[SYSTEMDS-3818] Fix parsing of indexing operations (scalar datatype)

This patch fixes an edge case of print(X[1,]) where the indexing
is mistakenly created with scalar data type because the print accepts
scalar. However, later we introduce print(toString(X[1,])). We now
simply make the parsing more robust as indexing is never scalar
other than forced by internal rewrites.

src/main/java/org/apache/sysds/parser/DMLTranslator.java

@@ -1686,11 +1686,13 @@ private Hop processIndexingExpression(IndexedIdentifier source, DataIdentifier t
 		if (target == null) {
 			target = createTarget(source);
 		}
+
 		//unknown nnz after range indexing (applies to indexing op but also
 		//data dependent operations)
 		target.setNnz(-1); 
 
-		Hop indexOp = new IndexingOp(target.getName(), target.getDataType(), target.getValueType(),
+		DataType dt = target.getDataType().isScalar() ? DataType.MATRIX : target.getDataType();
+		Hop indexOp = new IndexingOp(target.getName(), dt, target.getValueType(),
 			hops.get(source.getName()), ixRange[0], ixRange[1], ixRange[2], ixRange[3],
 			source.getRowLowerEqualsUpper(), source.getColLowerEqualsUpper());
 

src/test/scripts/functions/rewrite/RewriteNonScalarPrint.dml

@@ -36,8 +36,7 @@ else if(type==3){   # standard list case
     print(A_list)
 }
 else if(type==4){   # slice row from matrix
-    A_row = A[1,]
-    print(A_row)    # print(A[1,]) produces incorrect output
+    print(A[1,])
 }
 else if(type==5){   # slice column from matrix
     A_col = A[,1]