For Text/Varchar/Char/Bpchar columns, we should ignore generating StatsBuckets in DXL

Instead we should maintain NDVRemain and NullFreq to do Cardinality
Estimation.

Adding function to check if we need to create stats bucket in DXL

Function `FCreateStatsBucket` returns true if column data type is
not a text/varchar/char/bpchar type.
Signed-off-by: NEkta Khanna <ekhanna@pivotal.io>

src/backend/gpopt/translate/CTranslatorRelcacheToDXL.cpp

@@ -2366,32 +2366,42 @@ CTranslatorRelcacheToDXL::PimdobjColStats
 					&pdrgdatumHistValues, &iNumHistValues,
 					NULL, NULL);
 
-	// transform all the bits and pieces from pg_statistic
-	// to a single bucket structure
-	DrgPdxlbucket *pdrgpdxlbucketTransformed =
-			PdrgpdxlbucketTransformStats
-					(
-					pmp,
-					oidAttType,
-					dDistinct,
-					dNullFrequency,
-					pdrgdatumMCVValues,
-					pdrgfMCVFrequencies,
-					ULONG(iNumMCVValues),
-					pdrgdatumHistValues,
-					ULONG(iNumHistValues)
-					);
-
-	GPOS_ASSERT(NULL != pdrgpdxlbucketTransformed);
-
 	CDouble dNDVBuckets(0.0);
 	CDouble dFreqBuckets(0.0);
-	const ULONG ulBuckets = pdrgpdxlbucketTransformed->UlLength();
-	for (ULONG ul = 0; ul < ulBuckets; ul++)
-	{
-	   CDXLBucket *pdxlbucket = (*pdrgpdxlbucketTransformed)[ul];
-	   dNDVBuckets = dNDVBuckets + pdxlbucket->DDistinct();
-	   dFreqBuckets = dFreqBuckets + pdxlbucket->DFrequency();
+
+	// We only want to create statistics buckets if the column is NOT a text, varchar, char or bpchar type
+	// For the above column types we will use NDVRemain and NullFreq to do cardinality estimation.
+
+	if (CTranslatorUtils::FCreateStatsBucket(oidAttType))
+	{
+		// transform all the bits and pieces from pg_statistic
+		// to a single bucket structure
+		DrgPdxlbucket *pdrgpdxlbucketTransformed =
+		PdrgpdxlbucketTransformStats
+		(
+		 pmp,
+		 oidAttType,
+		 dDistinct,
+		 dNullFrequency,
+		 pdrgdatumMCVValues,
+		 pdrgfMCVFrequencies,
+		 ULONG(iNumMCVValues),
+		 pdrgdatumHistValues,
+		 ULONG(iNumHistValues)
+		 );
+
+		GPOS_ASSERT(NULL != pdrgpdxlbucketTransformed);
+
+		const ULONG ulBuckets = pdrgpdxlbucketTransformed->UlLength();
+		for (ULONG ul = 0; ul < ulBuckets; ul++)
+		{
+			CDXLBucket *pdxlbucket = (*pdrgpdxlbucketTransformed)[ul];
+			dNDVBuckets = dNDVBuckets + pdxlbucket->DDistinct();
+			dFreqBuckets = dFreqBuckets + pdxlbucket->DFrequency();
+		}
+
+		CUtils::AddRefAppend(pdrgpdxlbucket, pdrgpdxlbucketTransformed);
+		pdrgpdxlbucketTransformed->Release();
 	}
 
 	// there will be remaining tuples if the merged histogram and the NULLS do not cover
@@ -2406,16 +2416,12 @@ CTranslatorRelcacheToDXL::PimdobjColStats
  		dFreqRemain = std::max(CDouble(0.0), (1 - dFreqBuckets - dNullFrequency));
 	}
 
-	CUtils::AddRefAppend(pdrgpdxlbucket, pdrgpdxlbucketTransformed);
-
 	// free up allocated datum and float4 arrays
 	gpdb::FreeAttrStatsSlot(oidAttType, pdrgdatumMCVValues, iNumMCVValues, pdrgfMCVFrequencies, iNumMCVFrequencies);
 	gpdb::FreeAttrStatsSlot(oidAttType, pdrgdatumHistValues, iNumHistValues, NULL, 0);
 
 	gpdb::FreeHeapTuple(heaptupleStats);
 
-	pdrgpdxlbucketTransformed->Release();
-
 	// create col stats object
 	pmdidColStats->AddRef();
 	CDXLColStats *pdxlcolstats = GPOS_NEW(pmp) CDXLColStats

src/backend/gpopt/translate/CTranslatorUtils.cpp

@@ -2806,4 +2806,18 @@ CTranslatorUtils::UlNonSystemColumns
 	return ulNonSystemCols;
 }
 
+// Function to check if we should create stats bucket in DXL
+// Returns true if column datatype is not text/char/varchar/bpchar
+BOOL
+CTranslatorUtils::FCreateStatsBucket
+	(
+	OID oidAttType
+	)
+{
+	if (oidAttType != TEXTOID && oidAttType != CHAROID && oidAttType != VARCHAROID && oidAttType != BPCHAROID)
+		return true;
+
+	return false;
+}
+
 // EOF

src/include/gpopt/translate/CTranslatorUtils.h

@@ -410,6 +410,10 @@ namespace gpdxl
 			// return the count of non-system columns in the relation
 			static
 			ULONG UlNonSystemColumns(const IMDRelation *pmdrel);
+
+			// check if we need to create stats buckets in DXL for the column attribute
+			static
+			BOOL FCreateStatsBucket(OID oidAttType);
 	};
 }