>
 <td rowSpan="3">`instr`</td>
 <td rowSpan="3">Instr string operator</td>

integration_tests/src/main/python/string_test.py

@@ -28,6 +28,28 @@
 
 _regexp_conf = { 'spark.rapids.sql.regexp.enabled': 'true' }
 
+def test_string_decode_gbk_basic():
+    """Test GBK charset decoding with known Chinese characters, ASCII, mixed content, and nulls."""
+    data = [
+        (bytearray(b'\xc4\xe3\xba\xc3'),),              # 你好
+        (bytearray(b'\xca\xc0\xbd\xe7'),),              # 世界
+        (bytearray(b'Hello'),),                           # Pure ASCII
+        (bytearray(b'Hi\xc4\xe3\xba\xc3World'),),       # Mixed ASCII + Chinese
+        (bytearray(b''),),                                # Empty
+        (None,),                                          # Null
+        (bytearray(b'\xff\xff'),),                        # Invalid bytes
+        (bytearray(b'\x81'),),                            # Truncated lead byte
+    ]
+    assert_gpu_and_cpu_are_equal_collect(
+        lambda spark: spark.createDataFrame(data, "bin binary").selectExpr(
+            "decode(bin, 'GBK')"))
+
+def test_string_decode_gbk_random():
+    """Test GBK decoding with random binary data to exercise error handling paths."""
+    assert_gpu_and_cpu_are_equal_collect(
+        lambda spark: unary_op_df(spark, BinaryGen(min_length=0, max_length=50), length=2048)
+            .selectExpr("decode(a, 'GBK')"))
+
 def mk_str_gen(pattern):
     return StringGen(pattern).with_special_case('').with_special_pattern('.{0,10}')
 

sql-plugin/src/main/scala/com/nvidia/spark/rapids/GpuOverrides.scala

@@ -4184,7 +4184,8 @@ object GpuOverrides extends Logging {
   val expressions: Map[Class[_ <: Expression], ExprRule[_ <: Expression]] =
     commonExpressions ++ TimeStamp.getExprs ++ GpuHiveOverrides.exprs ++
         ZOrderRules.exprs ++ DecimalArithmeticOverrides.exprs ++
-        BloomFilterShims.exprs ++ InSubqueryShims.exprs ++ RaiseErrorShim.exprs ++
+        BloomFilterShims.exprs ++ StringDecodeShims.exprs ++
+        InSubqueryShims.exprs ++ RaiseErrorShim.exprs ++
         ExternalSource.exprRules ++ SparkShimImpl.getExprs
 
   def wrapScan[INPUT <: Scan](

sql-plugin/src/main/scala/com/nvidia/spark/rapids/objectsExpressions.scala

@@ -16,25 +16,69 @@
 
 package com.nvidia.spark.rapids
 
+import org.apache.spark.sql.catalyst.expressions.Expression
 import org.apache.spark.sql.catalyst.expressions.objects.StaticInvoke
-import org.apache.spark.sql.rapids.ExternalSource
+import org.apache.spark.sql.rapids.{ExternalSource, GpuStringDecode}
 
 /**
  * Meta class for overriding StaticInvoke expressions.
  * <br/>
- * When writing to partitioned table, iceberg needs to compute the partition values based on the 
- * partition spec using [[StaticInvoke]] expression.
+ * Handles two cases:
+ * - Iceberg partition value computation via StaticInvoke
+ * - Spark 4.0+ StringDecode (RuntimeReplaceable replaced by StaticInvoke)
  */
 class StaticInvokeMeta(expr: StaticInvoke,
   conf: RapidsConf,
   parent: Option[RapidsMeta[_, _, _]],
   rule: DataFromReplacementRule) extends ExprMeta[StaticInvoke](expr, conf, parent, rule) {
 
+  private val isStringDecode: Boolean = {
+    expr.staticObject.getName == "org.apache.spark.sql.catalyst.expressions.StringDecode" &&
+      expr.functionName == "decode"
+  }
+
+  private var charsetName: String = null
+
+  override val childExprs: Seq[BaseExprMeta[_]] = if (isStringDecode) {
+    // StringDecode StaticInvoke: decode(bin, charset, legacyCharsets, legacyErrorAction)
+    // Only wrap the first child (bin) as the GPU expression input
+    expr.arguments.take(1).map(GpuOverrides.wrapExpr(_, conf, Some(this)))
+  } else {
+    Seq.empty
+  }
+
   override def tagExprForGpu(): Unit = {
-    ExternalSource.tagForGpu(expr, this)
+    if (isStringDecode) {
+      tagStringDecode()
+    } else {
+      ExternalSource.tagForGpu(expr, this)
+    }
+  }
+
+  private def tagStringDecode(): Unit = {
+    if (expr.arguments.size < 2) {
+      willNotWorkOnGpu("StringDecode StaticInvoke has unexpected argument count")
+      return
+    }
+    // charset is the second argument, must be a foldable string literal
+    val charsetExpr = expr.arguments(1)
+    GpuOverrides.extractLit(charsetExpr).map(_.value) match {
+      case Some(cs: org.apache.spark.unsafe.types.UTF8String) if cs != null =>
+        charsetName = cs.toString.toUpperCase
+        if (charsetName != "GBK") {
+          willNotWorkOnGpu(s"only GBK charset is supported on GPU, got: $charsetName")
+        }
+      case _ =>
+        willNotWorkOnGpu("charset must be a string literal for GPU StringDecode")
+    }
   }
 
   override def convertToGpuImpl(): GpuExpression = {
-    ExternalSource.convertToGpu(expr, this)
+    if (isStringDecode) {
+      val bin = childExprs.head.convertToGpu().asInstanceOf[Expression]
+      GpuStringDecode(bin, charsetName)
+    } else {
+      ExternalSource.convertToGpu(expr, this)
+    }
   }
-}
+}

sql-plugin/src/main/scala/org/apache/spark/sql/rapids/stringFunctions.scala

@@ -30,6 +30,7 @@ import com.nvidia.spark.rapids.Arm._
 import com.nvidia.spark.rapids.RapidsPluginImplicits._
 import com.nvidia.spark.rapids.jni.{Arithmetic, RoundMode}
 import com.nvidia.spark.rapids.jni.CastStrings
+import com.nvidia.spark.rapids.jni.CharsetDecode
 import com.nvidia.spark.rapids.jni.GpuSubstringIndexUtils
 import com.nvidia.spark.rapids.jni.NumberConverter
 import com.nvidia.spark.rapids.jni.RegexRewriteUtils
@@ -2516,3 +2517,22 @@ case class GpuFormatNumber(x: Expression, d: Expression)
     }
   }
 }
+
+case class GpuStringDecode(bin: Expression, charsetName: String)
+  extends GpuUnaryExpression with ImplicitCastInputTypes with NullIntolerantShim {
+
+  override def child: Expression = bin
+
+  override def dataType: DataType = StringType
+
+  override def inputTypes: Seq[AbstractDataType] = Seq(BinaryType)
+
+  override def doColumnar(input: GpuColumnVector): ColumnVector = {
+    val charsetId = charsetName match {
+      case "GBK" => CharsetDecode.GBK
+      case other =>
+        throw new UnsupportedOperationException(s"Unsupported charset on GPU: $other")
+    }
+    CharsetDecode.decode(input.getBase, charsetId)
+  }
+}

sql-plugin/src/main/spark321/scala/com/nvidia/spark/rapids/shims/StringDecodeShims.scala

@@ -0,0 +1,74 @@
+/*
+ * Copyright (c) 2026, NVIDIA CORPORATION.
+ *
+ * Licensed 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.
+ */
+
+/*** spark-rapids-shim-json-lines
+{"spark": "321"}
+{"spark": "330"}
+{"spark": "330db"}
+{"spark": "331"}
+{"spark": "332"}
+{"spark": "332db"}
+{"spark": "333"}
+{"spark": "334"}
+{"spark": "340"}
+{"spark": "341"}
+{"spark": "341db"}
+{"spark": "342"}
+{"spark": "343"}
+{"spark": "344"}
+{"spark": "350"}
+{"spark": "350db143"}
+{"spark": "351"}
+{"spark": "352"}
+{"spark": "353"}
+{"spark": "354"}
+{"spark": "355"}
+{"spark": "356"}
+{"spark": "357"}
+{"spark": "358"}
+spark-rapids-shim-json-lines ***/
+package com.nvidia.spark.rapids.shims
+
+import com.nvidia.spark.rapids._
+
+import org.apache.spark.sql.catalyst.expressions.{Expression, StringDecode}
+import org.apache.spark.sql.rapids.GpuStringDecode
+
+object StringDecodeShims {
+  val exprs: Map[Class[_ <: Expression], ExprRule[_ <: Expression]] = Seq(
+    GpuOverrides.expr[StringDecode](
+      "Decodes binary data from a charset to a UTF-8 string",
+      ExprChecks.binaryProject(TypeSig.STRING, TypeSig.STRING,
+        ("bin", TypeSig.BINARY, TypeSig.BINARY),
+        ("charset", TypeSig.lit(TypeEnum.STRING), TypeSig.STRING)),
+      (a, conf, p, r) => new BinaryExprMeta[StringDecode](a, conf, p, r) {
+        private var charsetName: String = _
+        override def tagExprForGpu(): Unit = {
+          GpuOverrides.extractStringLit(a.charset) match {
+            case Some(cs) if cs != null =>
+              charsetName = cs.toUpperCase
+              if (charsetName != "GBK") {
+                willNotWorkOnGpu(s"only GBK charset is supported on GPU, got: $charsetName")
+              }
+            case _ =>
+              willNotWorkOnGpu("charset must be a string literal for GPU StringDecode")
+          }
+        }
+        override def convertToGpu(lhs: Expression, rhs: Expression): GpuExpression =
+          GpuStringDecode(lhs, charsetName)
+      })
+  ).map(r => (r.getClassFor.asSubclass(classOf[Expression]), r)).toMap
+}

sql-plugin/src/main/spark400/scala/com/nvidia/spark/rapids/shims/StringDecodeShims.scala

@@ -0,0 +1,35 @@
+/*
+ * Copyright (c) 2026, NVIDIA CORPORATION.
+ *
+ * Licensed 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.
+ */
+
+/*** spark-rapids-shim-json-lines
+{"spark": "400"}
+{"spark": "400db173"}
+{"spark": "401"}
+{"spark": "402"}
+{"spark": "411"}
+spark-rapids-shim-json-lines ***/
+package com.nvidia.spark.rapids.shims
+
+import com.nvidia.spark.rapids.ExprRule
+
+import org.apache.spark.sql.catalyst.expressions.Expression
+
+// Spark 4.0+ changed StringDecode to RuntimeReplaceable, which is replaced by
+// StaticInvoke(StringDecode.class, "decode", ...) during analysis.
+// GPU acceleration is handled in StaticInvokeMeta (objectsExpressions.scala).
+object StringDecodeShims {
+  val exprs: Map[Class[_ <: Expression], ExprRule[_ <: Expression]] = Map.empty
+}

tools/generated_files/330/operatorsScore.csv

@@ -263,6 +263,7 @@ StartsWith,4
 StaticInvoke,4
 StddevPop,4
 StddevSamp,4
+StringDecode,4
 StringInstr,4
 StringLPad,4
 StringLocate,4

tools/generated_files/330/supportedExprs.csv

@@ -570,6 +570,9 @@ Stack,S,`stack`,None,project,result,NA,NA,NA,NA,NA,NA,NA,NA,NA,NA,NA,NA,NA,NA,PS
 StartsWith,S, ,None,project,src,NA,NA,NA,NA,NA,NA,NA,NA,NA,S,NA,NA,NA,NA,NA,NA,NA,NA,NA,NA
 StartsWith,S, ,None,project,search,NA,NA,NA,NA,NA,NA,NA,NA,NA,PS,NA,NA,NA,NA,NA,NA,NA,NA,NA,NA
 StartsWith,S, ,None,project,result,S,NA,NA,NA,NA,NA,NA,NA,NA,NA,NA,NA,NA,NA,NA,NA,NA,NA,NA,NA
+StringDecode,S, ,None,project,bin,NA,NA,NA,NA,NA,NA,NA,NA,NA,NA,NA,NA,S,NA,NA,NA,NA,NA,NA,NA
+StringDecode,S, ,None,project,charset,NA,NA,NA,NA,NA,NA,NA,NA,NA,PS,NA,NA,NA,NA,NA,NA,NA,NA,NA,NA
+StringDecode,S, ,None,project,result,NA,NA,NA,NA,NA,NA,NA,NA,NA,S,NA,NA,NA,NA,NA,NA,NA,NA,NA,NA
 StringInstr,S,`instr`,None,project,str,NA,NA,NA,NA,NA,NA,NA,NA,NA,S,NA,NA,NA,NA,NA,NA,NA,NA,NA,NA
 StringInstr,S,`instr`,None,project,substr,NA,NA,NA,NA,NA,NA,NA,NA,NA,PS,NA,NA,NA,NA,NA,NA,NA,NA,NA,NA
 StringInstr,S,`instr`,None,project,result,NA,NA,NA,S,NA,NA,NA,NA,NA,NA,NA,NA,NA,NA,NA,NA,NA,NA,NA,NA

tools/generated_files/331/operatorsScore.csv

@@ -264,6 +264,7 @@ StartsWith,4
 StaticInvoke,4
 StddevPop,4
 StddevSamp,4
+StringDecode,4
 StringInstr,4
 StringLPad,4
 StringLocate,4