updated supported exprs files to include conv and added unit test

Signed-off-by: cindyyuanjiang <[email protected]>

core/src/main/resources/operatorsScore-databricks-aws.csv

@@ -75,6 +75,7 @@ CollectSet,2.45
 Concat,2.45
 ConcatWs,2.45
 Contains,2.45
+Conv,2.45
 Cos,2.45
 Cosh,2.45
 Cot,2.45

core/src/main/resources/operatorsScore-databricks-azure.csv

@@ -75,6 +75,7 @@ CollectSet,2.73
 Concat,2.73
 ConcatWs,2.73
 Contains,2.73
+Conv,2.73
 Cos,2.73
 Cosh,2.73
 Cot,2.73

core/src/main/resources/operatorsScore-dataproc-l4.csv

@@ -75,6 +75,7 @@ CollectSet,4.16
 Concat,4.16
 ConcatWs,4.16
 Contains,4.16
+Conv,4.16
 Cos,4.16
 Cosh,4.16
 Cot,4.16

core/src/main/resources/operatorsScore-dataproc-t4.csv

@@ -75,6 +75,7 @@ CollectSet,4.88
 Concat,4.88
 ConcatWs,4.88
 Contains,4.88
+Conv,4.88
 Cos,4.88
 Cosh,4.88
 Cot,4.88

core/src/main/resources/operatorsScore-emr-a10.csv

@@ -75,6 +75,7 @@ CollectSet,2.59
 Concat,2.59
 ConcatWs,2.59
 Contains,2.59
+Conv,2.59
 Cos,2.59
 Cosh,2.59
 Cot,2.59

core/src/main/resources/operatorsScore-emr-t4.csv

@@ -75,6 +75,7 @@ CollectSet,2.07
 Concat,2.07
 ConcatWs,2.07
 Contains,2.07
+Conv,2.07
 Cos,2.07
 Cosh,2.07
 Cot,2.07

core/src/main/resources/operatorsScore.csv

@@ -80,6 +80,7 @@ CollectSet,4
 Concat,4
 ConcatWs,4
 Contains,4
+Conv,4
 Cos,4
 Cosh,4
 Cot,4

core/src/main/resources/supportedExprs.csv

@@ -127,6 +127,10 @@ ConcatWs,S,`concat_ws`,None,project,result,NA,NA,NA,NA,NA,NA,NA,NA,NA,S,NA,NA,NA
 Contains,S, ,None,project,src,NA,NA,NA,NA,NA,NA,NA,NA,NA,S,NA,NA,NA,NA,NA,NA,NA,NA
 Contains,S, ,None,project,search,NA,NA,NA,NA,NA,NA,NA,NA,NA,PS,NA,NA,NA,NA,NA,NA,NA,NA
 Contains,S, ,None,project,result,S,NA,NA,NA,NA,NA,NA,NA,NA,NA,NA,NA,NA,NA,NA,NA,NA,NA
+Conv,NS,`conv`,This is disabled by default because GPU implementation is incomplete. We currently only support from/to_base values of 10 and 16. We fall back on CPU if the signed conversion is signalled via a negative to_base. GPU implementation does not check for an 64-bit signed/unsigned int overflow when performing the conversion to return `FFFFFFFFFFFFFFFF` or `18446744073709551615` or to throw an error in the ANSI mode. It is safe to enable if the overflow is not possible or detected externally. For instance decimal strings not longer than 18 characters / hexadecimal strings not longer than 15 characters disregarding the sign cannot cause an overflow.          ,project,num,NA,NA,NA,NA,NA,NA,NA,NA,NA,S,NA,NA,NA,NA,NA,NA,NA,NA
+Conv,NS,`conv`,This is disabled by default because GPU implementation is incomplete. We currently only support from/to_base values of 10 and 16. We fall back on CPU if the signed conversion is signalled via a negative to_base. GPU implementation does not check for an 64-bit signed/unsigned int overflow when performing the conversion to return `FFFFFFFFFFFFFFFF` or `18446744073709551615` or to throw an error in the ANSI mode. It is safe to enable if the overflow is not possible or detected externally. For instance decimal strings not longer than 18 characters / hexadecimal strings not longer than 15 characters disregarding the sign cannot cause an overflow.          ,project,from_base,NA,PS,PS,PS,PS,NA,NA,NA,NA,NA,NA,NA,NA,NA,NA,NA,NA,NA
+Conv,NS,`conv`,This is disabled by default because GPU implementation is incomplete. We currently only support from/to_base values of 10 and 16. We fall back on CPU if the signed conversion is signalled via a negative to_base. GPU implementation does not check for an 64-bit signed/unsigned int overflow when performing the conversion to return `FFFFFFFFFFFFFFFF` or `18446744073709551615` or to throw an error in the ANSI mode. It is safe to enable if the overflow is not possible or detected externally. For instance decimal strings not longer than 18 characters / hexadecimal strings not longer than 15 characters disregarding the sign cannot cause an overflow.          ,project,to_base,NA,PS,PS,PS,PS,NA,NA,NA,NA,NA,NA,NA,NA,NA,NA,NA,NA,NA
+Conv,NS,`conv`,This is disabled by default because GPU implementation is incomplete. We currently only support from/to_base values of 10 and 16. We fall back on CPU if the signed conversion is signalled via a negative to_base. GPU implementation does not check for an 64-bit signed/unsigned int overflow when performing the conversion to return `FFFFFFFFFFFFFFFF` or `18446744073709551615` or to throw an error in the ANSI mode. It is safe to enable if the overflow is not possible or detected externally. For instance decimal strings not longer than 18 characters / hexadecimal strings not longer than 15 characters disregarding the sign cannot cause an overflow.          ,project,result,NA,NA,NA,NA,NA,NA,NA,NA,NA,S,NA,NA,NA,NA,NA,NA,NA,NA
 Cos,S,`cos`,None,project,input,NA,NA,NA,NA,NA,NA,S,NA,NA,NA,NA,NA,NA,NA,NA,NA,NA,NA
 Cos,S,`cos`,None,project,result,NA,NA,NA,NA,NA,NA,S,NA,NA,NA,NA,NA,NA,NA,NA,NA,NA,NA
 Cos,S,`cos`,None,AST,input,NA,NA,NA,NA,NA,NA,S,NA,NA,NA,NA,NA,NA,NA,NA,NA,NA,NA

core/src/test/scala/com/nvidia/spark/rapids/tool/planparser/SqlPlanParserSuite.scala

@@ -976,6 +976,36 @@ class SQLPlanParserSuite extends BaseTestSuite {
     }
   }
 
+  test("conv is supported in ProjectExec") {
+    TrampolineUtil.withTempDir { parquetoutputLoc =>
+      TrampolineUtil.withTempDir { eventLogDir =>
+        val (eventLog, _) = ToolTestUtils.generateEventLog(eventLogDir,
+          "ProjectExprsSupported") { spark =>
+          import spark.implicits._
+          val df1 = Seq("10", "-10", "-1", "FFFFFFFFFFFFFFFF").toDF("value")
+          // write df1 to parquet to transform LocalTableScan to ProjectExec
+          df1.write.parquet(s"$parquetoutputLoc/testtext")
+          val df2 = spark.read.parquet(s"$parquetoutputLoc/testtext")
+          // conv should be part of ProjectExec
+          df2.select(conv(df2("value"), 16, 10))
+        }
+        val pluginTypeChecker = new PluginTypeChecker()
+        val app = createAppFromEventlog(eventLog)
+        assert(app.sqlPlans.size == 2)
+        val parsedPlans = app.sqlPlans.map { case (sqlID, plan) =>
+          SQLPlanParser.parseSQLPlan(app.appId, plan, sqlID, "", pluginTypeChecker, app)
+        }
+        val allExecInfo = getAllExecsFromPlan(parsedPlans.toSeq)
+        val wholeStages = allExecInfo.filter(_.exec.contains("WholeStageCodegen"))
+        assert(wholeStages.size == 1)
+        assert(wholeStages.forall(_.duration.nonEmpty))
+        val allChildren = wholeStages.flatMap(_.children).flatten
+        val projects = allChildren.filter(_.exec == "Project")
+        assertSizeAndSupported(1, projects)
+      }
+    }
+  }
+
   test("Parse SQL function Name in HashAggregateExec") {
     TrampolineUtil.withTempDir { eventLogDir =>
       val (eventLog, _) = ToolTestUtils.generateEventLog(eventLogDir, "sqlmetric") { spark =>